Luật Chơi Among Us

He really did hit the ground running. I worked for 48 years at the school and saw many interesting students graduate. Among them were some that went on to become scientists and arctic explorers. None of them ended up as a successful as Oleg, however.

By eighth grade, some of the kids were trying vodka. Oleg was doing business. My students told me about how Oleg had brought cosmetics back from the Baltics to sell. Oleg traded, exported and imported. At the time, neither his fellow students nor we teachers took these activities seriously.

Chapter 5

The Border Guard, not the Sports Club

When I was working at the mine, all I could think about was the coming Spring-because I hoped to be accepted, then, into the Army Sports Club. If I could not achieve that, then I would have to be a soldier. It was now that my coach Ivan Stepanovich failed me, for the first and probably the last time. I do not resent him for it now. Everything always works out fine in the end. He promised that he would get me a spot in the ASC, but there was only one vacancy. During the 1986 spring draft, it turned out I was not the only athlete born in 1967. There was another, the son of the director of the Novosibirsk ASC. Instead of selecting me, Oleg Tinkov, then-champion of the city and of the Kuznetsk Basin, winner of more than one competition-this son was admitted to the Novosibirsk ASC. He was admitted, too, even though I could have out-pedaled him with one leg.

The strong arm of the old-boy network has always held sway in Russia and always will. Thus I missed out. In April 1986, I was taken into the army and, instead of getting into the ASC, I ended up among the border troops, under the control of the USSR’s KGB. My career as a cyclist, which had held such great promise, ground to a halt. Everything that I did later-my return to sports in 2005-2006, my participation in races, the creation of the Tinkoff cycling team, and the cycling trips to Italy where we cover 4000 km in a month-is connected with the fact that my career was cut short at that time.

As we were being loaded into the railway coach, I looked out the window and saw the new spring growth. It reminded me of a song written by Vladimir Semyonovich Vysotsky:

Spring has just begun,

Most of the people are still indoors,

But I had to get out there, –

Then two arrived suddenly,

With a convoy, with a convoy.

“Get dressed,” they say, “come out!”

I begged the sergeant:

“Let me stay with the Spring!”

But the sergeant took me away. First, we took the train to Krasnoyarsk. From there we flew to Vlapostok and then took the bus to Nakhodka.

When we arrived at the unit, the dischargees took all our food from us. We did not resist. At first we were brisk with them. But they replied: “Let’s get you shaved, then we’ll talk”. We had our hair cut and then were sent to the sauna. Afterwards we were given submachine guns and then mounted Kalashnikov machine guns. This type of gun is very heavy but fires better than the others. It is hard to miss when using it, in contrast to the hand-held variety. We ran cross-country, but the 3-km sprints were easy for me: when I was cycling, after all, we would run 15-20 km during our winter practices.

We would sweat so hard from the heat and the pssure that the fabric of our uniforms would start to dissolve. Within three months they would tear. According to regulations we were supposed to get new clothing every six months, but ours had to be replaced more often. We had to run, jump, and blow things up. Explosions rumbled from all sides-flash! Bang! It is depssing to read about what goes on in the army nowadays. Serious training has been replaced by hazing. The goal then, was to make us into real border guards. If there had been a war we really would have protected our country. The army-or at least the border guard-was really and truly combat-ready.

The rumors about hazing in the Soviet army are highly exaggerated. Things were not all that bad. Believe it or not, we had nothing of the sort in the border guard. It all depended on your officers, their qualifications, and their approach to training. Sure, the army has its hierarchy: I washed the floors, of course, and the senior officers did not. But I was never-not once-punched in the two years that I was there. The worst I suffered was a shove or a minor kick in the butt if I was moving to slow. I was never beat up.

A lot of my friends ended up in the Afghan war and I might have been sent there myself. It was a choice between Afghanistan and the border. Our Fatherland decided that, being 190 cm (6′ 4″) in height, I would be of more use at the border. Three of my classmates who served in Afghanistan came home ruined men. The war taught them how to smoke marijuana and how to do a lot of other things. Up until 1986, when I was in Siberia, I did not know what drugs were. When I got back from the army, though, pot had become commonplace, distributed by veterans of that war. Heroin was next.

Healthy, young Siberian men were sent in bundles to the border and into Afghanistan. Less were taken from St. Petersburg and Moscow. I have heard terrible stories from friends about how much they were beaten and about how much they drank. The border guard was completely different. Everything was done with pcision and exactitude.

I was admitted to the sergeant training school. For six months, from April to October 1986, I trained to become junior sergeant. These were some of the most trying times of my life, taxing both physically and psychologically. I even considered suicide at one point. After that-until, perhaps, I was training for a race in 2005 (of which more later)-life was never so hard again.

A sergeant lives the good life: no washing floors, no kitchen duty and therefore no dirty dishwater. One day, though, I was stupid and told an officer to screw off. I had my sergeant’s badges ripped off and had to wash the kitchen floors and deal with the trash. This was unpleasant, to say the least, considering that I was in my second and last year of mandatory service. I did get my badges back, though, before being discharged.

Some of you may remember Mikhail Gorbachev’s famous trip to the Far East, to Nakhodka, Vlapostok, in June 1986. He even came to our unit, although I did not see him myself. Our unit was considered an elite one. It was called Nakhodka Independent Identity Check Point and was at the border at Vostochny Port. Because the country’s top official was coming, our combat training all but stopped for two months. Instead, we practiced for Gorbachev’s arrival. What did this involve?

I am not sure how things are today, but in the Soviet Army I was emotionally tormented. It was really a hard thing to take. One would get up in the morning and the clock would be ticking. Instead of socks, you would wrap cloth around your feet and pull on your boots. The blisters would burst and they would have to cut of the wrapping in the medical station. This happened to everyone. Stand, squat, get on the ground, push up! Silence, soldier! Shut up! My calluses got so thick from the binding and the canvas boots that today I can wear new shoes without the slightest discomfort.

My friend Oleg Kakovin, a guy from Leninsk, just could not let go. For three months, he kept fighting with the warrant officers. He would swear at them, but it was pointless. The system was built on the oppssion and destruction of the inpidual. Soldiers are rivets. The goal is to make the person into nothing. When there is no “I,” the color khaki pervades all. As the Soviet rock group Nautilus Pompilius sang: “I haven’t seen a scarier crowd than a crowd dressed in khaki.” Perhaps the same thing happens in western armies.

If my friend Oleg argued or told the sergeant to screw himself, the command would ring out: “Line up, platoon! Three kilometers running. March!” We’d all start running and tell Oleg to cut it out. Or we would be given the order to get up and all of us-except Oleg-would be on our feet. The punishment would be the same: a three-kilometer battle march. But when he refused to obey a third time and we were all forced to run three kilometers as a result, that made us angry. Everyone ran up to him and started kicking him. So you would give up resisting whether you liked it or not. We were disciplined as a group. You simply could not be mouthy or 30 people would get punished on your account. In the end you would suffer. This was very effective psychology. Thank God, I never used this approach in my business relations.

My responsibilities as a border guard included inspecting foreign ships as they were coming into port. We made sure that no one crossed the state border unnoticed and undocumented. It never happened. How crazy would you have to be to want to come illegally into the USSR? There were some that wanted to leave without permission, of course, but it was absurd to think that anyone would want to sneak in! Some Koreans, Japanese, and Chinese did come in legally. We would conduct a formal inspection. Going through the cabins, we saw foreign objects-jeans and magazines-and examined them closely.

This was how I learned my first few words in English:

“Open this.”

” Please, go one by one.”

“Show me your passport, please.”

“Please, open the door.“

In our unit there were two brothers from Bryansk, one of which was a snitch. I beat him up for tattling to our superior, Colonel Zubr, who was scary as hell. If you let one word slip, he would make you dig shit out of the toilet all day long. It was terror: you had to keep your mouth shut and your eyes on the floor. I left the army twenty-two years ago. But if I bumped into Zubr today, I would punch that pig’s face in. Imagine what you would have to put a person through to make him want to kick your head in, twenty years later? And I am not even a resentful guy. I have no hard feelings towards people who have screwed me over, once a year or so has gone by.

My second year of service was easier, but more hectic too. For the two years I was there I took a break from speculation. After all, it was impossible to sell stuff in the army. My military service marks the only period, two years in all, when I have no memory of doing business. My green shoulder marks kept me barricaded from it. I suddenly had new priorities. I wanted a border guard marks of excellence: level one, level two… The army sucked me in and I found myself working to build my military career, even considering staying on as a warrant officer.

Before I was discharged in April 1988, three of my friends and I applied to serve further. Later, I was even summoned. One of the officers convinced us: what was there to do as civilians? A wide-scale restructuring of the system had begun while we were gone. Everything was a mess. Factory workers were not getting their pay. By contrast, the army meant stability. We were promised full government support, food, and a monthly salary of 200‑250 rubles. Imagine how things would have gone if I had signed that contract! I would be a moron now, a senior warrant officer or captain, stationed in a place like Nikolayevsk-on-Amur.

Border guard captain Tinkov!

But the hand of God was at work. A voice inside me said, “No, Oleg. Go home.” I came to Captain Sayakhov and told him,

“Comrade Captain, I want to take back my application.”

“What! Are you nuts? We’ve already sent everything to Vlapostok and you’ve been approved.”

“I don’t want to be a warrant officer.”

The captain called me an idiot and I was dismissed.

At the same time, Providence protected me from membership in the Communist Party. If you wanted to become a warrant officer you had to be a communist. I even wrote a membership application.

After my discharge, two friends and I flew from Khabarovsk to Kemerovo. From there I needed to make my way home to Leninsk. Gorbachev’s battle with alcoholism was in full swing and it was nearly impossible to buy vodka. We waited in line, bought a bottle, and shared it between the three of us. I got sick, collapsed, and fell asleep right at the station in Kemerovo. I did not manage to get home that night. It was a strange picture: a soldier in full parade uniform with regalia, asleep under a bench in a puddle of vomit. Having slept it off, I went to the house of one of my co-dischargees who lived in Kemerovo. I got a brush from him, cleaned off my clothes, and left for Leninsk.

The army is simply bad, but I am happy that I enlisted. The experience hardened me physically and emotionally. The army is not for the faint-hearted and it is not for little girls. On the whole it was a negative experience and I would rather not go into too many details. Still, I would recommend military service to anyone: you really gain a lot from it and return, ultimately, a new person. It is easy for me to say this, I suppose, since I have already been through it myself. Believe it or not, I can discern to a high degree of accuracy if someone has been in the army-just by having a conversation with him. People who have served have a soundness of mind at their core. In a similar way, I can tell that a young woman is Russian based on a number of factors, no matter how well she speaks another language. I wrote about this in my blog once, which led to a lot of heated discussion.

Let me tell you an interesting story. On February 23, Fatherland Defenders’ day, which is, traditionally, a holiday when women congratulate men for being-well, men-about 200 people filed into our auditorium. I opened the event in the customary way, with a speech. I started with the words “A display of prowess.” I spoke to those gathered about the army, but they seemed perplexed or oblivious. So I asked: “Which of you guys served in the army, anyway?” There were about 150 men psent. I was expecting half of them to raise their hands-or maybe twenty, fifteen percent at least. Imagine how shocked I was when only three hands went up. I felt sick, quickly wrapped up my speech and left.

We have a lot of double standards in Russia. The spectacle of white-collar workers, none of whom served in the army, celebrating February 23 in Moscow offices-and with a bang-brings the word duplicity to mind. Either refrain from celebrating the holiday at all, or just call it “Men’s Day,” as though it were the male counterpart of Women’s Day on March 8. These days hardly anyone serves in the army. Who knows if this is good or bad?

One way or another, on May 28-Border Guard Day-I arrived back in Leninsk-Kuznetsky. This holiday is celebrated completely differently now than it was then. You can see crazy drunks in green berets running down the streets of Moscow and St. Petersburg, yelling, singing, and jumping in the fountains.

Dear reader, I solemnly declare that I am not one of them!

Even though I remember well our battle song:

A border guard must follow cruel laws,

We can’t sleep when others are asleep,

You and I, pal, we’re back on duty.

A border guard must follow cruel laws.

We were taught that the USSR border was sacred and untouchable. We had been warned about the hazing, but none of it happened in our unit. Sure, we washed the floors, while our senior officers did not, but I never got beat up in the two years that I was there. This letter contains the phrase “the doors of any educational institution are open to you.” I didn’t take it seriously. Oleg Ikonnikov, cyclist:

Oleg and I practiced together, went to training camps, and passed the qualifiers for Master of Sport candidacy. He was a very strong finisher. He’d win the criterium and the group races. He could keep a good pace and hold his own in a team when we went to Russia-wide competitions. Even though his age ought to have put him in the juniors he competed with the grown men. He could have won among the juniors, but he was put into a more senior team to fill in the gaps. If he had stuck to it, he would have achieved great heights. The conditions would have to have been different and the coaches more professional, like in Omsk or Kuibyshev. Whoever got into the Army Sports Club (in Omsk) ended up succeeding. If Oleg hadn’t stopped training, he could have been a champion. He has the talent.

It was hard to get into the ASC. You had to get to Omsk, first, then fall on your knees in front of the coach and beg him to take you. Apparently Oleg wasn’t too interested. The team in Omsk was strong. From there you would get into the Central ASC or into the All-Union team. Once the nineties came, getting into the Omsk ASC was your ticket into professional teams. Of course, back when Oleg and I were training, there was no such opportunity. The ASC in Novosibirsk, though, was run like in a village. There really was no good reason to go there, except as an excuse not to join the army. Coaches should do all they can to keep a hold of and promote athletes like Oleg. But we had to promote ourselves, otherwise nothing would work. But that’s supposed to be the coach’s job.

Chapter 6

There Will Be No Wildfire

I came home from the army certain that I would work in the mines.

I got a call from the Committee for State Security (KGB) immediately. They wanted to recruit me. Because the Border Guard was administered by the committee they told me, “You’re already one of us!” If you paid attention to what I said about the values my father instilled in me, however, you will understand why I politely declined.

I was going to be a miner like my dad! He had just retired, so I pd I could take his place in Kirov Mine. I went and put in an application to work there. But I also thought about how nice it would be to take a vacation beforehand.

I happened to bump into my homeroom teacher from school. She told me she was going to work as director of a Pioneers Camp and asked me if I would like to go with her to get some rest. There was a teachers’ college in our town, which ppared future pschool and primary school teachers. Before placement, the teachers had to complete internships as Pioneers’ counselors at a Young Builder camp belonging to a construction trust located in Leninsk-Kuznetsky. “You’re an athlete. Why not come teach phys ed?” my homeroom teacher asked. I agreed and worked there all summer. I would go to work in the mines in September.

Looking back, I think that June 1988 was the happiest month of my life. It turned out that there were only two men in the whole camp: myself and the art director. The artist painted posters with logos along the lines of “Pioneers ahead!” Unlike me, a good-looking chap recently “emancipated” from the border guard, he enjoyed no success of any kind among the women. If you count all the medics, management, and counselors, the male-to-female ratio at the camp was around 1 to 50. The impact was obvious. I felt like king of the camp! The gains in sexual experience were-fantastic! There were even catfights over me. I had money: that same thousand I had earned from selling the Colnago bike. I bought Hungarian champagne by the caseful-paying 5 rubles 50 kopeks per bottle-and kept it in my room, where we drank. I would get up in the morning to do my workout and the whole camp would laugh at me. The team-leaders understood everything and shouted at me: “Oleg, get some sleep!” They had heard me getting wasted with the girls all night. But I would yell back, “Do some push-ups!”

One day they asked me to lead a game called “Wildfire.” I did not know the rules. In order to get out of it, I had to have an affair with the senior camp counselor. She relented: “Fine. You don’t have to lead the game. Who even cares?” The Pioneers asked me,

“When are we gonna play wildfire, Mr. PE teacher?”

“There will be no wildfire,” I answered with confidence.

The sun, the river, wild berries, girls-what else does a recently discharged soldier need? I’d recommend that everyone coming home from the army work a summer as physical education teacher at a camp. For a soldier, it is as entertaining and romantic as it gets.

While I was at the camp I met a girl named Zhanna Pechorkina. She was doing her internship, working part time in the cafeteria as she ppared for medical school. When I saw her in the cafeteria, about three weeks after I had started working, I knew that my days of fooling around were over. It was love at first sight. She turned 17 that June. Given today’s standards, that seems very young, but this was not so in Soviet-era Siberia. At the time, it was considered normal to have your first child at 18. We went on walks in the forest holding hands. Ah, the romance! An innocent girl-my first true love.

We were inseparable and went to the city together to visit my parents. On June 28, 1988, we got on a yellow Ikarus bus departing from the central market in Leninsk-Kuznetsky and went to the village of Yegozovo. Everyone took a seat; we stood on the floor near the back and kissed. The bus drove at an immense speed and bumped wildly up and down. I wondered why the driver was going so fast. Suddenly there was a crash and a grating sound. I blacked out. The next thing I knew, I was lying on the steps of the bus, which had spun to a stop. Getting up, I saw that half of the bus was missing. The back part of the roof was torn off and the windows shattered. The bus sort of resembled one of those tourist buses in London or Paris. In a state of shock, I started calling for Zhanna. I climbed through the hole in the back of the bus where the window had been. I landed on the road and started looking for her. I found her in the ditch, her dress pulled up over her head so that all I could see were her bare legs and underwear. I told her, “What are you thinking? People can see everything.” I pulled her dress down from her face. And then I saw something that I hope never ever to see again in my life-my beloved girl with no head, in effect. Her casket stayed closed at the funeral.

I grabbed her hand, choked with memory. Then I felt hands grabbing me from behind and I heard someone say, “Get this one to the ambulance immediately!” My head started spinning once I got to the car. I spit and saw eight teeth fall onto the pavement.

What had happened in that fateful moment? As we were standing there, kissing, a KamAZ truck, driving too fast, hit the side of our bus. A pole broke loose from the force of the impact and I was thrown to the floor and onto the steps. I did not fly out of the bus and that saved me. Zhanna had been standing with her back to the pole, while I faced her. The pole just ripped through her head. She got the brunt of the blow, while I was hit with less force. Because Zhanna was six inches shorter than me, she got hit in the head, while I was struck in the teeth. Essentially, she saved my life by blocking the blow with her head. It is ironic that this happened just as we were kissing.

This was the first time that my life was spared. The Lord protected me… I was taken to the hospital. I underwent multiple surgeries. Investigators came and got information from me. I was totally devastated by this tragedy. What a thing for a twenty-year old man to suffer…

I could not look our mutual friends in the eye; I could not look at her parents or at buses or at the town. As Nautilus Pompilius sings: “I looked at these faces and couldn’t forgive them for being able to live without you.”

I had to leave Leninsk-Kuznetsky.

One day I ran into my friend and neighbor Yura, who lived across the way from me on Kooperativnaya Street. He told me that my other neighbor Vitya Starodubtsev had moved to Leningrad in order to attend school at the Mining Institute. Yura and Vitya explained that it was not that complicated. All I had to do was to get a paper from the mine saying that I had worked there. On top of that, I had already completed my military service. I was fascinated and inquired as to when they were accepting applications. It turned out that I had only one week left. My friend Edik Sozinov, who was still living in Leninsk, helped me. Quickly, we gathered all of the required doctor’s notes and I got a letter from the mine stating that I had worked there for nine months. Once I had all of my documents together and had put on my junior sergeant’s uniform, I got on the train and left for St. Petersburg to start school. To stay in Leninsk would have been unbearable.

About ten years ago, I met up with Edik. He told me,

“I remember how we took you to the train station. But no one believed you’d finish what you set out to do.”

I am not so sure that I believed it myself. I was absent for almost all of my last two years of high school. I then served two years in the border guard, which most likely did nothing for my intellectual capabilities. Who knows what kind of impudence it took to think that I could get into the Leningrad Mining Institute, the top university in Russia, established during the reign of Catherine the Great!

I returned from the army in 1988, grown up and hungry for sex. Looking back, I think that the happiest time of my life was spent working at the Pioneers’ Camp in June 1988 Lidia Irincheyevna Baturova, Oleg Tinkov’s homeroom teacher:

I met with Oleg’s class twenty-five years after their graduation. The kids told me what they had achieved. Fourteen of the students finished school with B’s and A’s, but none of them managed to achieve Oleg’s level of success.

When Oleg Tinkov’s generation was growing up, there were eleven mines and factories operating in Leninsk-Kuznetsky. Later, everything shut down. Circumstances during the stagnant Soviet period were stable at least: you finished school, attended an institute or technical school, got an education, and then went to work. Now these kids had taken their first grown-up steps at the end of the eighties, when the country was in turmoil. Few had managed to stand up to the revolution in our way of life.

All of the social facilities that existed then have closed down: the schools, pschools, and stadiums. The stadium and gym where Oleg grew up and trained as a cyclist have been demolished. There are only five mines in the city now and five of the other larger enterprises have been closed as well: the yarn plant, the light bulb plant, Kuzbasselement, Khimprom, and the clothing factory. Thousands of people were thrown overboard, in a manner of speaking. As a consequence, these people were unable to provide a good education for their children. The tragedy in small towns today consists in the fact that children have no opportunities for development and nowhere to go. I’m already teaching the children of my former students and I can say that it’s a rare thing to see someone achieve a level of education higher than trade or technical school. Only a very few inpiduals are capable of breaking free and going further. The children lack the finances and the motivation. Once Oleg had gotten on his feet, he came to Leninsk-Kuznetsky with his kids and brought them to see the school. His daughter Dasha, who had just returned from America, asked,

“Dad, I can’t believe kids actually go here.” The school is small, poorly maintained and has no funds. So Oleg decided to help. He was the first graduate of the school to donate money for repairs and equipment for his class. He wanted the kids to see that you can succeed through knowledge and schooling. I’m grateful to him for using gifts to encourage people in the right direction. His charity gave rise to a conundrum in the graduates’ minds: why is he able, but we are not? A whole movement was started. Everyone wanted to help out as much as they could.

The municipal school board received an official charitable donation in the amount of 150 thousand rubles. The bureaucrats decided to hold on to the money for a while in order to profit from it. I received a call from Oleg, who was in Italy at the time:

“Did you get the money?” My answer was,

“No.” He started swearing,

“Don’t just gape, find it!” He gets like that sometimes. I got in contact with the local criminal authorities, some of whom I had taught in school. Immediately, the money was found.

The municipal school board accounted for every kopek. And I had to keep Oleg just as informed as I was. I know my student. He can be very nice, but when it comes to money, he’s incredibly strict.

We ordered new furniture for the classroom, but it was taking forever for the delivery to come. September 1 was just around the corner. Once again, I had to involve the criminal element. These guys like Oleg a lot and respect him for helping the school. They drove to Kemerovo, where the furniture manufacturer is located. As a result, the furniture arrived the following day. Everything was assembled and set up over night.

When Oleg came, I showed him everything: the new windows, the newly laid linoleum floors, the desks and chairs, the board, the TV-VCR combo, the video camera and the audio library for geography class. It seemed like his mind was elsewhere, but he took note of everything and was interested in the details. He hadn’t just doled out the money like an aristocrat-he wanted to be sure that his money had been put to good use.

The next time he came to see what we’d done, everything had been set up. We reminisced about the new furniture that was brought to the school in the summer of 1980. We workers at the school, as well as the kids and parents, assembled everything ourselves. All of the chairs were still in tact, including the one that Oleg had put together and signed with his name over 20 years ago. We had his son Pasha sit on that chair.

This wasn’t the only time that Oleg helped the school board. We really wanted him to build a school, but he decided to build a playground instead, along with Natalya Vodyanova and Alexei Prilepsky. Good for him! I value his humanity. When he comes to visit, you never get the impssion that he’s stuck up or seeking attention. He always asks about everyone and takes an interest in how things are going and who needs help.

That last time we got together as a class, we noted with great sadness that six of the students have now passed away. All of them were Oleg’s good friends. Each of them went down a different path. Some got involved with organized crime and two girls drank themselves to death. Their male classmates can’t believe it.

Some of my earlier graduates fought in Afghanistan, and some of my more recent ones served in Chechnya. The have found it difficult to readjust to normal life.

To tell the truth, some of the groups of kids I’ve seen through to graduation ended up much worse off than Oleg’s class. In one class, for example, all but one of the boys served time. Many have died. The neighborhood where Oleg grew up became a hotbed of drugs. Thanks to sport, Oleg was able to catch hold of life and get further than these others.

My memories of Oleg are all good ones. He can be harsh and severe, but he always keeps his head about him. I really hope that he retained his grasp of, and sensitivity to, the situation at hand. This is a skill we lack. May everything be good for him.

Young people are living a modern life. You are insiders. We stand at the curb and can do little to affect what’s going on-except through you. Students are smarter than their teachers. We provide a base; we lay a foundation. What grows from this is up to the child. Every student, no matter who he or she is, is unique. As long as you don’t put too much pssure on kids, they will rise by themselves. My motto is “Teachers should bring up children so they can be learned from.”

Chapter 7

Change! We Wait for Change!

When I applied to the Leningrad Mining Institute, the physics professor took pity on me. At the exam I could not explain Newton’s second law. He looked at my sergeant’s shoulder marks, and my badges of excellence from the border guard and said,

“Do you promise to bring your physics up to snuff before the start of the semester?”

“I promise!”

“Okay, I’m giving you a C.”

“Thank you!”

I was really pleased, considering that I had already been given B’s in composition and math. I got in! It was a good thing that I had come dressed in uniform. Otherwise I would have ended up at Moskovsky Station waiting for a train to Leninsk-Kuznetsky-something I did not want at all. I cannot remember what that professor’s name was, but I would thank him, if I were to see him again, for offering me the chance to give life in Leningrad a shot.

Strangely enough, I did well in school. Because I had barely made it into the program, I had to promise myself that I would study hard. I sat in the first row, the best place for absorbing knowledge. I can still remember our math professor, Lobazin, and our physics professor, Mezentsev. If I could not understand something, I would come up to the teacher after class and ask him to explain further. My studies bore fruit: I was the first in my class to write my final exams and I was even one of only four students who passed physics on the first try.

At the institute I met the most cultured intellectual people. The best people in the country. The professors were the embodiment of intelligence. Their speech, their approach to the material, their love of liberty, and their ambition captivated me. It amazes me to think that such obviously anti-Soviet ps could end up teaching in a state-run university in the USSR. They criticized Soviet authority-some of them doing so indirectly, but the more courageous ones just telling it like it was. Some of them would make post-lecture announcements such as, “Remember, Nautilus Pompilius has a concert tonight.” The professors at the Mining Institute planted the seeds of nonconformism and inner freedom in me.

I do not know how things were in Moscow at that time, but in St. Petersburg everything was light and color. After its founding in 1703, the city was originally the freethinking capital of Russia. The Decembrist uprising of December 14, 1825, the workers’ revolt of January 9, 1905, and the 1917 revolution all happened there. It is not surprising that the years 1988-1990 saw an intensification of anti-Soviet sentiment in Leningrad. I just cannot understand why there has been no protest in St. Petersburg during the current stagnant period in our country.

Back then, freedom was in the air. I was really excited about all of it. I remember buying a badge reading, “Yegor, you’re wrong!” The slogan referred to Yegor Ligachev, who had interrupted one of Boris Yeltsin’s speeches, saying, “Boris, you’re wrong. It’s not just tactics that make us different now. Boris, you possess a great deal of energy, but it’s not creative energy. Rather, it’s destructive.” People decided that Yegor was wrong. We supported Yeltsin because we believed that he would save the country from communism.

Of course, it was Mikhail Sergeyevich Gorbachev who got the ball rolling. He had the strength and courage to take down the Soviet system from the inside. Of all the leaders of the USSR and Russia in the twentieth century, I respect him the most. He was a member of the communist mafia-there’s no other way to put it-and decided to fight against it. He was a rule-breaker. He broke down the very organization of which he was a part, in order to offer us freedom.

A lot of people say that Gorbachev had no choice-that everything would have unfolded exactly as it did regardless of his involvement, but I disagree. I think he had a choice. He could have tightened the bolts, as Andropov did in 1982-1983. But this was a person who had been elected as the General Secretary of the only political party in the country-and it turned out that he held democratic and liberal views. Gorbachev is the sharpest and greatest Russian politician ever. It is not for nothing that he has garnered praise and respect in the West. The only thing that he did wrong was the alcohol reform. Without a doubt, he will go down in history as the man who put an end to communism. Yeltsin will be remembered as Russia’s first psident. Lenin and Stalin will be remembered too, but with the least fondness. As for the others, my gut feeling is that their memory will be all but erased.

When he dies, we will cry. We will pay him our respects and remember what a good person he was. For now though, as long as he is still with us, I would like to expss my gratitude to him for conquering the communist dragon.

Mikhail Sergeyevich, THANK YOU! My deepest respect!

We still hear the voices of those idiots that praise the Soviet Union. What is there to praise, anyway? It is unfortunate that those in power in Russia today often appeal to the USSR. Some even anticipate its restoration. But that is something that I would most certainly not want to see. I would not want to see a specific quota of mohair scarves allocated to Dushanbe, Tajikistan, where the coldest it gets is plus 10. Nor would I want to see big shipments of plastic ice hockey sticks being sent to Tashkent, Uzbekistan. The Soviet economy was inoperable. The system’s collapse was just a matter of time. And then there was America dragging us into the arms race. But even if that had not happened, the crash would have been inevitable.

In the last few years we have seen a governmentalization of the economy; once again, today, we see socialism baring its teeth. Half of the economy now hangs on Gazprom, along with some soccer and hockey teams. We are treading water, in spite of Gorbachev’s breakthroughs and in spite of the work that Yeltsin did towards accelerating our country’s development. We have strayed from the path. We have taken one step forward and two steps back.

I would stress that my argument must be understood from an economic standpoint. I am no political scientist or politician. I do not know how many parties there are in the country. But I am certain that there are more than one. The economy must operate in accordance with capitalist mechanisms. It is the best approach that we have thought of. Take, for example, communist China’s economic successes. They are based on market principles. In this respect, their approach has not been socialist in the least.

In the summer of 1988, before I started my studies at the Mining Institute, the nineteenth Party Conference was held. At this conference, changes were announced, changes bearing on the Communist Party of the Soviet Union. It had become clear that communism was weak and could not last. The wind of change was blowing from the Baltic Sea, the Bay of Finland, and from Europe. The Scorpions would sing about the wind of change in the nineties. In Russia, the rock group Kino was already singing about it:

Change! Our hearts require it.

Change! Our eyes need it. In our laughter and tears and in our pulsing veins:

Change! We wait for change!

During my very first semester at the institute, our trigonometry professor told us about Anatoly Sobchak, a teacher at a nearby university. Sobchak’s star was only starting to rise. He was fighting for the most basic of western values: democracy, inpidual liberty, and private property.

At the end of the 1980s, we believed that everything would change and we would be the ones to make it happen. There is a reason why university students are considered key to revolutions. In spring 1989, a miracle happened: we elected Anatoly Alexandrovich Sobchak to the USSR parliament, repsenting the Vasilievsky Island District 47. I am glad that my vote played a part in that victory. Sobchak could not win in the first round, even though he had a majority of votes. But in the second round his win was unequivocal. One of the voting stations was in our dormitory in Building 5, Shkipersky Stream.

I fell in love with St. Petersburg. Vasilievsky Island, the buses full of foreigners, the imported goods, the colored lights, the wide prospects, the steamships-it was a country boy’s dream come true, especially after I had come so close to becoming a warrant officer. I just went crazy. I was in awe of the wealth of knowledge available at the mining institute, the statues by the entrance, the huge staircase leading down to the Neva River, the neighboring Baltic factory, where a reconstruction of the icebreaker Lenin was on display. I had only ever seen such things on TV. I would call my state upon arrival in Leningrad euphoric. It was as though I was high all time.

Ever since, I have had my own special relationship with the city. It was in this great, beautiful city that I grew up and became the person and the businessman I am today. It is disappointing to have to say it, but I am not a native Leningrader. And yet I am probably more of a patriot than many who were born there. Of course I consider myself Siberian, but after thirteen happy years in St. Petersburg, I am a most genuine Petersburger.

I went to university in this city, I met my future wife there, and this is where I started nearly all of my businesses. No city in the world has given me anything remotely like what this city has.

So I’ll quote Iosif Brodskoi:

There’s no country or graveyard,

Which I would pfer.

It’s on Vasilievsky Island

That I’ll be interred.

I associate the Leningrad of the late nineteen-eighties with the rock band Kino and its lead singer Viktor Tsoi. My Baptism

In December 1988, I went to Nikolsky Church to be baptized. The priest asked me,

“Do you know the Lord’s Prayer?”

“No.”

“Then I can’t baptize you. Go learn it.”

I memorized the prayer on the plane on one of my trips to Siberia. I still remember it:

“Our father, who art in heaven. Hallowed be thy name. Thy kingdom come, thy will be done on Earth, as it is in Heaven. Give us this day our daily bread. And forgive us our trespasses, as we forgive those who trespass against us. Lead us not into temptation, but deliver us from evil. Amen.”

On December 25, 1988-my birthday-the priest baptized me.

Chapter 8

The Mining Trade Institute

The Leningrad Mining Institute is the oldest technical school in Russia and is, for all intents and purposes, where international mining science originated. Because of this, at the end of the eighties, the Institute’s student body included people, not just from the socialist camp, but Americans and Western Germans as well. For the most part though, international students came, naturally, from “third-world” countries in Asia and Africa.

When they came back from holidays, they would bring merchandise with them. A lot of them flew through Berlin, while students from former French colonies (Algeria, Burkina Faso, Côte d’Ivoire) had stopovers in Paris. They all wanted to earn some extra cash. They came to Russia with jeans, perfume, and cassette tapes. After exams were over they would return home with their money in foreign currency-dollars, marks, or francs.

The foreigners were not good at selling on the street. Maybe they were just scared to do it. Instead they sold their goods to Russian speculators, like me. On the street, my profit would reach fifty to one hundred percent. I lived off this margin. To tell the truth, though, I did not save much as we loved to party in the dorms.

I soon realized that, in Siberia, goods that were in short supply could be sold for twice as much again. For instance, I could sell cosmetics kits in Leningrad for 25 rubles, while in Leninsk-Kuznetsky they sold for 50. Lipstick was 15 rubles in the city, but 25 in Siberia. Of course, when I could, then, I tried to sell in Siberia.

In Leninsk I would come to the shoe or yarn factory, which were staffed mostly by women. Because they would not let me in through reception, I had to climb through a window. The workers already knew that there was this guy named Oleg from Leningrad and that he traded in scarce, imported products. I was able to make even more money there than I would have if I had sold the stuff at the local market. This was because the women liked the idea of making their purchases from the comfort of their own place of employment. This is just another example of the importance of service in business.

(I still remember the lessons that I learned then. In particular, the savings program launched in 2010 by Tinkoff Credit Systems is based on the same principle: a bank repsentative comes to where you are when you want to open an account.)

Of course, sometimes, the merchandise turned out to be complete crap. Once I bought lipstick with glitter in it from some Gypsies on Staronevsky Prospect. Later I saw how they made it. They would take shiny chocolate wrappers, cut them into tiny pieces, and add them to the lipstick.

In the main, I sourced my merchandise from foreigners. Another source, however, was a fellow student at the Mining Institute, Igor Spiridonov. He was from Prokopyevsk in Kemerovo Province. Igor sold in small bulk: cosmetic kits came in full boxes; lipstick in blocks of 100 each; VHS cassettes in packs. I bought my first consignment in cash and sold the goods inpidually in Siberia. A week later, I bought more. Because Igor and I were fellow Siberians, he offered me a larger consignment and said I could pay him back after I had sold the product. In this way I made money off the difference with no investment.

At one point I was flying three or four times a month. I would load up a couple bags, buy a ticket to Kemerovo for 60 rubles, and then sell the goods in Siberia for twice as much or more than they would have cost in St. Petersburg. Of course my trips were not just about business. I also spent time with my friends, Edik Sozinov, Alexei Smirnov, Zhenya Brekhov, and Alexei Prilepsky. I even convinced the latter two to apply for university in Leningrad.

Later, I started bringing stuff back to sell. A group of Yugoslavian construction workers were building a hospital in Leninsk-Kuznetsky. They lived in what was called the Yugoslavian Village, in trailers. They brought German marks with them from Europe. You could not buy anything in Soviet stores with foreign currency, however-be it vodka or treats for the girls. Because the Siberians had no need for foreign money, I bought the marks from them at a ludicrously low price and sold them to speculators on Vasilievsky Island. If I remember correctly, they cost me five rubles each and I sold them for nine. Such was the Soviet Forex!

Most of the people studying at the Mining Institute were from regions where there was an active extraction industry. There were a lot of students from the Kuznetsk Basin, Don Basin (Donetsk, Chervonograd and Shakhty), Vorkuty, and Ukhty. There were a few students, too, from Slantsy and Yakutia, where diamonds are mined. I tried to stick with guys from Kemerovo Province; we were from the same area and I was used to trusting my own. I considered them more reliable and understanding.

But this approach almost backfired. Vitalik from Kemerovo, who was about five years older than me, got me involved in some shady dealings having to do with gold. And I crossed a few lines. I am ashamed to admit it, but it got to the point where I was taking part in some straight-up thievery. Thank God, I had the strength and soundness of mind to get away from these people. The Lord led me away. They wanted to expel me from the Institute. I lost so much. Worst of all, I lost my good name in the dormitory. The most important thing, though, is that I stopped hanging out with that crowd.

So why am I writing about this? None of us is perfect. Young men arriving in a new city are bound to get mixed up with bad apples. You have to try your best to avoid them, but if it is too late for that, then you have to have the strength to walk away. Now, I never judge people for their mistakes-remember that even Pinocchio got mixed up with the wrong crowd. But he showed what his character was like by breaking away. I was like Pinocchio in that story. I was led astray by their high life: the restaurants, discos, and strip chúng tôi was all so tempting. After all, before I came to Leningrad I had not even seen the inside of a restaurant, really.

One way or another, I decided that I would never become involved with crime. And although the article against speculation was only removed from the Criminal Code in 1991, it had been largely unenforced for a long while before then. Undoubtedly, I should have been more careful, but I was afraid of nothing in pursuit of the good life. I had to keep speculating.

Every day, during our long break after second period, we speculators would meet at the Mining Institute, in a wide square hall, which we called the “meeting spot.” People could get onto campus without documents and speculators from various neighborhoods, from places like Aprashka (Appraising Door) and Galyora (Gusting Door), came to the Institute to buy product. The meeting spot was a place of intense commercial activity. Items for sale included clothing, appliances, and electronics. Currency was also exchanged. Trade was evolving. At first clothing and perfume were the most sought after items; later, demand for electronic gear grew. For two years dual-cassette tape recorders were all the rage. We called them soapboxes.

We students made money any way we could. I would buy vodka at the store, during the day, and then sell it in the dorm, at night, for 20 rubles. Some people accused me of being an animal for this, but I disagree. If you do not go to the store, during the day, to get your vodka-and you want some at night-then you have to pay up. Nothing is free, including drink, when a sudden urge to have some sets in. My fellow students would get mad about it, but they would buy the vodka. One kid got a VCR from his parents and he used to charge a ruble to anyone who wanted to watch a movie in his room. All was right and fair: the VCR was an asset and assets should bring you profit. We would stay up all night watching movies starring Sylvester Stallone, Bruce Lee, and Arnold Schwarzenegger. We thought action movies were the height of cinematography.

I liked it in Leningrad, but I missed my friends in Leninsk dearly. In the winter after I had finished my first midterms, I almost made the biggest mistake of my life. There was a university transfer system in the Soviet Union, which allowed you to transfer to a more pstigious school after you had been accepted to a lesser one. In the winter of 1989 I went to the Kuznetsk Basin Polytechnical Institute in Kemerovo. Like the Mining Institute in Leningrad, it trained future mineworkers.

The young woman in the transfer department looked at me like I was an idiot.

“What! Are you stupid?”

“I’m sorry. What do you mean?”

“We have fifty students waiting on transfers from Kemerovo to Leningrad. What are you doing, man? Don’t screw around.”

She changed my mind; I withdrew my transfer documents. I feel like God was at work here too. That girl at the Institute could have taken all my papers without saying anything. I probably would have ended up working as an engineer or something in the mines in Leninsk!

Cosmetic kits cost 25 rubles in Leningrad. In Siberia, their price was double. One of my first investors, Oleg Korostelev, his wife Vera, Rina, and I in Morskoi Restaurant. Eduard Sozinov, a friend of Oleg’s from school:

Every time Oleg came to Leninsk-Kuznetsky, he would bring something to sell. It was the simplest way to make extra money. Before, this was called speculation; now we call it business. At the time, though, I thought it was a completely normal thing to do. He’d bring jeans and coats people had ordered-in small amounts, though. Mostly he sold cosmetics, however. Women go crazy over things like that and the stores didn’t carry anything. Lipstick and perfume sold like hotcakes, because the price was reasonable. Jeans, on the other hand, were something very few people could afford…

Igor Spiridonov, Oleg’s business partner during his university years:

I lived in a dormitory on Maly Prospect (Oleg lived on Shkipersky Stream). Oleg had good connections when it came to sales in Siberia. I knew where you could get stuff cheap in Leningrad. During our early days of speculation, the main products were clothing and toiletries. Later we started speculating on currency and electronics and started making grown-up money.

The first time Oleg came to my dorm on Maly Prospect, he told me that he was from Leninsk-Kuznetsky (we were practically neighbors, as I was born in Prokopyevsk in Kemerovo Province). He had heard from someone that I had merchandise for sale. A week later he came back and said he had sold everything. “Nice turnover,” I thought. Mostly Oleg bought cosmetic kits, VHS cassettes, and lipstick. Later on, like good neighbors, we agreed that he would take a bigger shipment of merchandise to Siberia and pay me when he got back.

Chapter 9

Gangster Stories

In Leninsk-Kuznetsky I saw some real tough gangsters. When I moved to Leningrad, I came into contact with athletes who called themselves gangsters. They were from Tambov, Kazan, and Vokruty, and because they did not know how to do anything else, they took up hustling. You would join a gang based on what part of the country you came from. Unlike in other cities, in St. Petersburg there were few people from the south. Chechens and Dagestanis played only a minor role there. The top guns were from Slavic gangs, which were dominated by former athletes. They were not really gangsters in an ordinary sense.

At the end of the eighties in St. Petersburg, if you were doing business then gangsters would inevitably become involved. This happened, for the first time, when I managed to sell a can of black caviar to some foreigners at the Pribaltiyskaya Hotel for fifty dollars-an insane amount of money at the time. It was the first time I had ever held a fifty-dollar bill in my hand. I nearly went nuts. To my dismay, the in-house gangsters saw me making the sale and decided that they deserved a cut. I had to escape through the restaurant kitchen, running past the frying pans with food cooking in them, just like in mafia movies.

The Mining Institute was under the protection of some guys from Vorkuty. They were big and aggressive and liked to rock it out at nightclubs. You would not say that they were well-structured, but they sure had biomass. The prostitutes on Vasilievsky Ostrov paid them for protection, as did the currency dealers and rich kids who sold matryoshka dolls. This was nothing serious, just old-fashioned racketeering. Now I realize that the gangsters did not make all that much money, but at the time they seemed super-rich, driving 2109 Ladas and eating out at restaurants.

My first encounter with the Vorkutians came when I was in my first year of university. I and our Komsomol rep, Vitya Cherkashin, were returning home from the pub across from Kazan Cathedral. We were a little tipsy when we got back to the dorm. We noticed the Vorkutians loitering there, as they often did. Their boxer, Igor, was harassing people and a few of the others stood nearby watching and laughing. Vitya and I were walking down the hall. They were walking towards us. I was certain that we were going to get punched and possibly kicked. If we pushed up against the wall, they would take exception; if we walked straight towards them, they would get mad. They would get pissed no matter what. Once we had reached them, the boxer took his stance.

“Whatchya gonna do, Tinky?” he sneered.

What was I to do? The beer and lack of options gave me the guts to act. I had nowhere to turn. I remember that I had been taking boxing at the Mining Institute for six months, at that point, and we had only worked through one punch-the right straight. I did not think about matters for long. I took up my position and followed through with the punch. It was a good hit, for me, but not so much for Igor. I got him in the jaw. Boxers know that this is the worse place to be hit-you can fall down immediately. I thought it was over, that they would kill me. Contrary to what I expected, though, his buddies, who were standing close by, opted not to get involved. None of them wanted to go down second. They just shit themselves! I shouted something along the lines of,

“That’s what’ll happen to every one of you!”-and withdrew.

Realizing I would be screwed if I stayed around, I ran out of the dormitory five minutes later, caught a cab, and went to my girlfriend’s place. She was studying economics and lived on Bolshaya Morskaya Street. The next day, after class, I came back to my room. I sat waiting, knowing they would be coming. The ringleader entered and said,

Suddenly, I was recognized all over and my reputation inched upward. One fine day during the long break, some strong men from Vorkuty, wearing black leather jackets, approached me.

“We’re going outside. We need to talk,” one of them said.

We went outside and stopped on the staircase in front of the chemistry department.

“So, you’re selling here?” one asked.

“Yeah, I’ve been trying to earn some money.”

“You’re going to have to pay us. You’ve got to feed the bros.”

“What does this have to do with me?”

“Listen, you! Are you looking for trouble?”

Of course I was already ppared for this moment.

“I couldn’t care less who you need to feed. I have a dad, mom and brother. They’re the only ones I owe anything to-no one else. If you harass me again, I’m writing a police statement.”

“Listen, what the hell is your problem? Don’t you know the rules?”

“I’m not interested in your rules. I set my own.”

“Okay, fine. What point is there talking to this piece of trash?” They threatened me and left.

Naturally, I never had to deal with them again and I kept on working.

Ever since then, I have understood that the dumb underdog gangsters are easily scared, while their leaders should be used; you might borrow money from them. They are rich and they have their head on their shoulders. Later on, I borrowed money from certain organizations, understanding clearly that they were controlled by people whose names were often mentioned in criminal histories. I took loans from them rather than from banks, at interest rates that were reasonable-to say the least. Should it make a difference where I get my loans? They had capital and I did not have the money I needed for various projects. And you would never hurt someone who owes you. No one would. They thought they were using me, but, in my view, they were the ones being used. Not many of those people are still alive today, although now and then I do see some of them around St. Petersburg. Now they have realized who was using who. After all, I was paying them at a fixed interest rate, but in the end their money earned me much, much more.

In the 1990s gangsters liked to follow a scheme called “raising hogs.” They would give money to an entrepneur, would get a share in the business, and then, when the company started to go under, they would milk the owner dry. Or kill him. I would never give gangsters a share in my business, because it always ended badly.

As long as he has something to hide, a businessman will always fall victim to extortion. For the moment, unfortunately, law-abiding businessmen are few and far between. A lot of people want to get rich in six months, buy a yacht and plane, and move to Monaco. In order to achieve this, they avoid paying taxes, or customs duty, and they bribe officials. They give extortionists something to work with.

My situation is different. I have been laboring hard for 20 years and yet I have not acquired anything extraordinary for myself. Compared to the average man, I am very rich, to be sure. But from the point of view of the richest, I am poverty-stricken. I am not accustomed to fast money and I am not willing to break the law in order to make a profit. I will not go against my own conscience. That is why I will not let anyone make my life difficult. It would be unfair. I will protect my rights by whatever means possible. As for those who steal from their country or from others, their lives should be made difficult. You must not forget, I have a home in St. Petersburg and a lot of my friends are high up now. I am respected. I receive offers of help as soon as I have a problem. A lot of people might say, “Well, I haven’t got any influential friends from St. Petersburg!” I am just saying that you must use your head and act in a way that protects you from harassment.

Again, I never got involved in any business with a really high profit margin that would be of interest to the mafia. A lot of my friends and other acquaintances have been killed, sometimes for no apparent reason. But I have never had bullets flying by my head-not even during those dark and dreary days when human life had lost nearly all its value.

I do have one story involving bullets, actually, but it has nothing to do with business. It was December 25, 1992, and I was celebrating my birthday in the Pribaltiyskaya Hotel. After dinner I invited all of my guests (there were eight or ten of us) to the dance club Eldorado in Karelia Hotel. It was controlled by thugs from the town of Tambov. A few of them sat a couple of tables away, giving us dirty looks. As the night wore on, most of the girls left the club. Finally, our wives were the only women left. One of the smaller thugs, who wore a cap, came up to Rina, pulled at her hand and said,

“Come on. Don’t ya wanna dance?”

I took hold of his hat, pulled it down over his face, and told him to you-know-what off. He hit me first, I hit back, and so the fight started. There were five of us and nine of them. The police put an end to the fight, but the gangsters went outside, got in their cars, and waited for us. The cops that worked at Karelia enjoyed some kind of relationship with the gangsters and maybe even got money from them. The policemen told us straight up,

“Guys, you’re screwed. You’ve got no chance. Get whoever protects you in here, otherwise you’re dead.”

The cops were slow to understand their pdicament. Half an hour later they realized that if we were killed, they would get in trouble too. In view of this, they offered to take us to the station. They backed a police van up to the exit and one by one we jumped inside. When they saw that we were going to get away with our offensive behavior, the gangsters drew their guns and jumped out of their BMWs.

The cops started shooting into the air.

“Everyone in your vehicles!”

When the van started moving, the mob cars followed us. They followed us all the way to the station. I was on the edge of my seat, as though I were in a movie. When we got to the police station, we were put in a cell. The police told us to wait until morning and then to call whoever it was that protected us to come pick us up. I really value my freedom, but I was totally fine with spending that night behind bars. By morning, the thugs had gone. We all went back to our homes and for the next couple of weeks tried not to stray outside.

Eduard Sozinov, a friend of Oleg’s from school:

The street fighting in Leninsk-Kuznetsky stopped after we were discharged from the army in 1988. The reason was the rising popularity of drugs. Within what seemed like moments, everyone united and became friends and brothers. At first grass and bud were the mainstay, but later on heroin made an appearance. During the early nineties, the shit infected the city and a lot of our peers died. Practically every single young adult used. No one tried to avoid it. At least everyone tried it once. I’m not sure about Moscow and St. Petersburg, but I think that drug-use thrives to this day across the country.

During those terrible, dark times, Oleg tried his best to stay away. The place was a cesspool of drugs and murder. For several years the shootings and funerals were incessant.

Oleg took pains to be careful. He made sure no one knew he was coming. One day, we were told some people were looking for him; it sounded as though they wanted to kill him. The boys had found out that he had money. They wanted to take everything they could from him. Probably the first order of business was to track him down, harass him for money, and then, based on his reaction, decide whether to put the squeeze on him further.

Oleg used to say that you could actually talk to the gangsters in St. Pete’s, but the ones back home wouldn’t listen, no matter what you said. He never even tried talking to them. Still, his trips were frequent, because he had to keep up his business in Leninsk. And he had to visit his parents. For safety’s sake, he avoided spending the night there. We found him a rental each time he came.

After the collapse of 1991, Oleg brought a large consignment back home. There was wine, vodka, and some kind of clothing-denim skirts, perhaps. The city bomb shelter was stuffed full. But there was a robbery and a lot was stolen. Oleg even went to the police, but they ended up finding nothing. It’s a good thing that Zhenya Brekhov and I had sold off some of the vodka and wine to various stores the day before.

Chapter 10

The Girl from Estonia

The dorm on Nalichnaya Street housed students from various faculties of the Mining Institute. I met an awesome girl there, at a dance, in April 1989. Here name was Ira. We danced and I fell in love. I always try to bring things to their logical conclusion. That night, though, it did not work out. The night was nearing its climax and I noticed that she had disappeared. She must have gone to her room. My search turned up nothing.

The next day I was walking by the math department and saw her (or was it someone else?).

“Ira! Hey!” I said with a start.

“I’m not Ira, I’m Rina.”

That is how, thanks to a random dance with a girl named Ira, I met my future wife, Rina. It was not until twenty years later that we were officially married-but more on that later. The next time we met was two months later, in June. I had gone into the grocery store at the corner of Gavanskaya Street and Shkipersky Stream to buy some sausage. As I stood in line, I noticed the same girl that I’d seen before, as I was walking by the math department-Rina.

I bought her a birch beverage for 11 kopeks and she had the indiscretion to tell me her room number. The following Saturday, I and my friend Edik, from Vorkutia, grabbed a bottle of wine and went to visit Rina on Nalichnaya Street. She had two female roommates and so Edik and I were happily outnumbered.

The Gavan Hotel had recently opened on Maly Prospect on Vasilievsky Island. It was not long before I took Rina there. We paid three rubles for entrance to the Hotel because it was part of the Intourist system and was not technically intended for Soviet citizens. You had to walk up to the glass door and show the doorman your open palm with a bill in it. He would open the door, take your money slyly, so that no one would notice, and let you into the hotel. There was a bar on the top floor. The barman Albert was an enterprising man in glasses. According to the menu, beer cost 55 kopeks, but everyone paid a ruble. Some people asked for change. When they came back for more beer, Albert would say, matter-of-factly,

“We haven’t got any beer.”

On this occasion, I got Rina really drunk and brought her back to my dorm. She gave in immediately, of course. It was not just anybody that could show a girl the kind of good time that I had. The moral of the story is simple: without money, you can accomplish nothing with women. I am kidding, of course. Rina is not materialistic. I took her to some cooperative restaurants, a few times, but later my money ran out. So Rina started taking me out! My sense was that she had money because she was from a fairly well-to-do Estonian family. According to her, however, it was because she was careful about how she spent her stipend. Whatever the case might have been, a girl paying my way was unacceptable. I started feeling shabby about it. I realized that the time had come to start making real money. I started putting twice the energy into my speculation business. My motive was simple: I wanted to take this beautiful girl to restaurants. The size of my consignments grew.

But I proved unable to get rich quick. I remained in the dorm and Rina moved in with me. It was we two, plus Andrei Pavlov from Kingisepp. Hungry days ensued. Andrei’s mom would bring him a sack of potatoes once a month and that is how we fed ourselves. I cannot stand potatoes to this day.

One day I stepped out of the kitchen to get some salt. When I came back, I discovered that someone had taken the whole pan of potatoes. There was an unwritten rule that said: make sure you stay with your potatoes during the last five minutes of frying-otherwise they will be stolen and later you will find your empty pan back in the kitchen. Sometimes people’s soup even went missing. There was no point in looking for it as 150 students lived on our floor alone.

There were bedbugs in the rooms. We would poison them, but they were never gone for long. Moving the beds away from the wall and into the center of the room afforded us some protection, but they would still climb up the walls, along the ceiling, and then fall on us from above, feeding on us once more. All of these domestic annoyances pushed me to do greater things. After all, I’d seen fortunate speculators who rented or bought their own apartments, drove their own cars, and were always going from restaurant to restaurant.

Sometime after I had finished my first term, I went to a regular store and bought some cans of red caviar at government prices. I got into a commuter train at Finlyandsky Station and got a ticket to Repino. I walked to Penaty Estate Museum where the Finnish tourists were filing out of the buses. I simply repeated the phrase “sata marka,” which means a hundred marks in Finnish. I quickly sold all the caviar for ten times what I had paid for it at the store.

After I pulled that off, I felt incredible. The business was easy and the profits huge. I told a kid in my dorm, Volodya, that there was money to be made. The next morning, we bought two whole cases of caviar and made the trip to Repino. After a couple days business with the Finns we found ourselves surrounded with 2106 Ladas with tinted windows. We did not know if it was the mob or the cops. Either way, since nothing good would come of us sticking around, Volodya and I started running in opposite directions. I raced along the tree line, tossing caviar into the bushes as I went. My hands were empty of cash. Even though I was an athlete, I could not outrun the officer, who wore a leather jacket. He caught up and twisted my arm behind my back, told me to pick up the jars that I had discarded, and took me to the Repino police department.

He took me to the special cases section, wrote me up, and confiscated my caviar. I sat across from that overstuffed cop, filling out the papers.

“You know what makes you lucky?” he asked

“What?”

“These problems you had today, they’re minor.”

“Minor? You caught me, didn’t you?”

“If you had been caught by the mobsters that control that spot, your problems would be much more significant. You were only here briefly. Now don’t come back.”

It seemed the cops were more afraid of the gangsters than I was. Maybe they were even getting a cut for protecting people that were essentially their superior officers. After these events, the Mining Institute received a letter saying that I was involved in the black market. For the second time they wanted to expel me. I’m not sure how they could let me leave to Poland with a service record like this: it must have been the lack of a unified information system.

I never again made the trip to Repino after this. In July, though, I received invaluable work experience in Soviet commerce. Nikolai Nikolayevich, the manager of the produce store on the Corner of Havana Street and Little Avenue, gave me a job selling fruit and vegetables at the stand. The kiosk still stands on that corner, next to the dairy store.

Our business was unique. You would weigh a kilogram of tomatoes. Then, before putting them in the bag, you would throw one of them under the table. Bananas, being both heavy and expensive, were especially profitable to tip in this way. Not stealing was not an option here. For instance, when a delivery would come in, they would tell us, “Here are a hundred kilos of tomatoes” and you would weigh them and there would be only ninety. But when you would say that some were missing they would always ask the same question: “Do you want to keep working here?” So really, you had to cheat-just another feature internal to the socialist system. To this day, when I go to the market, I always keep close watch on the shopkeepers’ fingers.

In August, Rina and I headed south with the money that I had “earned.”

Because I was only ever taken to Yevpatoria as a child, it was with pleasure that I took my love to the same small Crimean town. Memories of beach sex have blotted out all other recollections from the trip. Not surprising? Maybe not-except that we had sex during the crowded part of the day. We just covered ourselves with a blanket and assumed that nobody would notice what we were doing. As it turned out, we were mistaken.

At the same time, merchants from Moscow and Leningrad started sweeping up chainsaws and other electrical appliances and exporting them to eastern European countries. These products were still available for sale in the towns and villages of Kemerovo Province. I scooped them up with a view to selling them in Poland.

Rina came from Estonia to study at the Mining Institute. Estonia, though part of the USSR, was more like a foreign country. Rina Vosman, Oleg Tinkov’s wife:

Oleg was a Siberian guy, different from the others, unique. Life in Siberia is tough. I’m softer, more intelligent (laughs).

He was always different from the others-from the moment I met him. He wasn’t like anyone else. When I came to St. Petersburg, I was 20, a young, cute girl. And I knew a lot of people. But everything changed as soon as Tinkov came into my life. The last twenty years have flown by. Oleg has said that I’m from a rich family and that that’s why I had money kicking around. But it really was because I saved bit by bit. He loved to have a good time. When Tinkov got his stipend, everyone would have a good time. Every girl in the dorm would be in his room. He really loved girls ( laughs). There were girls named Mashka and Svetka and Lenka-all different kinds. He’d spend his whole stipend on champagne, then he’d eat fried potatoes or go hungry all month. But that’s how he’s always been: he has a big heart. As soon as I started coming to his dorm, the girls stayed away. It was the easiest thing in the world for me to achieve. Slowly I started moving my stuff in. When we lived in the dorm, we were poor. We had nothing to eat. After the third period in the day, we’d skip school and stand for three hours waiting in line to buy “blue birds.” That’s what we called the Soviet chickens due to their peculiar coloration. Fried potatoes and a three-liter jar of tomato juice-now that was a hearty meal! So we thought in those days, at least.

Chapter 11

Hello, Europe!

Rina’s parents lived in Estonia, while here maternal grandparents were in Szczecin, Poland. This made it easy for her to get into Poland. As for me, I had to get approval from various offices, the trade-union committee, the Communist youth league, and so on. Since Poland was still part of the Soviet bloc, the first time we went there, in 1989, we did not even have to apply for a foreign travel passport. Our Soviet ones were enough.

After we arrived at the home of Rina’s relatives in Warsaw, the first thing we did was head to Voskhodny Market, which means “Eastern Market” in English. We made the acquaintance of a Polish man there-Juliusz. He told us which goods from the USSR were in highest demand and so we started to bring these in.

In Poland, the price on anything with a power chord was three times higher than in the USSR. We would buy Raduga television sets in the Kozitsky Union Store on Maly Prospect on Vasilievsky Island. I would load them onto the train, disembark in Warsaw, sell them for 200 dollars apiece, and come home. Rina transported TV’s too. I would load them on the train in Leningrad and Juliusz would unload them in Warsaw.

In 1990, we made things more complicated. Rina spent the whole summer in Warsaw and I traveled back and forth. I flew to Siberia, bought Taiga chainsaws at various general stores for 200 rubles each, brought them with me to the airport in Kemerovo, paid for the excess baggage, and then flew on to Leningrad. From the station, I brought the saws to the room we rented in a co-op apartment on Gavanskaya Street. The next day, I was off to the station and, a twenty-four-hour train-ride later, I was in Warsaw. The logistics took up an awful lot of time. But it was totally worth it: in Poland we sold the saws for 200 dollars each, which was enough to buy another six or seven of them back in Russia.

From time to time, I’d fly in to Novosibirsk, hire a cab, and drive around to general and co-op stores, buying every electric appliance they had. In the cities, speculators had bought everything up, while in the villages, the stores were still stocked. Sometimes I would drop by my mom’s place in Leninsk-Kuznetsky for five minutes or so. She was always surprised because she thought I was in class.

One day Juliusz told us about a particular kind of business that the Poles liked conducting: taking cigarettes to Berlin. A pack cost one mark there, which was twice as much as in Poland. My Soviet passport allowed me to go to Poland, but not Germany. I took the risk and went with him anyway. I simply handed my passport to the German border guard, who decided he would not trouble me and put a red stamp in it.

In Berlin I was surprised by the stark contrast between capitalism and socialism, between West and East Berlin. It was at that time that they began tearing down the famous Berlin wall. I got on the S-Bahn train, which connected East Berlin with the West. It was like some crazy dream: like moving from a black and white movie into a colored one.

I got off the train at the Zoologische Garten Station, and found myself surrounded by the most delicious of aromas. There were little lights and flashing signs all around. In stalls along the street, you could buy all kinds of exotic fruit: kiwis, bananas, and pineapples. There was nothing like it in the USSR, nor in Poland. There, in West Berlin, I was finally set completely free from the illusions of communism and my father’s words-that capitalism is cool-were confirmed once and for all.

In Berlin, Rina and I had to sleep at the station. Once, while we were walking along the street, I saw a hotel with a sign out front stating that they charged 50 marks per night for a room. This may sound cheesy now, but I said,

“Trust me, Rina. A day will come when I will be making money and we’ll be able to stay in that hotel.”

Later, I stopped taking the risk of going to Berlin without a foreign travel passport. Rina started to go instead. Apart from cigarettes, skirts and shirts were big sellers. At the open-air market in Warsaw, we bought black Turkish skirts with belts as well as military shirts (faux denim) with tags reading, “US Army.” Rina is thin, so she would put on five or seven layers of shirts and skirts.

At the station in Berlin, Gypsies would spend mark upon mark to buy this crappy junk. It is a mystery where they sold it. After all, the quality was revolting. In any case, though, we made good money selling it. Within 15-20 minutes, the Gypsies would gobble everything up and Rina would board the train heading back to Warsaw.

Some of our imports from Europe included gas canisters, pistols, and cartridges. All of these things sold well in St. Petersburg. By the end of summer 1990, we had made a few thousand marks. I used the money to buy a computer, which I took with me on an LOT airlines flight to Leningrad. At Pulkovo airport everything might have come crashing down. A customs official took one look at my suspicious facial expssion and said,

“Would you mind stopping, sir?” I ptended I did not understand what it was he wanted. He was distracted and I managed to slip through.

After I sold the computer, I flew to Tyumen and bought my first Lada 2109. The color was called “wet asphalt.” It cost me somewhere in the range of 25,000 to 35,000 rubles. Lada aficionados will know what I mean when I say it had “long fenders.” The license plate had “TYU” on it, which meant that I was Tyumenian, automatically. I barely knew how to drive and so my friend Sergei Abakumov helped me to get the car back to Leningrad. As we were coming into the city he said that he was tired, so I got behind the wheel. You had to see my steering as we drove past Moskovsky Department Store to believe it! Somehow, though, I made it back to Vasilievsky Island.

Rina was not pleased:

“I slaved away all summer long and wore ten dresses at a time for you-and you went and bought a car?” And she was right. While we were in Europe we pinched pennies everywhere and we often went hungry. We did not want to spend our foreign currency. In Germany, for example, a kebab would cost a mark, while in the Soviet Union you could survive for a whole week on the same money. We would skip dinner and have sex instead. We went hungry so that we could make money. After all that-pig that I was-I went and bought a 2109. I am sorry, Rina! Remember, though, how it took us a mere three hours to drive that 2109 from Vasilievsky Island to your home in Kohtla-Jarve?

I spent all of our money on the car because I was sure that I would make more soon-which I did. In 1990 I met a man named Andrei Rogochov, who later started the Pyatyorochka retail chain and became the richest person in St. Petersburg. We started as equals, opening a company called LEK-kontakt. He held a 50 percent stake, while I shared the other fifty percent with the Pakhomov brothers (better known as the Ilyiches). My trips to Germany became more serious. I got a foreign travel passport. Rina now stayed home, happy to get treats, such as pineapples, from Europe.

I brought cash into Germany, but not altogether legally. I would hide it in a mattress or-no need to fret-in my own ass. I would then buy fairly large batches of printer cartridges and toner. Andrei was in charge of selling these in St. Petersburg.

Once, when I was taking our assets to Germany, I came close to losing everything. One night on the train, after the other passengers in my compartment had fallen asleep, I carefully opened a stretch of seam on the mattress, put the money inside, and sewed it back up. At customs, I had to roll up the mattress and wait for the officer. He caught me off guard when he said,

“All right, take out your money.”

“What money?”

“In the mattress.”

Catastrophe. I broke out in a cold sweat. The problem was not just that I might lose all the money. There would surely be a criminal investigation, as well, and I might even end up behind bars.

“I don’t have any money.”

“What do you mean, you don’t have any money? You do…”

The official started pinching and pulling at the mattress-touching the very place where the money was hidden. But he did not feel anything! He rolled the mattress up again and said, “You’re right, there’s nothing.”

What was that all about? Had one of the other passengers snitched? Was the officer bluffing?

You know what I think: it was God, protecting me once again from very serious trouble.

In Poland and Germany I honed my mastery of business. I bought Xerox toner in this store in Germany. On Meeting a Polish Man

One day, I wanted to get some zloty, but the currency exchange was already closed. An elderly Pole came up to me, and asked,

“Did you need something?”

“I wanted to changed some money. I have some German marks, but I need some zloty so that I can get something to eat.

I must have looked a bit unkempt, so the man took me into a bar and bought a sandwich and some tea.

“Are you Russian?” he asked

“Yes,” I replied.

He began telling me the story of the emancipation of Poland by Soviet forces. In 1944 the Poles revolted, but they could not get the support of the Red Army and the Germans crushed the revolt. According to his version of events, the Russians declined to help the Poles on purpose, in order to get rid of the dissidents of the day. He also talked of the violence committed by Russians against the local population. I had been raised to believe that we saved Europe, so this was a shock to me. Here I was, sitting in Europe, hungry; one of our “emancipated” Poles was feeding me and telling me about how evil we were.

“So why did you feed me?”

“I have no pjudice against you. You’re a poor hungry student. But you must know these facts.”

Now I understood that the same historical events may be interpted differently by different people. The Soviet version of these events was very different from the Polish one: Konstantin Rokossovsky, the commanding officer of the First Belorussian Front, who later became Poland’s National Defense Minister, asserted that the Polish uprising was in no way supported by the Red Army.

Rina Vosman, Oleg Tinkov’s wife:

We trekked to Poland to make money. Time after time, I made the trip from Warsaw to Berlin wearing military-style shirts that were supposed to look like they were made out of faded denim and ugly black skirts with gold-colored buckles and elastic waistbands. Oleg could not do it, because he did not have a foreign travel passport. The Gypsies would scoop everything up within 20 minutes of my arrival at the station and I would have to keep my wits about me to make sure that I was not ripped off in the frenzy. I failed to understand the business. Did someone need this junk somehow? The Gypsies paid in marks. I could not get my head around the fact that people in Germany paid two Deutschmarks for a Pepsi. To me, that seemed like crazy money. In Russia you could survive for quite a few days on two marks. That is why we packed sandwiches and water. We would do anything to hold onto those marks. One day I found myself in a stressful situation. Usually, the customs officers were men. They would simply look the other way, as it were, when faced with a women bundled up in clothes for sale. This time, however, the officer was a woman and she started to strip search me. The next thing I knew, they had taken Juliusz and me off the train. We had to spend the whole night on the platform. Some Germans walked by with dogs that sniffed at us. We had this uncanny feeling-as though it were 1943 again. We sat there until the sun came up. Then we took the next train back to Warsaw. We were lucky not to have had everything confiscated. We escaped with minor bruises, so to speak.

After our first trip to Europe, Oleg got a photocopier to bring home and sell. After our second trip, he brought back two of them. After our third, he was driving a “wet asphalt” colored number 9 Lada with long fenders.

Nikolai Nikitich Zhuravlev, former psident of Kuzbassprombank:

I met with a huge number of clients during my time at the bank. Oleg Tinkov left the best impssion of any of them. He came to me, told me what his business was about, and asked for a loan. I liked his reasoning, so we gave him a million rubles. He bought various goods with the money and then sold them inpidually. He even came to our bank to sell stuff. The girls loved that kind of shopping. So we started giving him more. Oleg was always careful to pay back his debts. Later he got into more serious business. He started opening stores that sold household appliances. In spite of our forty-year age difference, we became friends. I watched Oleg as he worked. He had clear and specific goals and he always got right down to business. He can talk to anyone and is good at building strong relationships, qualities that were given to him by nature.

It is amazing how good a worker he is. He is highly energetic and picks everything up as he goes along. That is why I was not surprised in the least when I heard that he had opened a pasta factory and later a brewery in St. Petersburg. He is curious. I have been with him at different meetings in the Central Bank and he always showed a keen interest in the inner workings of the financial industry.

To tell the truth, if there were fifty people like him in Russia, then they could keep the economy growing. I think Oleg would make a fine Minister of Finance.

Chapter 12

From the Soviet Union to Singapore

One rainy autumn day in 1990, I parked my No. 9, once again, across from the Institute. Our drilling-and-blasting professor parked his No. 1 in the next spot over. He looked at me and we went together into the lecture hall. What could he teach me? Well he could teach me about drilling and blasting. But when it came to making money, there was nothing he knew that I did not. In the end, I never wrote my final exams. I never crossed the finish line at the Institute.

This decision followed logically from my priorities at the time. Why had I started my studies? Well, my goal at the time had been to return later to Leninsk-Kuznetsky and to work as a section superintendent in one of the mines.

The peak of my career, then, would have been becoming mine director. If that had happened, my pay would have been 1000 rubles a month and I would have been given a Volga to drive. In my third year of university, however, I was already earning 10,000-15,000 rubles each month and the prospect of becoming a mine director held no attraction for me whatsoever.

Everything I do is based on economics. Sometimes, of course, I am motivated by charity, care, a desire to help, but I believe that if a person spends dozens of hours a month on something, he should reap the rewards. From that perspective, it seemed there was no point at all in continuing my studies at the Mining Institute. Moreover, with the help of the widely respected Novosibirsk businessman, Voldemar Basalayev, I and the Ilyiches had gotten into the car business. This business was nothing out of the ordinary, but the potential profits were huge and it would take some time to achieve them.

We came up with the idea of delivering cars by air. At the Chkalov factory we got soldiers to agree to take our cars on flights to Moscow or, less often, directly to Leningrad. Two cars could fit in an An-26. We paid the soldiers 5,000 rubles cash for each car and then drove them into the cargo hold. We stayed in them during the flight, which included a refueling stop in Chelyabinsk.

Every few days, my neighbors would be shocked to see me driving up to my building on Nakhimov Street (nearby the Pribaltiyskaya Hotel)-where Rina and I paid 500 rubles rent per month for an apartment with just one room and a kitchen-in a brand new car, either a 2108 or a 2109. We did not worry in the least about selling the cars at the market. Instead, we would just sell them at a slightly reduced price to people we knew-people who would actually go to the market and sell the cars there. I had around twenty cars registered under my name at the Department of Motor Vehicles.

Just think how ineffective the economic policies of the Soviet Union were. They would assemble a car in Tolyatti and ship it 2500 kilometers to Novosibirsk, via Ufa, Chelyabinsk, and Omsk. From there we would fly the car to Moscow and then drive the seven hundred kilometers from there to Leningrad. But we would still manage a huge profit margin. That is how inefficient the system was!

And I was not the least bit surprised that 1991 was the year in which the USSR collapsed. Events were unfolding rapidly. It was hard for me to keep on top of it all. Starting on August 19, a group of Communist Party hardliners put Gorbachev under house arrest at his summer cottage in Foros. Then they announced the creation of the USSR State Emergency Committee. The committee was made up of Vice President Gennady Yanayev, Prime Minister Valentin Pavlov, KGB chairman Vladimir Kryuchkov, Minister of Defense Dmitry Yazov, Minister of Internal Affairs Boris Pugo, First Deputy Chairman of the Defense Soviet Oleg Baklanov, Chairman of the Farmers’ Union Vasily Starodubtsev, and the President of the Association of State Enterprises and Industrial, Construction, Transport and Communication Facilities, Alexander Tizyakov. I remember how Yanayev’s hands were shaking when the state of emergency was declared. I could already see that the people who were trying to seize power had no control. None of them were particularly enthusiastic about their cause.

The people involved in the coup really believed that somehow the USSR could be saved. All they accomplished, however, was to ensure that its downfall was irreversible. The people were already taking big gulps of freedom and no one liked the bans that the SEC was trying to introduce. No one took to the streets in support of the committee. In contrast, the President of the RSFSR, Boris Yeltsin, who was leading the fight against the coup, garnered the support of hundreds of thousands. Thank heavens, the coup was over soon enough: on August 22 the members of the SEC were arrested and Mikhail Gorbachev returned to Moscow. Real power in Moscow had been transferred to Yeltsin though. The “parade of sovereign states” began: on August 24, Ukraine declared its independence, on 27 August, Moldova declared its sovereignty, Kyrgyzstan did so on August 31, and so on.

On September 6, the Presidium of the Supme Soviet of the RSFSR issued an order, renaming Leningrad Saint Petersburg. Of course, I was happy with this decision, as I had a clear understanding of the role Lenin played in Russia’s history.

It so happens that I spent my childhood in Leninsk-Kuznetsky, went to cycling camp in Leninabad, and later moved to Leningrad. Both my childhood and my youth, then, were connected to Lenin. When I was young, Soviet propaganda encouraged us to deify him. It was only at the end of the eighties, when I was in Leningrad, that I realized he was simply a Jewish weirdo who had made an agreement with the Petrograd bankers of the time and plunged the country into poverty, essentially destroying Russia and the Russians. He brought suffering upon the Russian people. We are still suffering. I would have had him burnt at the stake.

The last nail was hammered into the USSR’s coffin in Bialowieza Forest on December 8. Boris Yeltsin, Stanislav Shushkevich, and Leonid Kravchuk signed an agreement: “We, the Republic of Belarus, the Russian Federation (RSFSR), and the Ukraine, as founding states of the Union of Soviet Socialist Republics, hereinafter called the Supme Parties to the Agreement, having signed the Union Agreement of 1922, hereby declare the dissolution of the USSR as a subject of international law and as a geopolitical reality.”

Thus Gorbachev became the psident of a non-existent country. On December 25-on my birthday to be pcise-he retired.

By that time, the country’s economy was faltering badly and had reached a dead end: the government kept issuing unsecured money, which led to major deficits and a massive rate of inflation against the US dollar. The government simply could not go on regulating prices; regulation had bled itself dry. As we looked on, the ruble lost its value and respectability. Everyone was trying to get rid of any rubles that they had, buying foreign currency or goods. The shelves were empty. The whole country was collapsing. What was there to do? On November 6, Boris Yeltsin appointed Yegor Gaidar as deputy chairman of the economic policy committee of the RSFSR. Mr. Gaidar decided that shock tactics were the best medicine for the economy. On January 2, 1992, shoppers discovered that prices had increased enormously.

To tell the truth, however, these problems were no worry of mine. I kept my savings in dollars, but constantly turned them over. The value of the ruble against the dollar was shrinking faster than the prices of goods were growing. In January 2002, I was worth a ten thousand dollar wad of cash. I took that same pack of money with me on my first trip to Singapore. I really started making big money on that trip and, at the same time, my Tekhnoshok retail chain put down its roots.

Igor Sukhanov let me in on the Singapore idea. This man was a renowned speculator who went by the nickname Dushny, which means “soulful.” I bought computers and fax machines, which I was able to sell for a total of 30,000 dollars on the day of my return. I really liked this three-to-one business model and my trips to Singapore became very frequent indeed.

Each trip took a few days and a visa was required, but we had our tricks. I got some people at Aeroflot to give us a few stickers that were normally used to change the dates on tickets. In Singapore, we would stick them on our tickets to make it look like we were leaving the next day. They would let us into the country and then we would throw the stickers away. When our day of departure arrived, the customs officials would sometime noticed that we had exceeded the twenty-four hour visa-free period. They would put a red stamp in our passports, indicating our violation. Never once, however, did this lead to further problems. The officials in Singapore had the right to arrest us at the airport upon our third violation. In view of this fact, we always got new foreign travel passports after our second warning.

The idiocy of the Soviet system played into our hands. First of all, when one was leaving the country, it was possible to exchange 300 rubles for dollars at the government rate. In order to do that, we would have to stand in line for two or three hours at the Vneshekonombank on Gertsen Street. But it was not quite so simple. The wait-time was only two to three hours if you bought your place in line (another widespad phenomenon of the early nineties). Basically, there were people who made money by waiting in line outside, all night, in order to sell you their spot in line in the morning.

Secondly, because of the low dollar-exchange rate, business class tickets turned out to be incredibly cheap. In the West they would cost 1000 dollars, but in Russia you could get them for around 600 rubles. In other words, at the black market exchange rate of 15 rubles on the dollar, the cost of a business class ticket would work out to 40 bucks.

Igor Sukhanov gave me lessons on how to transport cash. Using Mr. Taya’s company, Future Systems Electronics, we would hand it over in Russia, then picked it up again upon our arrival in Singapore. This was an ideal method-particularly because it meant that I did not have to shove the bills through my back door as I had done on those earlier trips to Germany. We bought calculators, toner, photocopiers, computer parts, and even fax paper from Mr. Taya. If you could sell it in Russia, for a profit, we got it.

On the way back, I did not want to pay five dollars for every kilo in excess weight, so I would raise the scale from underneath with my foot. The important thing was to make sure that the needle remained stationary and did not jump around. One time, after I had held it up so carefully, the airline worker decided he was going to re-weigh the bag. I had no idea how hard I had been pushing up the first time. I cannot deny that, now and then, we would be asked to take a couple of steps back from the scale.

My approach to business was different from his. I always liked long cycles: because the prices there were much higher, I would sell my product in Kemerovo, Novokuznetsk, and Leninsk-Kuznetsky. For instance, I could get 2000 dollars in St. Petersburg for a computer that cost me 1000, while in Siberia the same computer cost 3000. Svetakov, however, liked fast cycles: he would pay 1000 for a computer in Singapore and sell it directly in Moscow for 1500. Each of us has our own approach. I do not like fast wholesale money, but try to squeeze every penny I can out of the process. Big markups are my weakness.

To be honest, it took some small kickbacks to get it done. The price depended on what you negotiated and on the amount of digits the calculators displayed (8, 10, 12 or 16). I did not like having to pay 14-25 percent just to get the money out. Plus, it was really risky business. I had to go to Moscow to get the rubles, in cash. Then, I would carry the bags with me on the train to St. Petersburg, on edge the whole way. Next, I would go to Vasilievsky Island and buy dollars from rich kids at the Gavan and Pribaltiyskaya Hotels. It was a ptty hodge-podge procedure! But I managed to find a way around it soon enough. I learned how to buy non-cash dollars for non-cash rubles, which I would then transfer directly to Singapore to pay for the hardware through joint ventures that were entitled to carry out wire transfers.

Before too much time had passed, I managed to close a very large calculator deal. Procurements at the yarn factory in Leninsk-Kuznetsky were done through a rather strange character. He contacted me himself and said he was looking to buy three thousand Aurora calculators. The record remains silent on the question why these yarn-spinning women needed such a massive number of calculators. But the enterprise was state-run, which meant that it did not really belong to anyone. This procurement worker was accountable to no one. I sold the calculators to him and earned a hundred grand in the process. I think it was purchases of this kind that led to the plant’s ultimate bankruptcy.

In those days, the sums I earned selling calculators were colossal. I was able to buy a two-bedroom apartment in a modern 137 series building on Korolyov Street, near Kommendant Airport. In order to establish my residency in St. Petersburg, I had to marry a local woman, Nina Iosifovna. Born in 1927, she was 40 years my senior. At the marriage office, everyone looked at us like we were crazy. When I gave a bouquet to the female marriage registrar, though, she smiled and said,

“I’ve got you guys pd out.”

Later I found a fake Leningrader husband for Rina as well. Here was another holdover from the Soviet system: even if you had money, you could not buy an apartment in a city unless you were registered there.

We bought a dog, a boxer, and started thinking about having kids. Every seven to ten days I would fly to Singapore. When possible, I would make two trips a week. It worked out, then, that I was in the air for fifty-six hours each week. And with every excursion, I doubled my capital.

My business grew by leaps and bounds. I could not fly to Singapore anymore, so I started shipping my merchandise on cargo planes. I would do the receiving and customs paperwork at Pulkovo Airport in St. Petersburg. Because of the immense pssure, my business methods became more civilized. In September 1992, Rina and I went to the Municipal Executive Committee to register a limited liability partnership, Petrosib. We chose a transparently honest name: I shipped electronics from St. Petersburg to Siberia. At the entrance to the building, which once housed the Kalininsky District Party Committee, we were told that, “You can register one of those businesses upstairs.”

After the failed coup, Yeltsin had outlawed the Soviet Communist Party and now big wax seals, labeled “sealed,” hung on the door. It was a criminal offense to tear off one of those seals. It is a real shame that, later on, Yeltsin betrayed himself by allowing the Communist Party to come into existence again. It would be better if those doors were still sealed and if people like Gennady Zyuganov, the current Communist Party leader, never had any say. Germany had forbidden Nazism; in consideration of its history, Russia should have made communism illegal as well. I am not really all that interested in politics, because it is irrational to bother about things that you have no control over. As a citizen, though, I am obliged to state my opinion.

Nikolai Nikitich Zhuravlev made a phone call from Kemerovo to Promstroibank in St. Petersburg and helped us to open both a US dollar and a rubles account for Petrosib. I hired an accountant, Nadezhda Ivanovna Turukhina. In this way, in the autumn of 1992, my operations became completely legal.

I had begun to get my bearings in Singapore and I switched suppliers, abandoning Future Systems Electronics for Cut Rate Electronics. This new company was headed up by an ethical Indian businessman, Ashok Vasmani, who everyone called Andy. Mister Andy. The stuff he sold may have been of slightly lower quality, but it was still cheaper.

One day, when I was buying yet another consignment of Record televisions, he asked me,

“Oleg, why don’t you get a container?”

“A container? How many TV’s is that?”

“Three hundred and twenty.”

“But that’s over sixteen thousand dollars. Plus you have to pay five thousand for the container. And then you have to wait forty days. I can’t take that much money out of circulation.”

“Correct, but when you send it by cargo, you’re paying five dollars per kilo. If you send them in a container, it’ll cost you almost half as much.”

I paid for half the container and convinced Andy to loan me the money for the other half. Forty days later, I was doing customs clearance at the St. Petersburg port. My partner, Andrei Surkov, and I unloaded the container and stored the 320 television sets at the Petrosib office at 10 Sadovaya Street. We put some on display in one of the rooms, set up some fake trees, and hung a digital clock on the wall.

We hung up a banner reading “Cheap Televisions” and instantly people started coming, asking questions and making purchases. The turnover was slow, of course, but our sales volumes increased and our profits grew with each TV that we sold.

The calculators made me tons of money and now, too, these TV sets. We got more of them, but it was getting harder and harder to sell them in St. Petersburg-even at a low price of $350. We started having them delivered to other regions. In Siberia, a TV cost 500 dollars. We registered more companies: Petrosib-Novosibirsk and Petrosib-Omsk. We used the Regional Supplier system for warehousing. This was highly profitable, as small-scale retailers from the district centers usually went to the Regional Suppliers-and our TV’s were right there. Two years later we did a count and were shocked: we had sold 300 containers of televisions!

Nikolai Nikitich Zhuravlev had given me my first money loan; Andy loaned me product worth much more. At one point, I owed him a million dollars. But he took the risk and trusted me and, in the end, both of us earned a good chunk of money in electronics sales.

Andy left the electronics business and is now the owner of one of Singapore’s biggest Indian restaurants. We remain close. My wife, some friends, and I flew to Indonesia recently via Singapore. We dropped by Andy’s place, tried his different dishes, and listened to eastern music.

Andy is a person who believed in me and helped me to build my career. He is another gift that fate bestowed upon my life. Everyone needs to meet someone along the way that believes in you. There is no other way to become a businessman. I sincerely hope that every person finds his or her own Andy.

1992 was a very difficult year for the country and a simply fantastic one for me. I was exhausted from the constant running around and now had the money to take a little break. Vyacheslav Butusov, a Russian singer, sings a song about America. Part of it goes like this: “It took a long time for us to learn to love your forbidden fruit.” As it turns out, he was absolutely right. As soon as I had the time to do it, I took off to the states.

I bought my first imported car, a Ford Orient, in 1992.

The Petrosib team next to the famous fake trees that I had purchased in Hong Kong.

Ashok Vasmani, nicknamed Andy, and Nikolai Nikitich Zhuravlev are people who helped me so much to do business in the early nineties. St. Pete’s speculators in Amsterdam: Igor Sukhanov, nicknamed Dushny, Igor Spiridonov, myself, and Oleg Korostelev. At first, San Francisco made no impssion on me. Valentina Vladimirovna, Oleg Tinkov’s mother:

When Oleg was training in the cycling team, he’d sometimes bring some stuff home with him, like scarves and arm warmers. I was worried, because I didn’t know where he was getting it. I got on his case. When he started doing business at the Institute, I didn’t get in his way. He was an adult now. He met Rina, studied, and made money on the side. One day he borrowed 150 rubles from me; he said he wanted to buy something. Later he made it back and sent me a wire transfer. But I sent him the money back. He needed it more because he was far from home. But in the end he dropped out of school after his third year and dedicated himself completely to business.

On fake trees

One of my clearest memories of my business in those days involved my trip to Hong Kong. Sankin introduced me to a former classmate, Max, who lived in Beijing. Max told me I should look into selling fake flowers and trees. I flew to Hong Kong to buy some. At that time it was still administered by Great Britain. All we did there was to buy a copy of the Yellow Pages, find a manufacturer, dial his number from a payphone, arrange an appointment, and go straight to the factory. It was not far from the airport. Leafing through the catalog, we called Moscow and found out that the same stuff cost five or six times as much there. My greed was my ruin. When I saw the potential for a huge net profit-sixfold!-I bought not one, but three containers. A fully loaded forty-foot container cost around twenty grand, so I paid sixty for the three.

On the one hand, I made the sixty back quickly, covering my investment. On the other hand, though, I had a lot to sell! So I started doing away with the plants in other ways. I took some to my house, sold some to my friends, and gave them to my employees as bonuses. Two years passed, but I just could not get rid of those trees and flowers!

So I called the head of our Kemerovo office, Svetlana Alexandrovna, and told her,

“Do something with these flowers!” A natural-borne salesperson, she had worked as commercial director at the Regional Supplier and she could sell just about anything. She is best described with a metaphor:

She will stop a horse in mid-gallop, enter a burning house, and convince someone inside to buy something!

She sold a bunch of the plants at a good price and then said,

“Oleg, I found a client that’s willing to buy everything we have.”

“How much?”

“They want a 60% discount.”

“Sold!” I did not want to have to deal with those stupid trees anymore. “Who’s buying, anyway?”

“The Kemerovo Funeral Home.”

There were so many flowers that, to this day, they are probably making wreaths from them for funerals in Kemerovo Province. On a side note, there is always good money to be made in the funeral business. A client who is under intense emotional pssure and on a very tight schedule will not try to talk you down. That is why funeral businesses drive their prices up.

Thus my business portfolio includes two strange deals in Kemerovo Oblast: the sale of three thousand calculators to a yarn factory and a truckload of fake flowers to a last-rites business. It is hard to know whether to laugh or cry.

Andrei Surkov, Oleg Tinkov’s partner at Tekhnoshok:

Our meeting was based on speculation. Oleg lived in a Mining Institute dorm on Shkipersky Stream, while my dorm was on Nalichnaya Street. We students would buy here, sell there. On this basis, some of us became business partners or even friends.

Later I got a job in a company, so I could understand what goes on when you don’t have to do all the running around the city with bags yourself. I wanted to see what it was like to run a more or less civilized business, in an office, with other people and some kind of organization. It was at that time that Oleg started actively pursuing business in Singapore. He would fly there to buy ink, toner, photocopiers, and calculators. He called me in 1991. We met up and went to the bathhouse, where he offered me a position as his junior partner, working with electronics. I agreed immediately, because I considered Oleg a good person and a competent businessman. At the beginning, our work at Petrosib consisted in the following: once a week Oleg would fly to Singapore for office supplies, calculators, and toner. Then he’d come back. A few days later, he would leave for Singapore again…

On selling liquor

For around a year Igor Spiridonov and I imported liquor into Russia. For around two years there was no extra fee for alcohol. One of the best products in those days was Royal, a type of hard liquor from Holland, but we did not know anyone there. There was, however, a small-scale plant in Hungary with which we knew how to do business. We ordered the liquor in Budapest and, once in Siberia, it sold well. Igor placed the orders and I was in charge of sales and payments. A half-liter bottle of liquor cost us between sixty and sixty-two cents, including delivery to Russia. A container held twenty-two thousand 500-mL bottles or seventeen 700-mL ones.

At first we got five thousand bottles, then ten thousand, then a container, then two, until we reached a maximum of ten containers. A funny thing happened with that last contract. We ordered five containers of 500-mL bottles of Dolce Vita and another five containing 700-mL bottles. The bottling plant mixed up the labels, but it did not matter. We sold the liquor with the wrong labels.

Chapter 13

Why Hello, America!

In the early nineties, in Russia, if you were a foreigner it was as if you had blue blood. It did not matter if you were a simple Italian plumber or an American mover. From our point of view, even foreigners coming to Russia on a tour package that cost them all they had seemed like billionaires. They wore Reebok or Nike sneakers and leather jackets, signs of great wealth during the breakup of the USSR. We called them businessmen. Now I understand that these were low-budget tourists, but in the midst of the rampant poverty, they seemed super rich. That is why everyone wanted to hang out with foreigners. Male university students chased after them, hoping to make a buck; women followed them around so that they could get into US-dollar bars like the ones at the Pribaltiyskaya or the Gavan Hotel. If they were really lucky, they might be taken to the Grand Hotel Yevropa. Better yet, they would get married and move away. Not that every story had a happy ending.

Every Wednesday at the Kirov Cultural Center there was a party for people over thirty. Those parties seemed really lame at the time and now, too, when I am over forty myself, they still seem like a silly idea. I do not know what possessed my classmate Sasha Sankin to go there. Maybe it seemed like it would be easier to meet a woman there and take her to a hotel-because people over thirty are more easy-going. What actually happened though was that he met an American woman over forty years old, they had sex at the dormitory-and she fell in love with him!

She was in love with a poor student twenty years her junior, who had moved to St. Petersburg from Tashkent! In the end she invited him to move to a small town called Santa Rosa, 30 miles from San Francisco, a typical Californian town with a population of about one hundred thousand. Not far away is the famous Wine Country, where there are thousands of wineries dotting the valleys of Napa, Sonoma, Alexander, Bennett, Dry Creek, and Russian River.

Santa Rosa lies along the Russian River; at its mouth, on the Pacific Ocean, stands the town of Fort Ross; it was the southernmost Russian colony during the early 19 ^th century. It was at Fort Ross that the Russian ships Yunona and Avos, made famous by Andrei Boznesensky’s and Alexei Rybnikov’s opera, made landfall. It is not a made-up story: in 1806, according to official records, the Russian aristocrat Nikolai Rezanov actually met and fell in love with Concepción Argüello, the daughter of the Spanish Governor.

The Russians left Fort Ross in 1841. From an economic point of view, there was no reason for them to stay there. In 1867, Alexander II sold Alaska to the Americans for 7.2 million dollars in gold, but the Russian colonies on the Pacific coast were not included in the transaction.

By a twist of fate, then, Sankin ended up in a place that had been historically Russian. He had been living there for a year already, but I had no idea; I was simply sitting in my Petrosib office, wearing a raspberry-red blazer. As soon as I found out he had moved there, I got hold of his telephone number. At that time, it was not easy to place a phone call to America. I went to the Central Post Office, waited in line, and got through to Sankin. It seemed miraculous-just as placing a phone call to Mars would now.

“Hi, Sasha! This is Oleg Tinkov. So you’re really in America? That’s awesome!”

“Hey, Oleg! Yeah, I’m slowly getting settled in here.”

“How’s your American wife?”

“We recently got porced…I got a Green Card and am official here now. I brought my dad here from Tashkent. I’m renting an apartment. I work a power lift at Friedman Brothers. We sell home hardware.”

“No way! Can I come visit you?”

“Fly on over. I’ll help you out when you first get here. You can stay at my place.”

Getting my visa was a headache and a half. I tried everything. For a small fee-or maybe out of the kindness of his heart (I don’t quite remember the details)-a friend from the Leningrad Army Sports Club hockey team set things up to appear as though I had been hired as support staff. My height, at six feet, four inches made me convincing. In December 1992, I came to the consulate for an interview.

“Okay, we’ll give you the visa, but how are you going to pay for living expenses?”

“I have a credit card.”

When I was in Singapore, I broke Russian law by opening an account at Citibank. I got a Visa Gold card. By then, my credit card history went back 18 years. This made an impssion on the consul: in St. Petersburg, out of a population of five million, there were perhaps a thousand people who had a card like that. How did someone who worked as part of a hockey team’s support staff end up with a Gold Visa? The consul refrained from asking and gave me another visa-in this case an American one.

I partied over New Year’s Eve. Then, in January 1993 I got into an Il-86 airplane on an Aeroflot flight from Sheremetyevo airport in Moscow to San Francisco via Anchorage. My surroundings shocked me: there were Jewish refugees, crying children, and a bunch of mesh bags. I was stunned by the smell of the San Francisco airport. Anyone who has flown in America knows that unique airport smell. There were a lot of iron doors and police shouting into megaphones:

“Go right!”

“Go left!”

The movie Gangs of New York with Leonardo Dicaprio reminds me of the imposing feeling I had then, at the beginning-the sense that America resembles a big prison. At the airport, if you are an outsider, they immediately make sure you know that everything is serious, that everything is under the government’s control. Big Brother is watching you! Sasha Sankin met me and drove me around San Francisco in an old Toyota. I was sleepy because of jetlag, but nevertheless we went to a bar to have a beer. At first glance, I did not like the city. It seemed strange, unintelligible, and unkempt. Today I think that it is the most European and the most beautiful city in the US. I spent close to 5 years there. If I ever decided to move to America permanently, I would settle in San Francisco.

We drove thirty miles to Sasha’s small house in Santa Rosa, past the famous Golden Gate Bridge, which I had often seen in Hollywood movies. The house really did look like it was made of cardboard, which is something people usually say about American houses. I was also surprised to note that Sasha’s father was very angry, aggressive, and bitter. Towards himself, towards Sasha, towards me-he resented everyone. They would get up at six in the morning and leave the house, making sure the heater was turned off, because they wanted to save money. The intense cold would wake me up. I realized I would not be able to sleep, so I got up right after they did. Welcome to capitalism!

When I arrived in America, I knew practically no English. I gawked at my surroundings, dumbfounded. It was hard to learn the language. Now, though, my speaking and writing skills are not all that bad. I make mistakes, but I doubt that my written Russian is much better.

Now, naturally, simply having a good time was not my only reason for being in America. I wanted to start something. That same January I went to a government office in Santa Rosa and registered a company, California Siberia Enterprise. Between the paperwork and getting a stamp made, the procedure took about an hour. Next I went to Kinko’s, where you can pay for office services. I leafed through the free templates and found an image of a Siberian Bear, which I decided to use as my company logo. Everything fell together: the bear symbolized Siberia and the yellow and green motif repsented California. I had a bunch of business cards printed, right away, which read:

California Siberia Enterprise Oleg Tinkov President

In 1993, for my part, I had a lot more than business on my mind. I was trying to find a way to stay in America. In order to get a Green Card (i.e. permanent residence), I had to go often to the INS (the United States Immigration and Naturalization Service), the authority that handled immigration.

In America, I became authentically Orthodox. I had already been baptized in Leningrad on December 25, my birthday, in 1988, but I seldom attended church. In Santa Rosa, a lot of local Russians gathered together at church. I found that I enjoyed going there as well. It was an outlet, the only place where I could speak Russian. The priest, congregants, and I would drink tea and eat crêpes after the service. I fell even deeper in love with Russian culture, the Orthodox religion, and the church. I was drawn to it.

Through the church I met a lot of “Old Russians,” descendents of White émigré families. Like me, most of them traced their roots to Siberia. Their ancestors had escaped the Bolsheviks by moving to Harbin, China. When China had its own revolution, they left by ship, traveling to Brazil and Venezuela, eventually settling in California. I hung out with these old timers who spoke three languages: Russian, Chinese, and English. I saw Russian ladies wearing veils. There, in the USA, in church, from the mouths of these old Russian immigrants, I heard the most articulate and beautiful Russian I had ever heard. In that atmosphere, I came to be even more convinced that Orthodoxy was my religion. Most importantly, however, I was able to meet a “different” kind of Russian, people that had not been affected by the Soviet system. They counseled me on how to adapt and took pity on me. One even gave me a mattress so that I could sleep better on Sankin’s floor. It was in America that I realized what Russia had lost.

In reality, the Russians had no relation whatsoever to the Russian Mafia-a popular topic of conversation in the States. There was a gang, for instance, active in San Francisco, that had been responsible for several murders. After they were caught, the newspaper printed a picture showing them with the Russian church in the background. The headline read, “Russian Mafia finally decapitated.” The article featured surnames such as Zimmerman and Lerner. This was offensive and insulting to the Old Russian intellectuals. Between the late eighties and early nineties, members of the noveau riche began immigrating to the US from the USSR, including Jews, Ukrainians and Moldovans, among others. Their Russian was grammatically incorrect and they hated the Russians, but the Americans still referred to them as Russians. The public automatically attributed all of their unsightly actions the “Russians” in general and the “Russian Mafia” in particular.

Let me talk a bit more about the so-called Russian Mafia. In 1993, shortly after I arrived, I went to the Russian restaurant StageCoach, which was a dance club on Saturdays. As usual, some of the local big-shots tried to pick a fight with me. They had watched a lot of post-Soviet movies were trying to look like the gangster characters in them. Really, they were trying to look like “brothers in arms” from their historical homeland, but in San Francisco they just looked cartoonish. There were some serious types there, mind you, like Pasha Ulder, whose brother was shot dead by the Chinese the night before I first met him.

When they started harassing me verbally at the bar, I was wearing black Versace from head to toe. I wore a diamond signet on my little finger and I had a scar on my face. In other words, by their standards I was a “dude” and maybe even a big-timer. I played along. I started talking like an ex-con. They decided I was one of them, befriended me and, in the end, they did not touch me.

And of course His Majesty Luck helped me out. A fellow Siberian, Nikolai Nikitich Zhuravlev, had come to visit me and, in accordance with Siberian tradition we decided that we wanted to visit a bathhouse. We found out that some Jewish immigrants ran a Sauna in downtown San Francisco, which was supposedly similar to a real Russian bathhouse. So off we went. The service was revolting, though, the place was a sanitary train-wreck, the temperature only reached 50 degrees, and so forth. It was a scandal. I started making demands and got into a battle of words with the owner’s wife. She turned out to be the sister of that same Pasha Ulder I mentioned above. On top of that, she was the girlfriend of my good friend from Odessa, the legendary Zorik. Zorik is an interesting specimen in that, even after having spent 25 years in the States, he still could not speak English at all. He is also well known for some interesting stories involving drunkenness and experimentation with drugs.

At the same time, however, Zorik is the most talented barber I have met in my entire life. He cuts hair without looking, very fast and with great confidence. I have known him for fifteen years and have never heard of him having an unsatisfied customer, man or woman. His shaving skills are to die for. If you are ever in San Francisco, make sure to visit him at the Backstage Salon on Green Street.

But let us get back to the sauna. After we left, the place burnt to the ground. The next morning, Pasha called me and said that, bro-to-bro, he realized that the proprietors had been in the wrong, but that he thought my reaction over the top. This was really and truly funny, but I did not try to set him straight. In the end I became a legend in San Francisco, and I never had any trouble with the local gangsters again.

In April, the long-awaited day arrived: I arrived at the airport in the ten-year-old red Ford that I had bought for four thousand, shaking with anticipation. I do not know how I drove her back to Santa Rosa. Can you imagine? Three months with no sex. We had a most authentic Parisian wedding and, 9 months later, on December 31, 1993, our first little miracle, Daria Tinkova, entered the world.

In the morning Sankin’s dad came into the room and told us with anger in his voice,

“We couldn’t sleep all night. Our walls are like cardboard. It’s over, get out of here.” He was also waiting for his wife. He was a very strong man somewhere between fifty and fifty-five years old, and his heart and other organs could not handle our sex-so he just kicked us out. There we were: Rina and I, the mattress, the fax machine, the Ford, and couple thousand dollars in our pocket. Where did we go? To the church of course. We only spent one night in the home of an acquaintance. Immediately, they helped us to rent a room for 300 dollars a month. Sankin and I stopped talking to each other because he had not stuck up for me when his dad kicked us out. Later he admitted that he had been in the wrong: sure we had kept everybody up, but that was not the right way to react… Without his help, I had been left with no interpter, in any case, and, as a result, my skill in English began to grow more quickly.

In the summer of 1993, I bought a house in Santa Rosa. An Armenian guy named Dzhavayan sold it to me. Like the Armenian he was, he just had to sell me something. So he sold me his house, which cost 120 thousand dollars. I paid 20 thousand up front and borrowed the rest from a bank. My monthly payment was 600 dollars. I bought a massive two-storey house that I really had no need for at all. I sold it later for less than I had paid for it and so lost money-but no matter. The important thing to notice is that this dark-skinned Armenian managed somehow to dump the place on me. I still cannot p out how he pulled it off.

While we were trying to get our bearings in America, things in Russia began to turn sour once again. President Yeltsin got into trouble with the Supme Soviet. The deputies were displeased with Yegor Gaidar’s reforms and blocked the initiatives attempted by the psident and government. The psident felt that, as guarantor of the Constitution, he did not have sufficient power to actually guarantee it. In the end, the banal power struggle stretched on for a year. On September 21, Yeltsin signed a decree Concerning Gradual Constitutional Reform in the Russian Federation, whereby parliament was dissolved and elections to the State Duma were set for December 11-12. The Supme Soviet, however, staged a protest, which ended with the White House being stormed on October 3-4. Soviet Chairman Ruslan Khasbulatov, Vice President Alexander Rutskoy, and some of Yeltsin’s other opponents were arrested.

We flew into Moscow on the very day that tanks were shooting at the White House. We watched the events live on CNN at the Olympic Penta Hotel where Andy, my partner from Singapore, had gotten a room. He got phone-call after phone-call from his friends that day, asking if he was okay. After a couple of days, Andy left, saying that he would never come to Russia again.

“It’s better if I send you containers. You guys have tanks shooting at houses there,” he explained.

The country was suffering true political and economic ruin. The quality of health care was on the decline and we just could not risk dealing with a Russian maternity clinic. An acquaintance recommended a clinic in downtown Prague. We took a train to Lviv, Ukraine, which turned out to be just as messed up as St. Petersburg. I was surprised to find that all of the restaurants there were closed, but that for a small bribe we were nevertheless allowed in for something to eat. We got back on the train, rode to Prague, and rented a small apartment, for pennies, not far from the maternity clinic.

Right before the New Year, on December 31 at 8 p.m. local time (10 p.m. in St. Petersburg), Rina gave birth to Daria Olegovna Tinkova. The end of yet another fortuitous year was marked by true happiness.

Zorik, from San Francisco, is the best hairdresser on the planet Earth. In this photograph, Daria Olegovna Tinkova is only 5 days old.

Chapter 14

This is not a Dream-this is Tekhnoshok!

At a certain point I realized that retail electronics sales had become more interesting. Selling televisions in bulk in Kemerovo, Novosibirsk, and Omsk brought in less and less money. Finally, I closed my branches in those cities. I simply had to have my own retail chain!

Around the same time, we opened a Sony store on Maly Prospect on Vasilievsky Island. Sony’s management was skeptical about the idea at first, but had no choice but to come to terms with it. First of all, at that time there were no legal grounds for complaints: after all, we would actually be selling Sony products. Secondly, we were buying the equipment from Sony’s official distribution network, but got it on the gray market, in Singapore. The Sony dealers were not happy about us opening the store: they felt that they were Sony, and we were not. In essence, we undermined the dealer system. Because Sony had no other choice, though, they actually helped us out in the end, providing design elements for the store, brochures, and slides.

On March 23, 1994 the newspaper Delovoy Peterburg (“St. Petersburg in Business”) published an article that ran as follows:

Andy came to the grand opening in spite of his fear over those shots fired at the White House.

“You are very ambitious,” he told me as we sat in a restaurant celebrating the store opening. At the time, the word “ambitious” had negative connotations in Russia. I asked,

“Andy, what do you mean?”

“You’ll go a long way.”

I liked this wording much better. Andy explained clarified his use of “ambition” and I realized that it actually denoted a really positive quality. Thank God, to be ambitious in Russia is no longer equivalent to being a scoundrel.

The Sony store sold twenty thousand dollars worth of product every day and the profit was phenomenal! Management’s main task was to sew money bags!

In Singapore, not only did I encounter delicious food, I also learned new ways of doing business. I discovered that many people could use a single phone number. This was incredibly apparent at Future Systems Electronics. I bought a Panasonic phone station. It had three lines for incoming calls as well as eight outgoing lines. We had grown used to Soviet calls, which all sounded exactly the same. Consequently, I particularly enjoyed changing my tone every day. And of course I made the secretary say “Hello, how may I be of service?” as secretaries did in Singapore where, for example, Future Systems ‘ secretary always answered the phone, saying “May I help you?” It is quite possible that we were the first company in St. Petersburg to offer assistance to our clients. In the chaotic nineties, people would be bowled over by this kind of treatment. Some just hung up the phone.

The office was up and running and I did not have to be psent constantly, so in the summer of 1994 we took little Dasha and flew to Santa Rosa, to our house on Little River Avenue. Soon afterwards, I met Alexander Koretsky, a descendent of the old-time Russian immigrants. Sasha helped me with my English and I gave him tips on Russian. Together we opened the Petrosib USA office. The office fulfilled the same function as Sankin’s home had done, back in 1993. We would find an interesting product and sent large shipments to my retail stores in St. Petersburg. We even hired a secretary and a couple of workers.

I liked America and did not rule out the possibility of settling there for good. I found an apartment in San Francisco, on famous Lombard Street, with an excellent view of the Golden Gate Bridge and of the whole city. Rina and I really felt at home in the flat and we decided to sell the house in Santa Rosa and buy the apartment.

But we did not end up being able to settle in the States. My retail business was growing and starting to bring in good profits. We had no choice but to work on it directly in Russia.

Realizing the prospects, in late 1994 (believe it or not) I decided to hire a lawyer and a marketing specialist. I had read someplace that the best way to go about recruiting this type of staff was through an agency, a head hunter, as such organizations were still know back then. I decided to give it a try.

My first experience exceeded my expectations. I got in contact with the company BusinessLink Personnel and put in a request for personnel to fill the two positions. This company had been founded right at the beginning of the nineties. When Procter and Gamble first entered the St. Petersburg market, staff at St. Petersburg University did a lot of the recruiting. The experience was a success and so the same people registered a business. They continue to achieve big things to this day. Working with the agency was unlike anything I had ever done before: I went to their office and they brought the candidates, one by one, into a special conference room where I interviewed each of them.

I was immediately impssed by a young guy who had just completed his law degree with honors at the university, in the same department where Putin and Medvedev had studied. The kid’s name was Sasha Kotin.

He turned out to be a diligent student. We began to restructure the company, borrowing money and doing up contracts according to correct protocol. This really helped me out in the end. If it had not been for Sasha and his astute legal moves, I really would have been screwed during the first crisis in 1998. It is likely that some very bad people would have taken our business away from us. Sasha fought for every penny, for every dollar, and for our company’s reputation. He was incredibly loyal, even though he did not have partner status. He was very conscientious, sharp, intelligent, intellectually astute, and an effective manager. Later, tragedy befell Sasha-but I will talk about that in a bit.

As with Sasha, I was incredibly lucky with the marketing specialist that I hired. BusinessLink secured the services of Samvel Avetisyan on our behalf. He had been working as a science consultant at the State Public Library on Fontanka River and was earning around 200 rubles per month. I believed he was the right person for the job and so offered him three hundred dollars a month-really good money at the time. Today, Samvel is a marketing hero. He has his own company, Arkhideya, which develops marketing concepts.

My first experience with hiring through the agency turned out to be a huge success. Prior to that, in accordance with Russian tradition, I had hired my friends, or acquaintances, or people that had been recommended to me. Both Samvel and Sasha demonstrated insane levels of efficiency, skill, and knowledge. This was especially clear when you compare their work with that of the managers I had hired on the strength of word of mouth endorsements. If you are serious about being a businessman, then sooner or later you are going to have to turn to a recruitment agency. This is because hiring strictly from among the people that you actually know is a surefire way of destroying your business. If you want to derail your work and lose your investment, do exactly that!

It all started on September 1, 1995. The city was drenched in Tekhnoshok. We booked as many billboards as well could and called all of the radio stations in order to secure as many spots as they would give us. We hung banners across Maly Prospect and Mayakovsky Street. Oleg Gusev filmed a commercial that was shown on the St. Petersburg TV station. This was around the time that Gusev was filming music videos for Pugachova and Kirkorov-most notably for the latter’s hit song, ” Zaika Moya ” (“My Baby”).

The song in our ad went like this:

There’s simply no need to bring a shopping bag when you come, Tekhnoshok will deliver anything you need, straight to your home.

It was utter nonsense, but within an hour everyone knew about us. There were line-ups outside the stores. We kept running out of product. The store’s daily revenue was 20 to 50 thousand dollars. It was as though we were printing money. Armored cars with men in bulletproof vests came and went, to and from the bank, picking up our cash. My thanks go out to Promstroibank and Vladimir Kogan, who believed in me and gave me a seven million dollar line of credit. This was serious money in Russia at the time: in 1995 our two stores achieved sales volume to the tune of 20 million dollars.

Business was going well and so we began to expand. We rented more space and opened a third store at Kommendantsky Airport, with a fourth on Moskovsky Prospect. By the beginning of 1996, we had a full-fledged chain: five stores in St. Petersburg, two each in Omsk and Kemerovo, and one in Novosibirsk.

It is not surprising that, at this point, I was not actually acquainted with all of the people who worked for me. In this respect, one interesting story concerns something that happened in October 1995, when I dialed an internal number and an unfamiliar voice answered the phone. Naturally I asked,

“Who is this?”

“Who are you?”

“I don’t quite get it. Come to my office, we should talk.”

It turns out I had been speaking to our new marketing manager, Vadim Stasovsky, who had been discovered and hired by Samvel. He had met everyone in the office, but did not know all that much about me. Vadim probably thought that this encounter would end badly for him. But we got acquainted and had a straightforward conversation.

As a marketing specialist, Vadim’s achievements were huge. We would send him to nearby stores and he would count the boxes our competitors were selling. He would stand for two hours in the freezing cold outside one store, then another two outside another, counting boxes and trying to get an idea of their price. Then he would ppare enormous analytical reports on their sales volume. We realized that Vadim was good with numbers, so we transferred him to finance. At that time corporate finance was a different world than it is today. For the most part there were two elementary types of calculation: addition and multiplication. Whoever was strongest in this regard would get into finance. Vadim has been on my team now for 15 years. He was involved in all my businesses and he has an enormous reservoir of experience. Today he works at Tinkoff Credit Systems.

In the meantime, business went well. At the end of 1995, Rina, Dasha, and I were able to move into a new apartment on Kamennostrovskoy Prospect. The apartment, which was located in that majestic building where the son of St. Petersburg Governor Yakovlev would later buy a flat, cost 250 thousand dollars. That was a giant leap for me at the time. When we started remodeling, we took our designer with us to San Francisco. We brought everything-from furniture and drywall to plates and spoons-from the USA. We filled three containers!

On December 31, we flew to Chamonix with Uniland ‘s Oleg Leonov. His wife came along too. Oleg was already a very adept skier. I, on the other hand, had only gotten started at 28 (do not think that 28 is too old). I simply fell in love with the mountains at that time and since then have made good progress. I also remember well a call that I received from Andrei Surkov on January 1, 1996. He wished us a happy New Year and told us the he had sold 100 thousand worth of product the day before-on New Year’s Eve. People bought psents and I grew richer.

From left to right: I, Igor Spiridonov, Andrei Surkov, and Alex Koretsky at the Hungarian stage of Formula 1 in 1995. That summer I decided to be a blonde.

My 28^th birthday celebration with the Tekhnoshok team on December 25, 1995.

The first big publication on me was published on August 13, 1996 in the newspaper Delovoy Peterburg.

Vadim Stasovsky, member of the board of directors at Tinkoff Credit Systems:

I started working at Petrosib in early October 1995 as a marketing manager. I was hired by Samvel Avetisyan. At that time Oleg spent most of his time in America and only came to Russia for six months of the year.

The Oleg of 15 years ago was very different from the Oleg that we know today. Back then, you could see him coming to the office from a mile away: he’s here! Everyone started running around and panicking. The situation changed by leaps and bounds. Now people still react to him in a particular way (especially those who’ve only known him for 2 or 3 or 5 years). As compared to the Oleg I knew 10 or 15 years ago, however, the change is very significant.

It was like a whirlwind. He did not shout-he just had a certain way of talking. He had a natural manner of speaking.

The six-month periods when that he was in the office had a very different feel to them than the six when he was away. He managed to maintain pssure on everyone though, too, even when he was a thousand miles away.

Interacting with Oleg outside the office, on the one hand, and interacting with him at work, on the other, leaves the impssion that you are dealing with two completely different people. At the same time, though, you have to keep in mind (and I only realized this many years after I had met him), that much of the time he is just messing around. He can shout and stomp his feet, but afterwards you realize that the emotions he had been expssing were simply not there. Oleg was just a good actor.

Vladimir Malyshov, editor of Delovoy Peterburg (St. Petersburg Business) in the 1990’s:

Tinkov is an unusual and fascinating Russian business p. He wants to be unique, attractive, and creative. And, without a doubt, he succeeds in these respects.

We first met at his first official event-the opening of the Sony store in St. Petersburg. Petrosib, his company at the time, had been selling mostly wholesale electronics. Then his priorities changed and Oleg opened the first store in the city that repsented a Japanese corporation. That evening, in a basement store on Vasilievsky Island, around 20-30 guests gathered. Oleg was really nervous. He was running around the room with orders, scolding his employees as well as the caterers. He greeting the guests, smiled at the pss, and, now and again, hid in the back offices with important ladies and gentlemen from the district administration. It was clear that the man wanted to stand before his guests and before the assembled journalists, in all his fame and glory, as the “owner of the first brand name Sony store in St. Petersburg.” In order to get the material that I needed, all I had to do was to ask him a few typical questions about his business (investments, return, Sony’s terms, etc.). But I did not want him to experience the nervousness that interviewees usually feel and I did not want journalists from competing publications to eavesdrop on our conversation.

The situation was so stressful I almost got into an argument with him. I introduced myself in the normal manner, saying,

“I’m Vladimir Malyshov, business editor for the newspaper Delovoy Peterburg.” Then I asked him, “When can we talk quietly?” He replied gladly,

“Let’s talk right now.” And then he turned, suddenly, to greet another guest, who took him off to talk with some other characters. They chatted boisterously as they walked quickly into one of the back rooms.

I was tired of it: time was passing and I hadn’t gotten a single snippet of information.

He was 26 then and had only just begun his difficult journey into the business elite.

Chapter 15

Premonitions of the Crisis

In 1996, political instability returned. The communists’ prospects for returning to power looked very realistic, particularly in consideration of Boris Yeltsin’s extremely low popularity ratings. But the state’s machinery worked solely towards getting him reelected as psident. The idea was simple: the message was not so much in favor of Yeltsin, but rather against Zyuganov, who was associated with the return of the Soviet Union. The slogans that arose in this context are still well-known today-“Vote with your heart!” “Vote or lose!” The Kremlin needed to get younger voters out to the polling stations, voters that would not want the communists back in power. And the Kremlin achieved its objectives. Young people started to worry about a communist victory. Just prior to the 1996 election, Anatoly Sobchak, mayor of St. Petersburg, told me,

“Two locomotives are flying at full speed towards one another. If they hit head on it will be a tragedy for the country. We need to redirect them. We can’t give the communists their revenge. Everyone must vote!”

To tell the truth, however, Sobchak did not emerge victorious in the municipal elections. In the first round of voting, on May 19, he defeated his former deputy, Vladimir Yakovlev (garnering 29% versus 21.6% of votes). But prior to the second round, the losing candidates had came out and spoke strongly against Sobchak. On Monday, June 3, I woke up and heard on the radio that Yakovlev had won the election by 1.7% of the vote. I immediately imagined how crushed Sobchak, who was such a stately and aristocratic person, must have felt. What state must he have been in as he left his house, got in his car, and drove to the Smolny? He would have been crushed, not because he had to turn of his blinker, surrender the government-owned Volvo 740, and get into a private car; he would have been crushed by the defeat of his liberal ideals.

Anatoly Alexandrovich Sobchak: a great man; a true democrat; a true patriot of his country; a man with truly righteous convictions. Indeed, I do not understand how he became the person that he is-in the context of the Soviet Union. Only freethinking St. Petersburg could have produced such a person in those times. In 1989, when I was still a student at the Mining Institute, I voted for him in the elections for the Supme Soviet of the USSR and, on June 12, 1991, I voted for him as mayor.

I met Anatoly Alexandrovich for the first time in 1996, at an Alla Pugachova concert. Afterwards we went to the Pribaltiyskaya Hotel and sat at the same table. We met a few times after that, too, and probably conversed for a total of ten minutes. On one occasion he even held Dasha. A real politician should always take kids in his arms and hold them: for every kid he picks up, he gets two votes, one from the dad and one from the mom. For my part, I know Sobchak’s daughter Xenia quite well. This might not please her too much, but I have to say that everything that is good in her, everything that I like about her, she got from her father. I can say for certain that he is one of only a few people whose depiction in the media has been completely true to what they are in reality-as with Richard Branson, for example, or myself.

It is such a pity that Sobchak was hounded after his defeat, with his sudden death in 2000, under suspicious circumstances, being the inevitable conclusion. Otherwise, he might still be alive today, continuing to bring much good to the country.

In 1996 Sobchak campaigned convincingly, so we voted for Yeltsin. On June 16, during the first round of votes, Yeltsin took 35.3% of the vote-3.3% more than Zyuganov. General Alexander Lebed came in third at 14.5%. In view of this achievement he was immediately granted the position of Secretary of Defense. Thus, in the second round of voting, the votes he would have received went to Yeltsin instead. That is how it worked out. On July 3, Zyuganov took 40.3%, while 53.8% voted for Yeltsin. Later on, it came out that Yeltsin had survived a heart attack in June, but they managed to keep this fact a secret until the elections were over.

Of course, I voted for Yeltsin, not just because of what Sobchak had to say about him, but also because of my hatred for communists and my respect towards Russia’s first psident. My feelings about him are deeply positive because he gave us the opportunity to at least try a taste of freedom. What Gorbachev started, Yeltsin deepened.

I recall the feelings I experienced between 1993 and 1998. Some might say that it was a time of anarchy and chaos. In my opinion, however, it was a time of freedom. My complaint against Yeltsin concerns a different matter entirely: he came under the influence of his daughter, as a result of which he went too far, distributing state property among people close to his family.

I, on the other hand, privatized nothing, but rather strived to develop my own businesses. By August 1996, I had garnered a lengthy article in the “My Business” column of Delovoy Peterburg. The article was written by journalist Volodya Malyshov. A number of quotes from it are still current today.

One example pertains to structuring a business:

It is impossible to operate successfully over the long term without structure-without people sitting in offices, filling out paperwork, holding meetings, working in the warehouse, working at the counter. That is why we build our structure-we hire the best specialists, we equip them with everything they need to work, and we open new stores. Right now people do not get us, but in 10 years, we will see what has become of the “two friends selling cordless phones” and what has become of us. They are moving fast, but who knows where they are going.

Another quote concerns professionalism:

If you can find the people, you will find the money. Unfortunately there are not very many genuine people out there. Now we only look for professionals. We need people with healthy ambitions, people whose goal is not simply to make a thousand dollars or more a month, but who really want to grow their career. When we manage to find people of that sort, then we see them grow up within a year… Often we have to say good-bye to friends-if they are not professional.

Or on business objectives:

Our original business philosophy was to work for profit. I am more impssed, not by the amount of product we sold, but by the amount that the company earned from these sales. For me, the indicator of a business’ success is net profit.

It was not for nothing that Volodya Malyshov wrote about Tekhnoshok and me. Between 1995 and 1996 our revenue had grown from 20 million to 40 million dollars. The competition, however, had grown fiercer. The chain Eldorado had come to St. Petersburg and commenced some severe low-balling. They offered incomphensible prices. How did they manage to survive? On the basis of which profit? On some driving-force items, the profit margin fell to 5-7%. This is next to nothing in retail, where the overhead-everything from salaries to rent-is so high.

In 1997, I felt the slowdown and began looking for opportunities to sell the chain. The company’s annual revenue had reached 60 million dollars and we were always hiring. At the end of the year, I threw a corporate party at the Olympia Club on Liteiny Prospect. I did not know half the people there and this made me afraid. I felt really big, but it was a scary feeling. Now I have my fifth business and I can say for certain that if I am in my office and find that I am faced with a bunch of people that I do not know, that I cannot feel, as it were, then that means it is about time to sell the business. In any case, I had already made up my mind to leave the electronics retail business and to start a pelmeni, where Isayev speaks German with no accent whatsoever. That is why we always speak Russian at home, even though we could speak English or Italian. Dasha is also strong in French.

Let me get back to my studies at Berkeley. They were incredibly difficult for me. Kostya, who knew English better than I, helped me to get the gist of a few assignments, but I completed them myself. My studies-and maybe even my life-however, were nearly cut short due to an accident that happened about a month after I started.

I had driven from Berkeley, riding my new Ducati Monster motorcycle, and I went to Kostya’s place for his birthday. We hung out and I had a couple drinks. I wanted to park the bike, but later decided that I would drive back carefully. In fact, I almost made it back, but on the second to last turn, coming along the winding road, I lost track of my speed and wiped out. The bike’s footrest hit the pavement at an enormous speed and I flew, spad like a butterfly, towards the ocean. I got to my feet and saw my pinky dangling in my glove. Full of adrenaline, I drove home and woke Rina. She was in shock. I myself dialed 911 and found out first-hand how terrifically well that system works.

They asked me to stay on the line, talked to me, and kept me psychologically stable. The ambulance, fire truck, and police car came flying in a mere five minutes later. They loaded me in and took me away. The police officer could smell the alcohol on my breath and tried to do some analyses for his report, but the young paramedic insisted that that should not be done: I was in critical condition and had to be taken to the operating room. The officer asked me how much I had had to drink and I replied with my usual answer,

“Two beers.”

Back home, they would have chopped off my finger, but in America two doctors spent the whole day restoring it. Today my fingers are all crooked and I am missing a knuckle on my left pinky. Thanks are due to the doctors and the young paramedic though. Once again, I was saved by my guardian angel. If it were not for him, a 36-hour jail sentence would have been waiting for me upon my exit from the hospital, for driving under the influence. I would have lost my insurance, my license, and so forth. Welcome to American democracy!

Please, never drive if you have been drinking alcohol!

So guys, go to the States to go to school. It is of the essence that you do so. Go for a few months, at least, like I did-but for intensive study-because business education is better there. For Americans business is like mother’s milk. It is a nation of salespeople, a country of entrepneurs. Americans understand business better than anyone else. So do not think twice about it: go to Berkeley or someplace else, but make sure it is in America. And, too, your education will enhance your knowledge of the English language immensely-another reason to chose the States.

In total, between January 1993 and June 2006, I lived in America for six years. And I realized that Americans and Russians are two of the closest nations around. If two poles are the same, they repel each other; consequently, we love and hate one another. It really is true that Russians are very similar to Americans!-even more so than they are to the British or Germans who live closer to us, in Europe.

Of course America never became my second home country, but it has had more influence on me than any other country, apart from Russia. It was there that I learned how to understand business, entrepneurship, and liberty-each of which is so lacking in Russia. In June 2009, I wrote in my blog that “I don’t like Americans, I don’t like America, but at the same time I love Americans. I don’t like Russia, I don’t like Russians, but at the same time I love Russians and I love Russia. These are two countries that have melded together, in me, in a contradictory way.”

If we do not study business, how can we become effective in it? On the one hand, it is a bad thing that there are so few professional businesspeople in our country. But on the other hand, you always have to look for opportunities where things are negative. If someone is not doing their job correctly, do it better and win out over him! Russia still has a lot of niches where you can develop a business. If Russian businesses were only 20 twenty percent as effective and smart as American businesses, then, considering our natural resources and our talented people ( Yes! Russians are a lot smarter and more talented than Americans), ours would be the number one country in the world instead of theirs.

While I was living in America, I attempted to get into forestry. Andrei Surkov became my minority partner. He was to work in Russia and I in the States.

What was the idea? I found out by chance that a cubic meter of our round timber cost ten dollars if you bought it directly from a forestry agency, while the Finns would pay thirty dollars for the same amount. Later I found out how much timber cost in America. It turned out that you could not import round timber, only sawn, dry lumber. Americans are very concerned about the environment and their country’s wild places and so they are worried about the introduction of beetles and other bugs.

In America the price for hard lumber-the oak and elm grown in the south of Russia, for example, in Krasnodar Krai-ranged from 1500 to 2000 dollars per cubic meter. In Russia, taking into account the procurement, transport from the south, milling and drying, shipping, fees, and delivery to America, the cost was 200 dollars per cubic meter. Wow!

We created a partnership with Nikolai Vladimirovich Kozlovsky, the owner of the St. Petersburg bank Finansovy Kapital, with a 50/50 split of ownership. He allotted us some land in Tosno and gave us a bank loan; we bought three American drying machines. Andrei’s job was to fly to Krasnodar and buy choice timber. It was cut at the factory and then he had it loaded into containers and sent to San Francisco.

It seemed like the perfect scheme, but we miscalculated with one thing-the Wild West, which bared its ugly teeth.

The lumber market in the States is very structured and has evolved in the course of many decades. As a rule, hardwood is used for kitchens in America. In California, for example there are around twenty manufacturers that use it. There are wholesale suppliers that import sawn lumber from the northern States or from Canada. At the same time, too, it is hard to tell where the Canadian company ends and the American one begins. They have an Anglo-Saxon friendship.

I immediately sold my first consignment of lumber for something close to twelve hundred dollars per cubic meter. Percentage-wise, the profit was colossal. But because the container did not hold very much, the profit was not so high in monetary terms. In order make a lot, you had to sell a lot. Naturally, being your run-of-the-mill greedy capitalist, I called Andrei and said: “Come on! Let’s see some normal volume!” But Andrei let me down a bit. He stopped controlling the quality-and Americans count the number of knots in a cubic meter very meticulously. As a result, the value can fall by ninety percent, from fifteen hundred to a hundred and fifty. At first he was sending top-grade lumber, which sold well. But then everyone realized that there was this new player on the market who was offering huge quantities at far lower prices. This made people very nervous. These distributors were just like the mafia. They came to an agreement with the kitchen-cabinet makers, who in turn stopped buying lumber from me directly. They said they did not need so much and that it would be better for us to go to the wholesalers. Now, when I went to the wholesalers, they complained about the quality, which really was not the best. They really got me down when they said they would not be buying my lumber anymore. Now I was really upset! In the end they offered to buy the wood, but at a price that was basically equal to cost. They made a rough estimate of what I had paid and started offering me two hundred dollars per cubic meter. Now I had fifty containers sitting in the port. At some point I realized that the cost of storing the containers was about the same as what they were offering me. In the end I managed to sell around ten containers. The other forty I just had to abandon. It was easier to simply leave them than to pay for their storage, reloading, and warehousing. I made a note of the loss. Nothing worked out in the States and we could not find any other market to sell in.

I still wonder if Russia exports sawn lumber overseas. But no: we have always exported round timber and we go on doing it. The European markets are under the protection of the state and the traders keep doing it against the wishes of foreigners, even though the governments of the importing countries continue to insist that the wood needs to be processed and not transported as timber. In Sweden and Finland the infrastructure already exists, the forestry business has been around for centuries, and the last thing they need is for the Russians to import product that would cost two to three times less. They are ready to buy the raw material and make things themselves. They will not even import our half-finished goods or products; thus completely finished products are absolutely out of the question-unless of course we built our own furniture stores there, as IKEA has done in Russia.

Neither the governments, nor the businessmen themselves want the Russians to come onto the scene because this puts pssure on prices and they lose their margin, income, and jobs. Even now there are no countries where it is possible to export any substantially processed wood product.

I have run businesses of varying degrees of success, but I always made money on them. Even in my restaurant business, which I do not consider to be particularly successful, I made good money. But probably every businessman has to live through one failure. This episode involving the lumber was probably my biggest failure. Taking into account the drying costs and the abandoned product, I lost somewhere in the range of one and a half to two million. Ten years ago that was a huge sum and even today it would be a lot.

It turned out to be hard to do business in America. I know that our oligarchs, like Mordashov, Deripaska, and Alekperov do not find it easy to work there either. Even Branson complained that America was the most difficult country for him. I was once again reminded of this in a far more personal way when I was selling my Tinkov beer in the U.S.A. For the American market, I changed the double “ff” to “v,” because surnames ending in two f’s look German, rather than Russian, to Americans. We bottled the beer behind our store in St. Petersburg and sold it in retail stores in California. I really understood then that the market was very difficult.

From that point on, Andrei Surkov and I stopped doing business together. We decided we would just stay friends and, thank goodness, we still are. If we had kept doing business together, we would have probably killed one other by now.

I was the only Russian in my Berkeley marketing class. My American friend, Jack Smith. This is the Ducati motorcycle that I crashed in 1999. In America they make a business out of everything, even when it comes to pictures of your kid in the maternity ward. These photos show Pasha, who was born on December 31, 1998.

This diploma lists all nine courses that I completed at Berkeley.

Konstantin Aristarkhov, member of the board of directors of Tinkov Credit Systems:

I first met Oleg in 1999 when we went skiing with a group of people near Lake Tahoe in Squaw Valley. There weren’t many recently arrived Russians, so we got to talking and it turned out that we had a lot of the same interests. We were both born in Siberia, I in Krasnoyarsk and Oleg in Kemerovo Province. We both skied. Oleg had served in Primorskoi Krai and I had served in Vlapostok.

I had come to the States to study, sponsored by Primorsklesprom, a forestry company where I had already worked for a while. So I helped Oleg a bit when he was involved in wood sales, when he and Andrei Surkov bought the saw and started processing lumber.

One day I ended up staying late at his house. He lived across the bay, while I lived downtown.

“Oleg, the boats aren’t running anymore. Let me borrow a car,” I asked him. He let me take his Porsche, which he had first taken to Russia, then brought back to America to sell. I drove home in it, parked it, and fell asleep. I live in a good neighborhood, but someone slashed the convertible roof during the night. It was probably a drunk or one of the bums that live under bridges in cardboard boxes. The car was empty; there was nothing in it. They just cut a hole in the roof, the morons. I was supposed to pay Oleg for the damage, but he acted nobly, saying,

“I see that you came and you’re trying to fix things yourself. I’m not taking money from you.”

Chapter 19

Abramovich’s First Daria

In 2001, a day came when Andrei Beskhmelnitsky, who managed Roman Abramovich’s food assets, along with Andrei Blokh (one of Abramovich’s first and main partners from back in his toy co-op days), were busy creating a holding company called Planet Management. The highly profitable but small business with the ptty name, Daria, sounded appealing to the oligarch.

Andrei Beskhmelnitsky is a unique inpidual. He is one of the most hardworking and meticulous managers that I have ever met. They hounded me for nearly six months, but in the end they gave me no other choice-there is no other way to put it-I simply had to sell this fast growing and high quality business.

On the one hand, the business brought in hundreds of thousands of dollars every day and so I was fine with it. On the other hand, though, the pelmeni market was worth a couple of million dollars a year and our share in it was already high. After I finished studying at Berkeley, I realized what market volume and share really are. If a market is large, then you can make good money even if your share is only three percent. But if the market is small, you have to be a powerful player. Now, naturally, it is really hard to grow your share if you are already the biggest player-the competition will always try to pinch a little piece of the pie. And here they were, in this case, talking me into selling my business for a couple tens of millions of dollars…

Abramovich came out to meet us and personally led us into his beautiful guest room. I could not help but stare at a massive black and white portrait hanging behind him in which Putin was wearing a kimono. The picture had obviously been retouched. On the table there was another photo of Putin, which was also quite strange. It seemed to me that this was some kind of sign or signal. Why did he have a picture of Putin in a kimono at work? Was it an expssion of respect for the psident or an implication that he, the 35-year-old Roman Arkadyevich Abramovich, was on joking terms with Putin? After all, it has been said that Abramovich is among the people who selected Yeltsin’s successor. At the same time, though, there was a certain degree of provocation in the photos. I still have not pd out the answer to this riddle.

I am a businessman and, as such, I have to have good intuition. There are a lot of incredibly unpleasant types among the oligarchy, but Abramovich made a very decent impssion on me. He is certainly no tool, as some are. But I cannot say that he is smart and scholarly. The saying, “be quiet and you’ll pass for smart,” describes him. In the half hour that we were there he said about four phrases: “Alright, okay. So what are you gonna do with the money once you’ve sold?” And his last words were, “All right guys, get him paid up.” That was it! Ellochka Lyudoyedka was involved in the socio-economic crisis in Pikalyova) told me about an empty property on Kazanskaya Street, in building 7, where there used to be a Zarya factory. The property was perfect for a fancy restaurant. It was 200 meters from Kazan Cathedral and 250 meters from Nevsky Prospect. It was on a quiet side street where it would be easy to park your car. Long ago it had been a Nobility Hall, but on January 1, 1842, the first savings bank in Russia was established there by Nicholas I, “with the goal of providing low income people of all professions with a means of saving in a reliable and profitable way.” If I had been in German Gref’s shoes, I would have bought the building for Sberbank. He is quite aware of it, as he helped me to rent it for use as a restaurant when he was chairman of the CPMC.

At the same time, the company Stanley, owned by Stanislav Bushnev, was also looking to get into the place. This kid is way behind the times. We all remember the turbid nineties. That is when he was born. Unfortunately, the mindset in the regions has changed at a much slower rate than in Moscow. I saw Bushnev in St. Petersburg recently, and he still walks around all grim and surrounded by bodyguards.

When I was trying to secure the property, Stas proclaimed,

“That place is mine.”

“But the papers show that it’s mine,” I replied.

It developed into a conflict and I told my friend Emma Vasilyevna Lavrinovich about the situation. She is still the director of Oktyabrsky Big Concert Hall. She knows all the politicians and businessmen in St. Petersburg. If she charged them a commission for everyone she knows, she would be one of the richest people in Russia. In 1996, she had introduced me to Alla Borisovna Pugachova. Now, however, she told me to get in touch with German Gref and to explain the situation to him.

Andrei Surkov attended the meeting. Gref finished listening to our story about how we wanted to build a restaurant, with a microbrewery attached, and signed a ten-year lease agreement, with far better terms than one usually sees in the city. We offered several times more than Stanley had offered. It is not enough to offer the government more money. You have to talk to a high-ranking official as well if you want to see the issue resolved.

I will tell you more: even today, the building’s fate has not been sealed. People are still arguing about it. I created and sold four businesses and fought to keep the restaurant in the same building the whole time. For the most part, however, the conflict does not concern the part of the building where the restaurant is located, but rather the Zarya Defense Plant. On March 17, 2004, Anatoly Ivanov, the plant’s director, was shot dead, along with his bodyguard. Ivanov had had lunch in my restaurant. Then, in the courtyard of his building, a contract killer opened fire on him with a Kalashnikov.

In downtown St. Petersburg the most fantastic things can occur within a few thousand square meters of real estate. In St. Petersburg, unfortunately, in contrast to Moscow, the legal mechanics of regulating real estate are very complicated. In the end it is still difficult to establish who the landlord of the building on Kazanskaya Street actually is. Currently a trust management institute is used for real estate objects, along with some other complex bureaucracies. Every building should have an owner though-either the city or a private owner. You cannot reinvent the wheel.

By spring the Germans had finished making the equipment and in May it arrived in St. Petersburg. Installation began. The question now arose of what kind of beer we would be making. Joost suggested a line-up of six standard varieties, along with four seasonal beers. This is a widespad scheme in German beer restaurants. I suggested that we name the beers according to color, however, as Russians have little sense of what terms like “pilsner,” “lager,” “porter,” and so forth actually mean.

– Platinum non-filtered (original gravity: 13%, alcohol content: 4.7%; light and caramel malt, hops, yeast, and water). I expssly asked Wachsmann to begin with pilsner, as it is my favorite kind of beer. Given a choice, I always buy either German or Czech pilsner (Budweiser, Pilsner Urquell, or the identical Plzensky Prazdroj).

– Platinum filtered (original gravity: 13%, alcohol content: 4.7%; light and caramel malt, hops, yeast, and water). We did not have much in the way of filtered beer-because I do not understand why anyone would want to drink filtered beer when non-filtered is available.

– White unfiltered (original gravity: 13%, alcohol content: 4.7%; light and wheat malt, hops, yeast, and water). I like wheat beer as well.

– Light unfiltered (original gravity: 11%, alcohol content: 4.0%; light, caramel and dark malt, hops, yeast, and water). This is a lager.

– Dark unfiltered (original gravity: 14%, alcohol content: 4.7%; light, melanoidin, dark and caramel malt). Porter.

– Non-alcoholic filtered (original gravity: 6%, alcohol content: 0.5%). It should be clear what we did here. We made a beer for people who like the taste, but cannot drink because they are behind the wheel, or for some other reason. The beer turned out to have pleasant bready notes and a hoppy aroma.

In order to offer more variety, we introduced a few seasonal beers. Two of them were:

– White Nights. I like this wheat beer even more than our usual “White”. We brewed it for the first time in St. Petersburg, which is famous for its long summer days with short “white nights” when the sun barely dips below the horizon. The beer was created especially for the summer: it is really light and, unlike traditional wheat beer, it is brewed using a top fermentation process, where the yeast is put on top instead of at the bottom.

– Winter Bock or Red (original gravity: 18%, alcohol content: 5.9 %). A beer with a wine-like flavor. For the strong at heart!

I set the opening for August 1, 1998. To promote the restaurant we gave away free food and drink for an entire day and night to all of our guests-something that was audacious for St. Petersburg at the time. A large number of the city’s restaurateurs came and were surprised to hear that I was anticipating daily revenue of ten to twenty thousand dollars. They thought that this would be impossible, given that their highly sophisticated restaurants only managed to pull in three or four grand. Supposedly, the record for the highest one-day sales volume, eight thousand dollars, was set by the Senate-Bar on Galernaya Street, where groups of foreigners were often taken and which U.S. President Bill Clinton himself visited in 1996.

From among the city’s administration came first vice governor Ilya Klebanov along with German Gref. German Oskarovich drank some beer, congratulated me, and said that he was moving to Moscow for work. On August 12, I heard on the news that he had been appointed first deputy to the Minister of State Property. When Vladimir Putin became acting psident, following the historic voluntary resignation of Boris Yeltsin on December 31, 1999, Gref was appointed head of the Center for Strategic Development, a body that was to come up with economic ideas for the new psident. As it turned out, Vladimir Putin must have felt that these ideas worked very well, considering that he appointed Gref Minister of Economic Development and Trade immediately following his election. He worked in that post for over seven years, until he managed to convince the psident that he would be better off working at Sberbank. The only minister to maintain his position longer was Alexei Kudrin, who had also started out in the St. Petersburg mayor’s office.

On one occasion, Vladimir Putin himself visited my restaurant, along with Vladimir Yakovlev. We drank some beer with them and they liked it. Putin said that he had had some beer in Germany and that Tinkoff was crap. But really, Tinkoff is a person and the beer is delicious! Thank goodness, Putin considers me a person. Bureaucrats, listen! Write this down! He drank my beer and he liked it. So you had better not cut me off on the road! This applies to law enforcement in particular. Think about it!

This is simply not true!

You can always find a niche. And you can do this yourself. After all, an entrepneur is a person who sees opportunities, a person who can ascertain what others cannot, who can perceive the positive in what seems at first glance to be a negative situation.

An entrepneur is an optimist by nature! Of course luck plays a significant role and, without a doubt, I am a lucky man. I always have been. But in order to make your luck work for you, you have to do something. Bear in mind that merely finding a good niche is not sufficient. You also need to choose the right people and motivate them in the right ways-materially and emotionally.

In St. Petersburg, in 1998, to have an in-house microbrewery attached to a restaurant was revolutionary.

The slogan “It’s one of a kind ” was not yet on the drawing board, but Tinkoff bottled beer was already my brainchild.

Valentina Vladimirovna, my mother, visiting in 1999 after Pasha was born.

Here I am in San Francisco with Dan Gordon, one of the co-founders of the Gordon Biersch chain of restaurants and microbreweries.

On Sushi

I saw it done in San Francisco and I just copied the practice. For Arkady Novikov, however, the idea did not work in St. Petersburg. And it did not work in the Moscow restaurant Sushi Vyosla, either. The assembly line approach is an option only for business lunches and for restaurants with a large output capacity.

Chapter 21

Moscow Sausage

Originally, I had not intended to create a restaurant chain. I had opened the St. Petersburg restaurant, in part to promote my beer brand, while dreaming of opening a full-fledged factory later on, and in part for myself -so that I would have someplace to go with my colleagues and friends after work. Randomly enough, I soon realized that this was not such a bad business after all. I started meeting lots of Muscovites who had fallen in love with the Restaurant on Kazanskaya Street in St. Petersburg. Of course, opening a restaurant in Moscow would be scary and expensive. The business world of St. Petersburg was not accustomed to paying forty thousand dollars a month for a property, when you could get the same place for eight thousand back home. I had my doubts, but the more praise I heard from Muscovites, the more I thought about opening a restaurant in Moscow. In 2000, then, when the crisis had eased slightly, I decided to enter this new territory.

Immediately, I felt the contrast between Moscow and St. Petersburg. There was extortion everywhere. Gimme, gimme, gimme! I can only imagine what goes on in big investment projects. Moscow is structured completely differently from St. Petersburg. Every step you take costs money; you have to pay tribute on everything. That is what you call the Byzantine Empire, my friends-Moscow, the capital of our homeland. Nothing of the sort ever happened in any of the other cities where we opened restaurants. But what can you expect from a city where the blood has already curdled? It is a city that has been ruled by the same man for 20 years now, a man whose wife, one of the richest people in Russia, is the only female billionaire (in dollars) in the country. This scenario would be impossible in any other civilized nation. An office-holder in a position like that would have stepped down from his post, at least. At most, he would have put a bullet in his own forehead.

Let me get back to the restaurant. We opened in Moscow in late 2001. All together the restaurant cost me two million dollars. In addition, we bought the property a year into our lease, which, as you can understand, ended up a very good investment, considering the growth in real estate prices.

People came to Protochny Alley, drank beer, and liked it. Everyone from Vladimir Zhirinovsky to Vagita Alekperova came to check it out.

I started coming to Moscow more often. You might say that I moved to the city, if you could say such a thing about a person who tries to spend no more than a few days at a time in any given country.

After all these years, though, I still do not feel like I love the city. Consequently, I agree with Bogdan Titomir’s song about Moscow: “Moscow is shit”.

I do not like the city at all. For me Moscow is one big office: huge, comfortable in places-an office, but not a home. When I fly into Sheremetyevo or Vnukovo airport, an interior switch flips to “work” position. When I am leaving Moscow, as soon as I get into the plane, it flips back to “rest.”

The city is not designed for family life; it is not pro-children. I realized this in 2001, when I took Rina and the kids to a restaurant, aptly called Hole in the Wall. Everyone looked at us as though we were enemies of the people. A hooker sat there with her legs crossed. Her facial expssion seemed to say: “Why in the world did you show up here with your kids?” It is not just that, though. The city, in general, was simply not built for living in.

There is always an issue about where to go for a stroll on the weekends. Do you leave town? It takes hours to get out of the city and the same amount of time to get back. The suburbs? Everything has been overhauled. They have not kept any of the old buildings and estates. The only decent place to go for a walk, perhaps, is Pokrovskoye-Streshnevo Park and only in winter. Overall, though, there is nowhere in Moscow that you can do it. You can do a lap around the Patriarch’s or Clear ponds, but no more.

Moscow is a city with completely bizarre architecture. Look at Khodynskoye Field. It is eclectic: round, square, tacky buildings (built, by the way, by Russia’s richest woman). It is totally absurd-it was an empty field. Why could they not have done what people do everywhere else in the world, constructing perpendicular and parallel streets, nice humane housing, and parks where you can go for a walk?

I went there to visit someone and it took me forever to find my way. All of the new Khodynskoye buildings, which were built five years ago, look as though they have been standing there for fifty. Alexander Kuzmin, the Chief Architect behind the project, is quite the character. Why does he not simply resign? What he has done is totally out of line, to put it mildly.

In St. Petersburg people are asking whether the Gazprom Tower might disturb the city’s harmony. Compare this with Moscow. The city has been snapped down the middle, trampled, and spat on.

Sure, Moscow reminds one of New York and, sure, it is a dynamic city. It is a good place to make some money too-I agree. But to live there is simply unbearable. That is why a lot of people send their children overseas to study, including the Moscow bureaucrats who do not believe in the city themselves. It is a dying city. Bulgakov wrote that Muscovites are good folk, but the apartment issue ruined them. After all, Woland (the Satan-p in Bulgakov’s The Master and Margarita) came to neither St. Petersburg, nor to Novosibirsk, but to Moscow-City of the Devil.

Why am I being so mean? It is because, when I wrote the foregoing, I had already been in Moscow for four weeks. Luckily I flew to Dubai the following day. A normal person has a home and an office. All of Moscow is my office. Unfortunately Luzhkov and his associates have made it so that you cannot stay in Moscow for long. How can you blame businesspeople, or the bureaucrats themselves, for sending their kids to other countries, if normal living conditions are completely lacking here? Sure, a big-time bureaucrat may be able to make himself a little Singapore in the suburbs, a gated 100-hectare estate with fields, woods, and animals. But what of the common man?

I love St. Pete’s; I dislike Moscow. And on the whole I feel okay about other major Russian cities. Novosibirsk, for example, is a very cozy city, even if it is big. There are problems with the infrastructure there, of course, but the city is interesting and suitable for living. I like it there, even though attempts were made to pvent me from opening a restaurant there in January 2003.

My restaurant in Samara, seating 275, opened in November 2002, a little earlier than the one in Novosibirsk. I created it in partnership with a local restaurateur, Alexander Terentyev, offering him a twenty-five percent share in exchange for his help in finding my way around the city and introducing me to the upper class.

It quickly became one of our most successful locations. Apparently this was because, although the city’s infrastructure is a complete wreck, its people are good.

For some reason the people of Samara and St. Petersburg are similar, like brothers. They have the exact same mindset. People befriend one another and entertain in their homes. In Moscow it is not common for people to go to one another’s houses. If you are invited, it means that there is going to be some kind of business-related standoff. A Muscovite is consequently a special kind of person, a severe type.

And the best, the most beautiful girls in the country live in Samara too. I do not know what caused it, but there appears to have been some kind of explosion there-environmental or demographic-and now every one of them is gorgeous. There are so many! It is a case where quantity and quality are not mutually exclusive.

By early 2003, we already had four restaurants. I went around to other cities, looking at how their markets were developing, trying to discern whether they were ready for us to come open a restaurant.

My memories of Nizhny Novgorod are very warm. The city is interesting and so are the people. I always enjoy my time there.

Kazan is a distinctive city. I can say a lot of nice things about its management. Despite a few Eastern frills, when it comes to attracting investment, everything is done there rationally and at a high quality. Dubai was taken as a reference point for Kazan. Even though there are some financial problems in Dubai, it would be stupid to deny that what they have done with their infrastructure is revolutionary-even if they did go a bit overboard.

Both Rustam Minnikhanov, the prime minister of Tatarstan, and the people at the mayor’s office in Kazan are on the right track. They have created an investor-friendly environment. Everything is understandable and pdictable, two critically important factors for investors. It is a good thing that Minnikhanov was appointed psident of Tatarstan in early 2010, rather than Mintimer Shaimiev.

The feelings I associate with the neighboring city of Ufa, where we also opened a restaurant, are less positive. There is more of a mess there. At least, that is how it was in 2003. Perhaps now, in 2010, Murtaza Rakhimov, the psident of Bashkiria, has done something to improve the investment climate.

I have bright memories of Yekaterinburg. Opening our restaurant there was a lot of fun. The governor of Sverdlovsk Province, Eduard Rossel, introduced me to his deputy and we assumed that we would not have any problems with the local authorities. But wait! Welcome to modern Russia! Who could have pdicted that relations between the municipal and regional administrations would be so tense? Our manager made a mistake by failing to hold any talks with the mayor at all. We brought in a bunch of musicians-Mikhail Boyarsky, Leonid Yarmolnik, Igor Kornelyuk, Mumiy Troll, De Phazz. Right at the climax of the performances there was a power failure and, in the meantime, the director of the power network was off at his summer cottage.

We could not simply disperse the crowd. At first we lit a bunch of candles. Then we worked things out with some military men, who pulled two generators up to the building. Everything worked out well end. The moral of this story is: never give up, always look for a solution and build a good team that will help you in the battle. People still remember the opening of our restaurant well, which is better promotion than we could have hoped for. It is not for nothing that the restaurant in Yekaterinburg is one of the best in our chain.

I have fond memories of Chuvashia as well. In 2003, I wanted to buy a brewery in Chebosary (which I will discuss in more detail in the next chapter). As Nikolai Fyodorov, the brewery’s psident, and I rode in the car, I experienced something that I had never experienced before and would never feel again. As we drove by some traffic cops, they saluted us.

By developing this most beautiful of cities, it can be made into a Mecca, both for Asian and Russian tourists.

With typical frankness, I declare that Kamchatka is the best place in Russia, if not in the whole world. I have never seen anything like it: volcanoes, geysers, bald mountains, snow, sudden weather changes. I went with a group of French, Germans, and Americans on a freeride skiing trip. Everyone was pleasantly surprised by the Jacuzzi-like hot springs that winter. It is a completely one-of-a-kind place. Thank God we did not sell it when we sold Alaska.

I like spending time in Sochi at Krasnaya Polyana. I do not recommend taking the chairlift, but as far as freeride skiing goes, it is one of the best places in the world, both in terms of snow quality the steep inclines. No wonder one of the stages of the freeride world cup is held there; even the best athletes are dumbstruck by it.

“Yes this really is an awesome place.” It is probable that Sochi will be the infrastructure capital of Russia now. After all, it is simply mandatory that we do everything really well in pparation for the 2014 Olympics.

There are a lot of beautiful, scenic places in Russia, inhabited by beautiful people. But our economy is upside-down. Everything flows to the center, to Moscow. This is very unfair and it is not right. So I am not in the least bit surprised that Muscovites are some of the least liked people in Russia. What does it mean that even St. Petersburg is more critically short of money than Moscow? This is especially apparent in the ambitions of Muscovite managers: they want salaries running to ps that their counterparts in St. Petersburg would never dream of demanding.

At the same time, managers in St. Petersburg are often more effective than those in Moscow. People from St. Petersburg are better workers. We are more effective, less narrow-minded. You can find examples in the business world, the political landscape, and in show business. We can see how St. Pete’s is at the top of its game in every category. The reasons are straightforward ones: on the one hand, it is the northern capital; on the other hand, there is less money there and consequently one has to put in more effort to make it. There is a parallel here with how boxers train: some use weights and others do not. That is why we are more effective managers than those one might find in other cities.

The same can be said with respect to artists. Shnur used to sing for a hundred dollars. He has seen it all. Or there is Mumiy Troll from Vlapostok who came, saw, and conquered. In my opinion, regional ambitions are always good. In Moscow, though, you can sing one song, or sell a mediocre product for 20 years straight, and still be successful. The market is like that; it is like a spoiled kid.

The competition in Moscow is very specific and in some areas there is none at all. Here I am judging from my own restaurant: at one point we had barely any competition and the restaurant made really good money. After a few years, though, our profits began to fall off and, by the time the most recent crisis rolled around, the restaurant was suffering. First of all, people started going out to eat less often. Second, there were now many similar restaurants all around Moscow. The Moscow public is easier to sway and tends to be less loyal than in a lot of places. The people love novelty. They are always looking for something better and so it can be quite hard to find and keep loyal customers. Whereas Muscovites are all about the new and the better, people in St. Petersburg visit the same restaurants decade after decade. That’s small-town Europe mentality for you! My favorite place is the best that there is, period!

Furthermore, the restaurant economy in Moscow is not market-based at all. A lot of restaurants were opened, but guest numbers did not grow, especially given the crisis. At the same time, millions of dollars were invested in these ventures. In America or Europe, eateries with such large investment volumes are doomed. In short, there, the market works.

Every day in San Francisco one restaurant opens and another closes. In other words, it would be impossible to eat at every restaurant that there is, even if you went to a new one every day. And if you see a restaurant at eighty percent capacity, in the evening, it means that it is going to shut down soon. Overhead is so high that you simply cannot keep a restaurant running unless it is packed all the time.

But in Russia, you see restaurants that remain virtually empty year after year. Why? Because the owner does not regard the place as a business. Rather, it is a status symbol. Or it is a way of keeping his wife busy. Or it is a place where he can sit in peace and quiet. Restaurants are designed with flaws built into them. And when a lot of the players do not play by the rules of the market, it is easy to see why a normal businessman, wanting to make some money, will have a hard time. That is why I do not recommend opening restaurants in Moscow at this time.

In the autumn of 2009, Aras and Emin Alagorov opened a restaurant called Nobu (the original Nobu was opened way back when by Nobu Matsuhisa and Robert De Niro) in Moscow. My family and I came there one day for lunch and we were the only guests in the whole place. It seems like that ought to have been a wake-up call, but this happened in the gorged Moscow of today. Back in the late nineties and in the early years of the new century, my restaurants were met with cheers across nearly all of Russia.

In 2001, I set up a Tinkoff restaurant in Moscow, investing unsparingly, because I wanted to keep the satiated public happy. I opened a 1300 square meter restaurant in Nizhny Novgorod on September 26, 2003. To the left is Joost Wachsmann, who sold me beer-making equipment for my restaurants. Sergei Kirienko, the psident’s authorized repsentative, came, along with his wife, to the opening of the restaurant in Nizhny Novgorod.

Chapter 22

Like No Other

The equipment at the St. Petersburg restaurant turned out to produce in such high volume that it was impossible to sell all of the beer that we had on tap. Thus Igor Sukhanov and I decided to buy a beer bottling line. I flew to Italy and ordered it, at a cost of a few hundred thousand dollars. A little later I bought out the twenty-five percent held by Igor and so became the restaurant’s sole owner. He left to work as a big boss then, as the deputy general director at Mezhregiongas.

The demand for bottled beer exceeded the supply. We could make a couple thousand bottles a day, but we needed ten times that. A bottle cost us thirty cents to produce and we sold them at wholesale for a dollar apiece. The price reflected a stock factor: if there was none left in the warehouse, we would raise the price, but if we had some, we would leave it alone. That is marketing in a nutshell.

We had insufficient product and so I started thinking about constructing a factory. I had some extra money at that point since I had recently sold my Daria pelmeni business. But I did not want to invest all of the proceeds in a factory, having decided to leave some for my family. So I approached Anton Bolshakov with the idea of building a factory. We had met back in 1999 and it still brings a smile to my face to remember how that came about.

When I was studying at Berkeley Igor Pastukhov called me and said,

In the evening, I got on an overnight train to St. Petersburg, saw how things were going at the Daria factory, and, immediately after that, went on to Pulkovo airport. When you are flying to America, it is as though you are going back in time. So I landed in San Francisco on Sunday evening, and on Monday morning went to school.

Every Monday we would talk about what we had done on the weekend. One of the students related how he had bought the Saturday edition of The Wall Street Journal and read it while sitting in a Starbucks. Someone else talked about skiing in Squaw Valley. The professor asked me,

“Oleg, where did you go?”

“I took a trip to Russia-to Moscow and St. Petersburg.”

“Are you kidding?”

“No, I had a business meeting in Moscow and then I checked in on my factory in St. Petersburg.”

It was ptty funny. The students and the professor were shocked by this Russian weirdo. But it was a good thing that I took that plane trip to Moscow and St. Petersburg. The fact that I knew Anton Bolshakov served me well in 2002.

He was the one I came to when I decided to build the beer factory.

“Anton, I have a beer restaurant in St. Petersburg, with a small bottling line. The beer is really taking off! We get several times more orders than our plant can fill-even though I have invested hardly anything so far in promoting the brand!”

“Interesting. Now what did you want?”

“I want to build a small brewery for twenty million dollars. I need a four million dollar loan. A factory of the size I envision would be able to brew twelve and a half million liters of beer per year. That’s three million bottles a month. It’s an intermediate option between a big plant and a microbrewery. I’ve got the blueprints.”

Anton believed in the idea and opened a line of credit for us. I have to give him credit: the man was able to distinguish between trash and treasure, grain and grass. And I am not the only one he has helped. On his initiative the bank started working with other capable businessmen.

Such considerations did not make me feel too much better though. I was doing the construction with borrowed money and every month the bill kept growing. I was really nervous. For the first time in my life I suffered from insomnia. It took me forever to fall asleep and I would wake up after two or three hours. If you do not sleep well, then during the day you feel like a piece of shit. In January 2003, just before the opening of the restaurant in Novosibirsk, I did not sleep for a night and a day. I felt so terrible that I left while the party was in full swing, right when Leningrad were starting their set. After I left, I wandered around the hotel. I could not sleep. I tried everything from hot milk to a warm bath. The birth of our third son Roma, on February 23, 2003, did not make matters any better. The only thing that worked was vodka. I could sleep normally after having some. I ended up having to go to a doctor specializing in sleep problems.

“What’s bothering you? What’s putting you on edge?”

“I’ll never finish building the brewery. I am afraid I’ll break the bank.”

“Once you’ve built it, then you’ll be able to sleep.”

The delay might have been even longer, if the construction had not been conducted at Stakhanovite. We’re already seeing this in the defaults that are just beginning to pile up.

Sergei answered me indirectly at a meeting with some journalists,

“I don’t know this man and I don’t want to know him, to tell you the truth. He makes strange claims and tells people to go places. Look at the clip. He told all developers to go someplace far away. For me that’s a strange position. There’s no one like him on the map. What has he done with his life? He has made Tinkoff Beer and that’s it.”

I could not resist the urge to respond to him-because it was a lie: I had done a lot more than just produce beer. We shot another video column in which I explained to Sergei that he was mistaken. After that, we did not attack one another again-particularly given Sergei’s problems with paying back his loans.

I’ve known Sergei for a long time and, no matter what’s gone between us, I think of him as a good guy. He is talented. Sometimes he goes a little too far, but on the whole he deserves credit. There was a huge pit in the middle of Moscow for years and years and monsters like Abramovich, Kerimov, and Luzhkov could not build anything there. When Sergei came to St. Petersburg as a young man, though, even if he was rude and obnoxious and not quite like everyone else, he just did what needed to be done.

From the point of view of business logic and common sense he did the wrong thing. He took out too many loans. And yet this is a problem that all developers face. They do not understand that sometimes we have stability, but that there are other times when we are faced with collapse. They simply follow the same formula, assuming that the price of a square meter will keep going up and up. If you build your business on such a formula, then sooner our later you’ll reap lamentable results.

That’s why, Sergei, I hope you return from Elba one day. For now, though, I’m afraid I have to say that your card is trumped. It’s as though you’re out in the ocean somewhere: if a shark swims past you, you’re safe, but once it’s circled you once, it’s all over. Freedman made his circle and this shark’s teeth are very sharp. You’ve fallen into his circle of interests. I can only sympathize. And I am not looking to dance on Sergei’s grave here. I really do feel bad for him. The more talented entrepneurs there are, the faster the economy will grow, which means that things will get a little better and life will become more pleasant. That’s why, Sergei, I wish you a triumphant return from Elba.

Of course, you’re going through tough times now, but hold tight. We business people are all waiting for you; we love you and we forgive all your foolish mistakes. We know that nobody’s perfect. At the end of the day everyone has his or her minuses and each of us has a skeleton in our closet.

And so, Seryoga, I wish you the best of luck!

Miracles never happen: you have to start with something small and push, push, pushLike me, Andrei Korkunov, got into the banking business after spending time in the food industry. . Andrei Rogachov, entrepneur:

With Leonid Shutov, who opened Bob Bob Ricard -one of the best restaurants in London.

Oleg is without a doubt one of the brightest entrepneurs in contemporary Russia. Entire generations of young risk-takers follow his example. Oleg Tinkov is our Richard Branson. He’s had a lot of successful projects. He knows how to choose one good idea from among a thousand. And then, too, he is able to bring the project to serious capitalization. He exhibits a rare combination of traits: he’s a charismatic leader, but at the same time he’s capable of bringing together team with technological savvy. I’m sure that we will see him undertaking more interesting projects in the future.

Around 1994, I went to visit Oleg on Sadovaya Street and was greeted by an unexpected scene. Oleg was scolding a foreigner. He was explaining to the man-I think he was American-that he had no idea how to run a business in Russia. “Stupid” was one of the milder epithets that he used. I asked Oleg whether he mightn’t be offending the American, but he optimistically replied that the American didn’t understand Russian and so it didn’t matter. I pd that Oleg had the American there as a stress reliever.

Why don’t Oleg and I take up joint-partnership on a project? Partners should supplement each other so that their team has stability. Maybe when I’m a little older-Oleg remains eternally young-we’ll start something together. He’d be the motor and I’d be the brakes.

Chapter 28

Once a Miner, Now a Banker

I started this book with a story about how our banking business was born. Recall that the final decision was made on November 18, 2005, on Necker Island. Little did I know the difficulties that awaited me after that.

On January 1, 2006, while I was spending the holidays with my family in New York before returning to San Francisco, Alex Koretsky flew from San Francisco to Moscow for the bank’s opening. Alex proved unable to work at anything like the miraculous rate that was required and, in the end, I had to go to Moscow myself to give the business a shove.

I hired old friends from Coruna (a company that I mentioned earlier), including general director Sergei Kim, to do the branding for the bank. Their price/idea/quality ratio is unparalleled in Russia. I gave them a few tasks: to develop a credit card and logo, to make a brand book, and so forth. Things progressed slowly but surely. We met sometime in June. On that occasion Sergei said,

At first I was surprised by this name, but then I realized that “Tinkoff” is a recognizable name, while “Credit Systems” sounds solid and would leave open the possibility of our offering services in addition to credit cards.

“Rustam, you know Camal very well. He’s going to head up my credit card business.” Rustam’s eyes nearly popped out of their sockets.

“I’ve invited Camal to work for me in Russia more than a couple times and he’s always refused. I can’t believe that he’ll come.”

For the next two weeks I waited for Camal Bouchie to develop and send me a business plan, but ultimately he called and declined our partnership. Rustam, who had realized that Camal was now willing to work in Russia, made him an offer that he could not refuse. According to my information, Rustam offered him more than twice as much as I had. But everything has worked out well in the end. I am glad that Rustam intercepted him and I understand why they ended up working together so effectively: they both love Eastern luxuries.

In the summer of 2006 I met with Maxim Chernushchenko, the deputy chairman of the board of Investsberbank, as well as with some other bank managers. Alexander Ponomarenko was already pparing the bank for sale to the Hungarian OTP Bank and the managers kept surveying the room. I suppose that is a normal reaction to outside threats because acquisitions always put management at risk.

In August, Alex Koretsky met with Georgy at Egon Zehnder ‘s offices. Alex liked him:

“The guy knows the trade. He’s into the idea of a model similar to Capital One. He said that we’d best not even think of buying IT-systems until we’ve hired either him or someone else with knowledge of the field.”

“That guy has pride issues, which is just what we need. How about I fly in to meet with him immediately?”

On September 18, 2006, Alex and I came to Egon Zehnder ‘s offices. Artyom Avdeyev said,

“Damn, Artyom, didn’t I ask you not to bring in people from Russian Standard? What are you, dumb?”

“Oleg, apart from people who’ve worked at Russian Standard, I couldn’t find anyone who met the requirements.”

“Why the hell have you brought me another person from Russian Standard? Is there really no one else available?”

“Georgy wasn’t a top manager at Russian Standard, so I think there’s nothing personal between him and Rustam, as there was with Chernushchenko.”

“Okay, I’ll call Rustam to talk about it.”

“Oleg, let’s not waste time,” said Georgy. “If I don’t fit into your plans because of my connection to Russian Standard, that’s no problem. If you need to talk to Rustam first, then let’s reschedule this meeting.”

“But why did you leave Investsberbank?”

“I got into an argument with a shareholder and he made the decision.”

“So you think it’s okay to fight with shareholders? A shareholder is God and King!”

“I learned my lesson.”

At the end of the meeting, I invited Georgy to come to our offices the following evening. We met in the conference room with Alex Koretsky, Kostya Aristarkhov, Ulyana Antonova, and Vadim Stasovsky, head of the legal department and the only person to work with me on running four of my businesses, beginning with Tekhnoshok. A consultant from MasterCard Advisors stood at the flip chart and talked about how payment systems are structured and how the cards work. I invited Georgy to sit at the table and, at the end of the psentation, all the managers started asking him questions. It was the same grilling that all of our key personnel had to submit to subsequently as well.

I hired him to work at the bank, which shared its offices-located on 1 ^st Street in Yamskoye Polye, directly across from where Golden Palace Casino now stands-with the Tinkoff Restaurant chain. We had just bought the license, so all we had was a desk, a chair, and an idea. That was it. We hired people to work exclusively on that idea. I am very grateful to our first ten employees. They had it harder than anyone else. Of course they worked to maintain my good name and they believed in me, but they still took a risk. Some had to give up old positions that they could not have back if their new jobs fell through. Others risked their own reputations.

The autumn of 2006 saw the most overheated labor market ever. People came, I would offer them a salary, they would call back two weeks later and tell me that they had been offered a better deal elsewhere. Nevertheless, in order to work with us, Artyom Yamanov and Stas Bliznyuk left the successful Raffeisenbank, which is one of the best run and most fortunate foreign banks in Russia. Kostya Aristarkhov was the head of a difficult department-debt collection. I had brought him with me from America. I have already mentioned that I am not keen on doing business with friends. I believe in friendships based on business, though, even if things do not work the other way around. But I would have to say that Kostya is the exception that proves the rule. We have been friends for many years-since 1999. In America, Russians hold each other close. In the States, we had spent some great times together and Kostya had shown his best qualities. Over the ten years of our friendship he has never once let me down. I hope that he can say the same thing about me.

He had gone through Far East University and has an American education. He is a fast learner and, most importantly, he has an enterprising spirit. He owned a construction company in the States, which sold and installed windows. He has everything we needed-loyalty, understanding of the market, and experience with the American business system. Moreover, he picks up new skills quickly. In the course of the three years that he has worked at the bank, he has built a good reputation for himself. He has built an incredibly effective, high quality, and technologically sophisticated debt collection department (one whose work draws on statistics, information volumes, and IT). The numbers speak for themselves. We know how difficult the market is at psent. There are more and more overdue payments, people are losing their jobs, or are simply inept at making their payments on time. At our bank, these indicators have remained relatively stable. In fact, they are the best in the industry.

Georgy Chesakov got to work on the IT platform immediately, building the network and guiding principles, and later got to work on the products. The rest of our staff began learning the ropes. We hired some high-end technicians, including Anatoly Makeshina from Zenit Bank. In this way we began to assemble the company-in our laps, essentially. I remember Georgy coming to me and saying,

He does not ask me that question any more.

Georgy Chesakov, chairman of the board of Tinkoff Credit Systems Bank:

We issued the first Tinkoff Platinum test cards in May 2007. That fall we began distributing them en masse.

Oleg is tough-skinned, explosive, and exuberant. But at the same time, he’s a lot more tolerant of resistance and complaints than my pvious employers, Rustam Tariko ( Russian Standard) and Alexander Ponomarenko ( Investsberbank)-even if, at first glance, it might seem like the opposite would be true.

I got a good impssion of Oleg back at the beginning of the banking project, in 2006, when he gathered the whole team together in his office. He pointed at the glass desk and said,

“I put my own money into this project and I’m ready to chew on this glass to ensure its success.” Later I quoted these words to candidates who asked if Oleg might shut down the project if it wasn’t working out. Another quote: “You need to have balls of steel to put 50 million dollars of your own money into a project like this.” I’m in awe of Oleg’s ability to think about what seem like completely unrelated things in business and in life and to make nontrivial, deep conclusions based on subtle observations. In those moments, you can’t help but think that you’d never thought about it before, but that his observation really was true. Oleg always remembers the basic economics of the industry and the project, quickly and correctly, and he is able to distinguish between important details and unimportant ones-and to disregard the latter. He’s not bound by ruling assumptions concerning the market. He’s able to attract and hold on to good people. He also knows how choose business ventures that have high margins of profit.

Konstantin Aristarkhov, member of the board of directors at Tinkoff Credit Systems Bank:

I learned from him how to be tougher, how to count pennies, and how to focus on and think about what’s actually important at a given moment in time.

I continued living my life, while Oleg returned to Russia to build a brewery. I felt a little awkward calling him: he’s always busy. But Oleg called me himself to find out how I was doing. Later he came back to America for a year. By that time I had my own businesses and I’d become more free and independent. We even had the time to take a trip together to Russia when he was closing a deal.

One day, when September was in full swing, we went to the Sanduny Bathhouse in Moscow. This was after Oleg had sold his beer company. Oleg said,

“Let’s go to Yalta right now. So we left straight away from the sauna and, wearing what we had on, flew to Yalta to swim in the Black Sea. We were back in Moscow two days later. In 2005 and 2006 we took quite a few trips together-around America as well as throughout Europe. Later Oleg offered me a job at the bank and I came. I had wanted him to hire me and had been waiting for it when he invited me. I sold my real estate and business before I left.

Chapter 29

Non-Standard Bankers

In early 2007 I began to realize that we needed someone to take responsibility for the bank’s management overall. I started looking and, as usual, I hired a recruitment agency, interviewing the candidates that they sent me. I probably met with ten hopefuls, which was not very many people and expensive. A lot of them were fairly strange. For instance, one particular nut job showed up-from Binbank I think-and asked for a salary of 1.5 million dollars.

In March I sat in the office and thought,

“Man-who will it be? Who will it be?” Suddenly I remembered an Englishman who worked at Visa. What was his name? John? Richard? Oliver? Yes: Oliver. Oliver Hughes. But he is a serious guy, I thought, the head of Visa. He’ll be a tough sell. Would he come to work at a bank that’s still taking losses? I started recalling the history of our relationship-which was not a simple one.”

As I have already written, I had approached Visa for the first time in the autumn of 2005, when I talked with Lou Naumovsky, the lead vice psident of Visa for Russia and the CIS. He is a Canadian of Russian heritage.

“Guys, I want to get into the credit card business.”

“What do you mean?” Lou glared at me.

“I’m totally serious. I want to distribute plastic cards and offer people credit.”

“Have you ever been in the business before?”

“No, but if you take Rustam Tariko as an example, he wasn’t in the business a few years ago and now he controls eighty percent of the market,” I replied.

“Oleg, why do you want to meet?”

“I want to offer you a job.”

“You know what? That sounds interesting. Let’s meet.”

I nearly fell off my chair. I thought he was joking. But another thought flashed through my mind: this Oliver Hughes is going to ask for ten million dollars a year.

We met at my office and came to an agreement really quickly. Maybe he was already planning on leaving Visa, which would explain why he was so quick to make a decision. He had worked there for nine years at that point, which is quite a long time. Or maybe there was something about me that he liked. Doubtless, he was attracted to the project itself and the people involved in it. He agreed immediately, in any case, at our first meeting. He said that he would be ready to start working in two months and he named his price. Oddly enough, I did not negotiate with him. Twenty minutes later we had our top manager.

On April 27, Oliver left his post as Visa repsentative for Russia and in June he started working at the bank. Prior obligations to his pvious employer entailed that he was not able to start working for us right away.

The news caused a storm in the market: people did not understand and are still unable to comphend why he left Visa and came to work with us. People often ask about it. In all honesty, though, people question it less, now that three years have gone by. Our results speak for themselves. At the time, however, everyone was shocked. A man who had worked for eight years as the head of Visa switched jobs to work at a small bank under the management of some sort of crazy person.

I agree with Marx’s assertion in Das Kapital that English managers are in charge in Russia. They were already in charge back in the nineteenth century. And it is true that if you see a good manager, he is usually an Englishman. The Russian people need Englishmen. There is still so much to do before our own managers reach maturity. I do not know what Skolkovo and other schools are up to but, objectively speaking, our achievements in the area of management are still mediocre. In the twenty years that I have been in business, I have seen a lot of managers and entrepneurs and I can make comparisons and discuss the issue. Anglo-Saxons-the British, Americans, and Canadians-are the best of the best. I like everything to have its own place on the shelf. In my cabinet, the vodka must be Russian, the cars German, the businessmen American, and the managers British.

Oliver’s wife is named Valmay. She was born in Wales. They have an interesting daughter, Maggie. Oliver used to be a punk. I even have a picture of him with a mullet. As I got to know him better, I realized that I had made the right decision: he is my kind of person. Just imagine: here was a foreigner who had spent 10 years in Russia with his British wife. I do not know very many foreigners who have spent that much time in Moscow without trading their foreign wife for a Russian one. I do not want to give a detailed account of why this happens. I am just stating a fact. I once told New Zealander Steven Jennings, head of Renaissance Capital (and I do not mean to jinx Oliver here) that he was the only foreigner living in Russia who lacked a Russian wife. Three months later he had porced his Canadian wife, Tina. I respect Oliver though. It is easy to come to Russia and find a Russian girl, aged twenty years, say, with long legs-a blonde or a brunette-who is simpler, more compliant, and easier to control than women back home. Oliver never took the easy path, however. He is a very conscientious, principled person. I liked him for his punk convictions and the fact that he is alive and real and not some kind of British bourgeois with a golden spoon in his mouth. I had been a nonconformist in my own youth, too-back when I had strange haircuts and wore badges with pictures of Viktor Tsoi on them and hung around the meeting place near the Kazan Cathedral in St. Petersburg. Oliver had traveled the world. He had lived in cheap motels, been to Vietnam, Uzbekistan, Azerbaijan, and Afghanistan. When I went to Morocco, he recommended a number of sights that I simply had to see. It seems like he has been everywhere. He does not choose the Malpes or Hawaii, either, but pfers places that are somewhat dangerous. He is interested in artifacts and has been on archeological digs in search of sphinxes in the Crimea. He is a very interesting inpidual. I am very happy and grateful to God that he has brought such good people into my life.

Oliver and I still work and grow together. There have been a couple of difficult moments over the past two years, but on the whole everything has run smoothly. I am pleased with him and think he is also pleased with our cooperation. Oliver is my partner, just like the other upper management at our bank. One of the conditions for growth in a company consists in having your top management as partners and so keeping them interested in growing capitalization and profits. Our top ten people are all shareholders in the company and this makes our model more sustainable. It is one of our strong points.

The Human Resources Department at Ernst & Young helped us develop this approach. In particular, Tim Carthy is a very talented HR specialist. I recommend him. We are very pleased with the results of our cooperation.

On May 15, 2007, all the bank’s systems were up and running. In terms of technology, we were ready to seize the credit card market. In May, we did a test mailing of 75,000 invitations to potential clients, most of whom were from Volgograd. The first response consisted in 1,500 card applications, some of which we approved. In the summer, we started mass mailings, sending out about 200,000 letters per month.

For some reason, their managing director, Julian Salisbury (who is now in charge of all of Europe and Asia), believed in me from the start. We met for the first time in February 2007 and in April we were already signing the Term Sheet, which consisted of pliminary agreements for the transaction. Next, they had lawyers review the document and, in September, just as the mortgage crisis was starting in the States, we closed the deal.

Goldman Sachs made a proposal for the purchase of a stake in a bank that had not yet issued a single card. That is to say, we were essentially nonexistent. They invested in our technology, in the team, and in our own faith. Indeed, ours is the only bank in Russia in which Goldman Sachs, the largest and most successful investment bank in the world, holds a stake. Julian said,

On September 1, 2007, I ended my vacation in Forte dei Marmi, boarded a plane, and flew to London. I attended a very brief meeting at Julian Salisbury’s office. He was very busy and Ion Dyagtoglu was the main person in charge of the transaction. We sat in the conference room and looked out the window. Ion said,

“The crisis has started.” When people talk about the crisis in Russia, for some reason they take 2008 as the starting point. For me, though, it is absolutely clear that it all began in the summer of 2007.

Ion said, “What’s going on in the States is a nightmare. Everyone’s defaulting on securities. The markets are falling. And so we can’t buy the stake for the amount we discussed before. You’ll have to lower the price.” I was furious. We fought intensely. But rage is useless in negotiations. Then Julian came in. I had brought Valentin Morozov, our financial director, with me. Six months later he ran away and joined the staff at Sberbank (he literally ran away-there is no other way to describe it). Valentin tried to bring the conversation around to a more rational tone, but I grew even more infuriated. What a waste of time!

While Nick and I ate lunch, he said,

“Listen, you have a good proposal, but I think that we’ll give you a little more and we’ll close this deal. We cannot do so, however, until closer to the New Year. If you want to move faster, you’ll have to accept their offer.” I appciated Nicholas’ honesty. It was the professional approach to take.

Two hours later I returned to the Goldman Sachs office. I said to Ion,

“Let’s make the deal-at the price you’re offering-but let’s break it into two payments. Ten percent now and five percent next June. If all the indicators line up with the business plan, then you’ll buy the rest and the price will compensate for the discount that I’m giving you today.”

He agreed.

Dear ladies and gentlemen! My fellow businesspeople! Remember that there is no such thing as a dead-end situation. There is always a way out. You have to offer an original proposal and usually you will find the approval you need. Of course you cannot disregard the role played by my meeting with Lehman Brothers. You must understand that I orchestrated a leak of that information. I still made a compromise though.

Why did the others not believe in us? Well, it is probably because the people at Goldman Sachs are smarter than most. Or maybe it was in the stars. I am very thankful to them and I know that they will earn huge money on their investment. Their stake is already worth much more than it was at the outset, in spite of the crisis. I am simply angry that the other companies I approached did not even want to talk to me.

In September, immediately following the closing of the deal, we began to send out mass invitations in earnest-in the millions. The response to our mail-outs was meager. People returned their applications incomplete or completed the forms incorrectly. Some included profanities, directed at me personally, in their applications. People tried to insert the cardboard mock-ups of the cards into bank machines; bankers from all across Russia called us and asked that we not send out any more of these “cards,” as they were getting stuck in ABM’s. Some clients, after getting the card, would go to the bank machine, withdraw the entire credit limit, and then throw the card in the nearest trashcan. It appeared that our direct marketing approach was connecting us with the most financially irresponsible people in the country.

Gradually, we began to acquire more clients. It was obvious that the money we had would not last long. The bank’s business consists of buying money, cheaply, and then selling it for more. Where could we borrow more though? We do not have offices. Consequently, we could not serve clients like a full-service bank. The capital markets left us with no options. We had to issue bonds. In the fall, though, it became clear that people with money pferred to keep it to themselves and that the Western markets had closed their doors to Russian borrowers entirely. The last few months of 2007 became a waking nightmare for us.

We set a date for October 23. Prior to the crisis, every time bonds were issued there was always a high level of oversubscription. It just looked bad if a company was unable to place an issue. We were able to sell only eighteen percent of the issue, in spite of a really attractive interest rate at eighteen percent annually. Objectively, the market situation was poor and it is unlikely that someone else could have sold bonds with greater success than we did. Nevertheless, our reputation took a hard blow. It was as if the investors did not believe in Tinkov.

We decided to complete the circulation at a later date. I called all of the financial workers that I knew and tried to convince them to take part in the circulation, but this achieved little. All together, we placed four hundred million rubles worth of bonds. The remaining six hundred million had to be placed on our own books, that is, we had to buy them out. The investors simply scattered.

This was a serious blow. I sat in the office at my round table, just crushed, and I cried. Of course, I am Siberian, a strong man, but I had tears running down my face. Why the hell were these bitches willing to buy shit? I did not understand it. A year later we saw all these shit retail and shit construction companies, which had been built on debt defaults.

An acquaintance of mine bought one hundred million dollars in bonds from each of ten companies, including mine. Out of the billion he invested, nine hundred million ended up in default. We were the only ones to return his money. Why were they buying from others, but not from us? What was with this attitude towards me? Why does everyone hate me? Why do people think that Tinkov is worthless? I did not understand and so I sat there and cried. The only person from the office who came in to see me was Oliver Hughes and I saw that his eyes were wet too. We just embraced and I said,

“Fuck it, we’ll win this war!” From that moment on I have always felt Oliver’s support and I hope that he feels mine as well-and we will continue this goddamn fight. We will prove to everyone out there that we are not in this business for nothing.

That fall I lost some friends. One of these was Alexander Vinokurov. Not only had he organized the bonds for me, he was also my friend and we had spent a lot of good times together. In London I had introduced him to Natalya Sindeyeva, co-owner of the radio station Seryebranaya Dozhd. I was at their wedding. They have a good family and I am happy for them, but my personal relationship with Sasha is unlikely to become friendly.

He called me once at one o’clock in the afternoon, when the trading day was wrapping up at 4 p.m., so I asked him:

“Sasha, buy a hundred million from us, or at least fifty. The investors are calling the brokers and telling them that KIT Finance, who organized the circulation, is not looking to buy, so they certainly do not want to buy any.”

“Oleg, I have a mortgage, you know that we put all our money into it.”

“Please, can’t you at least support us from a PR viewpoint. You can sell the bonds later.”

“Listen, I don’t believe in your idea. Credit cards are shit. You’re lying. You love to lie, but in reality things are not going well with you. And Goldman Sachs isn’t your partner, they’re just fronting you.”

When I heard this complete bullshit I hung up the phone. So I would like to say something here to Sasha: You were unable to buy fifty million rubles in bonds, but now you are broke and you owe the bank tens of billions. The problem was not that you failed to help out a friend. The problem is that you did not buy those bonds, when they were being sold by your bank.

During that circulation I lost another two or three people that had been friends, but I am thankful to the people that helped me. Anton Bolshakov bought one hundred million in bonds. Troika Dialog approached the issue professionally and Boris Jordan did as well. They believed in us, earned their annual eighteen percent, and received their money back exactly one year later. We were one of very few companies that borrowed money in 2006-2007 and then perfectly fulfilled our obligations on the one-year offer.

The idea behind entrepneurship is a simple one: if you do not take risks, you will never drink champagne. Here I am at the wedding of Alexander Vinokurov and Natalya Sindeyeva, whom I had introduced to one another in London. To my left are St. Petersburg entrepneur Alexander Aladushkin and actress Alla Dovlatova. Oliver Hughes, psident of Tinkoff Credit Systems Bank:

I was worried. The surname Vinokurov had become a red flag for me. A year later, in September 2008, justice was served. I opened the newspaper and read that he had lost it all. Did I gloat? To be honest, I did. At that moment it was really happy news for me.

The fact that Oleg approached this project as though it was normal, non-financial business was really the right thing to do. What difference does it make if your business is a bank or not? A bank is very much like any other business. The international and universal principles underlying the creation of businesses are the same: common sense, strategy, tactics, organizational structure, recruitment, and execution.

Konstantin Aristarkhov, member of the board of directors of Tinkoff Credit Systems Bank:

The original concept was to attract money from the debt markets and put that into our portfolio. It’s a totally normal task. But it became our Achilles’ heel, because the debt markets had died. This was a global problem. Even so, our very small startup attracted over two hundred million dollars during the deepest crisis since the Great Depssion. Huge sharks and small fish crashed and burned during the crisis, but we’re alive. We are an absolutely unique project. There are similar banks in the world, but the specifics of the countries, the legal requirements, the quality of the consumers, and the level of development of the banking sector, all play very important roles. Other projects may bear some superficial resemblance to ours, but they still differ. In my view, we are a bit like Capital One in its early stages, both in our approach and with respect to the technologies we use.

The fact that Oleg is able to organize, establish, fine tune, and give a burst of energy is indisputable. What’s unique about him is his meticulousness. If he enters into any sort of relationship, in any business, he’s always at the crest of the wave; he knows everything about it, he pays close attention to detail and he’s always aware of what’s going on. This is so, not just in his own work, but in everything that’s involved with the business he’s trying to build. It takes him only a moment to perceive every facet of a situation. I’ve never seen these one-of-a-kind qualities in anyone else. Plus, his mindset is undoubtedly Western, an awe-inspiring trait in a guy from Leninsk-Kuznetsky- notwithstanding the fact that he lived in St. Petersburg and worked the black market there. He picks up on everything in the blink of an eye and he correctly and appropriately evaluates and understands what is going on in the world-and why. A lot Russians that have spent time with foreigners have a habit of sucking up to them. Not so with Oleg. He’s straightforward and simple and has never sucked up to anyone in his life. He always judges people fairly, regardless of their gender, race, or ethnicity. He sees them through to the core. This quality also helps him in business. And if you look closely at him, you’ll see he doesn’t offend people. Sometimes he says things that may seem offensive, but on the whole, if he doesn’t have anything else against a person and sees his good qualities, he’ll never say an unkind word. If he does, in any case, instead of being offended, you need to listen to what he’s said and think about it and you’ll find there’s truth to his words.

For instance, he might say something sharply to his wife Rina, but she doesn’t even notice it any more. And even if he says something hurtful, you try, as you might with anyone, to imagine yourself in his shoes; you analyze what he said and realize that there was no offence in it at all. Usually it’s deserved. But if it wasn’t and you answer him constructively, staying on topic, and he agrees and admits he was wrong, then he might even apologize-if discreetly-ten times over. Because he knows his own personality and knows that he can flare up. He won’t keep fighting with you. He’s too classy to go on and on about something. If he’s at odds with you on some point of substance, all you can do is ask for an explanation. But if he himself is wrong, he backs down and doesn’t seem to find it difficult to admit that he’s made a mistake.

Chapter 30

How to Grow in a Crisis

We were barely able to place the bonds and things were still very bad. Late 2007 was the most difficult period that we have gone through to date. I had very little money left in the bank. We did not have money for operating costs or for expanding our portfolio. Through all kinds of craftiness and careful liquidity management, we got through that stage. But for a while the management were not getting their salaries.

In late December, Julian Salisbury of Goldman Sachs sent us a famous letter that is still kept in our office.

“We’re giving you more money. Use it as you have the other money we’ve given you. We believe in you, but it’s quite possible that this is the last money the bank will be able to attract. A serious crisis is coming.” That was the gist of the letter.

A lot of people ask me what it was that made TCS ready for the crisis. It is because we have partners who send us letters like that one. We never lived for today, but always lived for tomorrow. We saved money, penny by penny.

At the end of a nerve-wracking 2007, things eased up. The bank received money to feed into our credit card portfolio. Having skied through the holidays, we returned to the office at 1 Volokolamsky Proyezd, close to Pokrovskoye-Streshnevo Park. What did the year have in store for us? First of all, we needed to fulfill our obligations so that Goldman Sachs would buy the remaining five percent of the bank-at the highest option possible. Secondly, we needed to find more money and keep growing.

We could talk to Western partners with confidence, but the Russian market did not understand us. “Tinkov’s got a Bank? Ha ha ha! You’ll see what happens to them in a year.” “They don’t understand the banking business. You’re not stuffing pelmeni anymore!” “Sending cards in the mail is so last century. People don’t want to activate cards that they didn’t order.”

In one sense, we did not even try to prove the skeptics wrong. There are a lot of bankers out there that are still convinced that we hand out our cards to anyone and everyone. In reality we have never done anything of the kind. We have always sent out proposals, inviting people to become a client and then, only when the person has filled out and sent in the application, will he receive a card-and only if he is approved for it by the bank.

According to data registered in the turnover balance sheet of the bookkeeping accounts (form 101) of Tinkoff Credit Systems Bank, as of February 1, 2008, the bank’s total corporate credit portfolio amounted to 339 million rubles, while overdue debt on credits to legal entities amounted to 94.95 million rubles. These data are available on the Bank of Russia’s website. The percentage of corporate credits overdue as of February 1 was 28%, while overdue retail credits made up 1.58% of the total loans given out to inpiduals (886 million rubles).

Analysts note that these results are doubly surprising, given that Oleg Tinkov’s bank still psents itself as an exclusively retail-level institution. Now, however, it has been revealed that a third of TCS ‘s credit portfolio is made up of loans to legal entities. On top of this, the value of their overdue loans has beaten all records set in the consumer loan segment, where the levels of return have traditionally been high.

Hogwash! The story was not worth the paper it was printed on. In reality we had bought Khimmashbank, which had offered credit to companies and had begun expanding our portfolio with physical entities. Accordingly, the share of corporate credit in our portfolio had begun to shrink and, as a result, we were left with three bad loans, amounting to around one hundred million rubles-and so they remained unpaid. But when we bought the bank we knew that there were some problem loans and that legally speaking they were on the balance sheet. As far as the bank’s actual business went, this was not a problem at all. You must understand, however, that people read things diagonally; the see a negative headline and the words “30% of loans overdue” and think that things are not going well for the bank. In reality, though, the amount of overdue loans relative to the value of our entire portfolio was insignificant.

It is a good thing that foreigners do not read Russian newspapers. In the spring we held talks with a number of creditors. The money we had received from the syndicated loan was put into our credit portfolio. We wanted to keep growing. The proposals that we received from a few of the investment banks were quite ludicrous. They wanted a lot of shares at a low valuation of the bank and enormous interest rates on the debt instruments.

Vedomosti features a column entitled “Company of the Week.” In late June 2008, my bank was profiled as one such a company. Vasily Kudinov wrote in the column:

Of course, credit should not go to the Swedes alone, but to Oliver as well. All of the capital-attracting deals came about by way of a kind of road show. Oliver and I toured America extensively and then he traveled all around Europe on his own. Imagine telling the same story for two weeks, five or six times a day! You might wonder what the big deal is. Believe me, though, by the third day, even I, sitting beside him, was shocked. I did not speak much. Oliver gave the psentation in English, but even I got sick of listening to it. It is an exhausting and very serious job.

Having successfully placed the Eurobonds, we faced the crisis fully armed. In September 2008, when the banks were collapsing like houses of cards, we had around 130 million dollars in our accounts! Everyone was whining and complaining, but we tightened our belts as effectively as we could and kept on working. What had we done? Because we have no branches and therefore none of the associated costs, we simply scaled back our overhead for the mailings and cut other costs.

We started placing the money we made into our portfolio. On July 1, 2008 our credit portfolio was worth 2.5 billion rubles; by the first of October it had grown to 3.9 billion; and, as of January first, it had reached 4.8 billion rubles. In other words, we nearly doubled our portfolio at the peak of the crisis, thanks to the Eurobonds we had circulated.

Tinkoff Credit Systems expanded its credit portfolio in 2009 and earned nearly twenty million dollars in net profit. In addition to our usual credit cards, we have also begun issuing co-branded cards and debit cards. My lifelong dream has come true. I am a banker now. Sicily trip 2008: the Bank Tinkoff Credit Systems watched Team Tinkoff Credit Systems race live. On Corporate Culture

Beginning in November 2008, the bank became profitable, which the market did not expect. Between the purchase of the bank and our first profit, exactly two years had passed, which can be considered a good result. That is how life goes for us: we keep fighting and a few more people believe in us-but even more have yet to start believing. Their unbelief has filled me with anger, strength, and a desire to fight to prove them wrong. People do not believe in me? Well good for them. Now let me keep doing what I think needs to be done.

Once a year, I take our key team members on a trip abroad. We started the tradition back in the days of Daria, when we took trips to Bali and Hawaii. Usually we rent a huge villa so that we do not have to stay in a hotel. We rub elbows and feel the camaraderie. We take our wives and sometimes-less often-our children. The company pays for everything. In 2004 we took a memorable trip to Jamaica. We rented the villa where Ian Fleming lived and wrote James Bond. It’s a massive villa. (By the way, it’s not too expensive, and I’d recommend it; it was a lot cheaper than getting ten hotel rooms; there are housekeeping and food services.) Our wives spend time together and see who it is that their husbands spend time at work with, why they spend long nights away from home, and who they spend those nights with. Nothing brings a team closer together than company trips like these. We all have a rest and let our brains air out-although in the evenings we do sometimes do some business brainstorming. In all five of my businesses we’ve come up with good ideas while smoking cigars at villas or on our trip to Necker Island, after we sold the beer business in 2005.

In 2008 we went to Sicily and in 2009-2010 we went skiing in Verbier, where we rented out Richard Branson’s chalet. The last three trips have been with Tinkoff Credit Systems.

I like the expssions “There’s time for business and an hour for pleasure” and “He who rests well, works well.”

But company holidays should not be abused. There are companies that organize trips like these two or three times a year, trying to develop staff loyalty. When I was running Petrosib I did the same thing, taking my top managers and their wives all over the place. When I was skiing, so were they; when I was sailing, they were there with me. That’s not right, though. People need their personal space.

Chapter 31

Get out of the Restaurant!

After I sold my brewing business, I lost nearly all interest in my restaurant chain. In fact, I had wanted to sell the chain to the Belgians, but InBev, huge company that they are, had no need for it. In 2005, the restaurants reached their seventh year-which is quite old when you considered that I am typically interested in a business for between four and six years before I tire of it.

I began studying the market to see if it would be possible to sell the restaurants. I wanted to get twenty-five million dollars for them. In my opinion, this was a fair asking price. In 2004, the chain’s turnover was twenty million dollars. In the same year the chain had grown by fifty-four percent. It is clear that some of this growth was connected with two new restaurants that we had opened in Yekaterinburg and Sochi. Nevertheless, though, I sincerely believed that the chain’s revenue would reach one hundred million dollars within a few years, which is a significant number no matter what business you are running.

By early 2005, my chain consisted of eight restaurants: one each in Moscow, St. Petersburg, Samara, Novosibirsk, Ufa, Nizhniy Novgorod, Yekaterinburg, and Sochi. Two more locations were in development.

On March 21, 2005, we became an international chain when we sold a franchise in Almaty. The restaurant covered two thousand square meters and the brewery’s output capacity was 50,000 liters per month. Taken together, the three main floors and the summer veranda on the roof could seat three hundred guests at once! There were only 150 staff in this huge restaurant.

On June 11, Tinkoff Kazan opened. We rented a 1000-square-meter space in the former home of merchant Leonty Kekin, a significant historical site that was built in the early twentieth century. It took us five months and 1.7 million dollars to remodel the pmises and install equipment capable of producing 120,000 liters of beer per year.

Essentially, by selling the properties, I was pumping money out of the business. As a rule, I was selling for three times as much as I had paid in 2001-2004. I put all the money into pidends. Some people might say that it is not good policy to drain money like this, but I disagree. It is bad if it hurts the other shareholders, or if the company is carrying debt. I was the chain’s sole owner, however, and so I acted as I saw fit. The company leased back these same buildings, immediately. Our customers felt no change at all. At the same time, though, I was pulling cash out of the business.

In 2006, we expanded the chain by a mere two restaurants. These were different in that, instead of my being the investor, someone else was involved. Thus, for instance, the owner of the Tolyatti Restaurant in Rostov-on-Don was Restoria, which belongs to Samar businessman Alexander Terentyev.

Jumping ahead a bit, I will say that later on both franchises had to be closed. Franchising does not work in Russia-not for now at least-and we did not prove to be an exception in this regard. Moreover, these restaurants’ owners failed to fulfill their contracts. In short: the whole thing was a nightmare. This was so with Terentyev in particular: although our work together had always gone well at the Samara location, the Tolyatti Restaurant did not function well at all. He got some politicians involved and ended up in a difficult financial situation.

From the viewpoint of financial indicators, 2007 turned out to be the best in the chain’s history. Turnover amounted to 800 million rubles. The restaurants in Kazan, Sochi, Moscow, Samara, and Yekaterinburg were doing well.

Since then, only one new Tinkoff restaurant has been opened-at 23 Varshavsksaya Street in St. Petersburg. We invested close to forty million rubles in the project: a 1000-square-meter space seating 300. Of course, the idea was to have it designed in such a way as give it a different look and feel from the restaurant on Kazanskaya Street. The walls were finished in Cumaru wood, the bar in aged natural marble with brass ceramic fixtures. Tabletops were made of oak, and we hung projectors from the ceiling. We used crimped wire mesh, semi-transparent material, beer kegs, and faux gold. The restaurant ended up being beautiful and modern, but the financial indicators connected with it did not make us happy.

In 2007, Oliver Hughes, the psident of our bank, introduced me to his friend Gleb Davidyuk, a partner at the Mint Capital fund. Funds like this one deal in private equity; that is, they buy shares from non-public companies, help the companies to grow, and then sell their stake in an initial public offering or to a strategic investor.

At this point, Mint Capital had already invested in a number of Russian companies. These included, for example, UCMS, Fruzhé, Moné, A-Dept, Verysell, Maratex, Eleksnet, Gameland, ABBYY, Studio 2B, Advakom, ParallelGraphics, and jNetX. Mint Capital chooses companies that are undergoing active development, companies with annual revenue of ten to one hundred million dollars. Tinkoff met their criteria and they began to consider a partnership.

In August 2008, after having reviewed the idea, Mint Capital invested ten million dollars in the restaurant chain, buying twenty-six percent of the company’s shares for regional expansion. Maxim Sokov, who worked for Oleg Deripaska, became a minor partner. Valentin Morozov, the bank’s financial director, had introduced me to Maxim. What happened next, however, resembles a Great Septembrist Revolution affecting the consumer sphere as a whole and the restaurant industry in particular. In autumn 2008, the financial crisis began in the country. Companies began firing people and people began spending less-as a result of which other companies had to fire people as well. Of course, it is not as though everything got underway all at once: the crisis had been developing for a long time in the global economy. Already in 2007, we saw that it was becoming more and more difficult to borrow money from banks, both overseas and in Russia. The sector itself only began to feel the effects of the crisis domestically closer to the end of 2008.

The crisis put pssure on the company’s weak points. All of the management’s mistakes began coming to light. In the pceding years, I had stopped active involvement in the company and, in my absence, the managers did not do their jobs well. They began to get wrapped up in new construction projects, which remained at a standstill for long periods and which were undertaken in locales where restaurants were not supposed to have been built in the first place. They ppaid suppliers when they should not have done so. The quality of the food did not always meet Tinkoff ‘s standards. Servers began treating guests poorly.

Gleb and I decided to replace Alexei Yatsenko, the general director. The first candidate for the position was a guy from McDonald’s, while the second, Yevgeny Shalaginov, was already in the beer industry. After the interviews, I was more inclined towards the first candidate, while Gleb felt that, because his entrepneurial qualities were more highly developed, the second would be a better choice. In the end I gave in and we hired Shalaginov.

Later, we perceived that the company was in no state to rent properties at the old rates, so Gleb and I went to Troika Dialog to talk with Pavel Teplukhin.

“Pavel, you must understand that we won’t be able to continue suffering losses for long. Look at what’s happening in the world and in the country. It’s had a big impact on the restaurants. We’re losing money. If we close the chain, you’ll lose out on the rent we’re paying entirely. Let’s lower the rates.”

The negotiations ended with Troika agreeing to a substantial reduction in rent beginning on January 1, 2009. This enabled us to save money for the whole year, which is as it should be: in a crisis, property owners should be able to come to a compromise with their clients.

In February 2009, we were forced to close the restaurant in Rostov-on-Don. Why? After I had sold the beer business you could not find me in Russia. Consequently, it was not until much later that I learned of Alexei Yatsenko’s bad decision to open the restaurant in June 2006. The service industry has three rules: location, location, and location. Where a restaurant is located is of primary concern. The Don River’s north bank is the traditional recreational area for Rostovites, but we were renting 1500 square meters on the south bank. To make matters worse, the property was located in the commercial and office complex on Buyonovsky Prospect, in a district where the man in charge does not know how to do his job. Financial results there were mediocre at the best of times. They were even worse now that we were in the midst of a crisis.

In March 2009, we sold the Novosibirsk restaurant franchise, with all related equipment, to the investment company Blok, which is headed up by Voldemar Basalayev, the same man I had worked with in 1991, bringing cars to St. Petersburg from Siberia. The restaurant owed money to the landlord. Hence, we came out of the deal, basically, with nothing apart from income amounting to five or six percent of turnover, being the price paid by the franchisee for the use of the brand.

At various meetings of our board of directors we took these unpopular decisions: to close restaurants, to move our central offices from St. Petersburg to Moscow, and to cancel the construction of new restaurants in Moscow and Samara.

The company was in crisis mode. This was apparent, not only from the falling numbers, but from the flavor of our steaks, the smiles of our waitresses, and other small details. Russian consumers are extremely spoiled and so they immediately let us know what they thought of all this-by taking their rubles elsewhere.

Gleb, Marala, and I went to the Moscow restaurant on Protochny Alley to introduce them to their new director. Basically, what I said to them was,

“So you’re dissatisfied. Some people are on strike, but look at you. Where will the money come from to pay you guys if no one is eating in the restaurant? And why aren’t they coming? Because our food is shit. It used to be delicious, but now it’s shit. You’re the only ones who can earn a living for yourselves; I am not going to invest any more in the company; if it goes under, it goes under. Whoever is unwilling or unable to work can get the hell out.”

The food really had started to taste bad. On top of that, we were faced with delivery interruptions, as a result of which certain menu items were unavailable. Unhappy customers began tipping their servers less. The servers, in turn, stopped smiling and no longer made an effort to create a pleasant atmosphere in the restaurant. Basically, a crisis had descended upon the company.

In order to get out of it, the first thing we had to do was to pay all outstanding salaries, which we did. We also brought Andrei Shinkarenko of Deloitte on board as financial director. It was clear that our staff was back on track, but it was uncertain whether the business would survive. For my part, I grew more and more weary of the restaurants. They were taking up a lot of time and effort and that was taking a toll on my main business, Tinkoff Credit Systems. It sapped my strength and wore down my nerve-endings. Several times, Gleb Davidyuk and I came close to fighting in a manner from which there would have been no recovery.

But I have always had a fondness for acting irrationally. I think that I will feel more calm and balanced emotionally now, though, which is a lot more important than money to me. I’m tired of this business and, in reality, it never really belonged to me. The restaurants were supplementary to my factory and beer project-part of my marketing strategy.

I sold the brewery business long ago. I made a mistake by not selling it to the Makhmudov-Bokarev team two years ago, when its value was significantly higher. But we all make mistakes and I am no exception. I never considered myself a restaurateur and I couldn’t stand how detailed and procedural the business was. It wasn’t for me. So now I’ll feel good; I will not have to think about it anymore. I can concentrate on new things. Nevertheless, though, we really built one of the best chains in all of Russia. And this chain, headquartered in Moscow, had eleven locations, each of which adhered to the standards set in St. Petersburg back in 1998, and all of which were owned in their entirety by a single person! We covered four time zones and eleven provinces. I do not know of any other examples like that; but nothing is unique. The industry essentially consists of franchising and local partnerships. Times have changed now; there has been a paradigm shift, so to speak. The chain needs a new strategy, new strength, and new ideas. The new guys have all of these things.

I’d like to thank everyone who worked with me in building this chain! I remember you all and love you and I hate those who stole from me. I remember you too!

I still believe, in all sincerity, that the Tinkoff beer brewed in the restaurants-especially the unfiltered and unpasteurized varieties-is the best Russian beer there is.

I will continue to eat at my (?) restaurants. I wish all the best of luck to Gleb and Maxim. Guys, support them with your rubles.

Having sold the restaurants, I was finally able to breathe easy and focus all of my attention on the bank. The second St. Petersburg restaurant was closed after I had already left. As things turned out, its location was all wrong; the company was forced to write off its investment there. There was no electricity in the neighborhood and when a restaurant runs off a diesel generator its operational costs are huge.

My history in the restaurant industry taught me a great deal. I am convinced now that, in accordance with the laws of marketing, every business has its own life cycle. Perhaps you you’ve hear of “morning star,” “milking cow,” and “sunset.” Now the life cycle of my restaurant business was winding down. Of course, that does not mean that the new owners will not manage to breathe new life into it. They might well go on to introduce new ideas and marketing approaches, to freshen things up-as they say in the West.

I came to be convinced that the restaurant business was simply not for me. The business is a pain in the ass, consisting of a bunch of tiny details. Restaurateurs are a breed of their own, too. I respect them, but I feel sorry for them as well. Look at how famous and not-so-famous restaurateurs act and talk when they are in their own restaurants. Look at how they are constantly observing everything that is going on, how they do not really see you when they are talking to you, but rather see the kitchen, the service, and the tables being filled. These are sick people…

Second, the restaurants played a foundational role in the creation of the Tinkoff and Oleg Tinkov brands. People came to my restaurants, ate and drank beer, and so found out about me. Some people liked me and some did not; few were completely indifferent. This really helped us to grow. This was also key for our successful sale of the bottled beer business. I hope that this heightened profile will enable me to earn more in the future too.

The restaurant business demands attention to countless small details. This was not my thing, even though I made good money off the restaurants in the end. In 2009, I decided to “turn off the tap” and put an end to my involvement in the restaurant industry. With the girls at the opening of the restaurant in Sochi in 2004. With Konstantin Aristarkhov and Otar Kushanashvili at the opening of the restaurant in Tolyatti. Gleb Davidyuk, partner at Mint Capital:

If I want to drink some lively beer, then I go to Tinkoff, buy some “unfiltered platinum,” and order one of my three favorite dishes: dried smelt, calamari rings, or a meter of sausage.

This man has made a name for himself. He has created and sold a number of businesses. This is a man whose story is worthy of a book. He’s a living being with a head, two arms, and two legs that personifies a certain lifestyle. Does he fit in with Russian society’s social model? Not always, in my view. When society fails to see the behavior from you that it expects, then it becomes volatile in relation to you. So you have to learn how to control society’s attitude towards you, how to make it work for you and not against you. Oleg has recently spent a lot of time working to influence society’s feelings towards him-to render these favorable rather than harmful. Tinkov is a public businessman, which is a very rare thing in our country. When there is very little of some particular commodity in the market, then people’s interest in it is always great. There’s a deficit, to speak in Soviet terms. And you always want that thing of which there’s a deficit.

Our businessmen are outsiders in the western world. They are misunderstood as though they are aliens. But people try to understand Oleg and he does a lot to enable them to do so. I’m sure that Abramovich has not been completely understood, although he has recently come much closer to that. Abramovich just bought England, but he has not proven able to completely integrate into Western society-regardless of how much money he has. Oleg has a sportsman-like approach to business. On the one hand, it is a matter of constant forward movement until the end, until victory is achieved. On the other hand, if things are not working out it is better to leave the track. He’s a tall, strong athletic person. He’s always confident in himself. It is always easier to do business with assets like these. Oleg tries to mold circumstances to his favor, rather than depending on them.

We had the best possible dialog when, just as the crisis was starting in 2008, we went to Troika Dialog to talk about the lease agreement. Pavel Teplukhin probably remembers the conversation to this day. Oleg shone. A normal twenty-first century businessman would have slacked off and not known what to say. But Oleg explained the trends in the restaurant industry in a clear and straightforward manner, in layman’s terms, showing Pavel the prospects for his real estate trust if we came to an agreement.

Chapter 32

Patriarchy Forever!

Rina, Dasha, Pasha, and Roma are my family. They are a big motivator for me, as they would be for any normal person. Indeed, sometimes, they are my reason for getting up in the morning. But it would be false and stupid to claim that concern for my family is my only incentive. A normal man should be motivated by three things: sex, family, and ambition. If a person does not have these three motives, then he is not a man.

My kids are growing up to be good people. Anyone who knows them personally can attest to this. Of course, it is still hard to say what they will become. When your kids are little, their problems are little, but when they are big, their problems are big-this I know for a fact. Dasha is 16 and, from time to time, she does things that make our challenges with Pasha and Roma pale in comparison. Nevertheless, I am proud of my children. And I am very happy that I live with Rina, in particular, because she is the perfect mother.

They say sometimes that if a woman stays at home she does nothing and does not develop. This is a complete lie. Xenia Sobchak once told me that she does not like children. I think that this is the most horrible thing that a woman can say. Now, Xenia is still young, of course, and silly (in the good sense of the word), so it is forgivable.

A woman must love children. It is not a must that she remain at home. To force her to do so would also be extreme. In our case, though, that is just how things turned out: I have always been the one to earn the money and bring it home. Rina was pgnant once, a second time, and a third. We lived in America for a while and then in Italy, so that she simply never had an opportunity to work. But it would be silly to do as some of my friends have done, who purchased businesses for their wives and imagined that, in this way, they were taking care of them. There are so many examples of this sort of thing. She might be an architect-wife or PR-director-wife. We know all of these companies where people’s wives and lovers work.

Our whole family rides ordinary bicycles. We ride them along the seashore and make our way, like that, to crazy, phenomenal restaurants. We study Italian and horse around in the pool and on the beach. It is basically a normal family vacation. Sometimes, when we go to dinner, though, I run into acquaintances from Moscow.

“Oh, Oleg. We haven’t seen you for the whole month,” they say. But I think to myself,

“I’d go another month without seeing you. I was already sick of you in Moscow. I’m here with my family. I’m resting and I feel good.”

A businessman’s wife is of the utmost importance to him. Things have not changed since ancient times: the mother is the keeper of the hearth and must always keep the fire burning. In the beginning, we brought mammoth meat home and now it is cash-that is the only difference. I am very grateful to fate and God for having set me on the path to meeting Rina and for the fact that I live with her. Our example shows that Russians can get along with Estonians, even though relations between the two nations following the USSR’s collapse suggest the contrary. Not only has Rina always maintained the hearth, she has allowed me the freedom to take care of my businesses. She has never been burdensome. A man is free to act decisively when his home front is secure. When he knows that everything is well at home and that his family is there, waiting for him, he can leave and head out to the battlefield.

A lot of businessmen trade their old wives and lovers for new ones. Some of the oligarchs from Forbes magazine were never married at all. From my point of view, this is unhealthy. You have to have a wife. There has to be a hearth and a woman/mother guarding it. A woman is your home front. She is your salvation. She makes you what you are. I do not believe in doing big business without the support of a wife. Mikhail Prokhorov is an exception, but he is a talented and unique person.

Some may say that this emphasis on family is old-fashioned, but I could not care less about what is in fashion I associate the cave and the fire with women and the man with hunting and patriarchy. If things are not like that, then you have nothing. I believe in eternal values.

When I start cooperating with someone in business, I always look at what kind of wife he has. If he does not have a wife, then he has no base, no roots. I try to avoid doing business with such people. They are half-wits, as far as I am concerned. But if there is a woman in the picture, someone for whom everything is being done, then I see the person as reliable and orderly. There must be a clear, balanced structure.

I am a person with many positive and many negative traits.

A man’s family must love him and forgive his shortcomings. I am not a perfect person. But since I have no intention of getting into politics, I do not need to be perfect.

The latter entail various inconveniences and problems, but without them I would not be Oleg Tinkov. I am not superhuman. You only see superhuman people on television. If I were a lot richer, I would have a big villa with guards, a hundred-meter yacht, and a big airplane. I have all of that, though-only on a smaller scale. I pfer to be free and to live as I please. I do not want to watch my every step and bend to the strong of this world in order to increase my financial standing by a percentage point.

It is more important to me to spend time with my kids, to race up and down the passes of the Apuan Alps, and to ski down the wild slopes of the Savoy Alps. I do not enjoy thinking about what I ought to be saying to whom, or to worry about whether or not I am acquiring enemies. That would be too rational a life. And that is not my thing. I would go completely nuts if I thought like that. As it is, though, I sleep as soundly as can be. I do not owe anyone anything. I do not need to lie, nor to keep track of what my last lie consisted in. I live by my emotions. We only live once, after all. You do not get a second chance.

My problem is that I am a perfectionist. Some might say that by 42 years of age I should have calmed down and stopped seeing the world in black and white. But I have not calmed down. It might be because I am not as rich as I would like. I remain a maximalist. I do not like compromises and gray areas. My easily triggered temper does not make my life any easier either. If it was not for these demons, if I were more thoughtful, calm, and rational, I would have become a billionaire, in dollars, long ago-and I would have left Vladimir Lisin in the dust. But these things have always hindered me. I repent, I repent, I repent, but I am still this way.

It is not easy to say this, but I need to make another difficult statement: I have no friends in the literal sense of the word. If you have friends like that, then I envy you. Of course, in reality I do have somewhere between ten and fifteen friends, but these are not the sorts of friendship that you read about in books. Friendship is first and foremost a matter of self-sacrifice. I, however, am not ready to make sacrifices, nor would I accept the sacrifices of other people. Friendship requires time. But I am an entrepneur whose business is his life and so I only have enough time remaining for my family and for sports. Friendships are formed in the course of a lifetime and my constant movement around the world is not conducive to that.

But maybe having “friends at all costs” is not a necessity after all. In Russia, friendship has taken on an absolute value, thanks to our classics. In the Anglo-Saxon model, friendship is something more rational. I do not know which model is better. But I know that one only allows another into his or her inner world if that accords with the person’s own desire. If you do not let someone into your soul, then that person will not let you into his or hers either. This is a process of mutual exchange. In the same way, friendship that is one-sided ends quickly.

I probably only have one real friend-my wife Rina. It is a friendship that has been tested by time. We have been together since 1989-over twenty years. Our relationship is so honest that for a long time the thought of signing a marriage contract and having a wedding did not even cross our minds. But in order to protect the family economically, I decided it was time to get married and legalize our relationship.

The idea of having a wedding entered my mind in the winter of 2009, while I was turning the pedals of my Colnago Ferrari bicycle at the gym (I turned it into an exercise bike by attaching a cylinder under the back wheel). As I pedaled, I thought,

“In June, it will have been twenty years since I met Rina. We need to celebrate! Rina always wanted a white dress, but we never ended up having a wedding. If we wait much longer, we’ll actually be old before it happens. While the kids are little and while we’re young, we’ll look good in the photos. It’s time to get married!”

Instantly, other thoughts began popping up madly in my mind: Where? How? I have a lot of Italian friends, including Francesco and Patricia Gioffred, who own the huge castle Castello Di Tornano. I thought about the castle and about America. I thought about France, where we have a small flat with a view of the Eiffel Tower. No, it would not work. Forget it!

I kept pushing the pedals and the solution came all by itself. Why not make it Lake Baikal? I had never been there myself, but had heard a lot about it. I started talking with my friends. None of them had been there.

A lot of people ask where ideas come from. I was at the gym, riding an exercise bike; I got to thinking and then-Baikal.

All right, great!

The next morning I was already worried. I looked online. I called my friend Mikhail Slipenchuk (head of Metropol group). I had been introduced to him in 2008 by Oleg Anisimov, who then worked for Finance magazine, at a forum in Dubai. Now Oleg works in my bank as vice psident for marketing. Mikhail had mentioned that he had a few mining operations in Eastern Siberia, so I called him and he recommended Baikal’s Buryat shore.

The wedding was organized by Lena Surkova and Sveta Podolskaya. They own a party planning organization in St. Petersburg called Amusement City. (By the way, they are the best event organizers in the country!) Sveta Podolskaya’s husband, Stas, worked in my beer business and Lena’s husband, Andrei Surkov, worked with me to build the Tekhnoshok chain and also on our (unsuccessful) wood business. I called Sveta and Lena and said, “No one will do a better job. We’re old buddies.”

Ulan-Ude, the capital of Buryatia, lies 5500 kilometers from Moscow, a six-hour flight away.

The most important thing, friends, is to not grow old in heart,

To sing the song we wrote until we reach its end.

We have started on a long road, to a part of the taiga

You can only reach by plane.

Dear airplane, as you fly away,

Protect what is in your heart…

And under the airplane’s wings you can hear a song

Being sung by the green sea of the taiga.

On our way to Ulan Ude we stopped in Kemerovo, in my homeland. Aman Tuleyev gave us a cordial reception at the gubernatorial residence. My friend Alexei Prilepsky organized the reception. At one time I had encouraged Alexei to move to St. Petersburg. Now he is one of the major suppliers of mining equipment in the Kuznetsk Basin, constantly flying back and forth between Kemerovo and St. Petersburg.

While the guests were resting, I took Oliver Hughes and Stefano Feltrin, along with some other Italians, to my hometown of Leninsk-Kuznetsky. After this outing they came to understand me a lot better.

For five days, I was missing from the face of the earth. For Rina and I it was a true fairy tale. The wedding itself took place on June 12, 2009 and conformed to the usual Russian traditions-the matchmakers, the dowry, the guises, and all kinds of amusements. Rina’s dream came true: our children carried the train of her huge white dress down the aisle. We walked, holding hands, over the sand and rocks, since the wedding was in a marquee only a few meters from the Holy Baikal. We threw quite the party!

I was simply happy. My guests created a truly warm and sincere atmosphere. I would like to expss particular thanks to Valera Syutkin for hosting the event (for free!) and to my favorite Englishman, Bryan Ferry (I am happy that I invited him, rather than Valery Miladze, as I had originally wanted to do). I remembered-just in time-that his song, “Slave to Love,” from the movie 9 1/2 Weeks with Mickey Rourke and Kim Basinger, was a hit back when Rina and I first met. So twenty years later I decided that it would be our wedding tango.

I did everything I could (and I thank Richard Branson for his help too) to book Bryan Ferry. He came, sang “Slave to Love,” the love anthem, and Rina and I, wearing white, danced a beautiful dance on the shore of Lake Baikal. Yes, the dream came true.

It was unbelievably fun. And yet, even so, we were sad at times. It was a bona fide May 32 ^nd. Smile, gentlemen, smile! And think up holidays for yourselves, especially now, during this crisis!

A wedding, twenty years later, is awesome, I think. It is for real.

The lyrics to Yury Antonov’s song, “Twenty Year’s Later,” composed by Leonid Fadeyev, reflect my emotions perfectly:

I am thankful to fate

For the love that is given us.

I know I’ll need you,

Always you alone,

Just you alone.

I want us to be close,

In spite of all the years.

I want us to be close

Dasha, Pasha, and Roma: my children and my hope. Gucci spent 10 years with us before dying in March 2010. In 2001 I introduced my father and mother to America, a country that my father had respected during the Soviet era. I will not leave a fortune to them. I will pay for whatever level of education they choose to pursue. Beyond that, they should develop on their own. My children and I love our Villa Le Palme in Forte Dei Marmi. Rina has mixed feelings about it, however, since when we are there she has to do more housework than usual. About my Favorite Cities

Twenty years from now…

Of course, I am a rootless cosmopolitan. Where else would you find such a weirdo: born in Leninsk-Kuznetsky; lived there for eighteen years; spent two years in the Far East (a year in Nakohdka, in Primorsky Krai, plus another in Nikolayevsk-on-Amur in Khabarovsk Krai); thirteen years in St. Petersburg; six in the U.S.A.; a year in Italy; and now I reside mainly in Moscow. No wonder, then, that I am a man without roots. No wonder that the idea of a fatherland is foreign to me. I have a spot of homeland in Leninsk-Kuznetsky; my father’s grave is there on its ten-meter plot. But I am a man of the world. I like Americans, Italians, Frenchmen, and Russians. I like ordinary, good, capable people. I chose Forte Dei Marmi for one simple reason-the people there are pleasant and they feed us well.

Rina Vosman, Oleg Tinkov’s wife:

I do not understand why people go to France-where you pay people to be rude to you, offend you, and spit in your oysters. Only gluttons for punishment go there. There are two types of Russian people: those who like France and those who like Italy. I have noticed that I am usually friends with people of the second kind. And I do not really understand people who pfer France. My Italian friends are Patricia and Francesco who own the Castello di Tornano. I first met Patricia in the early nineties when she was head of Whirlpool and Petrosib was their dealer. We are still friends after all these years. She suggested that I get a home in Forte dei Marmi. It was important that an Italian make this suggestion and not a Russian. I love the place. It has come to be one of the most important places in my life. St. Petersburg, Forte dei Marmi, Leninsk-Kuznetsky, and San Francisco-I can draw a box around these four.

Oleg’s business successes are his achievements alone, but I played the part of guiding star, so to speak, his guiding force. I guess I have to say I motivated him. Or, to put it differently, he’d use me sometimes. I only found out after the fact that I was one of the stakeholders in Petrosib. He kept business information from me. Of course I was aware of what he was working on. For instance, when the factory was under construction, we drove to the site every Sunday to keep track of the progress. But he didn’t inform me of any problems. I think he did the right thing.

As the years passed he became calmer- much more so. In the nineties he was a total firework. There are still little bursts, now and then, but it is not a firework show. He’s always overflowing with energy; he did everything-and then some. Like Karlsson-on-the-Roof with his little motor! He was completely uncontrollable. I always tried to keep him in check. Maybe ours is the perfect marriage: opposites attract. He’s reliable and I didn’t make the wrong choice. He manages to keep his eye on me, on the children, and on his work. I never had to adapt to his needs. If there’s something that we don’t like, we always speak to each other frankly. Maybe that’s why we’ve been able to live together for twenty years. The first ten were really difficult, though; we’re completely different, after all. We celebrated our wedding in June 2009. By that time our other married friends had long since porced.

One Italian lady asked me,

Yevgeny Brekhov, friend of Oleg Tinkov:

“How did you live with someone you weren’t married to for twenty years?” But I felt completely comfortable. I didn’t need anyone’s stamp of approval.

We went over to Seryozha Ufimtsev’s house and I saw a man just back from the army sitting there. How did I know this about him? His dress was peculiar. I invited him to the movies and he began by turning me down, saying there’d be a fight there. I insisted that if he came with me there wouldn’t be any. That’s how I met Oleg Tinkov-a good, kind, unusual guy. Since then he’s told me two stories. One was about his girlfriend who was killed. He told me the second story, about Rina Vosman, his second total love, when he was in university. He really is completely lucky in life. If Oleg had stayed with his first girlfriend and the tragedy hadn’t happened, he would be working and living today in Leninsk-Kuznetsky. He would still be a miner in love. And he would never have met Rina-his true other half. Rina is the most important part of his life. She has helped Oleg to take a completely different stance on life and she has helped him to build it. She gave him his family. With Oleg, everything that happens, happens for Rina Vosman, the best woman in the world. I’m lucky that I can bear witness to the happiness of two people at once. They are a beautiful and very good couple.

Chapter 33

The Call Online

I believe that our future lies on the Internet. Consequently, I am interested in development in this area and in deeper integration with it. I want to prove that online business can be profitable. For the moment, few people earn decent money on the Internet. At our bank we have gambled on the possibility of attracting clients that are looking to open an account via the Internet. It is possible, however, that I will eventually have some other kind of business on the Web. For now, though, the Internet has more entertainment value for me than it does anything else. I enjoy maintaining my blog, tweeting, and experimenting with social networks. Let me tell you about how my blog came to be one of the most popular in Russia.

Hi!

Today I’m officially letting everyone know that I’ve started a blog. How do I see it?

1. I’m not afraid of mistakes. I’m from a small provincial town and I didn’t finish university.
2. I don’t want to do as my colleagues have done (you know who I mean), who brag about themselves in their blogs-in spite of the fact that their blogs are actually written by their assistant or a hired PR agency (respect, Yulia!).
3. This won’t be a blog; it will be my diary.

I’m not aiming for pulp fiction. Instead I’ll just write a couple lines here and there. A little of something good is best. 🙂

I hope you’ll forgive me if I don’t write all my silent letters, or if I leave them out entirely-it’s not that important.

Oleg

I wrote regularly and about everything: business, sex, music, politics, food, architecture, and so on. A year later, in spite of the odd errors in my writing, I had 15,000 friends and I was ranked in the top thirty in Yandex ‘s blog ratings, which is typically a very difficult thing for new bloggers to achieve.

In the course of the year I had been caught up in a number of Internet scandals. For instance, on September 29, 2009, I wrote a post in support of the construction of the Gazprom skyscraper in St. Petersburg. I had never had so much shit dumped on me before! People accused me of anything and everything: of stupidity, of poor taste, of brown-nosing… I will repeat my position though: I just like the tower. Both of my apartments in St. Petersburg have nice views. The first looks out on the Admiralteystvo, the second looks over the whole city. Once it is built, one will be able to see Okhta Center from there as well. I did not speak out in favor of the tower because I had been asked to by Matvienko, whom I had not seen since 2005, or by Miller, whom I have never seen in my life. I spoke out because I sincerely feel and-I am not afraid to say it-hold, from an aesthetic point of view, that building is beautiful. The project contributes to the development of a depssed part of the city, too, and will create jobs for local residents.

St. Petersburg is an excellent city, but it does have its weaknesses-that pseudo-intellectual arrogance that says: we will live in shit and we will not allow anyone to do anything about it.

Why not let them clean up the Okhta neighborhood, free the place from the rats and dirt? Why not let them build a ptty glass building, run the utilities and make a beautiful spire that I will be able to see from my building. Once upon a time a White Russian General who had immigrated to New York was asked why he was renting an apartment rather than buying one. His answer was that he already had an apartment-in St. Petersburg. I say the exact same thing. From my St. Petersburg apartment I can see everything from the mosque to St. Isaac’s Cathedral. I will look upon the tower at Okhta Center with pleasure too.

Some people felt that I had spoken out in favor of the tower because my business was not going well and because I therefore wanted to pay my respects to the city’s politicians. What a load of crap! I have never sucked up to politicians. Judge for yourselves: all those people you see on TV are accessible to me-all it would take would be a phone call. I know all of them and I feel no need to prove anything to them. I expss my own opinion on matters civil and aesthetic. I have never used political connections; I do not need to.

I cannot abide being wrongfully accused: I react sharply. Oleg Anisimov scolds me for this: why reply to an anonymous online user who has only five friends? And sometimes it fails to help at all. I lose my temper and write nasty things. I cannot stand to read bullshit.

People addressed a lot of crap to me after I invited Xenia Sobchak to the first episode of my Internet show, Business Secrets with Oleg Tinkov. They accused me of reducing the whole discussion to a conversation about sex. They complained that the program turned out to be mere sensationalism. They said that I was stupid. Few understood, however, that we opened the series with a scandal on purpose in order to make ourselves known. The people who were most offended were the first to watch subsequent episodes, which were devoted to business.

Our show has featured numerous guests: Mikhail Slipenchuk ( Metropol), Igor Ponomaryov ( Genser which did not lead to anything good. I am a little afraid because of these things.

Will I fight it? I never tried to twist myself into a Khodorkovsky caricature. I have great respect for the man, given what he did. I do not know whether he is guilty or not. Did the company he kept kill anyone? Possibly. Up until the moment he got on the plane he said that he didn’t care about anything and that he would fight. That’s why I consider him a great man and a hero. I admit that, like all the oligarchs, he did some shady things. But really, he suffered for what he did. And he acts nobly; he even spent six years in jail. And you should hear the interviews he gives! He never gets mad at anyone!

I would like to know for myself how you make decisions and take risks. It would be really interesting to see a discussion in the book concerning Russia’s business situation, its problems and ways of solving them. This won’t just be a biography, I hope, but a book that looks forward to the future.

When people tell me that I will never be like Khodorkovsky, I reply that, of course, I won’t be. He’s a monster of a man, a boulder with a big personality. I could never do what he has done. There are strong people and weak ones. Unfortunately, I am weak in this respect. That doesn’t mean that I am integrated into the system. To the contrary, I’m fighting it. I am a businessman and an entrepneur. I do my thing; I feed my family. Russia is my fatherland. Everyone knows that you do not chose where you come from. I am certainly not planning on getting involved in politics, creating a party and fighting-just as I do not plan on joining the United Russia political party.

It’s a tough question. We are talking about metaphysics here, so we have no material. It’s impossible after all to know why you fell in love with a particular girl. Something happened to you on a chemical or physical level and suddenly you were in love.

In order to make a decision, you have to make a quick analysis. Our brains are not fit with Pentiums but with much more awesome processors. We analyze quickly and make decisions quickly. I took a Socionics test and ended up in the Huxley category for intuition, my tendency to make fast choices and to find the right people.

When a person comes to an interview, how do you assess him or her?

I’m not ready to speculate on Russia’s future, as I am not a politician. To me the future looks fairly clear and steady. But I don’t expect there will be any major shifts to either side. Unfortunately, we will remain the earth’s source for natural resources. But at the same time our citizens’ prosperity will grow and maybe in the foreseeable future we will catch up to Eastern Europe. Realistically, though, there will be no breakthrough. Why? Because we just don’t have the education. More importantly, we don’t have any business education. We need to make substantive changes to our system for training personnel. The Soviet educational system has outlived its usefulness. The switch to innovative technologies is also important, but that’s not fundamental. The key is education.

Have you made mistakes?

I look the candidate in the eye. Of course I am not the Lord God and I can make mistakes, but I’m right most of the time. I choose people. If he’s a good person and, in addition to that, a good worker and manager (like Oliver Hughes, Georgy Chesakov, and Artyom Yamanov), you’ve achieved a success. All of our guys are supermen. On average I select three out of every ten. I evaluate people based on their human qualities. The most important thing is that they share my values-rebelliousness and joy in business.

Of course I have-though not very often. As a rule, I’ve made mistakes when I thought bad things about people and they turned out to be good. I’ve never had that happen to me the other way around. I don’t know if it’s a matter of stereotyping or what.

A lot of stereotypes are circulating in connection with me. “Now Tinkov, he’s a-” and the list starts. Damn! You guys don’t know me. Maybe when you finish reading this book you will have finally learned something about me. Guys, we don’t need stereotypes. Chichvarkin is a clown. Abramovich is a billionaire. All of these are stereotypes.

How do you feel about criticism?

Forgive me Lord. I repent. I was mistaken!

Just like any normal person: negatively. It is normal for a person to be averse to criticism. He may recognize that the criticism is fair, but on a subconscious level he will not like it. Nevertheless, criticism is a good thing and I like constructive criticism.

For example, if people tell me, “Oleg you’re such and such” and they back up their claim, then whether I reply or not, I give it some thought. But if they straight out insult me, then I just block them. I do not like it either when people write that I’m looking to win Matvienko’s favor just because I am in favor of the tower. It drives me up the wall because it is not true. I just like the tower project and I want to be able to see it out my window.

I welcome criticism. Please criticize me. I am a living person and not the God Almighty. I have so many shortcomings. …

Would you want to buy an island?

But I’m 42 years old and it would be impossible to reform me. I don’t have a media strategy. If I could be reformed, then I would long ago have become like Roman Abramovich or Xenia Sobchak. I might even have outdone them. I have always followed my gut, so things will turn out as they might. I’m a simple, Siberian felt boot. Oleg Anisimov-and my intuition-suggested I start a blog and so I did. My intuition, Oleg Anisimov, and Richard Branson suggested that I write a book and so I’m writing it. Sometimes I turn down offers to appear on TV, but there are other times when I agree. There is no strategy in any of this though.

Do you want your children to be in business?

To tell you the truth, yes. I have thought about it for a while. Branson influenced me in this regard as well. As with this book, Branson simply hammered the last nail in the coffin. I have always dreamt of acquiring my own island and creating a micro-state there with its own rules. Elsewhere people are always loading me down with all kinds of futility, a constitution and a passport. On the island, however, there would be 500, or better 200, of my citizens living there and I would create the Republic of Liberty. We would do as we pleased. Communism. I would be an emperor with no crown. I would buy the island only in order to make these people the happiest people ever.

What quality has hampered your life and your business?

Probably yes, but not necessarily. I would want my kids to be people, first and foremost. I like my sixteen-year-old daughter Dasha’s grasp of business. She’s very sensible. When it comes to Pasha it’s hard to say and the answer for Roma is no. The boys are doing their own thing. If they end up in business that would be great, but I won’t be discouraged if they don’t.

What do you like to do in your free time?

Life and business are one and the same thing for me. The only thing that has hampered me has been my short temper. Unfortunately my mother, Valentina Vladimirovna, passed this trait along to me. She was a poorly educated woman, hot tempered, provocative, and high strung. But I love her nonetheless, obviously…

You want to become really rich. Why?

I consider time “free” if I’m not spending it on business or on family. That said, my favorite pastime is sleep-or watching television. I feel that these are excellent hobbies. Sleep is a healthy thing to do. I don’t understand people who sleep three hours a night.

That is the hardest question you can ask a person. It involves higher philosophy. Do I want to become rich? Well, don’t you want to? Every normal person wants to be rich, healthy, and young.

Mezentsev, our great physics professor from the Mining Institute, asked one of my classmates to get up and he asked him,

“And how much money do you want?”

“Lots.” The professor replied,

“For some people a lot is not enough.” What a great answer. For some, having a villa and a pool in Tuscany would be their wildest dream come true. But for someone it would be jack squat-even if they had a villa on eight hectares with five pools and another ten villas scattered around the world. What does it mean to be rich? Who knows? Consider Bill Gates, Warren Buffet, and Abramovich. There is Polonsky. There is Tinkov. Each is rich in a different way. But the term is the same, “rich!” I became rich during my second year at the Mining Institute, after I moved from the dormitory into a room in a co-op flat for fifty rubles on the corner of Shkipersky Stream and Shevchenko Street. There were eight people in the apartment. Our drunken flat mates were constantly coming into our room. Once, our neighbor Auntie Masha stumbled and fell right into Rina’s and my bed. I got up and dragged her carefully back to her room. I was rich because all my friends lived in the dormitory, but I had my own room with my beloved girl. It is so important to have your own space. I got tired of living with-and banging my girlfriend in close proximity to-three other friends in our fifteen-meter room.

I do not want to do anything for free. It would not be right. It is White Anglo-Saxon Protestant logic-experience has shown that doing everything for money is the right thing to do. If people do something for free, based on friendship, then nothing will come of it. Or it will turn out poorly. But if it is done for money, then the result is different. I can do something for friendship’s sake once or twice, but then it does not work anymore. If a person spends more than an hour of his or her free time on something each day then it has to be paid for. Now, money is a tool. I need it in order to pay for my countless apartments all over the world and for my children’s clothing and education. My daughter’s education alone costs several tens of thousands of dollars a year and I have two more children growing up. My mission is to raise all three of them and send them on their way.

There are bankers who do not wear a suit. Here is an example.

Perhaps I will disappoint someone by saying this, but I will not leave my fortune to my children. The most that I will do will be to fund whatever level of education they want to achieve-and then I will allow them to continue developing in whatever way they see fit.

Oleg Anisimov, editor and founder of Finance magazine:

Sergei Galitsky, founder of the Magnit store chain, is an entrepneur whom I respect for the fact that he created a truly major business from scratch.

I suggested that Oleg Tinkov start a blog on Livejournal with a self-serving goal in mind: I wanted him to promote my magazine by publishing his columns and putting links on the site. I even told him we could post material on his behalf. But Oleg immediately rejected the offer and started writing himself.

He hasn’t always stepped too carefully, like any inexperienced blogger, but his sincerity has won people over: Oleg’s is one of the most popular blogs on the Internet. People see how he writes and communicates on his own, without the help of a blog secretary, so they’re willing to forgive his imperfections.

The bank’s blog has been a great help in 2010, when we started the deposit program at Tinkoff Credit Systems. At that time I already worked as the bank’s vice psident for marketing.

Chapter 34

A Revolution is our Prerogative

Let me get back to the bank. Due to funding problems, growth in the bank’s credit portfolio slowed in 2009: we had 5.2 billion as of April 1, 5.4 billion as of July 1, and 5.1 billion as of October 1. There was nothing we could do but to work better with the portfolio we had, increasing the quality of our risk analysis and improving our interactions with people who were failing to make payments.

In the bank we used a test and learn approach, based on the experience of Capital One, testing all kinds of different approaches to getting through to the client. Our method is somewhat like Japanese kaizen: we are constantly looking for the smallest improvements that we can make, which taken together yield big results. Here is a banal example: originally my signature was at the bottom of the letters that we sent. Later we started testing other signatures-Oliver Hughes’, Georgy Chesakov’s, and so forth. Some people might say that it makes no difference whose signature is at the bottom of the form, but we had to do this test to see which one of them worked better in practice.

We strove to improve our offer package (which goes in the envelope that is sent to the client), to find an optimum balance between being under the weight limit and achieving maximum effect with the mailing. The package has to interest the client, as does the text of offer. A client might stop reading at any moment, so it was important to maintain her interest and-assuming that she was in was in need of credit-to get her to fill out the form. Since March 2007, we have sent out close to thirty million letters. As a result, we have one of the best databases in Russia.

The largest bank in Russia, Sberbank, had absolutely nothing on the credit card market until only recently, and I looked for areas where we could cooperate.

German Gref is one of only a few Russian politicians towards whom my feelings have not grown cold. He went through fire and water when he worked for the St. Petersburg administration and in the Ministry of Economic Development. Sberbank was one of his less momentous jobs. Will he be able to remain an ordinary person and not lose track of his liberal values? He has the same Siberian-Petersburger values that I do. The next few years will tell. It’s difficult. The bank is big and it is no simple matter to cross a hedgehog with a water snake-no more so than teaching an elephant to dance.

Gref and I met, and I made a proposal aimed at securing his cooperation. In 2008, our credit card portfolio was bigger than Sberbank ‘s and the technology that we used was a generation ahead of theirs. More importantly, though, we knew how to attract customers and they didn’t. I said to him,

“Let’s cooperate. You know that magic word, ‘outsourcing’? We’ll candy-coat it for you. With our technology, if you don’t see the elephant dancing, at least it’ll be bouncing up and down.”

Gref understands that he is a big boy who can do anything. The Royal Bank of Scotland acts more competently, however-they gave Branson the right to accept deposits under his brand and they do not pide the additional earnings. As far as I know, Virgin Money ‘s net profit amounts to 60 million pounds sterling a year, half of which goes to the partner banks and half of which remains with Richard.

In order to succeed at Sberbank, Gref and his people need to change their mindset. They need to accept the ideas of partnerships and outsourcing. Alexander Lifshits, former Finance Minister, told us that sharing is important. The approach that begins with the attitude that “I’m big and don’t need anyone else” is behind the times. If we keep thinking like this, then it is unlikely that the elephant will dance in the near future.

My personal relationship with Gref is a good one. He stood up for businessmen more than a few times when he worked in the ministry. For instance, he defended Yevgeny Chichvarkin when the cops were attacking him. He was always a spokesman for liberal values in the government. My wish for him is that he remain the man that he is today and that he not be spoiled by the enormous power he now has.

Our banks may cooperate yet, but as of the spring of 2010, none of the required conditions is in place. In one way or another, Sberbank is not the only fish in the sea. Without a doubt, my bank is a project that I plan to sell someday and there will be other interested players when that time comes. As I said in Kommersant magazine in March 2010:

There is a possibility, then, that you will join forces with one of the existing players?

I understand, as do my minor stakeholders, that we are in this project whole-heartedly and that in the end it will become part of a large, global bank in which we will simply be the credit card department.

Are there any potential buyers? Who has expssed an interest in acquiring your bank? But have there been any concrete purchase proposals?

Of course not; it will be a sale. We will sell it and then they can integrate it with their bank.

There have been. I’ve listed a bunch of banks; we have received proposals from some of them.

My bank currently controls close to five percent of Russia’s credit card market. I am often asked, “Where are your cards? I’ve never seen any.” In all probability, this means that you are not supposed to see them, pcisely because you are not one of our target customers. Or you will see them soon: let us not forget that there are 140 million people living in Russia. I never saw Russian Standard cards either, when they were leading the market. Bear in mind, too, that Rustam and I work in the pmium consumer market, which is not always the most successful platform when it comes to building a profitable large-scale business.

So the model that my bank is following has been successfully implemented in a market as competitive as the States. In Russia, however, competition for customers is in its embryonic stage. How did our bank achieve serious success in the credit card market? This was thanks to our service. No one in our country, though, can imagine-or wants to imagine-providing customer service. Customers ought to find it convenient to use your services. If they do, then you will sell more-and for more money. People are always willing to pay for timesaving measures and for convenience. With these things in place, the bank’s profitability will line up with customer satisfaction.

It just so happens, I think, that my investors are smarter than those skeptics who bought up a bunch of securities before the crisis-securities, that is, which later defaulted. Why, then, am I the one that people write shit about online? These guys said, “I’d be better off buying securities from Yevrokommerts. They have a smart young psident, Grigory Karpovsky. He’s a financier. Tinkov, on the other hand, is a brewer.” But the next thing you know, Yevrokommerts defaulted and these investors’ balance had a gaping hole in it. I use this factoring company as an example because we were in direct competition with them when we were looking for investors in the debt market.

Give me even a single instance where I did not fulfill my obligations on a loan. All of my partners and investors are more than happy-everyone from Promstroibank to Goldman Sachs. Even the bondholders, who were really worried during the ruble’s devaluation period, got two warrants each at eighteen percent annually. They too are happy. It is always the more reliable, higher quality companies that end up having to restructure their obligations.

In cases like these, we need to consider the person standing behind the company and have a look at the company’s credit history. Someone that has been in business since 1989, someone that has built and sold four companies-such a person is trustworthy. Those who are afraid to lend to him are idiots-and that is all there is to it. They do not know how to pursue treasure. They do not know how to separate the wheat from the chaff. They continue to lend money to phony companies. Let them have their defaults; let them restructure their debt for years afterward. Smart investors will work with people who have been proving their worth for the last twenty years-and they will earn their eighteen percent. Building a business, like living a life, is not equivalent to crossing a field.

An entrepneur must always be read to plunge into battle. In March 2010, our company took a trip to Verbier. We did not merely ski, though. We also had business brainstorming sessions In the Steps of David Bowie?

As a banker I can confirm that one can invest confidently in someone that had his first business in 1989-1993. One can boldly loan money to such a person as well. People like this have gone through the toughest schooling out there-and their achievements speak for themselves. They survived in a time when a lot of businessmen dropped out of the race. They are the best of the best.

I have been thinking lately about issuing my own Ruble bonds @ Moscow. At one point, David Bowie set a very good example of the kind of thing that I have in mind (see http://en.wikipedia.org/wiki/Bowie_Bonds).

On November 7, 2009, I wrote all of the bank managers a letter. I give it in the original so that you can get in some English practice.

I think my name has more fame and credibility in Russia than even TCS Bank. According to some studies, the name Oleg Tinkov is in the top five inpiduals who come to mind when people think of Russian businessmen, beating Abramovich, Khodorkovsky and Deripaska.

【#4】Sensory Evaluation Of Food: Principles And Practices, Second Edition (Food Science Text Series)

Food Science Text Series

The Food Science Text Series provides faculty with the leading teaching tools. The Editorial Board has outlined the most appropriate and complete content for each food science course in a typical food science program and has identified textbooks of the highest quality, written by the leading food science educators.

Series Editor Dennis R. Heldman Editorial Board David A. Golden, Ph.D., Professor of Food Microbiology, Department of Food Science and Technology, University of Tennessee Richard W. Hartel, Professor of Food Engineering, Department of Food Science, University of Wisconsin Hildegarde Heymann, Professor of Food Sensory Science, Department of Food Science and Technology, University of California-Davis Joseph H. Hotchkiss, Professor, Institute of Food Science and Institute for Comparative and Environmental Toxicology, and Chair, Food Science Department, Cornell University Michael G. Johnson, Ph.D., Professor of Food Safety and Microbiology, Department of Food Science, University of Arkansas Joseph Montecalvo, Jr., Professor, Department of Food Science and Nutrition, California Polytechnic and State University-San Luis Obispo S. Suzanne Nielsen, Professor and Chair, Department of Food Science, Purdue University Juan L. Silva, Professor, Department of Food Science, Nutrition and Health Promotion, Mississippi State University

For further volumes: http://www.springer.com/series/5999

Harry T. Lawless · Hildegarde Heymann

Sensory Evaluation of Food Principles and Practices Second Edition

123

Harry T. Lawless Department of Food Science Cornell University Stocking Hall, Room 106 14853 Ithaca NY, USA

ISSN 1572-0330 ISBN 978-1-4419-6487-8 e-ISBN 978-1-4419-6488-5 DOI 10.1007/978-1-4419-6488-5 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2010932599 © Springer Science+Business Media, LLC 2010 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expssion of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

vi

Preface

Preface

vii

Peryam, David Stevens, Herb Meiselman, Elaine Skinner, Howard Schutz, Howard Moskowitz, Rose Marie Pangborn, Beverley Kroll, W. Frank Shipe, Lawrence E. Marks, Joseph C. Stevens, Arye Dethmers, Barbara Klein, Ann Noble, Harold Hedrick, William C Stringer, Roger Boulton, Kay McMath, Joel van Wyk, and Roger Mitchell. Ithaca, New York Davis, California

Harry T. Lawless Hildegarde Heymann

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Introduction and Overview . . . . . . . . . . . . . . . . . . . 1.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . 1.1.2 Measurement . . . . . . . . . . . . . . . . . . . . . 1.2 Historical Landmarks and the Three Classes of Test Methods . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Difference Testing . . . . . . . . . . . . . . . . . . . 1.2.2 Descriptive Analyses . . . . . . . . . . . . . . . . . 1.2.3 Affective Testing . . . . . . . . . . . . . . . . . . . 1.2.4 The Central Dogma-Analytic Versus Hedonic Tests . . . . . . . . . . . . . . . . . . . . . 1.3 Applications: Why Collect Sensory Data? . . . . . . . . . . . 1.3.1 Differences from Marketing Research Methods . . . 1.3.2 Differences from Traditional Product Grading Systems . . . . . . . . . . . . . . . . . . . 1.4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Physiological and Psychological Foundations of Sensory Function . . . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 2.2 Classical Sensory Testing and Psychophysical Methods 2.2.1 Early Psychophysics . . . . . . . . . . . . . 2.2.2 The Classical Psychophysical Methods . . . . 2.2.3 Scaling and Magnitude Estimation . . . . . . 2.2.4 Critiques of Stevens . . . . . . . . . . . . . . 2.2.5 Empirical Versus Theory-Driven Functions . 2.2.6 Parallels of Psychophysics and Sensory Evaluation . . . . . . . . . . . . . . . . . . . 2.3 Anatomy and Physiology and Functions of Taste . . . . 2.3.1 Anatomy and Physiology . . . . . . . . . . . 2.3.2 Taste Perception: Qualities . . . . . . . . . . 2.3.3 Taste Perception: Adaptation and Mixture Interactions . . . . . . . . . . . . . . . . . . 2.3.4 Inpidual Differences and Taste Genetics . . 2.4 Anatomy and Physiology and Functions of Smell . . . . 2.4.1 Anatomy and Cellular Function . . . . . . . .

1 1 1 3 4 5 6 7 8 10 13 15 16 17

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

19 19 20 20 21 23 25 25

. . . .

. . . .

. . . .

. . . .

26 27 27 30

. . . .

. . . .

. . . .

. . . .

30 33 34 34 ix

x

Contents

2.4.2 2.4.3 2.4.4 2.4.5

Retronasal Smell . . . . . . . . . . . . . . Olfactory Sensitivity and Specific Anosmia Odor Qualities: Practical Systems . . . . . Functional Properties: Adaptation, Mixture Suppssion, and Release . . . . . . . . . . 2.5 Chemesthesis . . . . . . . . . . . . . . . . . . . . . . 2.5.1 Qualities of Chemesthetic Experience . . . 2.5.2 Physiological Mechanisms of Chemesthesis 2.5.3 Chemical “Heat” . . . . . . . . . . . . . . 2.5.4 Other Irritative Sensations and Chemical Cooling . . . . . . . . . . . . . . . . . . . 2.5.5 Astringency . . . . . . . . . . . . . . . . . 2.5.6 Metallic Taste . . . . . . . . . . . . . . . . 2.6 Multi-modal Sensory Interactions . . . . . . . . . . . 2.6.1 Taste and Odor Interactions . . . . . . . . . 2.6.2 Irritation and Flavor . . . . . . . . . . . . . 2.6.3 Color-Flavor Interactions . . . . . . . . . . 2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Principles of Good Practice . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . 3.2 The Sensory Testing Environment . . . . . . 3.2.1 Evaluation Area . . . . . . . . . . 3.2.2 Climate Control . . . . . . . . . . 3.3 Test Protocol Considerations . . . . . . . . 3.3.1 Sample Serving Procedures . . . . 3.3.2 Sample Size . . . . . . . . . . . . 3.3.3 Sample Serving Temperatures . . 3.3.4 Serving Containers . . . . . . . . 3.3.5 Carriers . . . . . . . . . . . . . . 3.3.6 Palate Cleansing . . . . . . . . . . 3.3.7 Swallowing and Expectoration . . 3.3.8 Instructions to Panelists . . . . . . 3.3.9 Randomization and Blind Labeling 3.4 Experimental Design . . . . . . . . . . . . . 3.4.1 Designing a Study . . . . . . . . . 3.4.2 Design and Treatment Structures . 3.5 Panelist Considerations . . . . . . . . . . . 3.5.1 Incentives . . . . . . . . . . . . . 3.5.2 Use of Human Subjects . . . . . . 3.5.3 Panelist Recruitment . . . . . . . 3.5.4 Panelist Selection and Screening . 3.5.5 Training of Panelists . . . . . . . 3.5.6 Panelist Performance Assessment . 3.6 Tabulation and Analysis . . . . . . . . . . . 3.6.1 Data Entry Systems . . . . . . . . 3.7 Conclusion . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .

36 37 38

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

39 41 41 42 43

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

44 45 46 47 47 49 49 50 50

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

57 57 58 59 62 63 63 63 64 64 65 65 66 66 66 66 66 69 72 72 73 74 74 75 75 75 75 76 76

Contents

xi

4 Discrimination Testing . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Discrimination Testing . . . . . . . . . . . . . . . . . . . . . . 4.2 Types of Discrimination Tests . . . . . . . . . . . . . . . . . . 4.2.1 Paired Comparison Tests . . . . . . . . . . . . . . . 4.2.2 Triangle Tests . . . . . . . . . . . . . . . . . . . . . 4.2.3 Duo-Trio Tests . . . . . . . . . . . . . . . . . . . . 4.2.4 n-Alternative Forced Choice (n-AFC) Methods . . . 4.2.5 A-Not-A tests . . . . . . . . . . . . . . . . . . . . . 4.2.6 Sorting Methods . . . . . . . . . . . . . . . . . . . . 4.2.7 The ABX Discrimination Task . . . . . . . . . . . . 4.2.8 Dual-Standard Test . . . . . . . . . . . . . . . . . . 4.3 Reputed Strengths and Weaknesses of Discrimination Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Data Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Binomial Distributions and Tables . . . . . . . . . . 4.4.2 The Adjusted Chi-Square (χ2 ) Test . . . . . . . . . . 4.4.3 The Normal Distribution and the Z-Test on Proportion . . . . . . . . . . . . . . . . . . . . . 4.5 Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 The Power of the Statistical Test . . . . . . . . . . . 4.5.2 Replications . . . . . . . . . . . . . . . . . . . . . . 4.5.3 Warm-Up Effects . . . . . . . . . . . . . . . . . . . 4.5.4 Common Mistakes Made in the Interptation of Discrimination Tests . . . . . . . . . . . . . . . . Appendix: A Simple Approach to Handling the A, Not-A, and Same/Different Tests . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

79 79 80 80 83 84 85 85 87 88 88

5 Similarity, Equivalence Testing, and Discrimination Theory . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Common Sense Approaches to Equivalence . . . . . . . . . . 5.3 Estimation of Sample Size and Test Power . . . . . . . . . . . 5.4 How Big of a Difference Is Important? Discriminator Theory . . . . . . . . . . . . . . . . . . . . . . 5.5 Tests for Significant Similarity . . . . . . . . . . . . . . . . . 5.6 The Two One-Sided Test Approach (TOST) and Interval Testing . . . . . . . . . . . . . . . . . . . . . . . 5.7 Claim Substantiation . . . . . . . . . . . . . . . . . . . . . . . 5.8 Models for Discrimination: Signal Detection Theory . . . . . . 5.8.1 The Problem . . . . . . . . . . . . . . . . . . . . . . 5.8.2 Experimental Setup . . . . . . . . . . . . . . . . . . 5.8.3 Assumptions and Theory . . . . . . . . . . . . . . . 5.8.4 An Example . . . . . . . . . . . . . . . . . . . . . . 5.8.5 A Connection to Paired Comparisons Results Through the ROC Curve . . . . . . . . . . . . . . . 5.9 Thurstonian Scaling . . . . . . . . . . . . . . . . . . . . . . . 5.9.1 The Theory and Formulae . . . . . . . . . . . . . . . 5.9.2 Extending Thurstone’s Model to Other Choice Tests . . . . . . . . . . . . . . . . . . . . . .

101 101 103 104

88 89 89 90 90 92 92 94 97 97 98 99

105 108 110 111 111 112 112 113 114 116 116 116 118

xii

Contents

5.10

Extensions of the Thurstonian Methods, R-Index . . . 5.10.1 Short Cut Signal Detection Methods . . . . 5.10.2 An Example . . . . . . . . . . . . . . . . . 5.11 Conclusions . . . . . . . . . . . . . . . . . . . . . . Appendix: Non-Central t-Test for Equivalence of Scaled Data References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

119 119 120 120 122 122

6 Measurement of Sensory Thresholds . . . . . . . . . . . . . . . . . 6.1 Introduction: The Threshold Concept . . . . . . . . . . . . . . 6.2 Types of Thresholds: Definitions . . . . . . . . . . . . . . . . 6.3 Practical Methods: Ascending Forced Choice . . . . . . . . . . 6.4 Suggested Method for Taste/Odor/Flavor Detection Thresholds . . . . . . . . . . . . . . . . . . . . . . 6.4.1 Ascending Forced-Choice Method of Limits . . . . . 6.4.2 Purpose of the Test . . . . . . . . . . . . . . . . . . 6.4.3 Preliminary Steps . . . . . . . . . . . . . . . . . . . 6.4.4 Procedure . . . . . . . . . . . . . . . . . . . . . . . 6.4.5 Data Analysis . . . . . . . . . . . . . . . . . . . . . 6.4.6 Alternative Graphical Solution . . . . . . . . . . . . 6.4.7 Procedural Choices . . . . . . . . . . . . . . . . . . 6.5 Case Study/Worked Example . . . . . . . . . . . . . . . . . . 6.6 Other Forced Choice Methods . . . . . . . . . . . . . . . . . . 6.7 Probit Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 6.8 Sensory Adaptation, Sequential Effects, and Variability . . . . 6.9 Alternative Methods: Rated Difference, Adaptive Procedures, Scaling . . . . . . . . . . . . . . . . . . . . . . . 6.9.1 Rated Difference from Control . . . . . . . . . . . . 6.9.2 Adaptive Procedures . . . . . . . . . . . . . . . . . 6.9.3 Scaling as an Alternative Measure of Sensitivity . . . 6.10 Dilution to Threshold Measures . . . . . . . . . . . . . . . . . 6.10.1 Odor Units and Gas-Chromatography Olfactometry (GCO) . . . . . . . . . . . . . . . . . 6.10.2 Scoville Units . . . . . . . . . . . . . . . . . . . . . 6.11 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: MTBE Threshold Data for Worked Example . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

125 125 127 128

140 142 142 143 145

7 Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Some Theory . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Common Methods of Scaling . . . . . . . . . . . . . . . . 7.3.1 Category Scales . . . . . . . . . . . . . . . . . . 7.3.2 Line Scaling . . . . . . . . . . . . . . . . . . . . 7.3.3 Magnitude Estimation . . . . . . . . . . . . . . . 7.4 Recommended Practice and Practical Guidelines . . . . . . 7.4.1 Rule 1: Provide Sufficient Alternatives . . . . . . 7.4.2 Rule 2: The Attribute Must Be Understood . . . . 7.4.3 Rule 3: The Anchor Words Should Make Sense . 7.4.4 To Calibrate or Not to Calibrate . . . . . . . . . . 7.4.5 A Warning: Grading and Scoring are Not Scaling

149 149 151 152 152 155 156 158 159 159 159 159 160

. . . . . . . . . . . . .

. . . . . . . . . . . . .

129 129 129 130 131 131 131 133 133 134 136 136 137 137 138 140 140

Contents

xiii

7.5

Variations-Other Scaling Techniques . . . . . . . . . . . . . 7.5.1 Cross-Modal Matches and Variations on Magnitude Estimation . . . . . . . . . . . . . . . . . 7.5.2 Category-Ratio (Labeled Magnitude) Scales . . . . . 7.5.3 Adjustable Rating Techniques: Relative Scaling . . . 7.5.4 Ranking . . . . . . . . . . . . . . . . . . . . . . . . 7.5.5 Indirect Scales . . . . . . . . . . . . . . . . . . . . . 7.6 Comparing Methods: What is a Good Scale? . . . . . . . . . . 7.7 Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.7.1 “Do People Make Relative Judgments” Should They See Their Previous Ratings? . . . . . . 7.7.2 Should Category Rating Scales Be Assigned Integer Numbers in Data Tabulation? Are They Interval Scales? . . . . . . . . . . . . . . . . . 7.7.3 Is Magnitude Estimation a Ratio Scale or Simply a Scale with Ratio Instructions? . . . . . . . 7.7.4 What is a “Valid” Scale? . . . . . . . . . . . . . . . 7.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix 1: Derivation of Thurstonian-Scale Values for the 9-Point Scale . . . . . . . . . . . . . . . . . . . . . . . Appendix 2: Construction of Labeled Magnitude Scales . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Time-Intensity Methods . . . . . . . . . . . . . . . . . . . . . . 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 A Brief History . . . . . . . . . . . . . . . . . . . . . . . 8.3 Variations on the Method . . . . . . . . . . . . . . . . . . 8.3.1 Discrete or Discontinuous Sampling . . . . . . . 8.3.2 “Continuous” Tracking . . . . . . . . . . . . . . 8.3.3 Temporal Dominance Techniques . . . . . . . . . 8.4 Recommended Procedures . . . . . . . . . . . . . . . . . . 8.4.1 Steps in Conducting a Time-intensity Study . . . 8.4.2 Procedures . . . . . . . . . . . . . . . . . . . . . 8.4.3 Recommended Analysis . . . . . . . . . . . . . . 8.5 Data Analysis Options . . . . . . . . . . . . . . . . . . . . 8.5.1 General Approaches . . . . . . . . . . . . . . . . 8.5.2 Methods to Construct or Describe Average Curves 8.5.3 Case Study: Simple Geometric Description . . . 8.5.4 Analysis by Principal Components . . . . . . . . 8.6 Examples and Applications . . . . . . . . . . . . . . . . . 8.6.1 Taste and Flavor Sensation Tracking . . . . . . . 8.6.2 Trigeminal and Chemical/Tactile Sensations . . . 8.6.3 Taste and Odor Adaptation . . . . . . . . . . . . 8.6.4 Texture and Phase Change . . . . . . . . . . . . 8.6.5 Flavor Release . . . . . . . . . . . . . . . . . . . 8.6.6 Temporal Aspects of Hedonics . . . . . . . . . . 8.7 Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

160 160 162 164 165 166 167 168 168

169 169 169 170 171 172 174 179 179 180 182 182 183 184 185 185 186 186 187 187 188 189 192 193 193 194 194 195 195 196 197 198 198

xiv

Contents

9 Context Effects and Biases in Sensory Judgment . . . . . . . 9.1 Introduction: The Relative Nature of Human Judgment . 9.2 Simple Contrast Effects . . . . . . . . . . . . . . . . . 9.2.1 A Little Theory: Adaptation Level . . . . . . 9.2.2 Intensity Shifts . . . . . . . . . . . . . . . . 9.2.3 Quality Shifts . . . . . . . . . . . . . . . . . 9.2.4 Hedonic Shifts . . . . . . . . . . . . . . . . . 9.2.5 Explanations for Contrast . . . . . . . . . . . 9.3 Range and Frequency Effects . . . . . . . . . . . . . . 9.3.1 A Little More Theory: Parducci’s Range and Frequency Principles . . . . . . . . . . . 9.3.2 Range Effects . . . . . . . . . . . . . . . . . 9.3.3 Frequency Effects . . . . . . . . . . . . . . . 9.4 Biases . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.1 Idiosyncratic Scale Usage and Number Bias . 9.4.2 Poulton’s Classifications . . . . . . . . . . . 9.4.3 Response Range Effects . . . . . . . . . . . . 9.4.4 The Centering Bias . . . . . . . . . . . . . . 9.5 Response Correlation and Response Restriction . . . . 9.5.1 Response Correlation . . . . . . . . . . . . . 9.5.2 “Dumping” Effects: Inflation Due to Response Restriction in Profiling . . . . . 9.5.3 Over-Partitioning . . . . . . . . . . . . . . . 9.6 Classical Psychological Errors and Other Biases . . . . 9.6.1 Errors in Structured Sequences: Anticipation and Habituation . . . . . . . . . . . . . . . . 9.6.2 The Stimulus Error . . . . . . . . . . . . . . 9.6.3 Positional or Order Bias . . . . . . . . . . . . 9.7 Antidotes . . . . . . . . . . . . . . . . . . . . . . . . . 9.7.1 Avoid or Minimize . . . . . . . . . . . . . . 9.7.2 Randomization and Counterbalancing . . . . 9.7.3 Stabilization and Calibration . . . . . . . . . 9.7.4 Interptation . . . . . . . . . . . . . . . . . 9.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . 10.2 Uses of Descriptive Analyses . . . . . . . . . . . . 10.3 Language and Descriptive Analysis . . . . . . . . . 10.4 Descriptive Analysis Techniques . . . . . . . . . . R 10.4.1 Flavor Profile . . . . . . . . . . . . . . R 10.4.2 Quantitative Descriptive Analysis . . . R . . . . . . . . . . . . . 10.4.3 Texture Profile R 10.4.4 Sensory Spectrum . . . . . . . . . . . . 10.5 Generic Descriptive Analysis . . . . . . . . . . . . 10.5.1 How to Do Descriptive Analysis in Three Easy Steps . . . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

203 203 206 206 207 207 208 209 210

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

210 210 211 212 212 213 214 215 216 216

. . . . . . . . . . . .

217 218 218

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

218 219 219 219 219 220 221 222 222 223

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

227 227 228 228 231 231 234 237 238 240

. . . . . .

240

. . . . . . . . . .

. . . . . . . . . .

Contents

xv

10.5.2

Studies Comparing Different Conventional Descriptive Analysis Techniques . . . . . . 10.6 Variations on the Theme . . . . . . . . . . . . . . . . 10.6.1 Using Attribute Citation Frequencies Instead of Attribute Intensities . . . . . . . . . . . 10.6.2 Deviation from Reference Method . . . . . 10.6.3 Intensity Variation Descriptive Method . . . 10.6.4 Combination of Descriptive Analysis and Time-Related Intensity Methods . . . . 10.6.5 Free Choice Profiling . . . . . . . . . . . . 10.6.6 Flash Profiling . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

246 247

. . . . . . . . . . . . . . .

247 248 249

. . . .

. . . .

. . . .

. . . .

. . . .

249 249 252 253

Texture Evaluation . . . . . . . . . . . . . . . . . . . . . . . 11.1 Texture Defined . . . . . . . . . . . . . . . . . . . . . 11.2 Visual, Auditory, and Tactile Texture . . . . . . . . . . 11.2.1 Visual Texture . . . . . . . . . . . . . . . . . 11.2.2 Auditory Texture . . . . . . . . . . . . . . . 11.2.3 Tactile Texture . . . . . . . . . . . . . . . . . 11.2.4 Tactile Hand Feel . . . . . . . . . . . . . . . 11.3 Sensory Texture Measurements . . . . . . . . . . . . . 11.3.1 Texture Profile Method . . . . . . . . . . . . 11.3.2 Other Sensory Texture Evaluation Techniques 11.3.3 Instrumental Texture Measurements and Sensory Correlations . . . . . . . . . . . 11.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

259 259 262 262 262 264 268 270 270 272

. . . . . . . . . . . .

274 276 276

12

Color and Appearance . . . . . . . . . . . . . . . . . . 12.1 Color and Appearance . . . . . . . . . . . . . . . 12.2 What Is Color? . . . . . . . . . . . . . . . . . . . 12.3 Vision . . . . . . . . . . . . . . . . . . . . . . . 12.3.1 Normal Human Color Vision Variations 12.3.2 Human Color Blindness . . . . . . . . . 12.4 Measurement of Appearance and Color Attributes 12.4.1 Appearance . . . . . . . . . . . . . . . 12.4.2 Visual Color Measurement . . . . . . . 12.5 Instrumental Color Measurement . . . . . . . . . 12.5.1 Munsell Color Solid . . . . . . . . . . . 12.5.2 Mathematical Color Systems . . . . . . 12.6 Conclusions . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

283 283 284 285 286 286 286 286 289 293 293 294 299 299

13

Preference Testing . . . . . . . . . . . . . . . . . . 13.1 Introduction-Consumer Sensory Evaluation 13.2 Preference Tests: Overview . . . . . . . . . 13.2.1 The Basic Comparison . . . . . . 13.2.2 Variations . . . . . . . . . . . . . 13.2.3 Some Cautions . . . . . . . . . . 13.3 Simple Paired Preference Testing . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

303 303 305 305 305 306 306

11

. . . . . . .

. . . . . . .

. . . . . . .

xvi

Contents

13.3.1 Recommended Procedure . . . . . . . . . . . . . . . 13.3.2 Statistical Basis . . . . . . . . . . . . . . . . . . . . 13.3.3 Worked Example . . . . . . . . . . . . . . . . . . . 13.3.4 Useful Statistical Approximations . . . . . . . . . . 13.3.5 The Special Case of Equivalence Testing . . . . . . . 13.4 Non-forced Preference . . . . . . . . . . . . . . . . . . . . . . 13.5 Replicated Preference Tests . . . . . . . . . . . . . . . . . . . 13.6 Replicated Non-forced Preference . . . . . . . . . . . . . . . . 13.7 Other Related Methods . . . . . . . . . . . . . . . . . . . . . 13.7.1 Ranking . . . . . . . . . . . . . . . . . . . . . . . . 13.7.2 Analysis of Ranked Data . . . . . . . . . . . . . . . 13.7.3 Best-Worst Scaling . . . . . . . . . . . . . . . . . . 13.7.4 Rated Degree of Preference and Other Options . . . . 13.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix 1: Worked Example of the Ferris k-Visit Repeated Preference Test Including the No-Preference Option . . . . . . Appendix 2: The “Placebo” Preference Test . . . . . . . . . . . . . . . Appendix 3: Worked Example of Multinomial Approach to Analyzing Data with the No-Preference Option . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Acceptance Testing . . . . . . . . . . . . . . . . . . . . . . 14.1 Introduction: Scaled Liking Versus Choice . . . . . . 14.2 Hedonic Scaling: Quantification of Acceptability . . . 14.3 Recommended Procedure . . . . . . . . . . . . . . . 14.3.1 Steps . . . . . . . . . . . . . . . . . . . . . 14.3.2 Analysis . . . . . . . . . . . . . . . . . . . 14.3.3 Replication . . . . . . . . . . . . . . . . . 14.4 Other Acceptance Scales . . . . . . . . . . . . . . . . 14.4.1 Line Scales . . . . . . . . . . . . . . . . . 14.4.2 Magnitude Estimation . . . . . . . . . . . . 14.4.3 Labeled Magnitude Scales . . . . . . . . . 14.4.4 Pictorial Scales and Testing with Children . 14.4.5 Adjustable Scales . . . . . . . . . . . . . . 14.5 Just-About-Right Scales . . . . . . . . . . . . . . . . 14.5.1 Description . . . . . . . . . . . . . . . . . 14.5.2 Limitations . . . . . . . . . . . . . . . . . 14.5.3 Variations on Relative-to-Ideal Scaling . . . 14.5.4 Analysis of JAR Data . . . . . . . . . . . . 14.5.5 Penalty Analysis or “Mean Drop” . . . . . 14.5.6 Other Problems and Issues with JAR Scales 14.6 Behavioral and Context-Related Approaches . . . . . 14.6.1 Food Action Rating Scale (FACT) . . . . . 14.6.2 Appropriateness Scales . . . . . . . . . . . 14.6.3 Acceptor Set Size . . . . . . . . . . . . . . 14.6.4 Barter Scales . . . . . . . . . . . . . . . . 14.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . .

306 307 308 309 310 311 313 313 315 315 316 317 318 320 320 321 322 323 325 325 326 327 327 328 328 328 328 330 331 332 333 334 334 335 336 336 339 340 340 341 341 342 343 343 344

Contents

xvii

15

16

Consumer Field Tests and Questionnaire Design . . . . . . . . . . . 15.1 Sensory Testing Versus Concept Testing . . . . . . . . . . . . 15.2 Testing Scenarios: Central Location, Home Use . . . . . . . . 15.2.1 Purpose of the Tests . . . . . . . . . . . . . . . . . . 15.2.2 Consumer Models . . . . . . . . . . . . . . . . . . . 15.2.3 Central Location Tests . . . . . . . . . . . . . . . . 15.2.4 Home Use Tests (HUT) . . . . . . . . . . . . . . . . 15.3 Practical Matters in Conducting Consumer Field Tests . . . . . 15.3.1 Tasks and Test Design . . . . . . . . . . . . . . . . . 15.3.2 Sample Size and Stratification . . . . . . . . . . . . 15.3.3 Test Designs . . . . . . . . . . . . . . . . . . . . . . 15.4 Interacting with Field Services . . . . . . . . . . . . . . . . . 15.4.1 Choosing Agencies, Communication, and Test Specifications . . . . . . . . . . . . . . . . 15.4.2 Incidence, Cost, and Recruitment . . . . . . . . . . . 15.4.3 Some Tips: Do’s and Don’ts . . . . . . . . . . . . . 15.4.4 Steps in Testing with Research Suppliers . . . . . . . 15.5 Questionnaire Design . . . . . . . . . . . . . . . . . . . . . . 15.5.1 Types of Interviews . . . . . . . . . . . . . . . . . . 15.5.2 Questionnaire Flow: Order of Questions . . . . . . . 15.5.3 Interviewing . . . . . . . . . . . . . . . . . . . . . . 15.6 Rules of Thumb for Constructing Questions . . . . . . . . . . 15.6.1 General Principles . . . . . . . . . . . . . . . . . . . 15.6.2 Brevity . . . . . . . . . . . . . . . . . . . . . . . . . 15.6.3 Use Plain Language . . . . . . . . . . . . . . . . . . 15.6.4 Accessibility of the Information . . . . . . . . . . . 15.6.5 Avoid Vague Questions . . . . . . . . . . . . . . . . 15.6.6 Check for Overlap and Completeness . . . . . . . . . 15.6.7 Do Not Lead the Respondent . . . . . . . . . . . . . 15.6.8 Avoid Ambiguity and Double Questions . . . . . . . 15.6.9 Be Careful in Wording: Present Both Alternatives . . 15.6.10 Beware of Halos and Horns . . . . . . . . . . . . . . 15.6.11 Pre-test . . . . . . . . . . . . . . . . . . . . . . . . 15.7 Other Useful Questions: Satisfaction, Agreement, and Open-Ended Questions . . . . . . . . . . . . . . . . . . . 15.7.1 Satisfaction . . . . . . . . . . . . . . . . . . . . . . 15.7.2 Likert (Agree-Disagree) Scales . . . . . . . . . . . . 15.7.3 Open-Ended Questions . . . . . . . . . . . . . . . . 15.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix 1: Sample Test Specification Sheet . . . . . . . . . . . . . . Appendix 2: Sample Screening Questionnaire . . . . . . . . . . . . . . Appendix 3: Sample Product Questionnaire . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Qualitative Consumer Research Methods . . . . . . . 16.1 Introduction . . . . . . . . . . . . . . . . . . . 16.1.1 Resources, Definitions, and Objectives 16.1.2 Styles of Qualitative Research . . . . 16.1.3 Other Qualitative Techniques . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

349 349 351 351 352 353 354 355 355 355 356 358 358 359 360 360 362 362 362 363 364 364 364 364 365 365 365 365 366 366 366 366 367 367 367 367 368 370 371 374 378 379 380 380 380 382

xviii

Contents

16.2

Characteristics of Focus Groups . . . . . . . . . . . . . . . . . 383 16.2.1 Advantages . . . . . . . . . . . . . . . . . . . . . . 383 16.2.2 Key Requirements . . . . . . . . . . . . . . . . . . . 384 16.2.3 Reliability and Validity . . . . . . . . . . . . . . . . 384 16.3 Using Focus Groups in Sensory Evaluation . . . . . . . . . . . 385 16.4 Examples, Case Studies . . . . . . . . . . . . . . . . . . . . . 386 16.4.1 Case Study 1: Qualitative Research Before Conjoint Measurement in New Product Development 387 16.4.2 Case Study 2: Nutritional and Health Beliefs About Salt 387 16.5 Conducting Focus Group Studies . . . . . . . . . . . . . . . . 388 16.5.1 A Quick Overview . . . . . . . . . . . . . . . . . . 388 16.5.2 A Key Requirement: Developing Good Questions . . 389 16.5.3 The Discussion Guide and Phases of the Group Interview . . . . . . . . . . . . . . . . 390 16.5.4 Participant Requirements, Timing, Recording . . . . 391 16.6 Issues in Moderating . . . . . . . . . . . . . . . . . . . . . . . 392 16.6.1 Moderating Skills . . . . . . . . . . . . . . . . . . . 392 16.6.2 Basic Principles: Nondirection, Full Participation, and Coverage of Issues . . . . . . . . . 393 16.6.3 Assistant Moderators and Co-moderators . . . . . . . 394 16.6.4 Debriefing: Avoiding Selective Listening and Premature Conclusions . . . . . . . . . . . . . . 395 16.7 Analysis and Reporting . . . . . . . . . . . . . . . . . . . . . 395 16.7.1 General Principles . . . . . . . . . . . . . . . . . . . 395 16.7.2 Suggested Method (“Sorting/Clustering Approach”), also Called Classical Transcript Analysis . . . . . . . . . . . . . . . . . . . . . . . . 396 16.7.3 Report Format . . . . . . . . . . . . . . . . . . . . . 397 16.8 Alternative Procedures and Variations of the Group Interview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 16.8.1 Groups of Children, Telephone Interviews, Internet-Based Groups . . . . . . . . . . . . . . . . 398 16.8.2 Alternatives to Traditional Questioning . . . . . . . . 399 16.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 400 Appendix: Sample Report Group Report . . . . . . . . . . . . . . . . . 402 Boil-in-bag Pasta Project Followup Groups . . . . . . . . . . . . . . . 402 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404 17 Quality Control and Shelf-Life (Stability) Testing . 17.1 Introduction: Objectives and Challenges . . . 17.2 A Quick Look at Traditional Quality Control . 17.3 Methods for Sensory QC . . . . . . . . . . . 17.3.1 Cuttings: A Bad Example . . . . . . 17.3.2 In-Out (Pass/Fail) System . . . . . 17.3.3 Difference from Control Ratings . . 17.3.4 Quality Ratings with Diagnostics . . 17.3.5 Descriptive Analysis . . . . . . . . 17.3.6 A Hybrid Approach: Quality Ratings with Diagnostics . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

407 408 409 409 409 410 411 412 413

. . . . . . . . .

414

Contents

xix

17.3.7 The Multiple Standards Difference Test . . . . . . . Recommended Procedure: Difference Scoring with Key Attribute Scales . . . . . . . . . . . . . . . . . . . . . . . 17.5 The Importance of Good Practice . . . . . . . . . . . . . . . . 17.6 Historical Footnote: Expert Judges and Quality Scoring . . . . 17.6.1 Standardized Commodities . . . . . . . . . . . . . . 17.6.2 Example 1: Dairy Product Judging . . . . . . . . . . 17.6.3 Example 2: Wine Scoring . . . . . . . . . . . . . . . 17.7 Program Requirements and Program Development . . . . . . . 17.7.1 Desired Features of a Sensory QC System . . . . . . 17.7.2 Program Development and Management Issues . . . 17.7.3 The Problem of Low Incidence . . . . . . . . . . . . 17.8 Shelf-Life Testing . . . . . . . . . . . . . . . . . . . . . . . . 17.8.1 Basic Considerations . . . . . . . . . . . . . . . . . 17.8.2 Cutoff Point . . . . . . . . . . . . . . . . . . . . . . 17.8.3 Test Designs . . . . . . . . . . . . . . . . . . . . . . 17.8.4 Survival Analysis and Hazard Functions . . . . . . . 17.8.5 Accelerated Storage . . . . . . . . . . . . . . . . . . 17.9 Summary and Conclusions . . . . . . . . . . . . . . . . . . . Appendix 1: Sample Screening Tests for Sensory Quality Judges . . . . Appendix 2: Survival/Failure Estimates from a Series of Batches with Known Failure Times . . . . . . . . . . . . . . Appendix 3: Arrhenius Equation and Q10 Modeling . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

414

17.4

18

19

415 417 419 419 419 420 422 422 423 424 424 424 426 426 427 428 428 429 429 430 431

Data Relationships and Multivariate Applications . . . . . . . . . . 18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2 Overview of Multivariate Statistical Techniques . . . . . . . . 18.2.1 Principal Component Analysis . . . . . . . . . . . . 18.2.2 Multivariate Analysis of Variance . . . . . . . . . . . 18.2.3 Discriminant Analysis (Also Known as Canonical Variate Analysis) . . . . . . . . . . . . 18.2.4 Generalized Procrustes Analysis . . . . . . . . . . . 18.3 Relating Consumer and Descriptive Data Through Preference Mapping . . . . . . . . . . . . . . . . . . . . . . . 18.3.1 Internal Preference Mapping . . . . . . . . . . . . . 18.3.2 External Preference Mapping . . . . . . . . . . . . . 18.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

440 442 442 445 446

Strategic Research . . . . . . . . . . . . . . . 19.1 Introduction . . . . . . . . . . . . . . . 19.1.1 Avenues for Strategic Research 19.1.2 Consumer Contact . . . . . . . 19.2 Competitive Surveillance . . . . . . . . 19.2.1 The Category Review . . . . . 19.2.2 Perceptual Mapping . . . . . . 19.2.3 Multivariate Methods: PCA . . 19.2.4 Multi-dimensional Scaling . .

451 451 451 453 453 453 455 456 458

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

433 433 434 434 437 438 439

xx

Contents

19.2.5

Cost-Efficient Methods for Data Collection: Sorting . . . . . . . . . . . . . . . . . . . . . . 19.2.6 Vector Projection . . . . . . . . . . . . . . . . 19.2.7 Cost-Efficient Methods for Data Collection: Projective Mapping, aka Napping . . . . . . . . 19.3 Attribute Identification and Classification . . . . . . . . . 19.3.1 Drivers of Liking . . . . . . . . . . . . . . . . 19.3.2 The Kano Model . . . . . . . . . . . . . . . . 19.4 Preference Mapping Revisited . . . . . . . . . . . . . . . 19.4.1 Types of Preference Maps . . . . . . . . . . . . 19.4.2 Preference Models: Vectors Versus Ideal Points 19.5 Consumer Segmentation . . . . . . . . . . . . . . . . . . 19.6 Claim Substantiation Revisited . . . . . . . . . . . . . . 19.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 19.7.1 Blind Testing, New Coke, and the Vienna Philharmonic . . . . . . . . . . . . . . . . . . 19.7.2 The Sensory Contribution . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

459 460

. . . . . . . . . .

. . . . . . . . . .

461 462 462 463 464 464 464 465 467 468

. . . . . . . . .

468 469 469

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

473 473 474 475 476 478 478 478 479 480 482 482

. . . . .

. . . . .

. . . . .

484 485 485 486 487

. . . . . . . .

. . . . . . . .

. . . . . . . .

489 489 490 493 493 494 494 495

. . . . .

496

Appendix A Basic Statistical Concepts for Sensory Evaluation . . A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . A.2 Basic Statistical Concepts . . . . . . . . . . . . . . . . . A.2.1 Data Description . . . . . . . . . . . . . . . . A.2.2 Population Statistics . . . . . . . . . . . . . . . A.3 Hypothesis Testing and Statistical Inference . . . . . . . A.3.1 The Confidence Interval . . . . . . . . . . . . . A.3.2 Hypothesis Testing . . . . . . . . . . . . . . . A.3.3 A Worked Example . . . . . . . . . . . . . . . A.3.4 A Few More Important Concepts . . . . . . . . A.3.5 Decision Errors . . . . . . . . . . . . . . . . . A.4 Variations of the t-Test . . . . . . . . . . . . . . . . . . . A.4.1 The Sensitivity of the Dependent t-Test for Sensory Data . . . . . . . . . . . . . . . . . . A.5 Summary: Statistical Hypothesis Testing . . . . . . . . . A.6 Postscript: What p-Values Signify and What They Do Not A.7 Statistical Glossary . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix B Nonparametric and Binomial-Based Statistical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.1 Introduction to Nonparametric Tests . . . . . . . . . B.2 Binomial-Based Tests on Proportions . . . . . . . . . B.3 Chi-Square . . . . . . . . . . . . . . . . . . . . . . . B.3.1 A Measure of Relatedness of Two Variables B.3.2 Calculations . . . . . . . . . . . . . . . . . B.3.3 Related Samples: The McNemar Test . . . . B.3.4 The Stuart-Maxwell Test . . . . . . . . . . B.3.5 Beta-Binomial, Chance-Corrected Beta-Binomial, and Dirichlet Multinomial Analyses . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . . . .

Contents

xxi

B.4

Useful Rank Order Tests . . . . . . . . . . . . . . . . B.4.1 The Sign Test . . . . . . . . . . . . . . . . B.4.2 The Mann-Whitney U-Test . . . . . . . . . B.4.3 Ranked Data with More Than Two Samples, Friedman and Kramer Tests . . . . . . . . . B.4.4 Rank Order Correlation . . . . . . . . . . . B.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . B.6 Postscript . . . . . . . . . . . . . . . . . . . . . . . . B.6.1 Proof showing equivalence of binomial approximation Z-test and χ 2 test for difference of proportions . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .

499 499 500

. . . .

. . . .

501 502 503 503

. . . . . . . . . .

503 504

. . . .

Appendix C Analysis of Variance . . . . . . . . . . . . . . . . . . . C.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . C.1.1 Overview . . . . . . . . . . . . . . . . . . . . C.1.2 Basic Analysis of Variance . . . . . . . . . . . C.1.3 Rationale . . . . . . . . . . . . . . . . . . . . C.1.4 Calculations . . . . . . . . . . . . . . . . . . . C.1.5 A Worked Example . . . . . . . . . . . . . . . C.2 Analysis of Variance from Complete Block Designs . . . C.2.1 Concepts and Partitioning Panelist Variance from Error . . . . . . . . . . . . . . . . . . . . C.2.2 The Value of Using Panelists As Their Own Controls . . . . . . . . . . . . . C.3 Planned Comparisons Between Means Following ANOVA C.4 Multiple Factor Analysis of Variance . . . . . . . . . . . C.4.1 An Example . . . . . . . . . . . . . . . . . . . C.4.2 Concept: A Linear Model . . . . . . . . . . . . C.4.3 A Note About Interactions . . . . . . . . . . . C.5 Panelist by Product by Replicate Designs . . . . . . . . . C.6 Issues and Concerns . . . . . . . . . . . . . . . . . . . . C.6.1 Sensory Panelists: Fixed or Random Effects? . C.6.2 A Note on Blocking . . . . . . . . . . . . . . . C.6.3 Split-Plot or Between-Groups (Nested) Designs C.6.4 Statistical Assumptions and the Repeated Measures ANOVA . . . . . . . . . . . . . . . C.6.5 Other Options . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix D Correlation, Regression, and Measures of Association D.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . D.2 Correlation . . . . . . . . . . . . . . . . . . . . . . . . . D.2.1 Pearson’s Correlation Coefficient Example . . . D.2.2 Coefficient of Determination . . . . . . . . . . D.3 Linear Regression . . . . . . . . . . . . . . . . . . . . . D.3.1 Analysis of Variance . . . . . . . . . . . . . . D.3.2 Analysis of Variance for Linear Regression . . D.3.3 Prediction of the Regression Line . . . . . . . . D.3.4 Linear Regression Example . . . . . . . . . . .

. . . .

. . . . . . . .

. . . .

. . . . . . . .

. . . . . . . .

507 507 507 508 508 509 509 510

. . .

510

. . . . . . . . . . .

. . . . . . . . . . .

512 513 514 514 515 516 516 519 519 520 520

. . . . . . . . .

521 522 522

. . . . . . . . . .

525 525 527 528 529 529 530 530 530 531

. . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

xxii

Contents

D.4 D.5

Multiple Linear Regression . . . . . . . . . . . . . . Other Measures of Association . . . . . . . . . . . . D.5.1 Spearman Rank Correlation . . . . . . . . . D.5.2 Spearman Correlation Coefficient Example D.5.3 Cramér’s V Measure . . . . . . . . . . . . . D.5.4 Cramér Coefficient Example . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

531 531 531 532 532 533 533

Appendix E Statistical Power and Test Sensitivity . . . . . . . . E.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . E.2 Factors Affecting the Power of Statistical Tests . . . . . E.2.1 Sample Size and Alpha Level . . . . . . . . . E.2.2 Effect Size . . . . . . . . . . . . . . . . . . . E.2.3 How Alpha, Beta, Effect Size, and N Interact E.3 Worked Examples . . . . . . . . . . . . . . . . . . . . E.3.1 The t-Test . . . . . . . . . . . . . . . . . . . E.3.2 An Equivalence Issue with Scaled Data . . . E.3.3 Sample Size for a Difference Test . . . . . . . E.4 Power in Simple Difference and Preference Tests . . . . E.5 Summary and Conclusions . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

535 535 537 537 538 539 541 541 542 544 545 548 549

Appendix F Statistical Tables . . . . . . . . . . . . . . . . . . . . . . . . . Table F.A Cumulative probabilities of the standard normal distribution. Entry area 1-α under the standard normal curve from −∞ to z(1-α) . . . . . . . . . . . . . Table F.B Table of critical values for the t-distribution . . . . . . . . Table F.C Table of critical values of the chi-square (χ 2 ) distribution . . . . . . . . . . . . . . . . . . . . . . . . . Table F.D1 Critical values of the F-distribution at α = 0.05 . . . . . . Table F.D2 Critical values of the F-distribution at α = 0.01 . . . . . . Table F.E Critical values of U for a one-tailed alpha at 0.025 or a two-tailed alpha at 0.05 . . . . . . . . . . . . . . . . Table F.F1 Table of critical values of ρ (Spearman Rank correlation coefficient) . . . . . . . . . . . . . . . . . . . Table F.F2 Table of critical values of r (Pearson’s correlation coefficient) . . . . . . . . . . . . . . . . . . . . . . . . . Table F.G Critical values for Duncan’s multiple range test (p, df, α = 0.05) . . . . . . . . . . . . . . . . . . . . . . . Table F.H1 Critical values of the triangle test for similarity (maximum number correct as a function of the number of observations (N), beta, and proportion discriminating) . . . . . . . . . . . . . . . . . . . . . . . Table F.H2 Critical values of the duo-trio and paired comparison tests for similarity (maximum number correct as a function of the number of observations (N), beta, and proportion discriminating) . . . . . . . . . . Table F.I Table of probabilities for values as small as observed values of x associated with the binomial test (p=0.50) . . . . . . . . . . . . . . . . . . . . . . . .

551

552 553 554 555 556 556 557 558 559

560

561

562

Contents

xxiii

Table F.J

Critical values for the differences between rank sums (α = 0.05) . . . . . . . . . . . . . . . . . . . . . . . Critical values of the beta binomial distribution . . . . . . Minimum numbers of correct judgments to establish significance at probability levels of 5 and 1% for paired difference and duo-trio tests (one tailed, p = 1/2) and the triangle test (one tailed, p = 1/3) . . . . . . . . . . . . . . . . . . Minimum numbers of correct judgments to establish significance at probability levels of 5 and 1% for paired pference test (two tailed, p = 1/2) . . . . . . . . . . . . . . . . . . . . . . . . . . . Minimum number of responses (n) and correct responses (x) to obtain a level of Type I and Type II risks in the triangle test. Pd is the chance-adjusted percent correct or proportion of discriminators . . . . . . . . . . . . . . . . . . . . . . Minimum number of responses (n) and correct responses (x) to obtain a level of Type I and Type II risks in the duo-trio test. Pc is the chance-adjusted percent correct or proportion of discriminators . . . . . . . . . . . . . . . . . . . . . . d and B (variance factor) values for the duo-trio and 2-AFC (paired comparison) difference tests . . . . . . d and B (variance factor) values for the triangle and 3-AFC difference tests . . . . . . . . . . . . . . . . . Random permutations of nine . . . . . . . . . . . . . . . Random numbers . . . . . . . . . . . . . . . . . . . . . .

569 571 572

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

573

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

587

Table F.K Table F.L

Table F.M

Table F.N1

Table F.N2

Table F.O1 Table F.O2 Table F.P Table F.Q

563 564

565

566

567

567 568

Chapter 1

Introduction

Abstract In this chapter we carefully parse the definition for sensory evaluation, briefly discuss validity of the data collected before outlining the early history of the field. We then describe the three main methods used in sensory evaluation (discrimination tests, descriptive analysis, and hedonic testing) before discussing the differences between analytical and consumer testing. We then briefly discuss why one may want to collect sensory data. In the final sections we highlight the differences and similarities between sensory evaluation and marketing research and between sensory evaluation and commodity grading as used in, for example, the dairy industry.

Sensory evaluation is a child of industry. It was spawned in the late 40’s by the rapid growth of the consumer product companies, mainly food companies. . . . Future development in sensory evaluation will depend upon several factors, one of the most important being the people and their pparation and training. – Elaine Skinner (1989)

Contents Introduction and Overview . . . . . . . . . 1.1.1 Definition . . . . . . . . . . . . . . 1.1.2 Measurement . . . . . . . . . . . . 1.2 Historical Landmarks and the Three Classes of Test Methods . . . . . . . . . . . . . . 1.2.1 Difference Testing . . . . . . . . . . 1.2.2 Descriptive Analyses . . . . . . . . . 1.2.3 Affective Testing . . . . . . . . . . 1.2.4 The Central Dogma-Analytic Versus Hedonic Tests . . . . . . . . . . . . 1.3 Applications: Why Collect Sensory Data? . . 1.3.1 Differences from Marketing Research Methods . . . . . . . . . . . . . . 1.3.2 Differences from Traditional Product Grading Systems . . . . . . . . . . 1.4 Summary and Conclusions . . . . . . . . . References . . . . . . . . . . . . . . . . . . . 1.1

1.1 Introduction and Overview . . .

1 1 3

. . . .

4 5 6 7

. .

8 10

.

13

. . .

15 16 17

1.1.1 Definition The field of sensory evaluation grew rapidly in the second half of the twentieth century, along with the expansion of the processed food and consumer products industries. Sensory evaluation comprises a set of techniques for accurate measurement of human responses to foods and minimizes the potentially biasing effects of brand identity and other information influences on consumer perception. As such, it attempts to isolate the sensory properties of foods themselves and provides important and useful information to product developers, food scientists, and managers about the sensory characteristics of their products. The field was comphensively reviewed by Amerine, Pangborn, and Roessler in 1965, and more recent texts have been published by Moskowitz et al. (2006), Stone and Sidel (2004), and Meilgaard et al. (2006). These three later sources are practical works aimed at sensory specialists

H.T. Lawless, H. Heymann, Sensory Evaluation of Food, Food Science Text Series, DOI 10.1007/978-1-4419-6488-5_1, © Springer Science+Business Media, LLC 2010

1

2

in industry and reflect the philosophies of the consulting groups of the authors. Our goal in this book is to provide a comphensive overview of the field with a balanced view based on research findings and one that is suited to students and practitioners alike. Sensory evaluation has been defined as a scientific method used to evoke, measure, analyze, and interpt those responses to products as perceived through the senses of sight, smell, touch, taste, and hearing (Stone and Sidel, 2004). This definition has been accepted and endorsed by sensory evaluation committees within various professional organizations such as the Institute of Food Technologists and the American Society for Testing and Materials. The principles and practices of sensory evaluation involve each of the four activities mentioned in this definition. Consider the words “to evoke.” Sensory evaluation gives guidelines for the pparation and serving of samples under controlled conditions so that biasing factors are minimized. For example, people in a sensory test are often placed in inpidual test booths so that the judgments they give are their own and do not reflect the opinions of those around them. Samples are labeled with random numbers so that people do not form judgments based upon labels, but rather on their sensory experiences. Another example is in how products may be given in different orders to each participant to help measure and counterbalance for the sequential effects of seeing one product after another. Standard procedures may be established for sample temperature, volume, and spacing in time, as needed to control unwanted variation and improve test pcision. Next, consider the words, “to measure.” Sensory evaluation is a quantitative science in which numerical data are collected to establish lawful and specific relationships between product characteristics and human perception. Sensory methods draw heavily from the techniques of behavioral research in observing and quantifying human responses. For example, we can assess the proportion of times people are able to discriminate small product changes or the proportion of a group that expsses a pference for one product over another. Another example is having people generate numerical responses reflecting their perception of how strong a product may taste or smell. Techniques of behavioral research and experimental psychology offer guidelines as to how such measurement techniques should be employed and what their potential pitfalls and liabilities may be.

1 Introduction

The third process in sensory evaluation is analysis. Proper analysis of the data is a critical part of sensory testing. Data generated from human observers are often highly variable. There are many sources of variation in human responses that cannot be completely controlled in a sensory test. Examples include the mood and motivation of the participants, their innate physiological sensitivity to sensory stimulation, and their past history and familiarity with similar products. While some screening may occur for these factors, they may be only partially controlled, and panels of humans are by their nature heterogeneous instruments for the generation of data. In order to assess whether the relationships observed between product characteristics and sensory responses are likely to be real, and not merely the result of uncontrolled variation in responses, the methods of statistics are used to analyze evaluation data. Hand-in-hand with using appropriate statistical analyses is the concern of using good experimental design, so that the variables of interest are investigated in a way that allows sensible conclusions to be drawn. The fourth process in sensory evaluation is the interptation of results. A sensory evaluation exercise is necessarily an experiment. In experiments, data and statistical information are only useful when interpted in the context of hypotheses, background knowledge, and implications for decisions and actions to be taken. Conclusions must be drawn that are reasoned judgments based upon data, analyses, and results. Conclusions involve consideration of the method, the limitations of the experiment, and the background and contextual framework of the study. The sensory evaluation specialists become more than mere conduits for experimental results, but must contribute interptations and suggest reasonable courses of action in light of the numbers. They should be full partners with their clients, the end-users of the test results, in guiding further research. The sensory evaluation professional is in the best situation to realize the appropriate interptation of test results and the implications for the perception of products by the wider group of consumers to whom the results may be generalized. The sensory specialist best understands the limitations of the test procedure and what its risks and liabilities may be. A sensory scientist who is ppared for a career in research must be trained in all four of the phases mentioned in the definition. They must understand products, people as measuring instruments, statistical

1.1 Introduction and Overview

1.1.2 Measurement Sensory evaluation is a science of measurement. Like other analytical test procedures, sensory evaluation is concerned with pcision, accuracy, sensitivity, and avoiding false positive results (Meiselman, 1993). Precision is similar to the concept in the behavioral sciences of reliability. In any test procedure, we would like to be able to get the same result when a test is repeated. There is usually some error variance around an obtained value, so that upon repeat testing, the value will not always be exactly the same. This is especially true of sensory tests in which human perceptions are necessarily part of the generation of data. However, in many sensory test procedures, it is desirable to minimize this error variance as much as possible and to have tests that are low in error associated with repeated measurements. This is achieved by several means. As noted above, we isolate the sensory response to the factors of interest, minimizing extraneous influences, controlling sample pparation and psentation. Additionally, as necessary, sensory scientists screen and train panel participants. A second concern is the accuracy of a test. In the physical sciences, this is viewed as the ability of a test instrument to produce a value that is close to the “true” value, as defined by independent measurement from another instrument or set of instruments that have been appropriately calibrated. A related idea in the behavioral sciences, this principle is called the validity of a test. This concerns the ability of a test procedure to measure what it was designed and intended to measure. Validity is established in a number of ways. One useful criterion is pdictive validity, when a test result is of value in pdicting what would occur in another situation or another measurement. In sensory testing, for example, the test results should reflect the perceptions and opinions of consumers that might buy the product. In other words, the results of the sensory test should generalize to the larger population. The test results might correlate with instrumental measures, process or ingredient variables, storage factors, shelf life times,

3

or other conditions known to affect sensory properties. In considering validity, we have to look at the end use of the information provided by a test. A sensory test method might be valid for some purposes, but not others (Meiselman, 1993). A simple difference test can tell if a product has changed, but not whether people will like the new version. A good sensory test will minimize errors in measurement and errors in conclusions and decisions. There are different types of errors that may occur in any test procedure. Whether the test result reflects the true state of the world is an important question, especially when error and uncontrolled variability are inherent in the measurement process. Of primary concern in sensory tests is the sensitivity of the test to differences among products. Another way to phrase this is that a test should not often miss important differences that are psent. “Missing a difference” implies an insensitive test procedure. To keep sensitivity high, we must minimize error variance wherever possible by careful experimental controls and by selection and training of panelists where appropriate. The test must involve sufficient numbers of measurements to insure a tight and reliable statistical estimate of the values we obtain, such as means or proportions. In statistical language, detecting true differences is avoiding Type II error and the minimization of β-risk. Discussion of the power and sensitivity of tests from a statistical perspective occurs in Chapter 5 and in the Appendix. The other error that may occur in a test result is that of finding a positive result when none is actually psent in the larger population of people and products outside the sensory test. Once again, a positive result usually means detection of a statistically significant difference between test products. It is important to use a test method that avoids false positive results or Type I error in statistical language. Basic statistical training and common statistical tests applied to scientific findings are oriented toward avoiding this kind of error. The effects of random chance deviations must be taken into account in deciding if a test result reflects a real difference or whether our result is likely to be due to chance variation. The common procedures of inferential statistics provide assurance that we have limited our possibility of finding a difference where one does not really exist. Statistical procedures reduce this risk to some comfortable level, usually with a ceiling of 5% of all tests we conduct.

4

1 Introduction

(Lawless and Klein, 1989). An important function of sensory scientists in an academic setting is to provide consulting and resources to insure that quality tests are conducted by other researchers and students who seek to understand the sensory impact of the variables they are studying. In government services such as food inspection, sensory evaluation plays a key role (York, 1995). Sensory principles and appropriate training can be key in insuring that test methods reflect the current knowledge of sensory function and test design. See Lawless (1993) for an overview of the education and training of sensory scientists-much of this piece still rings true more than 15 years later.

1.2 Historical Landmarks and the Three Classes of Test Methods The human senses have been used for centuries to evaluate the quality of foods. We all form judgments about foods whenever we eat or drink (“Everyone carries his own inch-rule of taste, and amuses himself by applying it, triumphantly, wherever he travels.”-Henry Adams, 1918). This does not mean that all judgments are useful or that anyone is qualified to participate in a sensory test. In the past, production of good quality foods often depended upon the sensory acuity of a single expert who was in charge of production or made decisions about process changes in order to make sure the product would have desirable characteristics. This was the historical tradition of brewmasters, wine tasters, dairy judges, and other food inspectors who acted as the arbiters of quality. Modern sensory evaluation replaced these single authorities with panels of people participating in specific test methods that took the form of planned experiments. This change occurred for several reasons. First, it was recognized that the judgments of a panel would in general be more reliable than the judgments of single inpidual and it entailed less risk since the single expert could become ill, travel, retire, die, or be otherwise unavailable to make decisions. Replacement of such an inpidual was a nontrivial problem. Second, the expert might or might not reflect what consumers or segments of the consuming public might want in a product. Thus for issues of product quality and overall appeal, it was safer (although often more time consuming and expensive)

1.2 Historical Landmarks and the Three Classes of Test Methods

to go directly to the target population. Although the tradition of informal, qualitative inspections such as benchtop “cuttings” persists in some industries, they have been gradually replaced by more formal, quantitative, and controlled observations (Stone and Sidel, 2004). The current sensory evaluation methods comprise a set of measurement techniques with established track records of use in industry and academic research. Much of what we consider standard procedures comes from pitfalls and problems encountered in the practical experience of sensory specialists over the last 70 years of food and consumer product research, and this experience is considerable. The primary concern of any sensory evaluation specialist is to insure that the test method is appropriate to answer the questions being asked about the product in the test. For this reason, tests are usually classified according to their primary purpose and most valid use. Three types of sensory testing are commonly used, each with a different goal and each using participants selected using different criteria. A summary of the three main types of testing is given in Table 1.1.

1.2.1 Difference Testing The simplest sensory tests merely attempt to answer whether any perceptible difference exists between two types of products. These are the discrimination tests or simple difference testing procedures. Analysis is usually based on the statistics of frequencies and proportions (counting right and wrong answers). From the test results, we infer differences based on the proportions of persons who are able to choose a test product correctly from among a set of similar or control products. A classic example of this test was the triangle procedure, used in the Carlsberg breweries and in the Seagrams distilleries in the 1940s (Helm and Trolle,

5

1946; Peryam and Swartz, 1950). In this test, two products were from the same batch while a third product was different. Judges would be asked to pick the odd sample from among the three. Ability to discriminate differences would be inferred from consistent correct choices above the level expected by chance. In breweries, this test served primarily as a means to screen judges for beer evaluation, to insure that they possessed sufficient discrimination abilities. Another multiple-choice difference test was developed at about the same time in distilleries for purposes of quality control (Peryam and Swartz, 1950). In the duo-trio procedure, a reference sample was given and then two test samples. One of the test samples matched the reference while the other was from a different product, batch or process. The participant would try to match the correct sample to the reference, with a chance probability of one-half. As in the triangle test, a proportion of correct choices above that expected by chance is considered evidence for a perceivable difference between products. A third popular difference test was the paired comparison, in which participants would be asked to choose which of two products was stronger or more intense in a given attribute. Partly due to the fact that the panelist’s attention is directed to a specific attribute, this test is very sensitive to differences. These three common difference tests are shown in Fig. 1.1. Simple difference tests have proven very useful in application and are in widespad use today. Typically a discrimination test will be conducted with 25-40 participants who have been screened for their sensory acuity to common product differences and who are familiar with the test procedures. This generally provides an adequate sample size for documenting clear sensory differences. Often a replicate test is performed while the respondents are psent in the sensory test facility. In part, the popularity of these tests is due to the simplicity of data analysis. Statistical tables derived from the binomial distribution give the minimum number of correct responses needed to conclude statistical

Table 1.1 Classification of test methods in sensory evaluation Class Question of interest

Type of test

Panelist characteristics

Discrimination

Are products perceptibly different in any way

“Analytic”

Descriptive

How do products differ in specific sensory characteristics How well are products liked or which products are pferred

“Analytic”

Screened for sensory acuity, oriented to test method, sometimes trained Screened for sensory acuity and motivation, trained or highly trained Screened for products, untrained

Affective

“Hedonic”

6

1 Introduction

Fig. 1.1 Common methods for discrimination testing include the triangle, duo-trio, and paired comparison procedures.

significance as a function of the number of participants. Thus a sensory technician merely needs to count answers and refer to a table to give a simple statistical conclusion, and results can be easily and quickly reported.

the consensus of a panel was likely to be more reliable and accurate than the judgment of a single inpidual. Second, it provided a means to characterize the inpidual attributes of flavor and provide a comphensive analytical description of differences among a group of products under development. Several variations and refinements in descriptive analysis techniques were forthcoming. A group at the General Foods Technical Center in the early 1960s developed and refined a method to quantify food texture, much as the flavor profile had enabled the quantification of flavor properties (Brandt et al., 1963, Szczesniak et al., 1975). This technique, the Texture Profile method, used a fixed set of force-related and shape-related attributes to characterize the rheological and tactile properties of foods and how these changed over time with mastication. These characteristics have parallels in the physical evaluation of food breakdown or flow. For example, perceived hardness is related to the physical force required to penetrate a sample. Perceived thickness of a fluid or semisolid is related in part to physical viscosity. Texture profile panelists were also trained to recognize specific intensity points along each scale, using standard products or formulated pseudo-foods for calibration. Other approaches were developed for descriptive analysis problems. At Stanford Research Institute in the early 1970s, a group proposed a method for descriptive analysis that would remedy some of the R apparent shortcomings of the Flavor Profile method

1.2 Historical Landmarks and the Three Classes of Test Methods

Table 1.2 Descriptive evaluation of cookies-texture attributes Phase Attributes Word anchors Surface

First bite

First chew Chew down

Residual

Roughness Particles Dryness Fracturability Hardness Particle size Denseness Uniformity of chew Moisture absorption Cohesiveness of mass Toothpacking Grittiness Oiliness Particles Chalky

Smooth-rough None-many Oily-dry Crumbly-brittle Soft-hard Small-large Airy-dense Even-uneven None-much Loose-cohesive None-much None-much Dry-oily None-many Not chalky-very chalky

7

a uniform and controlled manner, typical of an analytical sensory test procedure. For example, the first bite may be defined as cutting with the incisors. The panel for such an analysis would consist of perhaps 10- 12 well-trained inpiduals, who were oriented to the meanings of the terms and given practice with examples. Intensity references to exemplify scale points are also given in some techniques. Note the amount of detailed information that can be provided in this example and bear in mind that this is only looking at the product’s texture-flavor might form an equally detailed sensory analysis, perhaps with a separate trained panel. The relatively small number of panelists (a dozen or so) is justified due to their level of calibration. Since they have been trained to use attribute scales in a similar manner, error variance is lowered and statistical power and test sensitivity are maintained in spite of fewer observations (fewer data points per product). Similar examples of texture, flavor, fragrance, and tactile analyses can be found in Meilgaard et al. (2006).

8

1 Introduction

the food product (the stimulus). However, since a sensory perception involves the necessary interaction of a person with a stimulus, it should be apparent that similar test methods are necessary to characterize this person-product interaction.

1.2.4 The Central Dogma-Analytic Versus Hedonic Tests

Fig. 1.2 The 9-point hedonic scale used to assess liking and disliking. This scale, originally developed at the U.S. Army Food and Container Institute (Quartermaster Corps), has achieved widespad use in consumer testing of foods.

The central principle for all sensory evaluation is that the test method should be matched to the objectives of the test. Figure 1.3 shows how the selection of the test procedure flows from questions about the objective of the investigation. To fulfill this goal, it is necessary to have clear communication between the sensory test manager and the client or end-user of the information. A dialogue is often needed. Is the important question whether or not there is any difference at all among the products? If so, a discrimination test is indicated. Is the question one of whether consumers like the new product better than the pvious version? A consumer acceptance test is needed. Do we need to know what attributes have changed in the sensory characteristics of the new product? Then a descriptive analysis procedure is called for. Sometimes there are multiple objectives and a sequence of different tests is required (Lawless and Claassen, 1993). This can psent problems if all the answers are required at once or under severe time pssure during competitive product development. One of the most important jobs of the sensory specialist in the food industry is to insure a clear understanding and specification of the type of information needed by the end-users. Test design may require a number of conversations, interviews with different people, or even written test requests that specify why the information is to be collected and how the results will be used in making specific decisions and subsequent actions to be taken. The sensory specialist is the best position to understand the uses and limitations of each procedure and what would be considered appropriate versus inappropriate conclusions from the data. There are two important corollaries to this principle. The sensory test design involves not only the selection of an appropriate method but also the selection of appropriate participants and statistical analyses. The three classes of sensory tests can be pided into two types, analytical sensory tests including discrimination

1.2 Historical Landmarks and the Three Classes of Test Methods

9

Fig. 1.3 A flowchart showing methods determination. Based on the major objectives and research questions, different sensory test methods are selected. Similar decision processes are made in panelist selection, setting up response scales, in choosing experimental designs, statistical analysis, and other tasks in designing a sensory test (reprinted with permission from Lawless, 1993).

and descriptive methods and affective or hedonic tests such as those involved in assessing consumer liking or pferences (Lawless and Claassen, 1993). For the analytical tests, panelists are selected based on having average to good sensory acuity for the critical characteristics (tastes, smells, textures, etc.) of products to be evaluated. They are familiarized with the test procedures and may undergo greater or lesser amounts of training, depending upon the method. In the case of descriptive analysis, they adopt an analytical frame of mind, focusing on specific aspects of the product as directed by the scales on their questionnaires. They are asked to put personal pferences and hedonic reactions aside, as their job is only to specify what attributes are psent in the product and at what levels of sensory intensity, extent, amount, or duration. In contrast to this analytical frame of mind, consumers in an affective test act in a much more integrative fashion. They perceive a product as a whole pattern. Although their attention is sometimes captured by a specific aspect of a product (especially if it is a bad, unexpected, or unpleasant one), their reactions to the product are often immediate and based on the integrated pattern of sensory stimulation from the product and expssed as liking or disliking. This occurs without a great deal of thought or dissection of the product’s specific profile. In other words, consumers are effective at rendering impssions based on the integrated pattern of perceptions. In such consumer

tests, participants must be chosen carefully to insure that the results will generalize to the population of interest. Participants should be frequent users of the product, since they are most likely to form the target market and will be familiar with similar products. They possess reasonable expectations and a frame of reference within which they can form an opinion relative to other similar products they have tried. The analytic/hedonic distinction gives rise to some highly important rules of thumb and some warnings about matching test methods and respondents. It is unwise to ask trained panelists about their pferences or whether they like or dislike a product. They have been asked to assume a different, more analytical frame of mind and to place personal pference aside. Furthermore, they have not necessarily been selected to be frequent users of the product, so they are not part of the target population to which one would like to generalize hedonic test results. A common analogy here is to an analytical instrument. You would not ask a gas chromatograph or a pH meter whether it liked the product, so why ask your analytical descriptive panel (O’Mahony, 1979). Conversely, problems arise when consumers are asked to furnish very specific information about product attributes. Consumers not only act in a non-analytic frame of mind but also often have very fuzzy concepts about specific attributes, confusing sour and bitter tastes, for example. Inpiduals often differ markedly

10

in their interptations of sensory attribute words on a questionnaire. While a trained texture profile panel has no trouble in agreeing how cohesive a product is after chewing, we cannot expect consumers to provide pcise information on such a specific and technical attribute. In summary, we avoid using trained panelists for affective information and we avoid asking consumers about specific analytical attributes. Related to the analytic-hedonic distinction is the question of whether experimental control and pcision are to be maximized or whether validity and generalizability to the real world are more important. Often there is a tradeoff between the two and it is difficult to maximize both simultaneously. Analytic tests in the lab with specially screened and trained judges are more reliable and lower in random error than consumer tests. However, we give up a certain amount of generalizability to real-world results by using artificial conditions and a special group of participants. Conversely, in the testing of products by consumers in their own homes we have not only a lot of real-life validity but also a lot of noise in the data. Brinberg and McGrath (1985) have termed this struggle between pcision and validity one of “conflicting desiderata.” O’Mahony (1988) has made a distinction between sensory evaluation Type I and Type II. In Type I sensory evaluation, reliability and sensitivity are key factors, and the participant is viewed much like an analytical instrument used to detect and measure changes in a food product. In Type II sensory evaluation, participants are chosen to be repsentative of the consuming population, and they may evaluate food under more naturalistic conditions. Their emphasis here is on pdiction of consumer response. Every sensory test falls somewhere along a continuum where reliability versus real-life extrapolation are in a potential tradeoff relationship. This factor must also be discussed with end-users of the data to see where their emphasis lies and what level of tradeoff they find comfortable. Statistical analyses must also be chosen with an eye to the nature of the data. Discrimination tests involve choices and counting numbers of correct responses. The statistics derived from the binomial distribution or those designed for proportions such as chi-square are appropriate. Conversely, for most scaled data, we can apply the familiar parametric statistics appropriate to normally distributed and continuous data, such as means, standard deviations, t-tests, analysis of variance. The choice of an appropriate statistical test is not

1 Introduction

always straightforward, so sensory specialists are wise to have thorough training in statistics and to involve statistical and design specialists in a complex project in its earliest stages of planning. Occasionally, these central principles are violated. They should not be put aside as a matter of mere expediency or cost savings and never without a logical analysis. One common example is the use of a discrimination test before consumer acceptance. Although our ultimate interest may lie in whether consumers will like or dislike a new product variation, we can conduct a simple difference test to see whether any change is perceivable at all. The logic in this sequence is the following: if a screened and experienced discrimination panel cannot tell the difference under carefully controlled conditions in the sensory laboratory, then a more heterogeneous group of consumers is unlikely to see a difference in their less controlled and more variable world. If no difference is perceived, there can logically be no systematic pference. So a more time consuming and costly consumer test can sometimes be avoided by conducting a simpler but more sensitive discrimination test first. The added reliability of the controlled discrimination test provides a “safety net” for conclusions about consumer perception. Of course, this logic is not without its pitfalls-some consumers may interact extensively with the product during a home use test period and may form stable and important opinions that are not captured in a short duration laboratory test, and there is also always the possibility of a false negative result (the error of missing a difference). MacRae and Geelhoed (1992) describe an interesting case of a missed difference in a triangle test where a significant pference was then observed between water samples in a paired comparison. The sensory professional must be aware that these anomalies in experimental results will sometimes arise, and must also be aware of some of the reasons why they occur.

1.3 Applications: Why Collect Sensory Data? Human perceptions of foods and consumer products are the results of complex sensory and interptation processes. At this stage in scientific history, perceptions of such multidimensional stimuli as conducted

1.3 Applications: Why Collect Sensory Data?

by the parallel processing of the human nervous system are difficult or impossible to pdict from instrumental measures. In many cases instruments lack the sensitivity of human sensory systems-smell is a good example. Instruments rarely mimic the mechanical manipulation of foods when tasted nor do they mimic the types of peri-receptor filtering that occur in biological fluids like saliva or mucus that can cause chemical partitioning of flavor materials. Most importantly, instrumental assessments give values that miss an important perceptual process: the interptation of sensory experience by the human brain prior to responding. The brain lies interposed between sensory input and the generation of responses that form our data. It is a massively parallel-distributed processor and computational engine, capable of rapid feats of pattern recognition. It comes to the sensory evaluation task complete with a personal history and experiential frame of reference. Sensory experience is interpted, given meaning within the frame of reference, evaluated relative to expectations and can involve integration of multiple simultaneous or sequential inputs. Finally judgments are rendered as our data. Thus there is a “chain of perception” rather than simply stimulus and response (Meilgaard et al., 2006). Only human sensory data provide the best models for how consumers are likely to perceive and react to food products in real life. We collect, analyze, and interpt sensory data to form pdictions about how products have changed during a product development program. In the food and consumer products industries, these changes arise from three important factors: ingredients, processes, and packaging. A fourth consideration is often the way a product ages, in other words its shelf life, but we may consider shelf stability to be one special case of processing, albeit usually a very passive one (but also consider products exposed to temperature fluctuation, light-catalyzed oxidation, microbial contamination, and other “abuses”). Ingredient changes arise for a number of reasons. They may be introduced to improve product quality, to reduce costs of production, or simply because a certain supply of raw materials has become unavailable. Processing changes likewise arise from the attempt to improve quality in terms of sensory, nutritional, microbiological stability factors, to reduce costs or to improve manufacturing productivity. Packaging changes arise from considerations of product stability or other quality factors, e.g., a certain

11

amount of oxygen permeability may insure that a fresh beef product remains red in color for improved visual appeal to consumers. Packages function as carriers of product information and brand image, so both sensory characteristics and expectations can change as a function of how this information can be carried and displayed by the packaging material and its print overlay. Packaging and print ink may cause changes in flavor or aroma due to flavor transfer out of the product and sometimes transfer of off-flavors into the product. The package also serves as an important barrier to oxidative changes, to the potentially deleterious effects of light-catalyzed reactions, and to microbial infestations and other nuisances. The sensory test is conducted to study how these product manipulations will create perceived changes to human observers. In this sense, sensory evaluation is in the best traditions of psychophysics, the oldest branch of scientific psychology, that attempts to specify the relationships between different energy levels impinging upon the sensory organs (the physical part of psychophysics) and the human response (the psychological part). Often, one cannot pdict exactly what the sensory change will be as a function of ingredients, processes, or packaging, or it is very difficult to do so since foods and consumer products are usually quite complex systems. Flavors and aromas depend upon complex mixtures of many volatile chemicals. Informal tasting in the lab may not bring a reliable or sufficient answer to sensory questions. The benchtop in the development laboratory is a poor place to judge potential sensory impact with distractions, competing odors, nonstandard lighting, and so on. Finally, the nose, eyes, and tongue of the product developer may not be repsentative of most other people who will buy the product. So there is some uncertainty about how consumers will view a product especially under more natural conditions. Uncertainty is the key here. If the outcome of a sensory test is perfectly known and pdictable, there is no need to conduct the formal evaluation. Unfortunately, useless tests are often requested of a sensory testing group in the industrial setting. The burden of useless routine tests arises from overly entrenched product development sequences, corporate traditions, or merely the desire to protect oneself from blame in the case of unexpected failures. However, the sensory test is only as useful as the amount of reduction in uncertainty that occurs. If there is no uncertainty, there

12

is no need for the sensory test. For example, doing a sensory test to see if there is a perceptible color difference between a commercial red wine and a commercial white wine is a waste of resources, since there is no uncertainty! In the industrial setting, sensory evaluation provides a conduit for information that is useful in management business decisions about directions for product development and product changes. These decisions are based on lower uncertainty and lower risk once the sensory information is provided. Sensory evaluation also functions for other purposes. It may be quite useful or even necessary to include sensory analyses in quality control (QC) or quality assurance. Modification of traditional sensory practices may be required to accommodate the small panels and rapid assessments often required in online QC in the manufacturing environment. Due to the time needed to assemble a panel, ppare samples for testing, analyze and report sensory data, it can be quite challenging to apply sensory techniques to quality control as an on-line assessment. Quality assurance involving sensory assessments of finished products are more readily amenable to sensory testing and may be integrated with routine programs for shelf life assessment or quality monitoring. Often it is desirable to establish correlations between sensory response and instrumental measures. If this is done well, the instrumental measure can sometimes be substituted for the sensory test. This is especially applicable under conditions in which rapid turnaround is needed. Substitution of instrumental measurements for sensory data may also be useful if the evaluations are likely to be fatiguing to the senses, repetitive, involve risk in repeated evaluations (e.g., insecticide fragrances), and are not high in business risk if unexpected sensory problems arise that were missed. In addition to these product-focused areas of testing, sensory research is valuable in a broader context. A sensory test may help to understand the attributes of a product that consumers view as critical to product acceptance and thus success. While we keep a wary eye on the fuzzy way that consumers use language, consumer sensory tests can provide diagnostic information about a product’s points of superiority or shortcomings. Consumer sensory evaluations may suggest hypotheses for further inquiry such as exploration of new product opportunities. There are recurrent themes and enduring problems in sensory science. In 1989, the ASTM Committee

1 Introduction

1.3 Applications: Why Collect Sensory Data?

(4)

(5)

(6)

(7)

control in multi-plant manufacturing situations, as well as international product development and the problem of multiple sensory testing sites and panels. Environmental and biochemical factors. Kamen recognized that pferences may change as a function of the situation (food often tastes better outdoors and when you are hungry). Meiselman (1993) questioned whether sufficient sensory research is being performed in realistic eating situations that may be more pdictive of consumer reactions, and recently sensory scientists have started to explore this area of research (for example, Giboreau and Fleury, 2009; Hein et al., 2009; Mielby and Frøst, 2009). Resolving discrepancies between laboratory and field studies. In the search for reliable, detailed, and pcise analytical methods in the sensory laboratory, some accuracy in pdicting field test results may be lost. Management must be aware of the potential of false positive or negative results if a full testing sequence is not carried out, i.e., if shortcuts are made in the testing sequence prior to marketing a new product. Sensory evaluation specialists in industry do not always have time to study the level of correlation between laboratory and field tests, but a prudent sensory program would include periodic checks on this issue. Inpidual differences. Since Kamen’s era, a growing literature has illuminated the fact that human panelists are not identical, interchangeable measuring instruments. Each comes with different physiological equipment, different frames of reference, different abilities to focus and maintain attention, and different motivational resources. As an example of differences in physiology, we have the growing literature on specific anosmias-smell “blindnesses” to specific chemical compounds among persons with otherwise normal senses of smell (Boyle et al., 2006; Plotto et al., 2006; Wysocki and Labows, 1984). It should not be surprising that some olfactory characteristics are difficult for even trained panelists to evaluate and to come to agreement (Bett and Johnson, 1996). Relating sensory differences to product variables. This is certainly the meat of sensory science in industrial practice. However, many product developers do not sufficiently involve their sensory specialists in the underlying research questions.

13

They also may fall into the trap of never ending sequences of paired tests, with little or no planned designs and no modeling of how underlying physical variables (ingredients, processes) create a dynamic range of sensory changes. The relation of graded physical changes to sensory response is the essence of psychophysical thinking. (8) Sensory interactions. Foods and consumer products are multidimensional. The more sensory scientists understand interactions among characteristics such as enhancement and masking effects, the better they can interpt the results of sensory tests and provide informed judgments and reasoned conclusions in addition to reporting just numbers and statistical significance. (9) Sensory education. End-users of sensory data and people who request sensory tests often expect one tool to answer all questions. Kamen cited the simple dichotomy between analytical and hedonic testing (e.g., discrimination versus pference) and how explaining this difference was a constant task. Due to the lack of widespad training in sensory science, the task of sensory education is still with us today, and a sensory professional must be able to explain the rationale behind test methods and communicate the importance and logic of sensory technology to non-sensory scientists and managers.

14 Table 1.3 Contrast of sensory evaluation consumer tests with market research tests

1 Introduction Sensory testing with consumers Participants screened to be users of the product category Blind-labeled samples-random codes with minimal conceptual information Determines if sensory properties and overall appeal met targets Expectations based on similar products used in the category Not intended to assess response/appeal of product concept Market research testing (concept and/or product test) Participants in product-testing phase selected for positive response to concept Conceptual claims, information, and frame of reference are explicit Expectations derived from concept/claims and similar product usage Unable to measure sensory appeal in isolation from concept and expectations

designed to make the product conceptually appealing (e.g., bringing attention to convenience factors in pparation). In a sensory test all these potentially biasing factors are stripped away in order to isolate the opinion based on sensory properties only. In the tradition of scientific inquiry, we need to isolate the variables of interest (ingredients, processing, packaging changes) and assess sensory properties as a function of these variables, and not as a function of conceptual influences. This is done to minimize the influence of a larger cognitive load of expectations generated from complex conceptual information. There are many potential response biases and task demands that are entailed in “selling” an idea as well as in selling a product. Participants often like to please the experimenter and give results consistent with what they think the person wants. There is a large literature on the effect of factors such as brand label on consumer response. Product information interacts in complex ways with consumer attitudes and expectancies (Aaron et al., 1994; Barrios and Costell, 2004; Cardello and Sawyer, 1992; Costell et al., 2009; Deliza and MacFie, 1996; Giménez et al., 2008; Kimura et al., 2008; Mielby and Frøst, 2009; Park and Lee, 2003; Shepherd et al., 1991/1992). Expectations can cause assimilation of sensory reactions toward what is expected under some conditions and under other conditions will show contrast effects, enhancing differences when expectations are not met (Siegrist and Cousin, 2009; Lee et al., 2006; Yeomans et al., 2008; Zellner et al., 2004). Packaging and brand information will also affect sensory judgments (Dantas et al., 2004; Deliza et al., 1999; Enneking et al., 2007). So the apparent resemblance of a blind sensory test and a fully concept-loaded market research test are quite illusory. Corporate management needs to be reminded of this important distinction. There continues to be

1.3 Applications: Why Collect Sensory Data?

1.3.2 Differences from Traditional Product Grading Systems A second arena of apparent similarity to sensory evaluation is with the traditional product quality grading systems that use sensory criteria. The grading of agricultural commodities is a historically important influence on the movement to assure consumers of quality standards in the foods they purchase. Such techniques were widely applicable to simple products such as fluid milk and butter (Bodyfelt et al., 1988, 2008), where an ideal product could be largely agreed upon and the defects that could arise in poor handling and processing gave rise to well-known sensory effects. Further impetus came from the fact that competitions could be held to examine whether novice judges-in-training could match the opinions of experts. This is much in the tradition of livestock grading-a young person could judge a cow and receive awards at a state fair for learning to use the same criteria and critical eye as the expert judges. There are noteworthy differences in the ways in which sensory testing and quality judging are performed. Some of these are outlined in Table 1.4. The commodity grading and the inspection tradition have severe limitations in the current era of highly processed foods and market segmentation. There are fewer and fewer “standard products” relative to the wide variation in flavors, nutrient levels (e.g., low fat), convenience pparations, and other choices that

15

line the supermarket shelves. Also, one person’s product defect may be another’s marketing bonanza, as in the glue that did not work so well that gave us the ubiquitous post-it notes. Quality judging methods are poorly suited to research support programs. The techniques have been widely criticized on a number of scientific grounds (Claassen and Lawless, 1992; Drake, 2007; O’Mahony, 1979; Pangborn and Dunkley, 1964; Sidel et al., 1981), although they still have their proponents in industry and agriculture (Bodyfelt et al., 1988, 2008). The defect identification in quality grading emphasizes root causes (e.g., oxidized flavor) whereas the descriptive approach uses more elemental singular terms to describe perceptions rather than to infer causes. In the case of oxidized flavors, the descriptive analysis panel might use a number of terms (oily, painty, and fishy) since oxidation causes a number of qualitatively different sensory effects. Another notable difference from mainstream sensory evaluation is that the quality judgments combine an overall quality scale (psumably reflecting consumer dislikes) with diagnostic information about defects, a kind of descriptive analysis looking only at the negative aspects of products. In mainstream sensory evaluation, the descriptive function and the consumer evaluation would be clearly separate in two distinct tests with different respondents. Whether the opinion of a single expert can effectively repsent consumer opinion is highly questionable at this time in history.

Table 1.4 Contrast of sensory evaluation tests with quality inspection Sensory testing Separates hedonic (like-dislike) and descriptive information into separate tests Uses repsentative consumers for assessment of product appeal (liking/disliking) Uses trained panelists to specify attributes, but not liking/disliking Oriented to research support Flexible for new, engineered, and innovative products Emphasizes statistical inference for decision making, suitable experimental designs, and sample sizes Quality inspection Used for pass-fail online decisions in manufacturing Provides quality score and diagnostic information concerning defects in one test Uses sensory expertise of highly trained inpiduals May use only one or very few trained experts Product knowledge, potential problems, and causes are stressed Traditional scales are multi-dimensional and poorly suited to statistical analyses Decision-making basis may be qualitative Oriented to standard commodities

16

1 Introduction

1.4 Summary and Conclusions Sensory evaluation comprises a set of test methods with guidelines and established techniques for product psentation, well-defined response tasks, statistical methods, and guidelines for interptation of results. Three primary kinds of sensory tests focus on the

existence of overall differences among products (discrimination tests), specification of attributes (descriptive analysis), and measuring consumer likes and dislikes (affective or hedonic testing). Correct application of sensory technique involves correct matching of method to the objective of the tests, and this requires good communication between sensory specialists and

Methods Selection

yes Consumer Acceptability Question? no Choose from: Preference/choice Ranking Rated Acceptability

no go to Panel Setup

Sensory Analytical Question? yes

yes

Simple Same/different Question?

re-open discussion of objectives

no

no

Choose from: Overall difference tests n-alternative forced choice Rated difference from control

go to Panel Setup Nature of Difference Question? yes

Choose from: descriptive analysis techniques or modifications

go to Panel Setup

Fig. 1.4 A sensory evaluation department may interact with many other departments in a food or consumer products company. Their primary interaction is in support of product research and development, much as marketing research supports the

References

end-users of the test results. Logical choices of test participants and appropriate statistical analyses form part of the methodological mix. Analytic tests such as the discrimination and descriptive procedures require good experimental control and maximization of test pcision. Affective tests on the other hand require use of repsentative consumers of the products and test conditions that enable generalization to how products are experienced by consumers in the real world. Sensory tests provide useful information about the human perception of product changes due to ingredients, processing, packaging, or shelf life. Sensory evaluation departments not only interact most heavily with new product development groups but may also provide information to quality control, marketing research, packaging, and, indirectly, to other groups throughout a company (Fig. 1.4). Sensory information reduces risk in decisions about product development and strategies for meeting consumer needs. A wellfunctioning sensory program will be useful to a company in meeting consumer expectations and insuring a greater chance of marketplace success. The utility of the information provided is directly related to the quality of the sensory measurement. . . . , sensory food science stands at the intersection of many disciplines and research traditions, and the stakeholders are many (Tuorila and Monteleone, 2009). Quantities derive from measurement, ps from quantities, comparisons from ps, and victory from comparisons (Sun Tzu – The Art of War (Ch. 4, v.18)).

References Aaron, J. I., Mela, D. J. and Evans, R. E. 1994. The influence of attitudes, beliefs and label information on perceptions of reduced-fat spad. Appetite, 22(1), 25-38. Adams, H. 1918. The Education of Henry Adams. The Modern Library, New York. Amerine, M. A., Pangborn, R. M. and Roessler, E. B. 1965. Principles of Sensory Evaluation of Food. Academic, New York. ASTM E1958. 2008. Standard guide for sensory claim substantiation. ASTM International, West Conshohocken, PA. ASTM. 1989. Sensory evaluation. In celebration of our beginnings. Committee E-18 on Sensory Evaluation of Materials and Products. ASTM, Philadelphia. Barrios, E. X. and Costell, E. 2004. Review: use of methods of research into consumers’ opinions and attitudes in food research. Food Science and Technology International, 10, 359-371.

18 Giboreau, A. and Fleury, H. 2009. A new research platform to contribute to the pleasure of eating and healthy food behaviors through academic and applied food and hospitality research. Food Quality and Preference, 20, 533-536 Giménez, A., Ares, G. and Gámbaro, A. 2008. Consumer attitude toward shelf-life labeling: does it influence acceptance? Journal of Sensory Studies, 23, 871-883. Hein, K. A., Hamid, N., Jaeger, S. R. and Delahunty, C. M. 2009. Application of a written scenario to evoke a consumption context in a laboratory setting: effects on hedonic ratings. Food Quality and Preference. doi:10.1016/j.foodqual.2009.10.003 Helm, E. and Trolle, B. 1946. Selection of a taste panel. Wallerstein Laboratory Communications, 9, 181-194. Jones, L. V., Peryam, D. R. and Thurstone, L. L. 1955. Development of a scale for measuring soldier’s food pferences. Food Research, 20, 512-520. Kamen, J. 1989. Observations, reminiscences and chatter. In: Sensory Evaluation. In celebration of our Beginnings. Committee E-18 on Sensory Evaluation of Materials and Products. ASTM, Philadelphia, pp. 118-122. Kimura, A., Wada, Y., Tsuzuki, D., Goto, S., Cai, D. and Dan, I. 2008. Consumer valuation of packaged foods. Interactive effects of amount and accessibility of information. Appetite, 51, 628-634. Lawless, H. T. 1993. The education and training of sensory scientists. Food Quality and Preference, 4, 51-63. Lawless, H. T. and Claassen, M. R. 1993. The central dogma in sensory evaluation. Food Technology, 47(6), 139-146. Lawless, H. T. and Klein, B. P. 1989. Academic vs. industrial perspectives on sensory evaluation. Journal of Sensory Studies, 3, 205-216. Lee, L., Frederick, S. and Ariely, D. 2006. Try it, you’ll like it. Psychological Science, 17, 1054-1058. MacRae, R. W. and Geelhoed, E. N. 1992. Preference can be more powerful than detection of oddity as a test of discriminability. Perception and Psychophysics, 51, 179-181. Meilgaard, M., Civille, G. V. and Carr, B. T. 2006. Sensory Evaluation Techniques. Fourth Second edition. CRC, Boca Raton. Meiselman, H. L. 1993. Critical evaluation of sensory techniques. Food Quality and Preference, 4, 33-40. Mielby, L. H. and Frøst, M. B. 2009. Expectations and surprise in a molecular gastronomic meal. Food Quality and Preference. doi:10.1016/j.foodqual.2009.09.005 Moskowitz, H. R., Beckley, J. H. and Resurreccion, A. V. A. 2006. Sensory and Consumer Research in Food Product Design and Development. Wiley-Blackwell, New York. Moskowitz, H. R. 1983. Product Testing and Sensory Evaluation of Foods. Food and Nutrition, Westport, CT. Oliver, T. 1986. The Real Coke, The Real Story. Random House, New York. O’Mahony, M. 1988. Sensory difference and pference testing: The use of signal detection measures. Chpater 8 In: H. R. Moskowitz (ed.), Applied Sensory Analysis of Foods. CRC, Boca Raton, FL, pp. 145-175.

1 Introduction O’Mahony, M. 1979. Psychophysical aspects of sensory analysis of dairy products: a critique. Journal of Dairy Science, 62, 1954-1962. Pangborn, R. M. and Dunkley, W. L. 1964. Laboratory procedures for evaluating the sensory properties of milk. Dairy Science Abstracts, 26, 55-121. Park, H. S. and Lee, S. Y. 2003. Genetically engineered food labels, information or warning to consumers? Journal of Food Products Marketing, 9, 49-61. Peryam, D. R. and Swartz, V. W. 1950. Measurement of sensory differences. Food Technology, 4, 390-395. Plotto, A., Barnes, K. W. and Goodner, K. L. 2006. Specific anosmia observed for β-ionone, but not for α-ionone: Significance for flavor research. Journal of Food Science, 71, S401-S406. Shepherd, R., Sparks, P., Belleir, S. and Raats, M. M. 1991/1992. The effects of information on sensory ratings and pferences: The importance of attitudes. Food Quality and Preference, 3, 1-9. Sidel, J. L., Stone, H. and Bloomquist, J. 1981. Use and misuse of sensory evaluation in research and quality control. Journal of Dairy Science, 61, 2296-2302. Siegrist, M. and Cousin, M-E. 2009. Expectations influence sensory experience in a wine tasting. Appetite, 52, 762-765. Skinner, E. Z. 1989. (Commentary). Sensory evaluation. In celebration of our beginnings. Committee E-18 on Sensory Evaluation of Materials and Products. ASTM, Philadelphia, pp. 58-65. Stone, H. and Sidel, J. L. 2004. Sensory Evaluation Practices, Third Edition. Academic, San Deigo. Stone, H., Sidel, J., Oliver, S., Woolsey, A. and Singleton, R. C. 1974. Sensory evaluation by quantitative descriptive analysis. Food Technology 28(1), 24, 26, 28, 29, 32, 34. Sun Tzu (Sun Wu) 1963 (trans.), orig. circa 350 B.C.E. The Art of War. S.B. Griffith, trans. Oxford University. Szczesniak, A. S., Loew, B. J. and Skinner, E. Z. 1975. Consumer texture profile technique. Journal of Food Science, 40, 1253-1257. Tuorila, H. and Monteleone, E. 2009. Sensory food science in the changing society: opportunities, needs and challenges. Trends in Food Science and Technology, 20, 54-62. Wysocki, C. J. and Labows, J. 1984. Inpidual differences in odor perception. Perfumer and Flavorist, 9, 21-24. Yeomans, M. R., Chambers, L., Blumenthal, H. and Blake, A. 2008. The role of expectation in sensory and hedonic evaluation: The case of salmon smoked ice-cream. Food Quality and Preference, 19, 565-573. York, R. K. 1995. Quality assessment in a regulatory environment. Food Quality and Preference, 6, 137-141. Zellner, D. A., Strickhouser, D. and Tornow, C. E. 2004. Disconfirmed hedonic expectations produce perceptual contrast, not assimilation. The American Journal of Psychology, 117, 363-387.

Chapter 2

Physiological and Psychological Foundations of Sensory Function

Abstract This chapter reviews background material underpinning sensory science and sensory evaluation methodologies. Basic and historical psychophysical methods are reviewed as well as the anatomy, physiology, and function of the chemical senses. The chapter concludes with a discussion of multi-modal sensory interactions. There is no conception in man’s mind which hath not at first, totally or in parts, been begotten upon by the organs of sense. -Thomas Hobbes, Leviathan (1651)

Contents 2.1 2.2

2.3

2.4

2.5

Introduction . . . . . . . . . . . . . . . . . Classical Sensory Testing and Psychophysical Methods . . . . . . . . . . . . . . . . . . . 2.2.1 Early Psychophysics . . . . . . . . . . 2.2.2 The Classical Psychophysical Methods . . . . . . . . . . . . . . . 2.2.3 Scaling and Magnitude Estimation . . . 2.2.4 Critiques of Stevens . . . . . . . . . . 2.2.5 Empirical Versus Theory-Driven Functions . . . . . . . . . . . . . . . 2.2.6 Parallels of Psychophysics and Sensory Evaluation . . . . . . . . . . . . . . Anatomy and Physiology and Functions of Taste . . . . . . . . . . . . . . . . . . . 2.3.1 Anatomy and Physiology . . . . . . . . 2.3.2 Taste Perception: Qualities . . . . . . . 2.3.3 Taste Perception: Adaptation and Mixture Interactions . . . . . . . . . . . . . . 2.3.4 Inpidual Differences and Taste Genetics . . . . . . . . . . . . . . . Anatomy and Physiology and Functions of Smell . . . . . . . . . . . . . . . . . . . 2.4.1 Anatomy and Cellular Function . . . . . 2.4.2 Retronasal Smell . . . . . . . . . . . 2.4.3 Olfactory Sensitivity and Specific Anosmia . . . . . . . . . . . . . . . 2.4.4 Odor Qualities: Practical Systems . . . . 2.4.5 Functional Properties: Adaptation, Mixture Suppssion, and Release . . . . . . . . Chemesthesis . . . . . . . . . . . . . . . . 2.5.1 Qualities of Chemesthetic Experience . . . . . . . . . . . . . .

2.5.2

19 20 20 21 23 25 25

Physiological Mechanisms of Chemesthesis . . . . . . . . . . . 2.5.3 Chemical “Heat” . . . . . . . . . . 2.5.4 Other Irritative Sensations and Chemical Cooling . . . . . . . . . . . . . . 2.5.5 Astringency . . . . . . . . . . . . . 2.5.6 Metallic Taste . . . . . . . . . . . . 2.6 Multi-modal Sensory Interactions . . . . . 2.6.1 Taste and Odor Interactions . . . . . . 2.6.2 Irritation and Flavor . . . . . . . . . 2.6.3 Color-Flavor Interactions . . . . . . 2.7 Conclusions . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . .

. .

42 43

. . . . . . . . .

44 45 46 47 47 49 49 50 50

26

2.1 Introduction 27 27 30 30 33 34 34 36 37 38 39 41 41

In order to design effective sensory tests and provide insightful interptation of the results, a sensory professional must understand the functional properties of the sensory systems that are responsible for the data. By a functional property, we mean a phenomenon like mixture interactions such as masking or suppssion. Another example is sensory adaptation, a commonly observed decrease in responsiveness to conditions of more or less constant stimulation. In addition, it is useful to understand the anatomy and physiology of the senses involved as well as their functional limitations. A good example of a functional limitation is the threshold or minimal amount of a stimulus needed for perception. Knowing about the anatomy of the senses

H.T. Lawless, H. Heymann, Sensory Evaluation of Food, Food Science Text Series, DOI 10.1007/978-1-4419-6488-5_2, © Springer Science+Business Media, LLC 2010

19

20

2

Physiological and Psychological Foundations of Sensory Function

can help us understand how consumers and panelists interact with the products to stimulate their senses and by what routes. Different routes of smelling, for example, are the orthonasal or sniffing route, when odor molecules enter the nose from the front (nostrils), versus retronasal smell, when odor molecules pass into the nose from the mouth or from breathing out, and thus have a reversed airflow pathway from that of external sniffing. Another basic area that the sensory professional should have as background knowledge involves the sensory testing methods and human measurement procedures that are the historical antecedents to the tests we do today. This is part of the science of psychophysics, the quantification and measurement of sensory experiences. Psychophysics is a very old discipline that formed the basis for the early studies in experimental psychology. Parallels exist between psychophysics and sensory evaluation. For example, the difference test using paired comparisons is a version of the method used for measuring difference thresholds called the method of constant stimuli. In descriptive analysis with trained panels, we work very hard to insure that panelists use singular uni-dimensional scales. These numerical systems usually refer to a single sensory continuum like sweetness or odor strength and are thus based on changes in perceived intensity. They do not consider multiple attributes and fold them into a single score like the old-quality grading methods. Thus there is a clear psychophysical basis for the attribute scales used in descriptive analysis. This chapter is designed to provide the reader some background in the sensory methods of psychophysics. A second objective is to give an overview of the structure and function of the chemical senses of taste, smell, and the chemesthetic sense. Chemesthesis refers to chemically induced sensations that seem to be at least partly tactile in nature, such as pepper heat, astringency, and chemical cooling. These three senses together comprise what we loosely call flavor and are the critical senses for appciating foods, along with the tactile, force, and motion-related experiences that are part of food texture and mouthfeel. Texture is dealt with in Chapter 11 and color and appearance evaluations in Chapter 12. The auditory sense is not a large part of food perception, although many sounds can be perceived when we eat or manipulate foods. These provide another sense modality to accompany and reinforce our texture perceptions, as in the case of crisp

or crunchy foods, or the audible hissing sound we get from carbonated beverages (Vickers, 1991). One growing area of interest in the senses concerns our human biopersity, differences among people in sensory function. These differences can be due to genetic, dietary/nutritional, physiological (e.g., aging), or environmental factors. The research into the genetics of the chemical senses, for example, has experienced a period of enormous expansion since the first edition of this book. The topic is too large and too rapidly changing to receive a comphensive treatment here. We will limit our discussion of inpidual differences and genetic factors to those areas that are well understood, such as bitter sensitivity, smell blindness, and color vision anomalies. The sensory practitioner should be mindful that people exist in somewhat different sensory worlds. These differences contribute to the persity of consumer pferences. They also limit the degree to which a trained panel can be “calibrated” into uniform ways of responding. Inpidual differences can impact sensory evaluations in many ways.

2.2 Classical Sensory Testing and Psychophysical Methods 2.2.1 Early Psychophysics The oldest branch of experimental psychology is that of psychophysics, the study of relationships between physical stimuli and sensory experience. The first true psychophysical theorist was the nineteenth century German physiologist, E. H. Weber. Building on earlier observations by Bernoulli and others, Weber noted that the amount that a physical stimulus needed to be increased to be just perceivably different was a constant ratio. Thus 14.5 and 15 ounces could be told apart, but with great difficulty, and the same could be said of 29 and 30 ounces or 14.5 and 15 drams (Boring, 1942). This led to the formulation of Weber’s law, generally written nowadays as I/I = k

(2.1)

where I is the increase in the physical stimulus that was required to be just discriminably different from some starting level, I. The fraction, I/I, is sometimes called the “Weber fraction” and is an index of

2.2 Classical Sensory Testing and Psychophysical Methods

how well the sensory system detects changes. This relationship proved generally useful and provided the first quantitative operating characteristic of a sensory system. Methods for determining the difference threshold or just-noticeable-difference (j.n.d.) values became the stock in trade of early psychological researchers. These methods were codified by G. T. Fechner in a book called Elemente der Psychophysik (Elements of Psychophysics) in 1860. Fechner was a philosopher as well as a scientist and developed an interest in Eastern religions, in the nature of the soul, and in the Cartesian mind-body dichotomy. Fechner’s broader philosophical interests have been largely overlooked, but his little book on sensory methods was to become a classic text for the psychology laboratory. Fechner also had a valuable insight. He realized that the j.n.d. might be used as a unit of measurement and that by adding up j.n.d.s one could construct a psychophysical relationship between physical stimulus intensity and sensory intensity. This relationship approximated a log function, since the integral of 1/x dx is proportional to the natural log of x. So a logarithmic relationship appeared useful as a general psychophysical “law:” S = k log I

(2.2)

where S is sensation intensity and I is once again the physical stimulus intensity. This relationship known as Fechner’s law was to prove a useful rule of thumb for nearly 75 years, until it was questioned by acoustical researchers who supplanted it with a power law (see Section 2.2.3).

21

with a particular type of measured response of sensory systems. The method of limits was well suited to determine absolute or detection thresholds. The method of constant stimuli could be used to determine difference thresholds and the method of adjustment to establish sensory equivalence. In the method of limits the physical stimulus is changed by successive discrete steps until a change in response is noted. For example, when the stimulus is increasing in intensity, the response will change from “no sensation” to “I detect something.” When the stimulus is decreasing in intensity, at some step the response will change back to “no sensation.” Over many trials, the average point of change can be taken as the person’s absolute threshold (see Fig. 2.1). This is the minimum intensity required for detection of the stimulus. Modern variations on this method often use only an ascending series and force the participants to choose a target sample among alternative “blank” samples at each step. Each concentration must be discriminated from a background level such as plain water in the case of taste thresholds. Forced-choice methods for determining thresholds are discussed in detail in Chapter 6. In the method of constant stimuli, the test stimulus is always compared against a constant reference level (a standard), usually the middle point on a series of physical intensity levels. The subject’s job is to respond to each test item as “greater than” or “less

2.2.2 The Classical Psychophysical Methods Fechner’s enduring contribution was to assemble and publish the details of sensory test methods and how several important operating characteristics of sensory systems could be measured. Three important methods were the method of limits, the method of constant stimuli (called the method of right and wrong cases in those days), and the method of adjustment or average error (Boring, 1942). The methods are still used today in some research situations and variations on these methods form part of the toolbox of applied sensory evaluation. Each of the three methods was associated

CONCENTRATION

16.0 8.0 4.0 2.0 1.0 0.5 .25 A1

D1

A2

D2

A3

D3

TRIAL

Fig. 2.1 An example of the method of limits. The circled reversal points would be averaged to obtain the person’s threshold. A: ascending series. D: descending series. In taste and smell, only ascending series are commonly used to pvent fatigue, adaptation or carry-over of persistent sensations.

22

2

Physiological and Psychological Foundations of Sensory Function

Frequency (percent) judged “sweeter” than the standard

80

60

40 UDL

20 standard

2

4

6

8

10

12

14

16

18

SUCROSE CONCENTRATION (%)

Fig. 2.2 A psychometric function derived from the Method of Constant Stimuli, a repeated series of paired comparisons against a constant (standard) stimulus, in this case 10% sucrose. Frequency of judgments in which the comparison stimulus is

judged sweeter than the standard are plotted against concentration. The difference threshold is determined by the concentration difference between the standard and the interpolated 75% (or 25%) point. UDL: Upper difference limen (threshold).

than” the standard. Many replications of each intensity level are psented. The percentage of times the response is “greater than” can be plotted as in Fig. 2.2. This S-shaped curve is called a psychometric function (Boring, 1942). The difference threshold was taken as the difference between the 50 and 75% points interpolated on the function. The method of constant stimuli bears a strong resemblance to current techniques of paired comparison, with two exceptions. One point of difference is that the method was geared toward interval estimation, rather than testing for statistically significant differences. That is, the technique estimated points on the psychometric function (25, 50, and 75%) and researchers were not concerned with statistical significance of difference tests. Also, a range of comparison stimuli were tested against the standard and not just a single paired comparison of products. The third major method in classical psychophysics was the method of adjustment or average error. The subject was given control over a variable stimulus like a light or a tone and asked to match a standard in brightness or loudness. The method could be used to determine difference thresholds based on the variability of the subject over many attempts at matching, for example, using the standard deviation as a measure

of difference threshold. A modern application is in measuring sensory tradeoff relationships. In this type of experiment the duration of a very brief tone could be balanced against a varying sound pssure level to yield a constant perception of loudness. Similarly, the duration of a flash of light could be traded off against its photometric intensity to create a constant perceived brightness. For very brief tones or brief flashes, there is summation of the intensity over time in the nervous system, so that increasing duration can be balanced against decreasing physical intensity to create a constant perception. These methods have proven useful in understanding the physiological response of different senses to the temporal properties of stimuli, for example, how the auditory and visual systems integrate energy over time. Adjustment methods have not proven so useful for assessing sensory equivalence in applied food testing, although adjustment is one way of trying to optimize an ingredient level (Hernandez and Lawless, 1999; Mattes and Lawless, 1985). Pangborn and co-workers employed an adjustment method to study inpidual pferences (Pangborn, 1988; Pangborn and Braddock, 1989). Adding flavors or ingredients “to taste” at the benchtop is a common way of initially formulating

2.2 Classical Sensory Testing and Psychophysical Methods

products. It is also fairly common to make formula changes to produce approximate sensory matches to some target, either a standard formula or perhaps some competitor’s successful product. However, the method as applied in the psychophysics laboratory is an unwieldy technique for the senses of taste and smell where elaborate equipment is needed to provide adjustable stimulus control. So methods of equivalency adjustment are somewhat rare with food testing.

23

they are a log scale of sound pssure relative to a reference (db = 20 log (P/P0 ) where P is the sound pssure and P0 is the reference sound pssure, usually a value for absolute threshold). However, discrepancies were observed between decibels and loudness proportions. Instead, Stevens found with the direct magnitude estimation procedure that loudness was a power function of stimulus intensity, with an exponent of about 0.6. Scaling of other sensory continua also gave power functions, each with its characteristic exponent (Stevens, 1957, 1962). Thus the following relationship held:

2.2.3 Scaling and Magnitude Estimation S = kI n or log S = n log I + log k A very useful technique for sensory measurement has been the direct application of rating scales to measure the intensity of sensations. Historically known as the “method of single stimuli,” the procedure is highly cost efficient since one stimulus psentation yields one data point. This is in contrast to a procedure like the method of constant stimuli, where the psentation of many pairs is necessary to give a frequency count of the number of times each level is judged stronger than a standard. Rating scales have many uses. One of the most common is to specify a psychophysical function, a quantitative relationship between the perceived intensity of a sensation and the physical intensity of the stimulus. This is another way of describing a dose-response curve or in other words, capturing the input-output function of a sensory system over its dynamic range. The technique of magnitude estimation grew out of earlier procedures in which subjects would be asked to fractionate an adjustable stimulus. For example, a subject would be asked to adjust a light or tone until it seemed half as bright as a comparison stimulus. The technique was modified so that the experimenter controlled the stimulus and the subject responded using (unrestricted) numbers to indicate the proportions or ratios of the perceived intensities. Thus if the test stimulus was twice as bright as the standard, it would be assigned a number twice as large as the rating for the standard and if one-third as bright, a number one-third as large. An important observation in S. S. Stevens’ laboratory at Harvard was that the loudness of sounds was not exactly proportional to the decibel scale. If Fechner’s log relationship was correct, rated loudness should grow in a linear fashion with decibels, since

(2.3)

where n was the characteristic exponent and k was a proportionality constant determined by the units of measurement. In other words, the function formed a straight line in a log-log plot with the exponent equal to the slope of the linear function. This was in contrast to the Fechnerian log function which was a straight line in a semilog plot (response versus log physical intensity). One of the more important characteristics of a power function is that it can accommodate relationships that are expanding or positively accelerated while the log function does not. The power function with an exponent less than one fits a law of diminishing returns, i.e., larger and larger physical increases are required to maintain a constant proportional increase in the sensation level. Other continua such as response to electric shocks and some tastes were found to have a power function exponent greater than one (Meiselman, 1971; Moskowitz, 1971; Stevens, 1957). A comparison of power functions with different exponents is shown in Fig. 2.3. Many sensory systems show an exponent less that one. This shows a compssive energy relationship that may have adaptive value for an organism responding to a wide range of energy in the environment. The range from the loudest sound one can tolerate to the faintest audible tone is over 100 dB. This repsents over 10 log units of sound energy, a ratio of 10 billion to one. The dynamic range for the visual response of the eyes to different levels of light energy is equally broad. Thus exponents less than one have ecological significance for sensory systems that are tuned to a broad range of physical energy levels.

24

2

Physiological and Psychological Foundations of Sensory Function

Psychological Magnitude, S

Electric Shock

Apparent Length n=1

Brightness n1 n=1 n

1 2

(4.3)

The verbal form of the alternative hypothesis is that the population would be correct (saying that AB and BA pairs are different and that AA and BB pairs are the same) more than half the time. The results of the paired difference test will only indicate whether the panelists could significantly discriminate between the samples. Unlike the paired directional test, no specification or direction of difference is indicated. In other words, the sensory scientist will only know that the samples are perceptibly different but not in which attribute(s) the samples differed. An alternative analysis is psented in the Appendix to this chapter, where each panelist sees an identical pair (AA or BB) and one test pair (AB or BA) in randomized sequence.

4.2.2 Triangle Tests In the triangle test, three samples are psented simultaneously to the panelists, two samples are from the same formulation and one is from the different formulation. Each panelist has to indicate either which sample is the odd sample or which two samples are most similar. The usual form of the score sheet asks the panelist to indicate the odd sample. However, some sensory specialists will ask the panelist to indicate the pair of similar samples. It probably does not matter which question is asked. However, the sensory specialist should not change the format when re-using panelists since they will get confused. See Fig. 4.3 for a sample score sheet. Similarly to the paired difference test the panelist must be trained to understand the task as described by the score sheet. The null hypothesis for the triangle test states that the long-run probability (Pt ) of making a correct selection when there is no perceptible difference between the samples is one in three (H0 :Pt = 1/3). The alternative hypothesis states that the probability that the underlying population will make the correct decision

Fig. 4.3 Example of a triangle score sheet.

84

4

1 3

(4.4)

This is a one-sided alternative hypothesis and the test is one tailed. In this case there are six possible serving orders (AAB, ABA, BAA, BBA, BAB, ABB) which should be counterbalanced across all panelists. As with the difference paired comparison, the triangle test allows the sensory specialist to determine if two samples are perceptibly different but the direction of the difference is not indicated by the triangle test. Again, the sensory scientist will only know that the samples are perceptibly different but not in which attribute(s) the samples differed.

Discrimination Testing

Again, the panelists should be trained to perform the task as described by the score sheet correctly. Duo-trio tests allow the sensory specialist to determine if two samples are perceptibly different but the direction of the difference is not indicated by the duo-trio test. In other words, the sensory scientist will only know that the samples are perceptibly different but not in which attribute(s) the samples differed. There are two formats to the duo-trio test, namely the constant reference duo-trio test and the balanced reference duo-trio test. From the point of view of the panelists the two formats of the duo-trio test are identical (see Figs. 4.4a and b), but to the sensory specialist the two formats differ in the sample(s) used as the reference. 4.2.3.1 Constant Reference Duo-Trio Test

1 2

(4.5)

In this case, all panelists receive the same sample formulation as the reference. The constant reference duo-trio test has two possible serving orders (RA BA, RA AB) which should be counterbalanced across all panelists. The constant reference duo-trio test seems to be more sensitive especially if the panelists have had prior experience with the product (Mitchell, 1956). For example, if product X is the current formulation (familiar to the panelists) and product Z is a new reformulation then a constant reference duo-trio test with product X as reference would be the method of choice.

Before starting please rinse your mouth with water and expectorate. There are three samples in each of the two duo−trio sets for you to evaluate. In each set, one of the coded pairs is the same as the reference. For each set taste the reference first. Then taste each of the coded samples in the sequence psented, from left to right. Take the entire sample in your mouth. NO RETASTING. Circle the number of the sample which is most similar to the reference. Do not swallow any of the sample or the water. Expectorate into the container provided. Rinse your mouth with water between sets 1 and 2.

Fig. 4.4a Example of a constant reference duo-trio score sheet.

Set 1

Reference

2

Reference

4.2 Types of Discrimination Tests

85

Fig. 4.4b Example of a balanced reference duo-trio score sheet.

Reference

2

Reference

reference and the other half of the panelists receive the other sample formulation as the reference. In this case, there are four possible serving orders (RA BA, RA AB, RB AB, RB BA) which should be counterbalanced across all panelists. This method is used when both products are prototypes (unfamiliar to the panelists) or when there is not a sufficient quantity of the more familiar product to perform a constant reference duo-trio test.

differ in the specified dimension and which sample is higher in perceived intensity of the specified attribute. The danger is that other sensory changes will occur in a food when one attribute is modified and these may obscure the attribute in question. Another version of the n-AFC asks panelists to pick out the weakest or strongest in overall intensity, rather than in a specific attribute. This is a very difficult task for panelists when they are confronted with a complex food system.

4.2.5 A-Not-A tests There are two types of A-not-A tests referenced in the literature. The first and the more commonly used version has a training phase with the two products followed by monadic evaluation phase (Bi and Ennis, 2001a, b), we will call this the standard A-not-A test. The second version is essentially a sequential paired difference test or simple difference test (Stone and Sidel, 2004), which we will call the alternate A-notA test. The alternate A-not-A test is not frequently used. In the next section we will discuss the alternate A-not-A test first since the statistical analysis for this version is similar to that of the paired comparison discrimination test. The statistical analyses for the various standard A-not-A tests are based on a different theory and somewhat more complex and will be discussed later.

4.2.5.1 Alternate A-Not-A test This is a sequential same/difference paired difference test where the panelist receives and evaluates the first

86

4

Fig. 4.5 Example of a three-alternative forced choice score sheet.

relevant to the objective of the study. However, the differences in color or shape or size have to be very subtle and only obvious when the samples are psented simultaneously. If the differences are not subtle the panelists are likely to remember these and they will make their decision based on these extraneous differences.

4.2.5.2 Standard A-Not-A Test Panelists inspect multiple examples of products that are labeled “A” and usually also products that are labeled “not-A.” Thus there is a learning period. Then once the training period has been completed the panelists receive samples one at a time and are asked whether each one is either A or not-A. As discussed by Bi and Ennis (2001a) the standard A-not-A test potentially has four different designs. For the monadic A-not-A test the panelist, after the training phase, is psented with a single sample (either A or not-A). In the paired A-not-A version the panelist, after completion of the training phase, is psented with a pair of samples, sequentially (one A and one not-A, counter balanced across panelists). In the replicated monadic A-not-A version the panelist, after completion of training, receives a series of samples of either A or not-A but not both. This version is rarely used in practice. Lastly, in the replicated mixed A-not-A version the panelist, after completion of training, receives a series of A and not-A samples. Each of these different formats requires different statistical models and using an inappropriate model could lead to a misleading conclusion. As described by Bi and Ennis (2001a) “The statistical models for the A-Not A method are different from that of other discrimination methods such as the m-AFC, the triangle, and the duo-trio methods.” “Pearson’s and McNemar’s chi-square statistics with one degree of freedom can be used for the

4.2 Types of Discrimination Tests

standard A-Not A method while binomial tests based on the proportion of correct responses can be used for the m-AFC, the triangle, and the duo-trio methods. The basic difference between the two types of difference tests is that the former involves a comparison of two proportions (i.e., the proportion of “A” responses for the A sample versus that for the Not A sample) or testing independence of two variables (sample and response) while the latter is a comparison of one proportion with a fixed value (i.e., the proportion of correct responses versus the guessing probability)”. Articles by Bi and Ennis (2001a, b) clearly describe data analysis methods for these tests.. Additionally, the article by Brockhoff and Christensen (2009) describes a R-package called SensR (http://www.cran.rproject.org/package=sensR/) that may be used for the data analyses of some Standard A-not-A tests. The data analyses associated with the standard A-not-A tests are beyond the scope of this textbook, but see the Appendix of this chapter which shows the application of the McNemar chi-square for a simple A-not-A test where each panelist received one standard product (a “true” example of A) and one test product. Each is psented separately and a judgment is collected for both products.

87

4.2.6.2 The Harris-Kalmus Test

4.2.6 Sorting Methods In sorting tests the panelists are given a series of samples and they are asked to sort them into two groups. The sorting tests can be extremely fatiguing and are not frequently used for taste and aroma sensory evaluation but they are used when sensory specialists want to determine if two samples are perceptibly different in tactile or visual dimensions. The sorting tests are statistically very efficient since the long-run probability of the null hypotheses of the sorting tests can be very small. For example, the null hypothesis of the two-out-of-five test is 1 in 10 (P2/5 = 0.1) and for the Harris-Kalmus test the null hypothesis is 1 in 70 (P4/8 = 0.0143). These tests are discussed below. 4.2.6.1 The Two-Out-of-Five Test The panelists receive five samples and are asked to sort the samples into two groups, one group should contain the two samples that are different from the other

88

4

Discrimination Testing

4.2.8 Dual-Standard Test The dual standard was first used by Peryam and Swartz (1950) with odor samples. It is essentially a duo-trio test with two reference standards-the control and the variant. The two standards allow the panelists to create a more stable criterion as to the potential difference between the samples. The potential serving orders for this test are R(A) R(B) , AB, R(A) R(B) BA, R(B) R(A) AB, R(B) R(A) BA. The probability of guessing the correct answer by chance is 0.5 and the data analyses for this test are identical to that of the duo-trio test. Peryam and Swartz felt quite strongly that the technique would work best with odor samples due to the relatively quick recovery and that the longer recovery associated with taste samples would pclude the use of the test. The test was used by Pangborn and Dunkley (1966) to detect additions of lactose, algin gum, milk salts, and proteins to milk. O’Mahony et al. (1986) working with lemonade found that the dual-standard test elicited superior performance over the duo-trio test. But O’Mahony (personal communication, 2009) feels that this result is in error, since the panelists were not instructed to evaluate the standards prior to each pair evaluation and therefore the panelists were probably reverting to a 2-AFC methodology. This would be in agreement with Huang and Lawless (1998) who studied sucrose additions to orange juice and they did not find superiority in performance between the dual standard and the duo-trio or the ABX tests.

4.3 Reputed Strengths and Weaknesses of Discrimination Tests If the batch-to-batch variation within a sample formulation is as large as the variation between formulations

4.4 Data Analyses

then the sensory specialist should not use triangle or duo-trio tests (Gacula and Singh, 1984). In this case the paired comparison difference test could be used but the first question that the sensory specialist should ask is whether the batch-to-batch variation should not be studied and improved prior to any study of new or different formulations. The major weakness of all discrimination tests is that they do not indicate the magnitude of the sensory difference(s) between the sample formulations. As the simple discrimination tests are aimed at a yes/no decision about the existence of a sensory difference, they are not designed to give information on the magnitude of a sensory difference, only whether one is likely to be perceived or not. The sensory specialist should not be tempted to conclude that a difference is large or small based on the significance level or the probability (pvalue) from the statistical analysis. The significance and p-value depend in part upon the number of panelists in the test as well as the inherent difficulty of the particular type of discrimination test method. So these are no acceptable indices of the size of the perceivable difference. However, it is sensible that a comparison in which 95% of the judges answered correctly has a larger sensory difference between control and test samples than a comparison in which performance was only at 50% correct. This kind of reasoning works only if a sufficient number of judges were tested, the methods were the same, and all test conditions were constant. Methods for interval level scaling of sensory differences based on proportions of correct discriminations in forced choice tests are discussed further in Chapter 5 as Thurstonian scaling methods. These methods are indirect measures of small differences. They are also methodologically and mathematically complex and require certain assumptions to be met in order to be used effectively. Therefore we feel that the sensory specialist is wiser to base conclusions about the degree of difference between samples on scaled (direct) comparisons, rather than indirect estimates from choice performance in discrimination tests. However, there are alternative opinions in the sensory community and we suggest that interested parties read Lee and O’Mahony (2007). With the exception of the 2-AFC and 3-AFC tests the other discrimination tests also do not indicate the nature of the sensory difference between the samples. The major strength of the discrimination tests is that the task that the panelists perform is quite simple and

89

intuitively grasped by the panelists. However, it is frequently the very simplicity of these tests that lead to the generation of garbage data. Sensory specialists must be very aware of the power, replication, and counterbalancing issues associated with discrimination tests. These issues are discussed later in this chapter.

4.4 Data Analyses The data from discrimination tests may be analyzed by any of the following statistical methods. The three data analyses are based on the binomial, chi-square, or normal distributions, respectively. All these analyses assume that the panelists were forced to make a choice. Thus they had to choose one sample or another and could not say that they did not know the answer. In other words, each panelist either made a correct or incorrect decision, but they all made a decision.

4.4.1 Binomial Distributions and Tables The binomial distribution allows the sensory specialist to determine whether the result of the study was due to chance alone or whether the panelists actually perceived a difference between the samples. The following formula allows the sensory scientists to calculate the probability of success (of making a correct decision; p) or the probability of failure (of making an incorrect decision; q) using the following formula. P(y) =

n! py pn−y y!(n − y)!

(4.6)

where n = total number of judgments y = total number of correct judgments p = probability of making the correct judgment by chance In this formula, n! describes the mathematical factorial function which is calculated as n×(n-1)× (n-2). . .×2×1. Before the widespad availability of calculators and computers, calculation of the binomial formula was quite complicated, and even now

90

4

it remains somewhat tedious. Roessler et al. (1978) published a series of tables that use the binomial formula to calculate the number of correct judgments and their probability of occurrence. These tables make it very easy to determine if a statistical difference were detected between two samples in discrimination tests. However, the sensory scientist may not have these tables easily available thus he/she should also know how to analyze discrimination data using statistical tables that are more readily available. We abridged the tables from Roessler et al. (1978) into Table 4.3. Using this table is very simple. For example, in a duo-trio test using 45 panelists, 21 panelists correctly matched the sample to the reference. In Table 4.3, in the section for duo-trio tests, we find that the table value for 45 panelists at 5% probability is 29. This value is larger than 21 and therefore the panelists could not detect a difference between the samples. In a different study, using a triangle test, 21 of 45 panelists correctly identified the odd sample. In Table 4.3, in the section for triangle tests, we find that the table value for 45 panelists at 5% probability is 21. This value is equal to 21 and therefore the panelists could detect a significant difference between the samples at the alpha probability of 5%.

Discrimination Testing

O2 = observed number incorrect choices E1 = expected number of correct choices E1 is equal to total number of observations (n) times probability (p) of a correct choice, by chance alone in a single judgment where p = 0.100 for the two-out-of-five test p = 0.500 for duo-trio, paired difference, paired directional, alternate A-not-A tests p = 0.333 for triangle tests E2 = expected number of incorrect choices E2 is equal to total number of observations (n) times probability (q) of an incorrect choice, by chance alone in a single judgment where q = 1−p q = 0.900 for the two-out-of-five test q = 0.500 for duo-trio, paired difference, paired directional, alternate A-not-A tests, ABX tests q = 0.667 for triangle tests The use of discrimination tests allows the sensory scientist to determine whether two products are statistically perceived to be different, therefore the degrees of freedom equal one (1). Therefore, a χ 2 table using df = 1 should be consulted, for alpha (α) at 5% the critical χ2 value is 3.84. For other alpha levels consult the chi-square table in the Appendix.

4.4.2 The Adjusted Chi-Square (χ2 ) Test

4.4.3 The Normal Distribution and the Z-Test on Proportion The chi-square distribution allows the sensory scientist to compare a set of observed frequencies with a matching set of expected (hypothesized) frequencies. The chi-square statistic can be calculated from the following formula (Amerine and Roessler, 1983), which includes the number -0.5 as a continuity correction. The continuity correction is needed because the χ 2 distribution is continuous and the observed frequencies from discrimination tests are integers. It is not possible for one-half of a person to get the right answer and so the statistical approximation can be off by as much as 1/2, maximally. χ = 2

The sensory specialist can also use the areas under the normal probability curve to estimate the probability of chance in the results of discrimination tests. The tables associated with the normal curve specify areas under the curve (probabilities) associated with specified values of the normal deviate (z). The following two formulae (Eqs. (4.8) and (4.9)) can be used to calculate the z-value associated with the results of a specific discrimination test (Stone and Sidel, 1978):

z=

+ 1.645(1.5) (X/N)(1−X/N) N

(5.7) Here is a worked example. Suppose we do a triangle test with 60 panelists, and 30 get the correct answer. We can ask the following questions: What is the best estimate of the number of discriminators? The proportion of discriminators? What is the upper 95% confidence interval on the number of discriminators? What is the confidence interval on the proportion of discriminators? Finally, could we conclude that there is significant similarity, based on a maximum allowable proportion of discriminators of 50%? The solution is as follows: Let X = number correct, D = number of discriminators, so X = 30 and N = 60. We have 1.5(30)-0.5(60) = D or 15 discriminators, or 25% of our judges detecting the difference, as our best estimate.

110

5 Similarity, Equivalence Testing, and Discrimination Theory

The standard error is given by

1.5

(30/60)

(6.4)

variation was used by Stevens et al. (1988) in a landmark paper on the inpidual variability in olfactory thresholds. In this case, five correct pairs were required to score the concentration as correctly detected, and this performance was confirmed at the next highest concentration level. The most striking finding of this study was that among the three inpiduals tested 20 times, their inpidual thresholds for butanol, pyridine, and phenylethylmethylethyl carbinol (a rose odorant) varied over 2,000- to 10,000-fold in concentration. Variation within an inpidual was as wide as the variation typically seen across a population of test subjects. This surprising result suggests that day-to-day variation in olfactory sensitivity is large and that thresholds for an inpidual are not very stable (for an example, see Lawless et al., 1995). More recent work using extensive testing of inpiduals at each concentration step suggests that these estimates of variability may be high. Walker et al. (2003) used a simple yes/no procedure (like the A, not-A test, or signal detection test) with 15 trials of targets and 15 trials of blanks at each concentration level. Using a model for statistical significant differences between blank and target trials, they were able to get sharp gradients for the inpidual threshold estimates. In summary, an ascending forced-choice method is a reasonably useful compromise between the need to pcisely define a threshold level and the problems encountered in sensory adaptation and observer fatigue when extensive measurements are made. However, the user of an ascending forced-choice procedure should be aware of the procedural choices that can affect the obtained threshold value. The following choices will affect the measured value: the number of alternatives (both targets and blanks), the stopping rule, or the number of correct steps in a row required to establish a threshold, the number of replicated correct trials required at any one step, and the rule to determine at what level of concentration steps the threshold value is assigned. For example, the inpidual threshold might be assigned at the lowest level correct, the geometric mean between the lowest level correct and highest level incorrect. Other specific factors include the chosen step size of concentration units (factors of two or three are common in taste and smell), the method of averaging or combining replicated ascending runs on the same inpidual and finally the method of averaging or combining group data. Geometric means are commonly used for the last two purposes.

136

6 Measurement of Sensory Thresholds

6.7 Probit Analysis

7.0

97.5 95 (90)

90 (80)

(85)

6.0

Cumulative Percentage

80 70

(57)

60 5.0

50

(40)

40

PROBIT

It is often useful to apply some kind of transformation or graphing method to the group data to linearize the curve used to find the 50% point in a group. Both the psychometric curve that repsents the behavior of an inpidual in multiple trials of a threshold test and the cumulative distribution of a group will resemble an S-shaped function similar to the cumulative normal distribution. A number of methods for graphing such data are shown in the ASTM standard E-1432 (ASTM, 2008b). One simple way to graph the data is simply to plot the cumulative percentages on “probability paper.” This p-printed solution provides a graph in which equal standard deviations are marked off along the ordinate, effectively stretching the percentile intervals at the ends and compssing them in the midrange to conform to the density of the normal distribution. Another way to achieve the straightening of the S-shaped response curve is to transform the data by taking z-scores. Statistical packages for data analysis often provide options for transformation of the data. A related method was once widely used in threshold measurement, called Probit analysis (ASTM, 2008b; Dravnieks and Prokop, 1975; Finney, 1971). In this approach, the inpidual points are transformed relative to the mean value, pided by the standard deviation and then a constant value of +5 is added to translate all the numbers to positive values for convenience. A linear fitted function can now be interpolated at the value of 5 to estimate the threshold as in Fig. 6.6. The conversion (to a z-score +5) tends to make an S-shaped curve more linear. An example of this can be found in the paper by Brown et al. (1978), using data from a multiple paired test. First the percent correct is adjusted for chance. Then the data are transformed from the percent correct (across the group) at each concentration level by conversion to z-scores and a constant of 5 is added. The mean value or Probit equal to 5 can be found by interpolation or curve fitting. An example of this technique for estimating threshold from a group of 20 panelists is shown in Meilgaard et al. (1991) and in ASTM (2008b). Probit plots can be used for any cumulative proportions, as well as ranked data and analysis of inpiduals who are more extensively tested than in the 3-AFC method example shown earlier.

30 (18)

20 10

4.0 (6)

5 2.5

3.0 1

2 4 8 16 32 64 Threshold Concentration (log scale)

Fig. 6.6 An example of Probit analysis. Numbers in parentheses are the cumulative percentages of panelists reaching threshold at each concentration step. Note the uneven scale on the left axis. Probits mark off equal standard deviations and are based on the z-score for any proportion plus the constant, 5. Interpolation at the 50% or Probit 5.0 gives the threshold value.

6.8 Sensory Adaptation, Sequential Effects, and Variability Inpidual variability, both among a group of people and within an inpidual over repeated measurements psents a challenge to the idea that the threshold is anything like a fixed value. For example, stable olfactory thresholds of an inpidual are difficult to measure. The test-retest correlation for inpidual’s olfactory threshold is often low (Punter, 1983). Even within an inpidual, threshold values will generally decrease with practice (Engen, 1960; Mojet et al., 2001; Rabin and Cain, 1986), and superimposed upon this practice effect is a high level of seemingly random variation (Stevens et al., 1988). Inpiduals may become sensitive to odorants to which they were formerly anosmic, apparently through simple exposure (Wysocki et al., 1989). Increased sensitivity as a function of exposure may be a common phenomenon among women of childbearing age (Dalton et al., 2002; Diamond et al., 2005). Sensory adaptation and momentary changes in sensitivity due to sequences may have occurred in

6.9 Alternative Methods: Rated Difference, Adaptive Procedures, Scaling

the experiments of Stevens et al. (1988) and could have contributed to some instability in the measurements. As pdicted by sequential sensitivity analysis (Masuoka et al., 1995; O’Mahony and Odbert, 1985) the specific stimulus sequence will render discrimination more or less difficult. After the stronger of two stimuli, an additional second strong stimulus psented next may be partially adapted and seem weaker than normal. Stevens et al. remarked that sometimes subjects would get all five pairs correct at one level with some certainty that they “got the scent” but lost the signal at the next level before getting it back. This report and the reversals of performance in threshold data are consistent with adaptation effects temporarily lessening sensitivity. The sensory impssion will sometimes “fade in and out” at levels near threshold. In attempts to avoid adaptation effects, other researchers have gone to fewer psentations of the target stimulus. For example, Lawless et al. (1995) used one target among three blank stimuli, a 4-AFC test that has appeared in pvious studies (e.g., Engen, 1960; Punter, 1983). This lowers the chance performance level and lessens the potential adaptation at any one concentration step. To guard against the effects of correct guessing, threshold was taken as the lowest concentration step with a correct choice when all higher concentrations were also correct. Thresholds were measured in duplicate ascending runs in a test session, and a duplicate session of two more ascending runs was run on a second day. Correlations across the four ascending series ranged from 0.75 to 0.92 for cineole and from 0.51 to 0.92 for carvone. For carvone, thresholds were better duplicated within a day (r = 0.91 and 0.88) than across days (r from 0.51 to 0.70). This latter result suggests some drift over time in odor thresholds, in keeping with the variability seen by Stevens et al. (1988). However, results with this ascending method may not be this reliable for all compounds. Using the ascending 4-AFC test and a sophisticated olfactometer, Punter (1983) found median retest correlations for 11 compounds to be only 0.40. The sense of taste may fare somewhat better. In a study of electrogustometric thresholds with ascending paired tests requiring five correct responses, retest correlations for an elderly population were 0.95 (Murphy et al., 1995). In many forced-choice studies, high variability in smell thresholds is also noted across the testing pool.

137

Brown et al. (1978) stated that for any test compound, a number of insensitive inpiduals would likely be seen in the data set, when 25 or more persons were tested to determine an average threshold. Among any given pool of participants, a few people with otherwise normal smell acuity will have high thresholds. This is potentially important for sensory professionals who need to screen panelists for detection of specific flavor or odor notes such as defects or taints. In an extensive survey of thresholds for branched-chain fatty acids, Brennand et al. (1989) remarked that “some judges were unable to identify the correct samples in the pairs even in the highest concentrations provided” and that ” panelists who were sensitive to most fatty acids found some acids difficulty to perceive” (p. 109). Wide variation in sensitivity was also observed to the common flavor compound, diacetyl, a buttery-smelling by-product of lactic bacteria fermentation (Lawless et al., 1994). Also, simple exposure to some chemicals can modify specific anosmia and increase sensitivity (Stevens and O’Connell, 1995).

6.9 Alternative Methods: Rated Difference, Adaptive Procedures, Scaling 6.9.1 Rated Difference from Control Another practical procedure for estimating threshold has involved the use of ratings on degree-of-difference scales, where a sample containing the to-be-recognized stimulus is compared to some control or blank stimulus (Brown et al., 1978; Lundahl et al., 1986). Rated difference may use a line scale or a category scale, ranging from no difference or “exact same” to a large difference, as discussed in Chapter 4. In these procedures ratings for the sensory difference from the control sample will increase as the intensity of the target gets stronger. A point on the plot of ratings versus concentration is assigned as threshold. In some variations on this method, a blind control sample is also rated. This provides the opportunity to estimate a baseline or false alarm rate based on the ratings (often nonzero) of the control against itself. Identical samples will often get nonzero difference estimates due to the moment-to-moment variability in sensations.

138

of the inpidual thresholds, due to the larger number of observations, another oddity of using statistical significance to determine the threshold. Instead of using statistical significance as a criterion, Marin et al. determined the point of maximum curvature on the dose-response curve as the threshold. Such an approach makes sense from consideration of the general form of the dose-response (psychophysical) curve for most tastes and odors. Figure 6.7 shows a semi-log plot for the Beidler taste equation, a widely applied dose-response relationship in studies of the chemical senses (see Chapter 2). This function has two sections of curvature (change in slope, i.e., acceleration) when plotted as a function of log concentration. There is a point at which the response appears to be slowly increasing out from the background noise and then rises steeply to enter the middle of the dynamic range of response. The point of maximum curvature can be estimated graphically or determined from curve fitting and finding the maximum rate of change (i.e., maximum of the second derivative) (Marin et al., 1991).

Psychophysical function (semi-hyperbolic) 100

Response (percent of maximum)

In one application of this technique for taste and smell thresholds, a substance was added in various levels to estimate the threshold in a food or beverage. In each inpidual trial, three samples would be compared to the control sample with no added flavor-two adjacent concentration steps of the target compound and one blind control sample (Lundahl et al., 1986). Samples were rated on a simple 9-point scale, from zero (no difference) to eight (extreme difference). This provided a comparison of the control to itself and a cost-effective way of making three comparisons in one set of samples. Since sample concentrations within the three rated test samples were randomized, the procedure was not a true ascending series and was dubbed the “semi-ascending paired difference method.” How is the threshold defined in these procedures? One approach is to compare the difference ratings for a given level with the difference ratings given to the control sample. Then the threshold can be based on some measure of when these difference ratings perge, such as when they become significantly different by a t-test (see Brown et al., 1978). Another approach is simply to subtract the difference score given to the blind control from the difference score give to each test sample and treat these adjusted scores as a new data set. In the original paper of Lundahl et al. (1986), this latter method was used. In the analysis, they performed a series of ttests. Two values were taken to bracket the range of the threshold. The upper level was the first level yielding a significant t-test versus zero, and the lower level was the nearest lower concentration yielding a significant ttest versus the first. This provided an interval in which the threshold (as defined by this method) lies between the two bracketing concentrations. One problem with this approach is that when the threshold is based on the statistical significance of t-statistics (or any such significance test), the value of threshold will depend upon the number of observations in the test. This creates a nonsensical situation where the threshold value will decrease as a function of the number of panelists used in the test. This is an irrelevant variable, a choice of the experimenter, and has nothing to do with the physiological sensitivity of the panelist or the biological potency of the substance being tested, a problem recognized by Brown et al. (1978) and later by Marin et al. (1991). Marin et al. also pointed out that a group threshold, based on a larger number of observations than an inpidual threshold, would be lower than the mean

6 Measurement of Sensory Thresholds

80

60

R = (Rmax C) / (K + C) K = 1000, Rmax = 100

40

20

zone of inflection = threshold range

0 0

2

4 log Concentration

6

8

Fig. 6.7 Beidler curve.

6.9.2 Adaptive Procedures Popular methods for threshold measurement for visual and auditory stimuli have been procedures in which the next stimulus intensity level to be tested depends

6.9 Alternative Methods: Rated Difference, Adaptive Procedures, Scaling

SAMPLE UDTR RECORD CONCENTRATION STEP

CONCENTRATION STEP

the part of the respondent. Psychophysical researchers have found ways to undo this sequential dependence to counteract observer expectancies. One example is the double random staircase procedure (Cornsweet, 1962) in which trials from two staircase sequences are randomly intermixed. One staircase starts above the threshold and descends, while the other starts below the threshold and ascends. On any given trial, the observer is unaware which of the two sequences the stimulus is chosen from. As in the simple staircase procedure, the level chosen for a trial depends upon detection or discrimination in the pvious trial, but of that particular sequence. Further refinements of the procedure involve the introduction of forced-choice (Jesteadt, 1980) to eliminate response bias factors involved in simple yes/no detection. Another modification to the adaptive methods has been to adjust the ascending and descending rules so that some number of correct or incorrect judgments is required before changing intensity levels, rather than the one trial as in the simple staircase (Jesteadt, 1980). An example is the “up down transformed response” rule or UDTR (Wetherill and Levitt, 1965). Wetherill and Levitt gave an example where two positive judgments were required before moving down, and only one negative judgment at a given level before moving up. An example is shown in Fig. 6.9. Rather than estimating the 50% point on a traditional psychometric function, this more stringent requirement now tends to converge on the 71% mark as an average of the peaks and valleys in the series. A forced choice can be added to an adaptive procedure. Sometimes

SAMPLE STAIRCASE RECORD

64 32 16 8 4

R 2 R R 2

3

4

R

R

1

1

5

139

6

7

R 8

9

10

TRIAL

Fig. 6.8 Staircase example.

TRIAL

Fig. 6.9 Staircase example.

140

6.9.3 Scaling as an Alternative Measure of Sensitivity Threshold measurements are not the only way to screen inpiduals for insensitivity to specific compounds like PTC or to screen for specific anosmia. Do thresholds bear any relation to suprathreshold responding? While it has been widely held that there is no necessary relationship between threshold sensitivity and suprathreshold responding (Frijters, 1978; Pangborn, 1981), this assertion somewhat overstates the case. Counter-examples of good correlations can be seen in tests involving compounds like PTC where there are insensitive groups. For example, there is a -0.8 correlation between simple category taste intensity ratings for PTC and the threshold, when the rated concentration is near the antimode or center between the modes of a bimodal threshold frequency distribution (Lawless, 1980). Thus ratings of a carefully chosen level can be used for a rapid screening method for PTC taster status (e.g., Mela, 1989). Similar results have been noted for smell. Berglund and Högman (1992) reported better reliability of suprathreshold ratings than threshold determinations in screening for olfactory sensitivity. Stevens and O’Connell (1991) used category ratings of perceived intensity as well as qualitative descriptors as a screening tool before threshold testing for specific anosmia. Threshold versus rating correlations were in the range of -0.6 for cineole, -0.3 for carvone, and -0.5 for diacetyl (Lawless et al., 1994, 1995). The correlations were obtained after subtraction of ratings to a blank stimulus, in order to correct for differences in scale usage. Thus there is a moderate negative correlation of sensitivity and rated intensity when one

6 Measurement of Sensory Thresholds

examines the data across a highly variable group as is the case with specific anosmia or tasting PTC bitterness. The correlation is negative since higher thresholds indicate lower sensitivity and thus lower rated intensity.

6.10 Dilution to Threshold Measures 6.10.1 Odor Units and Gas-Chromatography Olfactometry (GCO) In this section, several applied methods will be described that make use of the threshold concept in trying to determine the sensory impact of various flavors and odor materials. The first group of methods concerns the olfactory potency of volatile aroma compounds as they are found in foods or food components. The issue here becomes one of not just threshold determination, but determination of both the threshold and the actual concentration psent in a food sample. The ratio of these concentrations (actual concentration to threshold concentration) can help indicate whether or not a given flavor substance is likely to contribute to the overall sensory impssion in a food. These ratios are commonly called “odor units.” The second much older method is similar in logic and was developed to determine the point at which the irritative or heat sensations from pepper compounds would be first detectable when diluted to a given extent, the Scoville procedure. Both of these techniques then use dilution-to-threshold as a measure of sensory impact. When a complex natural product like a fruit extract is analyzed for its chemical components, hundreds or even thousands of chemicals may be identified, many of which have odor properties. The number of potential flavor compounds identified in any product seems only to be limited by the resolution and sensitivity of the current methods in analytical chemistry. These methods are always improving leading to longer and longer lists of possible contributing flavor materials (Piggott, 1990). Flavor scientists need to find a way to narrow the list or to separate those compounds which are most likely contributing to the overall flavor from those compounds that are psent in

6.10 Dilution to Threshold Measures

such low concentrations that they are probably not important. Obviously, a sensory-based method is needed in conjunction with the analytical chemistry to provide a bioassay for possible sensory impact (Acree, 1993). Thresholds can be useful in addressing this kind of problem. The reasoning goes that only those compounds that are psent in the product in concentrations above their threshold are likely to be contributors to the flavor of the product. There are a number of potential flaws in this thinking discussed below, but for now let us see how this can be put to use. Given a concentration C psent in a natural product, a dimensionless quantity can be derived by piding that concentration by the threshold concentration Ct , and the ratio C/Ct defines the number of odor units (or flavor units) for compounds assessed by smell. According to this logic, only those compounds with odor units greater than one will contribute to the aroma of the product. This reasoning is extended sometimes to include the idea that the greater the number of odor units, the greater the potential contribution. However, it is now widely recognized that the odor unit is a concentration multiple and not a measure of subjective magnitude. Only direct scaling methods can assess the actual magnitude of sensation above threshold and the psychophysical relationship between concentration and odor intensity (Frijters, 1978). Furthermore, this idea ignores the possibility of subthreshold additivity or synergy (Day et al., 1963). A closely related group of chemical compounds might all be psent below their inpidual thresholds, but together could stimulate common receptors so as to produce an above-threshold sensation. Such additivity is not pdicted by the odor unit approach and such a group of compounds could be missed in dilution analysis. Nonetheless, thresholds provide at least one isointense reference point on the dose response curve, so they have some utility as a measure of potency used to compare different odor compounds. In analyzing a food, one could look up literature values for all the identified compounds in the product in one of the published compendia of thresholds (e.g., ASTM, 1978; van Gemert, 2003). If the concentration in the product is determined, then the odor unit value can be calculated by simply piding by threshold. However, it is important to remember that the literature values for thresholds depend upon the method and the medium

141

of testing. Unless the same techniques are used and the same medium was used as the carrier (rarely the case) the values may not be necessarily comparable for different compounds. A second approach is to actually measure the dilutions necessary to reach threshold for each compound, starting with the product itself. This necessitates the use of a separatory procedure, so that each compound may be inpidually perceived. The combination of gas chromatography with odor port sniffing of a dilution series is a popular technique (Acree, 1993). Various catchy names have been applied to such techniques in the flavor literature, including Aroma Extract Dilution Analysis (for examples, see Guth and Grosch, 1994; Milo and Grosch, 1993; Schieberle and Grosch, 1988), CHARM analysis (Acree et al., 1984) or more generically, gas chromatography olfactometry or GCO. The basis of these techniques is to have subjects respond when an odor is perceived when sniffing the exit port during a GC run. In recent years, the effluent has been embedded in a cooled, humidified air stream to improve the comfort of the observer and to increase sensory resolution of the eluting compounds. Over several dilutions, the response will eventually drop out, and the index of smell potency is related to the reciprocal of the dilution factor. The sniffer’s responses occur on a time base that can be cross-referenced to a retention index and then the identity of the compound can be determined by a combination of retention index, mass spectrometry, and aroma character. In practice, these techniques considerably shorten the list of potential aroma compounds contributing to flavor in a natural product (e.g., Cunningham et al., 1986). The method has also been used as an assessment technique for measuring the sensitivity of human panelists, as opposed to being a tool to determine the sensory impact of flavor compounds (Marin et al., 1988). In this approach, the gas chromatograph once again serves as an olfactometer. Known compounds can be psented as a dilution series of mixtures. Variation in inpidual thresholds can readily be assessed for a variety of compounds since they can be combined in a GC run as long as they have different retention times, i.e., do not co-elute on the column of choice. The potential use of GCO for screening panelists, for assessing odor responses in general, and for assessing specific anosmias has also been attempted (Friedrich and Acree, 2000; Kittel and Acree, 2008).

142

6.10.2 Scoville Units Another example of a dilution method is the traditional Scoville procedure for scoring pepper heat in the spice trade. This procedure was named for W. Scoville who worked in the pharmaceutical industry in the early twentieth century. He was interested in the topical application of spice compounds like pepper extracts as counterirritants, and he needed to establish units that could be used to measure their potency. His procedure consisted of finding the number of dilutions necessary for sensations to disappear and then using this number of dilutions as an estimate of potency. In other words, potency was defined as a reciprocal threshold. A variation of this procedure was adopted by the Essential Oil Association, British Standards Institution, International Standards Organization, American Spice Trade Association (ASTA), and adopted as an Indian standard method (for a review, see Govindarajan, 1987). The procedure defined units of pungency as the highest dilution at which a definite “bite” would be perceived and thus contains instructions consistent with a recognition threshold. Scoville units were dilution factors, now commonly given as mL/g. ASTA Method 21 (ASTA, 1968) is widely used and contains some modifications in an attempt to overcome problems with the original Scoville procedure. In brief, the method proceeds as follows: Panelists are screened for acuity relative to experienced persons. Dilution schedules are provided which simplify calculations of the eventual potency. Solutions are tested in 5% sucrose and negligible amounts of alcohol. Five panelists participate and concentrations are given in ascending order around the estimated threshold. Threshold is defined as the concentration at which three out of five judges respond positively. This method is difficult in practice and a number of additional variations have been tried to improve on the accuracy and pcision of the method (Govindarajan, 1987). Examples include the following: (1) substitution of other rules for 3/5, e.g., mean + SD of 20-30 judgments, (2) use of a triangle test, rather than simple yes/no at each concentration, (3) requiring recognition of pungency (Todd et al., 1977), (4) reduction of sucrose concentration in the carrier solution to 3%, and (5) use of a rating scale from 1 (definitely not detectable) to 6 (definitely detectable). This latter

6 Measurement of Sensory Thresholds

modification defined a detection threshold at mean scale value of 3.5. Almost all these methods specify mandatory rest periods between samples due to the long lasting nature of these sensations. The measurements are still difficult. One problem is that capsaicin, the active heat principle in red pepper, is prone to desensitize observers within a session and also regular consumers of hot spices also become less sensitive, leading to wide inpidual differences in sensitivity among panelists (Green, 1989; Lawless et al., 1985). Alternative procedures have been developed based on rating scales with fixed physical references (Gillette et al., 1984) and these have been endorsed by ASTM (2008c) as standard test methods. These rating scale procedures show good correlations with instrumental measures of capsaicin content in pepper samples and can be cross-referenced to Scoville units for those who pfer to do business in the traditional units (Gillette et al., 1984).

6.11 Conclusions Threshold measurements find three common uses in sensory analysis and flavor research. First, they can be used to compare the sensitivities of different panelists. Second, they can be used as an index of the biological potency of a flavor compound. Third, they can provide useful information regarding the maximum tolerable levels of an off-flavor or taint. A variety of different techniques have been used to find thresholds or have employed the threshold concept in practical flavor work. Examples of different threshold methods are given in Table 6.4. In spite of their practical applications, the usefulness of threshold measures is often questioned in sensory evaluation. One criticism is that thresholds are only one point on an intensity function and thus they do not tell us anything about abovethreshold responding. There are some good examples in which thresholds do not pdict or do not correlate very well with suprathreshold responses. For example, patients irradiated for cancer may lose their sense of taste temporarily and thresholds return to normal long before suprathreshold responsiveness is recovered (Bartoshuk, 1987). However, as we have seen both in the case of PTC tasting and in specific anosmias, insensitive inpiduals (as determined by their threshold) will also tend to be less responsive above

Appendix

143

Table 6.4 Threshold procedures Method Citations/examples

Response

Threshold

Ascending forced choice Ascending forced choice

ASTM E-679-79 Stevens et al. (1988)

3-AFC 2-AFC, 5 replicates

Semi-ascending paired difference

Lundahl et al. (1986) Rated difference from control

Adaptive, up-down transformed response rule

Reed et al. (1995)

2-AFC

Double Random Staircase CHARM analysis

Cornsweet (1962) Acree et al. (1986)

Yes/No Yes/No

Geometric mean of inpidual threshold Lowest correct set with confirmation at next concentration t-test for significant difference from zero with all blank trials (blind control scores) subtracted Average of responses ascending after one incorrect, descending after two correct at each level Average of reversal points Nonresponse on descending concentration runs (implied)

threshold. These correlations are strong when comparing inpiduals of very different sensitivities, but the threshold-suprathreshold parallel may not extend to all flavor compounds. A more complete understanding of the whole dynamic range of dose-response, as found in scaling studies, would be more informative. Other shortcomings need to be kept in mind by sensory workers who would use threshold measurements for making product decisions. First, thresholds are statistical constructs only. Signal detection theory warns us that the signal and noise perge in a continuous fashion and discontinuities in perception may be an idealized construction that is comfortable but unrealistic. There is no sudden transition from nondetection to 100% detection. Any modern concept of threshold must take into account that a range of values, rather than a single point is involved in the specification. Thresholds depend upon the conditions of measurement. For example, as the purity of the diluent increases, the threshold for taste will go down. So a measure of the true absolute threshold for taste (if such a thing existed) would require water of infinite purity. The threshold exists not in this abstract sense, but only

as a potentially useful construct of our methods and informational requirements. Finally, because of the problems mentioned above, sensory professionals need to keep the following principles firmly in mind when working with threshold procedures: First, changes in method will change the obtained values. Literature values cannot be trusted to extend to a new product or a new medium or if changes are made in test procedure. Next, threshold distributions do not always follow a normal bell curve. There are often high outliers and possibly cases of insensitivity due to inherited deficits like specific anosmia (Amoore, 1971; Brown et al., 1978; Lawless et al., 1994). Threshold values for an inpidual are prone to high variability and low test-retest reliability. An inpidual’s threshold measure on a given day is not necessarily a stable characteristic of that person (Lawless et al., 1995; Stevens et al., 1988). Practice effects can be profound, and thresholds may stabilize over a period of time (Engen, 1960; Rabin and Cain, 1986). However, group averaged thresholds are reliable (Brown et al., 1978; Punter, 1983) and provide a useful index of the biological activity of a stimulus.

Appendix: MTBE Threshold Data for Worked Example

Panelist 1 2 3 4 5 6

Concentration (μg/L) 2 3.5 + o o o o + o o + o o o

6 + o o o + +

10 + + o o o +

18 + + o o + +

30 + + o o + +

60 + + + + o +

100 + + + + + +

BET 4.6 7.7 42 42 77 4.6

log(BET) 0.663 0.886 1.623 1.623 1.886 0.663

144

Panelist 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Prop. Corr.

6 Measurement of Sensory Thresholds

Concentration (μg/L) 2 3.5 6 o + + + + o o o + o o o + o + o o o + + + + + + o + + o o + + o + o o + + + + + o + + + + + + + + + o + + + o + + o + + + o o o o + o + o o + + + o o + + + + + + o o o o o o o o o + + + + o + + + o o o + + + + o + o o o o o + + o + + + + + o + o o + o o o o o o + o o o + + + + + + + o o o o + o o o o o + 0.44 0.49 0.61

10 o + o o + o + + o o + + + + + + o + + + o + o o + + o o + + + + o o + + o o o + o + + + + + + o + o + 0.58

18 + + + o + + + + + o + + + o + + + + o o + + + + o o + o o o + + + + + + o + + + o o o + + + + o + o o 0.65

30 + + + o + o + + + o + + + + + + + + + + o + + + o o o o + + + + + + + + o o + + + + + + + + + + + + + 0.77

60 + + + o + + + + + + + + + o + + + + + + o + + + + o + + + + + + + + + + + + + o + + + + + + + + + + o 0.89

100 + + + + + o + + + o + + + o + + + + + + + o + + o + + o o + + + + + + + + + + + o + + + + + + + + + + 0.86

BET 13 7.7 13 77 4.6 132 1.4 1.4 13 132 4.6 4.6 1.4 132 1.4 1.4 13 1.4 23 23 77 132 13 13 132 77 42 132 132 23 1.4 1.4 13 13 1.4 7.7 42 42 13 77 132 23 23 4.6 7.7 1.4 1.4 23 4.6 23 77 Mean (log(BET)) 10ˆ1.154=

log(BET) 1.114 0.886 1.114 1.886 0.663 2.121 0.146 0.146 1.114 2.121 0.663 0.663 0.146 2.121 0.146 0.146 1.114 0.146 1.362 1.362 1.886 2.121 1.114 1.114 2.121 1.886 1.623 2.121 2.121 1.362 0.146 0.146 1.114 1.114 0.146 0.886 1.623 1.623 1.114 1.886 2.121 1.362 1.362 0.663 0.886 0.146 0.146 1.362 0.663 1.362 1.886 1.154 14.24

References

References Acree, T. E. 1993. Bioassays for flavor. In: T. E. Acree and R. Teranishi (eds.), Flavor Science, Sensible Principles and Techniques. American Chemical Society Books, Washington, pp. 1-20. Acree, T. E., Barnard, J. and Cunningham, D. G. 1984. A procedure for the sensory analysis for gas chromatographic effluents. Food Chemistry, 14, 273-286. American Spice Trade Association 1968. Pungency of capsicum spices and oleoresins. American Spice Trade Association Official Analytical Methods, 21.0, 43-47. Amoore, J. E. 1971. Olfactory genetics and anosmia. In: L. M. Beidler (ed.), Handbook of Sensory Physiology. Springer, Berlin, pp. 245-256. Amoore, J. E. 1979. Directions for pparing aqueous solutions of primary odorants to diagnose eight types of specific anosmia. Chemical Senses and Flavor, 4, 153-161. Amoore, J. E.,Venstrom, D. and Davis, A. R. 1968. Measurement of specific anosmia. Perceptual and Motor Skills, 26, 143-164. Antinone, M. A., Lawless, H. T., Ledford, R. A. and Johnston, M. 1994. The importance of diacetyl as a flavor component in full fat cottage cheese. Journal of Food Science, 59, 38-42. ASTM. 1978. Compilation of Odor and Taste Threshold Values Data. American Society for Testing and Materials, Philadelphia. ASTM. 2008a. Standard practice for determining odor and taste thresholds by a forced-choice ascending concentration series method of limits, E-679-04. Annual Book of Standards, Vol. 15.08. ASTM International, Conshocken, PA, pp. 36-42. ASTM. 2008b. Standard practice for defining and calculating inpidual and group sensory thresholds from forcedchoice data sets of intermediate size, E-1432-04. ASTM International Book of Standards, Vol. 15.08. ASTM International, Conshocken, PA, pp.82-89. ASTM. 2008c. Standard test method for sensory evaluation of red pepper heat, E-1083-00. ASTM International Book of Standards, Vol. 15.08. ASTM International, Conshocken, PA, pp. 49-53. Ayya, N. and Lawless, H. T. 1992. Qualitative and quantitative evaluation of high-intensity sweeteners and sweetener mixtures. Chemical Senses, 17, 245-259. Bartoshuk, L. M. 1987. Psychophysics of taste. American Journal of Clinical Nutrition, 31, 1068-1077. Bartoshuk, L. M., Murphy, C. and Cleveland, C. T. 1978. Sweet taste of dilute NaCl: Psychophysical evidence for a sweet stimulus. Physiology and Behavior, 21, 609-613. Berglund, B. and Högman, L. 1992. Reliability and validity of odor measurements near the detection threshold. Chemical Senses, 17, 823-824. Blakeslee, A. F. 1932. Genetics of sensory thresholds: Taste for phenylthiocarbamide. Proceedings of the National Academy of Sciences, 18, 120-130. Brennand, C. P., Ha, J. K. and Lindsay, R. C. 1989. Aroma properties and thresholds of some branched chain and other minor volatile fatty acids occurring in milkfat and meat lipids. Journal of Sensory Studies, 4, 105-120.

145 Boring, E. G. 1942. Sensation and Perception in the History of Experimental Psychology. Appleton-CenturyCrofts, New York, pp. 41-42. Brown, D. G. W., Clapperton, J. F., Meilgaard, M. C. and Moll, M. 1978. Flavor thresholds of added substances. American Society of Brewing Chemists Journal, 36, 73-80. Bufe, B., Breslin, P. A. S., Kuhn, C., Reed, D., Tharp, C. D., Slack, J. P., Kim, U.-K., Drayna, D. and Meyerhof, W. 2005. The molecular basis of inpidual differences in phenylthiocarbamide and propylthiouracil bitterness perception. Current Biology, 15, 322-327. Cain, W. S. 1976. Olfaction and the common chemical sense: Some psychophysical contrasts. Sensory Processes, 1, 57-67. Cain, W. S. and Murphy, C. L. 1980. Interaction between chemoreceptive modalities of odor and irritation. Nature, 284, 255-257. Cornsweet, T. M. 1962. The staircase method in psychophysics. American Journal of Psychology, 75, 485-491. Cunningham, D. G., Acree, T. E., Barnard, J., Butts, R. M. and Braell, P. A. 1986. CHARM analysis of apple volatiles. Food Chemistry, 19, 137-147. Dale, M. S., Moylan, M. S., Koch, B. and Davis, M. K. 1997. Taste and odor threshold determinations using the flavor profile method. Proceedings of the American Water Works Association Water-Quality Technology Conference, Denver, CO. Dalton, P., Doolittle, N. and Breslin, P. A. S. 2002. Genderspecific induction of enhanced sensitivity to odors. Nature Neuroscience, 5, 199-200. Day, E. A., Lillard, D. A. and Montgomery, M. W. 1963. Autooxidation of milk chúng tôi Effect on flavor of the additive interactions of carbonyl compounds at subthreshold concentrations. Journal of Dairy Science, 46, 291-294. Diamond, J., Dalton, P., Doolittle, N. and Breslin, P. A. S. 2005. Gender-specific olfactory sensitization: Hormonal and cognitive influences. Chemical Senses, 30(suppl 1), i224-i225. Dravnieks, A. and Prokop, W. H. 1975. Source emission odor measurement by a dynamic forced-choice triangle olfactometer. Journal of the Air Pollution Control Association, 25, 28-35. Engen, T. E. 1960. Effects of practice and instruction on olfactory thresholds. Perceptual and Motor Skills, 10, 195-198. Finney, D. J. 1971. Probit Analysis, Third Edition. Cambridge University, London. Friedrich, J. E. and Acree, T. E. 2000. Design of a standard set of odorants to test inpiduals for specific anosmia. Frontiers in Flavour Science [Proc. 9th Weurman Flavour Res. Symp.], 230-234. Frijters, J. E. R. 1978. A critical analysis of the odour unit number and its use. Chemical Senses and Flavour, 3, 227-233. Gillette, M., Appel, C. E. and Lego, M. C. 1984. A new method for sensory evaluation of red pepper heat. Journal of Food Science, 49, 1028-1033. Green, B. G. 1989. Capsaicin sensitization and desensitization on the tongue produced by brief exposures to a low concentration. Neuroscience Letters, 107, 173-178. Govindarajan, V. S. 1987. Capsicum – production, technology, chemistry and quality. Part III. Chemistry of the color, aroma and pungency stimuli. CRC Critical Reviews in Food Science and Nutrition, 24, 245-311.

146 Guth, H. and Grosch, W. 1994. Identification of the character impact odorants of stewed beef juice by instrumental analysis and sensory studies. Journal of Agricultural and Food Chemistry, 42, 2862-2866. Harris, H. and Kalmus, H. 1949. The measurement of taste sensitivity to phenylthiourea (P.T.C.). Annals of Eugenics, 15, 24-31. Harvey, L. O. 1986. Efficient estimation of sensory thresholds. Behavior Research Methods, Instruments and Computers, 18, 623-632. James, W. 1913. Psychology. Henry Holt and Co., New York. Jesteadt, W. 1980. An adaptive procedure for subjective judgments. Perception and Psychophysics, 28(1), 85-88. Kittel, K. M and Acree, T. E. 2008. Investigation of olfactory deficits using gas-chromatography olfactometry. Manuscript submitted, available from the authors. Lawless, H. T. 1980. A comparison of different methods used to assess sensitivity to the taste of phenylthiocarbamide PTC. Chemical Senses, 5, 247-256. Lawless, H. T. 2010. A simple alternative analysis for threshold data determined by ascending forced-choice method of limits. Journal of Sensory Studies, 25, 332-346. Lawless, H. T., Antinone, M. J., Ledford, R. A. and Johnston, M. 1994. Olfactory responsiveness to diacetyl. Journal of Sensory Studies, 9(1), 47-56. Lawless, H. T., Rozin, P. and Shenker, J. 1985. Effects of oral capsaicin on gustatory, olfactory and irritant sensations and flavor identification in humans who regularly or rarely consumer chili pepper. Chemical Senses, 10, 579-589. Lawless, H. T., Thomas, C. J. C. and Johnston, M. 1995. Variation in odor thresholds for l-carvone and cineole and correlations with suprathreshold intensity ratings. Chemical Senses, 20, 9-17. Linschoten, M. R., Harvey, L. O., Eller, P. A. and Jafek, B. W. 1996. Rapid and accurate measurement of taste and smell thresholds using an adaptive maximum-likelihood staircase procedure. Chemical Senses, 21, 633-634. Lundahl, D. S., Lukes, B. K., McDaniel, M. R. and Henderson, L. A. 1986. A semi-ascending paired difference method for determining the threshold of added substances to background media. Journal of Sensory Studies, 1, 291-306. Masuoka, S., Hatjopolous, D. and O’Mahony, M. 1995. Beer bitterness detection: Testing Thurstonian and sequential sensitivity analysis models for triad and tetrad methods. Journal of Sensory Studies, 10, 295-306. Marin, A. B., Acree, T. E. and Barnard, J. 1988. Variation in odor detection thresholds determined by charm analysis. Chemical Senses, 13, 435-444. Marin, A. B., Barnard, J., Darlington, R. B. and Acree, T. E. 1991. Sensory thresholds: Estimation from dose-response curves. Journal of Sensory Studies, 6(4), 205-225. McBurney, D. H. and Collings, V. B. 1977. Introduction to Sensation and Perception. Prentice-Hall, Englewood Cliffs, NJ. Meilgaard, M., Civille, G. V. and Carr, B. T. 1991. Sensory Evaluation Techniques, Second Edition. CRC Press, Boca Raton, FL. Mela, D. 1989. Bitter taste intensity: The effect of tastant and thiourea taster status. Chemical Senses, 14, 131-135. Milo, C. and Grosch, W. 1993. Changes in the odorants of boiled trout (Salmo fario) as affected by the storage of the raw

6 Measurement of Sensory Thresholds material. Journal of Agricultural and Food Chemistry, 41, 2076-2081. Mojet, J., Christ-Hazelhof, E. and Heidema, J. 2001. Taste perception with age: Generic or specific losses in threshold sensitivity to the five basic tastes. Chemical Senses 26, 854-860. Morrison, D. G. 1978. A probability model for forced binary choices. American Statistician, 32, 23-25. Murphy, C., Quiñonez, C. and Nordin, S. 1995. Reliability and validity of electrogustometry and its application to young and elderly persons. Chemical Senses, 20(5), 499-515. O’Mahony, M. and Ishii, R. 1986. Umami taste concept: Implications for the dogma of four basic tastes. In: Y. Kawamura and M. R. Kare (eds.), Umami: A Basic Taste. Marcel Dekker, New York, pp. 75-93.. O’Mahony, M. and Odbert, N. 1985. A comparison of sensory difference testing procedures: Sequential sensitivity analysis and aspects of taste adaptation. Journal of Food Science, 50, 1055-1060. Pangborn, R. M. 1981. A critical review of threshold, intensity and descriptive analyses in flavor research. In: Flavor ’81. Walter de Gruyter, Berlin, pp. 3-32. Piggott, J. R. 1990. Relating sensory and chemical data to understand flavor. Journal of Sensory Studies, 4, 261-272. Prescott, J., Norris, L., Kunst, M. and Kim, S. 2005. Estimating a “consumer rejection threshold” for cork taint in white wine. Food Quality and Preference, 18, 345-349. Punter, P. H. 1983. Measurement of human olfactory thresholds for several groups of structurally related compounds. Chemical Senses, 7, 215-235. Rabin, M. D. and Cain, W. S. 1986. Determinants of measured olfactory sensitivity. Perception and Psychophysics, 39(4), 281-286. Reed, D. R., Bartoshuk, L. M., Duffy, V., Marino, S. and Price, R. A. 1995. Propylthiouracil tasting: Determination of underlying threshold distributions using maximum likelihood. Chemical Senses, 20, 529-533. Saliba, A. J., Bullock, J. and Hardie, W. J. 2009. Consumer rejection threshold for 1,8 cineole (eucalyptol) in Australian red wine. Food Qualithy and Preference, 20, 500-504. Schieberle, P. and Grosch, W. 1988. Identification of potent flavor compounds formed in an aqueous lemon oil/citric acid emulsion. Journal of Agricultural and Food Chemistry, 36, 797-800. Stevens, D. A. and O’Connell, R. J. 1991. Inpidual differences in threshold and quality reports of subjects to various odors. Chemical Senses, 16, 57-67. Stevens, D. A. and O’Connell, R. J. 1995. Enhanced sensitivity to adrostenone following regular exposure to pemenone. Chemical Senses, 20, 413-419. Stevens, J. C., Cain, W. S. and Burke, R. J. 1988. Variability of olfactory thresholds. Chemical Senses, 13, 643-653. Stocking, A. J., Suffet, I. H., McGuire, M. J. and Kavanaugh, M. C. 2001. Implications of an MTBE odor study for setting drinking water standards. Journal AWWA, March 2001, 95-105. Todd, P. H., Bensinger, M. G. and Biftu, T. 1977. Determination of pungency due to capsicum by gas-liquid chromatography. Journal of Food Science, 42, 660-665. Tuorila, H., Kurkela, R., Suihko, M. and Suhonen. 1981. The influence of tests and panelists on odour detection

References thresholds. Lebensmittel Wissenschaft und Technologie, 15, 97-101. USEPA (U.S. Environmental Protection Agency). 2001. Statistical analysis of MTBE odor detection thresholds in drinking water. National Service Center for Environmental Publications (NSCEP) # 815R01024, available from http://nepis.epa.gov. Van Gemert, L. 2003. Odour Thresholds. Oliemans, Punter & Partners, Utrecht, The Netherlands. Viswanathan, S., Mathur, G. P., Gnyp, A. W. and St. Peirre, C. C. 1983. Application of probability models to threshold determination. Atmospheric Environment, 17, 139-143.

147 Walker, J. C., Hall, S. B., Walker, D. B., Kendall-Reed, M. S., Hood, A. F. and Nio, X.-F. 2003. Human odor detectability: New methodology used to determine threshold and variation. Chemical Senses, 28, 817-826. Wetherill, G. B. and Levitt, H. 1965. Sequential estimation of points on a psychometric function. British Journal of Mathematical and Statistical Psychology, 18(1), 1-10. Wysocki, C. J., Dorries, K. M. and Beauchamp, G. K. 1989. Ability to perceive androstenone can be acquired by ostensibly anosmic people. Proceedings of the National Academy of Science, USA, 86, 7976-7989.

Chapter 7

Scaling

Abstract Scaling describes the application of numbers, or judgments that are converted to numerical values, to describe the perceived intensity of a sensory experience or the degree of liking or disliking for some experience or product. Scaling forms the basis for the sensory method of descriptive analysis. A variety of methods have been used for this purpose and with some caution, all work well in differentiating products. This chapter discusses theoretical issues as well as practical considerations in scaling. The vital importance of knowing the properties and limitations of a measuring instrument can hardly be denied by most natural scientists. However, the use of many different scales for sensory measurement is common within food science; but very few of these have ever been validated. . . . -(Land and Shepard, 1984, pp. 144-145)

Contents 7.1 7.2 7.3

7.4

7.5

7.6

Introduction . . . . . . . . . . . . . . . . . Some Theory . . . . . . . . . . . . . . . . Common Methods of Scaling . . . . . . . . . 7.3.1 Category Scales . . . . . . . . . . . . 7.3.2 Line Scaling . . . . . . . . . . . . . 7.3.3 Magnitude Estimation . . . . . . . . . Recommended Practice and Practical Guidelines . . . . . . . . . . . . . . . . . . 7.4.1 Rule 1: Provide Sufficient Alternatives . . . . . . . . . . . . . . 7.4.2 Rule 2: The Attribute Must Be Understood . . . . . . . . . . . . 7.4.3 Rule 3: The Anchor Words Should Make Sense . . . . . . . . . . . . . . . . . 7.4.4 To Calibrate or Not to Calibrate . . . . . 7.4.5 A Warning: Grading and Scoring are Not Scaling . . . . . . . . . . . . . . . . Variations-Other Scaling Techniques . . . . 7.5.1 Cross-Modal Matches and Variations on Magnitude Estimation . . . . . . . . . 7.5.2 Category-Ratio (Labeled Magnitude) Scales . . . . . . . . . . . . . . . . 7.5.3 Adjustable Rating Techniques: Relative Scaling . . . . . . . . . . . . . . . . 7.5.4 Ranking . . . . . . . . . . . . . . . 7.5.5 Indirect Scales . . . . . . . . . . . . Comparing Methods: What is a Good Scale? . . . . . . . . . . . . . . . . . . . .

Issues . . . . . . . . . . . . . . . . . . . . 7.7.1 “Do People Make Relative Judgments” Should They See Their Previous Ratings? . . . . . . . . . . . . . . . 7.7.2 Should Category Rating Scales Be Assigned Integer Numbers in Data Tabulation? Are They Interval Scales? . . . . . . . . . 7.7.3 Is Magnitude Estimation a Ratio Scale or Simply a Scale with Ratio Instructions? . . . . . . . . . . . . . 7.7.4 What is a “Valid” Scale? . . . . . . . . 7.8 Conclusions . . . . . . . . . . . . . . . . . Appendix 1: Derivation of Thurstonian-Scale Values for the 9-Point Scale . . . . . . . . . . . . . Appendix 2: Construction of Labeled Magnitude Scales . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . 7.7

149 151 152 152 155 156 158 159 159 159 159 160 160 160 162 164 165 166 167

168

168

169

169 169 170 171 172 174

7.1 Introduction People make changes in their behavior all the time based on sensory experience and very often this involves a judgment of how strong or weak something feels. We add more sugar to our coffee if it is not sweet enough. We adjust the thermostat in our home if it is too cold or too hot. If a closet is too dark to find your shoes you turn the light on. We apply more force to

H.T. Lawless, H. Heymann, Sensory Evaluation of Food, Food Science Text Series, DOI 10.1007/978-1-4419-6488-5_7, © Springer Science+Business Media, LLC 2010

149

150

chew a tough piece of meat if it will not disintegrate to allow swallowing. These behavioral decisions seem automatic and do not require a numerical response. But the same kinds of experiences can be evaluated with a response that indicates the strength of the sensation. What was subjective and private becomes public data. The data are quantitative. This is the basis of scaling. The methods of scaling involve the application of numbers to quantify sensory experiences. It is through this process of numerification that sensory evaluation becomes a quantitative science subject to statistical analysis, modeling, pdiction, and hard theory. However, as noted in the quote above, in the practical application of sensory test methods, the nature of this process of assigning numbers to experiences is rarely questioned and deserves scrutiny. Clearly numbers can be assigned to sensations by a panelist in a variety of ways, some by mere categorization, or by ranking or in ways that attempt to reflect the intensity of sensory experience. This chapter will illustrate these techniques and discuss the arguments that have been raised to substantiate the use of different quantification procedures. Scaling involves sensing a product or stimulus and then generating a response that reflects how the person perceives the intensity or strength of one or more of the sensations generated by that product. This process is based on a psychophysical model (see Chapter 2). The psychophysical model states that as the physical strength of the stimulus increases (e.g., the energy of a light or sound or the concentration of a chemical stimulus) the sensation will increase in some orderly way. Furthermore, panelists are capable of generating different responses to indicate these changes in what they experience. Thus a systematic relationship can be modeled of how physical changes in the real world result in changing sensations. Scaling is a tool used for showing differences and degrees of difference among products. These differences are usually above the threshold level or justnoticeable difference. If the products are very similar and there is a question of whether there is any difference at all, the discrimination testing methods are more suitable (Chambers and Wolf, 1996). Scaling is usually done in one of the two scenarios. In the first, untrained observers are asked to give responses to reflect changes in intensity and it is psumed that (1) they understand the attribute they are asked to scale, e.g., sweetness and (2) there is no need to train or calibrate them to use the scale. This is the kind of scaling done to study a

7

Scaling

Fig. 7.1 The two processes involved in scaling. The first, physiological process is the psychophysical translation of energy in the outside world into sensation, i.e., conscious experience. The second is the translation of that experience into some response. The psychophysical process can be modified by physiological processes such as adaptation and masking. The judgment function can be modified by cognitive processes such as contextual effects, number usage, and other response biases.

7.2 Some Theory

7.2 Some Theory Measurement theory tells us that numbers can be assigned to items in different ways. This distinction was popularized by S. S. Stevens, the major proponent of magnitude estimation (1951). At least four ways of assigning numbers to events exist in common usage. These are referred to as nominal scaling, ordinal scaling, interval scaling, and ratio scaling. In nominal scaling, numbers are assigned to events merely as labels. Thus gender may be coded as a “dummy variable” in statistical analysis by assigning a zero to males and a one to females; no assumption is made that these numbers reflect any ordered property

151

of the sexes. They merely serve as convenient labels. The meals at which a food might be eaten could be coded with numbers as categories-one for breakfast, two for lunch, three for supper, and four for snacks. The assignment of a number for analysis is merely a label, a category or pigeonhole. The appropriate analysis of such data is to make frequency counts. The mode, the most frequent response, is used as a summary statistic for nominal data. Different frequencies of response for different products or circumstances can be compared by chi-square analysis or other nonparametric statistical methods (Siegel, 1956; see Appendix B). The only valid comparisons between inpidual items with this scale is to say whether they belong to the same category or to different ones (an equal versus not equal decision). In ordinal scaling, numbers are assigned to recognize the rank order of products with regard to some sensory property, attitude, or opinion (such as pference). In this case increasing numbers assigned to the products repsent increasing amounts or intensities of sensory experience. So a number of wines might be rank ordered for perceived sweetness or a number of fragrances rank ordered from most pferred to least pferred. In this case the numbers do not tell us anything about the relative differences among the products. We cannot draw conclusions about the degree of difference perceived nor the ratio or magnitude of difference. In an analogy to the order of runners finishing in a race, we know who placed first, second, third, etc. But this order does indicate neither the finishing distances between contestants nor the differences in their elapsed times. In general, analyses of ranked data can report medians as the summary statistic for central tendency or other percentiles to give added information. As with nominal data, nonparametric statistical analyses (see Appendix B) are appropriate when ranking is done (Siegel, 1956). The next level of scaling occurs when the subjective spacing of responses is equal, so the numbers repsent equal degrees of difference. This is called interval-level measurement. Examples in the physical sciences would be the centigrade and Fahrenheit scales of temperature. These scales have arbitrary zero points but equal pisions between values. The scales are inter-convertible through a linear transformation, for example, ◦ C = 5/9 (◦ F-32). Few scales used in sensory science have been subjected to tests that would help establish whether they achieved an interval level of measurement and yet this level is often assumed.

152

7

Scaling

instructed to assign numbers in relative proportions that reflect the strength of their sensations (Stevens, 1956). However, ratio instructions are easy to give, but whether the scale has ratio properties in reflecting a person’s actual subjective experiences is difficult to determine, if not impossible. Because of these different measurement types with different properties, the sensory professional must be careful about two things. First, statements about differences or ratios in comparing the scores for two products should not be made when the measurement level is only nominal or ordinal. Second, it is risky to use parametric statistics for measurements that reflect only frequency counts or rankings (Gaito, 1980; Townsend and Ashby, 1980). Nonparametric methods are available for statistical analyses of such data.

7.3 Common Methods of Scaling Several different scaling methods have been used to apply numbers to sensory experience. Some, like magnitude estimation, are adapted from psychophysical research, and others, like category scaling have become popular through practical application and dissemination in a wide variety of situations. This section illustrates the common techniques of category scales, line marking, and magnitude estimation. The next section discusses the less frequently used techniques of hybrid category-ratio scales, indirect scales, and ranking as alternatives. Two other methods are illustrated. Intensity matching across sensory modalities, called cross-modality matching, was an important psychophysical technique and a pcedent to some of the category-ratio scales. Finally, adjustable rating techniques in which panelists make relative placements and are able to alter their ratings are also discussed.

7.3.1 Category Scales Perhaps the oldest method of scaling involves the choice of discrete response alternatives to signify increasing sensation intensity or degrees of liking and/or pference. The alternatives may be psented in a horizontal or vertical line and may offer choices of integer numbers, simple check boxes, or word phrases. Examples of simple category scales are shown in

7.3 Common Methods of Scaling

153

A) INTENSITY 1

2

3

4

5

6

7

8

9 Strong

Weak

B) Oxidized

not noticeable trace, not sure faint slight mild moderate definite strong very strong

D) Sweetness

Extremely sweet

Not sweet at all

E) Sweetness

R

Weaker

Stronger

F) Hedonic scale for children

Super Good

Really Good

Good

Maybe Good or Maybe Bad

Bad

Really Bad

Super Bad

Fig. 7.2 Examples of category scales. (a) a simple integer scale for sensation strength (after Lawless and Malone, 1986b); (b) a verbal scale for degree of oxidized flavor (after Mecredy et al., 1974; (c) a verbal scale for degree of difference from some reference or control sample (after Aust et al., 1985), (d) a simple

check-box scale for perceived intensity; (e) a simple check-box scale for difference in intensity from some reference sample, marked R (after Stoer and Lawless, 1993); (f) a facial scale suitable for use with children, after Chen et al. (1996).

Fig. 7.2. The job of the consumer or panelist is to choose the alternative that best repsents their reaction or sensation. In a category scale the number of alternative responses is limited. Seven to 15 categories are commonly used for intensity scaling depending upon the application and the number of gradations that the panelists are able to distinguish in the products. As panel training progresses, perceptual discrimination

of intensity levels will often improve and more scale points may be added to allow the panel to make finer distinctions. A key idea is to psent an easily understandable word like “sweetness” and ask the participant to evaluate the perceived intensity of that attribute. A second important factor concerns the verbal labels that appear along the alternatives. At the very least, the low and high ends of the scale must be labeled

154

with words that make sense, e.g., “not sweet at all” to “extremely sweet.” A wide variety of these scales have been used. A common version is to allow integer numerical responses of approximately nine points (e.g., Lawless and Malone, 1986a, b). Further gradations may be allowed. For example, Winakor et al. (1980) allowed options from 1 to 99 in rating attributes of fabric handfeel. In the Spectrum method (Meilgaard et al., 2006) a 15-point category scale is used, but allows intermediate points in tenths, rendering it (at least in theory) a 150point scale. In hedonic or affective testing, a bipolar scale is common, with a zero or neutral point of opinion at the center (Peryam and Girardot, 1952). These are often shorter than the intensity scales. For example, in the “smiling face” scale used with children, only three options may be used for very young respondents (Birch et al., 1980, 1982), although with older children as many as nine points may be used (Chen et al., 1996; Kroll, 1990). Lately there has been a move away from using labels or integers, in case these may be biasing to subjects. People seem to have favorite numbers or tendencies to use some numbers more than others (e.g., Giovanni and Pangborn, 1983). A solution to this problem is to use an unlabeled check-box scale as shown in Fig. 7.2. In early applications of category scaling, the procedure specifically instructed subjects to use the categories to repsent equal spacing. They might also be instructed to distribute their judgments over the available scale range, so the strongest stimulus was rated at the highest category and the weakest stimulus at the lowest category. This use of such explicit instructions surfaces from time to time. An example is in Anderson’s (1974) recommendation to show the subject specific examples of bracketing stimuli that are above and below the anticipated range of items in the set to be judged. A related method is the relative scaling procedure of Gay and Mead (1992) in which subjects place the highest and lowest stimuli at the scale endpoints (discussed below). The fact that there is an upper boundary to the allowable numbers in a category scaling task may facilitate the achievement of a linear interval scale (Banks and Coleman, 1981). A related issue concerns what kind of experience the high end anchor refers to. Muñoz and Civille (1998) pointed out that for descriptive analysis panels, the high end-anchor phrase could refer to different situations. For example, is the term “extremely sweet”

7

Scaling

7.3 Common Methods of Scaling

7.3.2 Line Scaling

155

a) very poor

excellent

AROMA

b) moderate

weak

strong

Pepper Heat

c)

O threshold

slight

moderate

strong

Sweetness

d) weaker

reference

stronger

overall opinion

e) dislike moderately

f)

X

least liked

neither

like moderately

X most liked

Fig. 7.3 Examples of line-marking scales: (a) with endpoints labeled (after Baten, 1946); (b) with indented “goal posts” (after Mecredy et al., 1974); (c) with additional points labeled as in ASTM procedure E-1083 (ASTM, 2008b); (d) a line for ratings relative to a reference as in Stoer and Lawless (1993); (e) hedonic scaling using a line; (f) the adjustable scale of Gay and Mead (1992) as pictured by Villanueva and D Silva (2009).

An important historical trend to use lines in descriptive analysis was instrumental in the popularization of line marking. Stone et al. (1974) recommended the use of line marking for Quantitative Descriptive Analysis (QDA), then a relatively new approach to specifying the intensities of all important sensory attributes. It was important to have a scaling method in QDA that approximated an interval scale, as analysis of variance was to become the standard statistical technique for comparing products in descriptive analysis. The justification for the application in QDA appears to rest on the pvious findings of Baten regarding the sensitivity of the method and the writings of Norman Anderson on functional measurement theory (Anderson, 1974). In his approach, Anderson used indented end anchors (see Weiss, 1972, for another example). Anderson also showed his subjects examples of the high and

156

7

Scaling

with a cursor that moves up and down via the computer mouse. Time-intensity methods are reviewed more fully in Chapter 8.

7.3.3 Magnitude Estimation 7.3.3.1 The Basic Techniques A popular technique for scaling in psychophysical studies has been the method of magnitude estimation. In this procedure, the respondent is instructed to assign numbers to sensations in proportion to how strong the sensation seems. Specifically, the ratios between the numbers are supposed to reflect the ratios of sensation magnitudes that have been experienced. For example, if product A is given the value of 20 for sweetness intensity and product B seems twice as sweet, B is given a magnitude estimate of 40. The two critical parts of the technique are the instructions given to the participant and the techniques for data analysis. Two primary variations of magnitude estimation have been used. In one method, a standard stimulus is given to the subject as a reference and that standard is assigned a fixed value such as 10. All subsequent stimuli are rated relative to this standard, sometimes called a “modulus.” It is often easier for panelists if the reference (i.e., the item used as the modulus) is chosen from somewhere near the middle of the intensity range. In the other variation of magnitude estimation, no standard stimulus is given and the participant is free to choose any number he or she wishes for the first sample. All samples are then rated relative to this first intensity, although in practice people probably “chain” their ratings to the most recent items in the series. Because people can choose different ranges of numbers in this “non-modulus” magnitude estimation, the data have to be treated to bring all judgments into the same range, an extra step in the analysis. Variations on magnitude estimation and guidelines for data analysis are found in ASTM Standard Test Method E 1697-05 (ASTM, 2008a). In the psychophysical laboratory, where magnitude estimation has found its primary usage, generally only one attribute is rated at a time. However, rating multiple attributes or profiling has been used in taste studies (McBurney and Bartoshuk, 1973; McBurney and Shick, 1971; McBurney et al., 1972) and this can

7.3 Common Methods of Scaling

naturally be extended to foods with multiple taste and aromatic attributes. Magnitude estimation has not been used very often for descriptive analysis, but in principle there is no reason why it could not be used for that purpose. Participants should be cautioned to avoid falling into pvious habits of using bounded category scales, e.g., limited ranges of numbers from zero to ten. This may be a difficult problem with pviously trained panels that have used a different scaling method, as people like to stick with a method they know and feel comfortable with. Panelists who show such behavior may not understand the ratio nature of the instructions. It is sometimes useful with a new panelist to have the participant perform a warm-up task to make sure they understand the scaling instructions. The warm-up task can involve estimation of the size or area of different geometric ps (Meilgaard et al., 2006) or the length of lines (McBurney et al., 1972). Sometimes it is desired to have panelists rate multiple attributes at the same time or to break down overall intensity into specific qualities. If this kind of “profiling” is needed, the geometric ps can include areas with different shading or the lines can be differently colored. A practice task is highly recommended so that the sensory scientist can check on whether the participant understands the task. Values of zero are allowed in this method as some of the products may in fact have or no sensation for a given attribute (like no sweetness in our example). Of course, the rating of zero should not be used for the reference material. While the value of zero is consistent with common sense for products with no sensation of some attributes, it does complicate the data analysis as discussed below. 7.3.3.2 Instructions The visual appearance of the ballot in magnitude estimation is not critical; it is the instructions and the participant’s comphension of the ratio nature of the judgments that are important. Some ballots even allow the subject/participant to view all pvious ratings. Here are sample instructions for the use of magnitude estimation with a reference sample or modulus with a fixed number assigned to it: Please taste the first sample and note its sweetness. This sample is given the value of “10” for its sweetness

157 intensity. Please rate all other samples in proportion to this reference. For example, if the next sample is twice as sweet, assign it a value of “20”, if half as sweet, assign it a value of “5” and if 3.5 times as sweet, assign in a value of 35. In other words, rate the sweetness intensity so that your numbers repsent the ratios among the intensities of sweetness. You may use any positive numbers including fractions and decimals.

The other major variation on this method uses no reference. In this case the instructions may read as follows: Please taste the first sample and note its sweetness. Please rate all other samples relative to this reference, applying numbers to the samples to repsent the ratios of sweetness intensity among the samples. For example, if the next sample was twice as sweet, you would give it a number twice as big as the rating assigned to the first sample, if half as sweet, assign it a number half as big and if 3.5 times as sweet, assign it a number 3.5 times as big. You may use any positive numbers including fractions and decimals.

7.3.3.3 Data Treatment In non-modulus methods, participants will generally choose some range of numbers they feel comfortable with. The ASTM procedure suggests having them pick a value between 30 and 100 for the first sample, and avoiding any number that seems small. If participants are allowed to choose their own number range, it becomes necessary to re-scale each inpidual’s data to bring them into a common range before statistical analysis (Lane et al., 1961). This will pvent subjects who choose very large numbers from having undue influence on measures of central tendency (means) and in statistical tests. This rescaling process has been referred to as “normalizing” (ASTM, 2008a) although it has nothing to do with the normal distribution or Z-scores. A common method for rescaling proceeds as follows: (1) Calculate the geometric mean of each inpidual’s ratings across their data set. (2) Calculate the geometric mean of the entire data set (of all subjects combined). (3) For each subject, construct a ratio of the grand geometric mean of the entire data set to each person’s geometric mean. The value of this ratio provides a post hoc inpidual rescaling factor for each subject. In place of the grand geometric mean,

158

any positive numerator may also be chosen in constructing this factor, e.g., a value of 100. (4) Multiply each data point for a given person by their inpidual rescaling factor. Do this for all participants using their own inpidual re-scaling factors. These re-scaled data are then analyzed. Note that due to the extra data treatment step in this method, it is simpler to use the modulus-based variation with a standard reference item. Magnitude estimation data are often transformed to logs before data analysis (Butler et al., 1987; Lawless, 1989). This is done primarily because the data tend to be positively skewed or log-normally distributed. There tends to be some high outlying values for any given sample. Perhaps this is not surprising because the scale is open-ended at the top, and bounded by zero at the bottom. Transformation into log data and/or taking geometric means psents some problems, however, when the data contain zeros. The log of zero is undefined. Any attempt to take a geometric mean by calculating the product of N items will yield a zero on multiplying. Several approaches have been taken to this problem. One is to assign a small positive value to any zeros in the data, perhaps one-half of the smallest rating given by a subject (ASTM, 2008a). The resulting analysis, however, will be influenced by this choice. Another approach is to use the median judgments in constructing the normalization factor for non-modulus methods. The median is less influenced by the high outliers in the data than the arithmetic mean.

7.3.3.4 Applications For practical purposes, the method of magnitude estimation may be used with trained panels, consumers, and even children (Collins and Gescheider, 1989). However, the data do tend to be a bit more variable than other bounded scaling methods, especially in the hands of untrained consumers (Lawless and Malone, 1986b). The unbounded nature of the scale may make it especially well suited to sensory attributes where an upper boundary might impose restrictions on the panelists’ ability to differentiate very intense sensory experiences in their ratings. For example, irritative or painful sensations such as chili pepper intensity might all be rated near the upper bound of a category scale

7

Scaling

of intensity, but an open-ended magnitude estimation procedure would allow panelists more freedom to differentiate and report variations among very intense sensations. With hedonic scaling of likes and dislikes, there is an additional decision in using magnitude estimation scaling. Two options have been adopted in using this technique, one employing a single continuum or unipolar scale for amount of liking and the other applying a bipolar scale with positive and negative numbers plus a neutral point (Pearce et al., 1986). In bipolar magnitude scaling of likes and dislikes, positive and negative numbers are allowed in order to signify ratios or proportions of both liking and disliking (e.g., Vickers, 1983). An alternative to positives and negatives is to have the respondent merely indicate whether the number repsents liking or disliking (Pearce et al., 1986). In unipolar magnitude estimation only positive numbers (and sometimes zeros) are allowed, with the lower end of the scale repsenting no liking and higher numbers given to repsent increasing proportions of liking (Giovanni and Pangborn, 1983; Moskowitz and Sidel, 1971). It is questionable whether a unipolar scale is a sensible response task for the participant, as it does not recognize the fact that a neutral hedonic response may occur, and that there are clearly two modes of reaction, one for liking and one for disliking. If one assumes that all items are on one side of the hedonic continuum-either all liked to varying degrees or all disliked to varying degrees then the one-sided scale makes sense. However, it is a rare situation with foods or consumer product testing in which at least some indifference or change of opinion was not visible in at least some respondents. So a bipolar scale fits common sense.

7.4 Recommended Practice and Practical Guidelines Both line scales and category scales may be used effectively in sensory testing and consumer work. So we will not expend much effort in recommending one of these two common techniques over another. Some practical concerns are given below to help the student or practitioner avoid some potential problems. The category-ratio or labeled magnitude scales may facilitate comparisons among different groups, and this issue is discussed below in Section 7.5.2.

7.4 Recommended Practice and Practical Guidelines

7.4.1 Rule 1: Provide Sufficient Alternatives One major concern is to provide sufficient alternatives to repsent the distinctions that are possible by panelists (Cox, 1980). In other words, a simple 3-point scale may not suffice if the panel is highly trained and capable of distinguishing among many levels of the stimuli. This is illustrated in the case of the flavor profile scale, which began with five points to repsent no sensation, threshold sensation, weak, moderate, and strong (Caul, 1957). It was soon discovered that additional intermediate points were desirable for panelists, especially in the middle range of the scale where many products would be found. However, there is a law of diminishing returns in allowing too many scale points-further elaboration allows better differentiation of products up to a point and then the gains diminish as the additional response choices merely capture random error variation (Bendig and Hughes, 1953). A related concern is the tendency, especially in consumer work, to simplify the scale by eliminating options or truncating endpoints. This brings in the danger caused by end-use avoidance. Some respondents may be reluctant to use the end categories, just in case a stronger or weaker item may be psented later in the test. So there is some natural human tendency to avoid the end categories. Truncating a 9-point scale to a 7-point scale may leave the evaluator with what is functionally only a 5-point scale for all practical purposes. So it is best to avoid this tendency to truncate scales in experimental planning.

7.4.2 Rule 2: The Attribute Must Be Understood Intensity ratings must be collected on an attribute which the participants understand and about which they have a general consensus and agreement as to its meaning. Terms like sweetness are almost universal but a term like “green aroma” might be interpted in different ways. In the case of a descriptive panel, a good deal of effort may be directed at using reference standards to illustrate what is meant by a specific term. In the case of consumer work, such training is not done, so if any intensity ratings are collected, they must be about simple terms about which people generally

159

agree. Bear in mind that most early psychophysics was done on simple attributes like the loudness of a sound or the heaviness of a weight. In the chemical senses, with their perse types of sensory qualities and fuzzy consumer vocabulary, this is not so straightforward. Other problems to avoid include mixing sensation intensity (strength) and hedonics (liking), except in the just-right scale where this is explicit. An example of a hedonically loaded sensory term is the adjective “fresh.” Whatever this means to consumers, it is a poor choice for a descriptive scale because it is both vague and connotes some degree of liking or goodness to most people. Vague terms are simply not actionable when it comes to giving feedback to product developers about what needs to be fixed. Another such vague term that is popular in consumer studies is “natural.” Even though consumers might be able to score products on some unknown basis using this word, the information is not useful as it does not tell formulators what to change if a product scores low. A similar problem arises with attempting to scale “overall quality.” Unless quality has been very carefully defined, it cannot be scaled.

7.4.3 Rule 3: The Anchor Words Should Make Sense In setting up the scales for descriptive analysis or for a consumer test, the panel leader should carefully consider the nature of the verbal end anchors for each scale as well as any intermediate anchors that may be needed. Should the scale be anchored from “very weak” to “very strong” or will there be cases in which the sensory attribute is simply not psent? If so, it makes sense to verbally anchor the bottom of the scale with “not at all” or “none.” For example, a sweetness scale could be anchored with “not at all sweet” and a scale for “degree of oral irritation” could be anchored using “none.”

160

7.4.5 A Warning: Grading and Scoring are Not Scaling In some cases pseudo-numerical scales have been set up to resemble category scales, but the examples cut

7

Scaling

across different sensory experiences, mixing qualities. An example is the pseudo-scale used for baked products, where the number 10 is assigned for perfect texture, 8 for slight dryness, 6 for gumminess, and 4 if very dry (AACC, 1986). Gumminess and dryness are two separate attributes and should be scaled as such. This is also an example of quality grading, which is not a true scaling procedure. When the numbers shift among different sensory qualities, this violates the psychophysical model for scaling the intensity of a single attribute. Although numbers may be applied to the grades, they cannot be treated statistically, as the average of “very dry ” (4) and slightly dry (8) is not gummy (6) (see Pangborn and Dunkley, 1964, for a critique of this in the dairy grading arena). The numbers in a quality-grading scheme do not repsent any kind of unitary psychophysical continuum.

7.5 Variations-Other Scaling Techniques An important idea in scaling theory is the notion that people may have a general idea of how weak or strong sensations are, and that they can compare different attributes of a product for their relative strength, even across different sensory modalities. So, for example, someone could legitimately say that this product tastes much more salty than it is sweet. Or that the trumpets are twice as loud as the flutes in a certain passage in a symphony. Given that this notion is correct, people would seem to have a general internal scale for the strength of sensations. This idea forms the basis for several scaling methods. It permits the comparison of different sensations cross-referenced by their numerical ratings, and even can be used to compare word responses. Methods derived from this idea are discussed next.

7.5.1 Cross-Modal Matches and Variations on Magnitude Estimation The method of magnitude estimation has a basis in earlier work such as fractionation methods and the method of sense ratios in the older literature (Boring,

7.5 Variations-Other Scaling Techniques

161

1942) where people would be asked to set one stimulus in a given sensation ratio to another. The notion of allowing any numbers to be generated in response to stimuli, rather than adjusting stimuli to repsent fixed numbers appeared somewhat later (Moskowitz, 1971; Richardson and Ross, 1930; Stevens, 1956). An important outcome of these studies was the finding that the resulting psychophysical function generally conformed to a power law of the following form:

brightness of the lights about in the same proportions as the loudness of the sounds. Stevens (1969) proposed that these experiments could validate the power law, since the exponents of the cross-modality matching function could be pdicted from the exponents derived from separate scaling experiments. Consider the following example: For one sensory attribute (using the log transform, Eq. (7.2)),

R = kI n

log R1 = n1 log I1 + log k1

(7.1)

and for a second sensory attribute

or after log transformation: log(R) = n log(I) + log(k)

(7.3)

(7.2)

where R was the response, e.g., perceived loudness (mean or geometric mean of the data) and I was the physical stimulus intensity, e.g., sound pssure, and k was a constant of proportionality that depends upon the units of measurement. The important characteristic value of any sensory system was the value for n, the exponent of the power function or slope of the straight line in a log-log plot (Stevens, 1957). The validity of magnitude estimation then came to hang on the validity of the power law-the methods and resulting functions formed an internally consistent theoretical system. Stevens also viewed the method as providing a direct window into sensation magnitude and did not question the idea that these numbers generated by subjects might be biased in some way. However, the generation of responses is properly viewed as combining at least two processes, the psychophysical transformation of energy into conscious sensation and the application of numbers to those sensations. This response process was not given due consideration in the early magnitude estimation work. Responses as numbers can exhibit nonlinear transformations of sensation (Banks and Coleman, 1981; Curtis et al., 1968) so the notion of a direct translation from sensation to ratings is certainly a dangerous oversimplification. Ratio-type instructions have been applied to other techniques as well as to magnitude estimation. A historically important psychophysical technique was that of cross-modality matching, in which the sensation levels or ratios would be matched in two sensory continua such as loudness and brightness. One continuum would be adjusted by the experimenter and the other by the subject. For example, one would try to make the

log R2 = n2 log I2 + log k2

(7.4)

Setting R1 = R2 in cross-modality matching gives n1 log I1 + log k1 = n2 log I2 + log k2

(7.5)

and rearranging, log I1 = (n2 /n1 ) log I2 + a constant

(7.6)

If one plots log I1 against log I2 from a crossmodality matching task, the slope of the function can be pdicted from the ratio of the slopes of the inpidual exponents (i.e., n2 /n1 , which you can derive from two separate magnitude estimation tasks). This pdiction holds rather well for a large number of compared sensory continua (Steven, 1969). However, whether it actually provides a validation for the power law or for magnitude estimation has been questioned (e.g., Ekman, 1964). For practical purposes, it is instructive that people can actually take disparate sensory continua and compare them using some generalized notion of sensory intensity. This is one of the underpinnings of the use of a universal scale in the Spectrum descriptive procedure (Meilgaard et al., 2006). In that method, different attributes are rated on 15-point scales that can (in theory) be meaningfully compared. In other words, a 12 in sweetness is twice as intense a sensation as a 6 in saltiness. Such comparisons seem to makes sense for tastes and flavors but may not cut across all other modalities. For example, it might seem less sensible to compare the rating given for the amount of chocolate chips in a cookie to the rating given for the cookie’s hardness-these seem like quite different experiences to quantify.

162

7

Scaling

or line scale? The idea of scaling word phrases takes shape in the labeled magnitude scales discussed next.

7.5.2 Category-Ratio (Labeled Magnitude) Scales A group of hybrid techniques for scaling has recently enjoyed some popularity in the study of taste and smell, for hedonic measurement and other applications. One of the problems with magnitude estimation data is that it does not tell in any absolute sense whether sensations are weak or strong, only giving the ratios among them. This group of scales attempts to provide ratio information, but combines it with common verbal descriptors along a line scale to provide a simple frame of reference. They are referred to as category-ratio scales, or more recently, labeled magnitude scales. They all involve a horizontal or vertical line with deliberately spaced labels and the panelists’ task is to make a mark somewhere along the line to indicate the strength of their perception or strength of their likes or dislikes. In general, these labeled line scales give data that are consistent with those from magnitude estimation (Green et al., 1993). An unusual characteristic of these scales is the verbal high endanchor phrase, which often refers to the “strongest imaginable.” The technique is based on early work by Borg and colleagues, primarily in the realm of perceived physical exertion (Borg, 1982, 1990; see Green et al., 1993). In developing this scale, Borg assumed that the semantic descriptors could be placed on a ratio scale and that they defined the level of perceptual intensity and that all inpiduals experienced the same perceptual range. Borg suggested that for perceived exertion, the maximal sensation is roughly equivalent across people for this sensory “modality” (Marks et al., 1983). For example, it is conceivable that riding a bicycle to the point of physical exhaustion produces a similar sensory experience for most people. So the scale came to have the highest label referring to the strongest sensation imaginable. This led to the development of the labeled magnitude scale (LMS) shown in Fig. 7.4. It is truly a hybrid method since the response is a vertical linemarking task but verbal anchors are spaced according to calibration using ratio-scaling instructions (Green

7.5 Variations-Other Scaling Techniques

Strongest imaginable

Very strong

(95.5)

(50.12)

Strong

(33.1)

Moderate

(16.2)

Weak Barely detectable

163

LMS

(5.8) (1.4)

Fig. 7.4 The labeled magnitude scale (LMS) of Green et al. (1993).

2007; Lawless et al., 2010a, b, c). A growing number of similar scales have been developed for various applications including oral pleasantness/unpleasantness (the “OPUS” scale, Guest et al., 2007), perceived satiety (the “SLIM” scale, Cardello et al., 2005), clothing fabric comfort (the “CALM” scale, Cardello et al., 2003), and odor dissimilarity (Kurtz et al., 2000). All of these scales depend upon a ratio scaling task to determine the spacing of the verbal descriptors and almost all use a Borg-type high end-anchor phrase. Others will surely be developed. Instructions to participants have differed in the use of these scales. In the first application of the LMS, Green et al. (1993) instructed subjects to first choose the most appropriate verbal descriptor, and then to “fine tune” their judgment by placing a mark on the line between that descriptor and the next most appropriate one. In current practice less emphasis may be placed on the consideration of the verbal labels and instructions may be given to simply make a mark “anywhere” on the line. A common observation with the hedonic versions of the scale is that some panelists will mark at or very near a verbal descriptor, seeming to use it as a category scale (Cardello et al., 2008; Lawless et al., 2010a). The proportion of people displaying this behavior may depend upon the physical length of the line (and not the instructions or examples that may be shown) (Lawless et al., 2010b). Results may depend in part on the nature of the high end-anchor example and the frame of reference of the subject in terms of the sensory modality they are thinking about. Green et al. (1996) studied the application of the LMS to taste and odor, using descriptors for the upper bound as “strongest imaginable” taste, smell, sweetness, etc. Steeper functions (a smaller response range) were obtained when mentioning inpidual taste qualities. This appears to be due to the omission of painful experiences (e.g., the “burn of hot peppers”) from the frame of reference when sensations were scaled relative to only taste. The steepening of the functions for the truncated frame of reference is consistent with the idea that subjects expanded their range of numbers as seen in other scaling experiments (e.g., Lawless and Malone, 1986b). The fact that subjects appear to adjust their perceptual range depending on instructions or frame of reference suggests that the scales have relative and not absolute properties, like most other scaling methods. Cardello et al. (2008) showed that the hedonic version of the scale (the LAM

164 Greatest imaginable like

Most imaginable

(+100)

Scaling (+100)

Like Extremely

(+74.2)

extremely

(+75.2)

Like Very Much

(+56.1)

very

(+56.1)

Like Moderately

(+36.2)

moderately

(+33.9)

Like Slightly

(+11.2)

slightly weakly

(+19.7) (+15.0)

PLEASANT

Fig. 7.5 Affective labeled magnitude scales, including the LAM scale (Cardello and Schutz, 2004) and the OPUS scale (Guest et al., 2007).

7

barely detectable Neither like nor dislike

(+6.5) (0)

(0)

(+6.5)

(-10.6)

barely detectable weakly slightly

(+15.0) (+19.7)

Dislike moderately

(-31.9)

moderately

(+33.9)

Dislike Very much

(-55.5)

very

(+56.1)

Dislike extremely

(-75.5)

extremely

(+75.2)

Greatest imaginable dislike

(-100)

LAM scale) will also show such range effects. A compssed range of responses is obtained when the frame of reference is greatest imaginable like (dislike) for an “experience of any kind” rather than something more delimited like “foods and beverages.” Apparently the compssion is not very detrimental to the ability of the LAM scale to differentiate products (Cardello et al., 2008, but see also Lawless et al., 2010a). Is this a suitable method for cross-subject comparisons? To the extent that Borg’s assumptions of common perceptual range and the similarity of the high end-anchor experience among people are true, the method might provide one approach to valid comparisons of the ratings among different respondents. This would facilitate comparisons of clinical groups or patients with sensory disorders or genetically different inpiduals such as anosmics, PTC/PROP taster groups. Bartoshuk and colleagues (1999, 2003,

UNPLEASANT

Dislike slightly

Most imaginable

(-100)

OPUS 2004a, b, 2006) have argued that the labeled magnitude scales should anchor their endpoints to “sensations of any kind” as such a reference experience would allow different inpiduals to use the scale in similar ways/and thus facilitate inter-inpidual comparisons. Scales with this kind of high end anchor have been termed “generalized” labeled magnitude scales (or gLMS). However, the sensory evaluation practitioner should be aware of the compssion effects that can occur with this kind of scale, which could potentially lead to lessened differentiation among products.

7.5.3 Adjustable Rating Techniques: Relative Scaling A few methods have been tried that allow consumers or panelists to change their ratings. An example is

7.5 Variations-Other Scaling Techniques

165

7.5.4 Ranking Another alternative to traditional scaling is the use of ranking procedures. Ranking is simply ordering the products from weakest to strongest on the stated

166

7.5.5 Indirect Scales A conservative approach to scaling is to use the variance in the data as units of measurement, rather than the numbers taken at face value. For example, we could

7

Scaling

ask how many standard deviations apart are the mean values for two products. This is a different approach to measurement than simply asking how many scale units separate the means on the response scale. On a 9-point scale, one product may receive mean rating of seven, and another nine, making the difference two scale units. If the pooled standard deviation is two units, however, they would only be one unit apart on a variability-based scale. As one example, Conner and Booth (1988) used both the slope and the variance of functions from just-right scales to derive a “tolerance discrimination ratio.” This ratio repsents a measure of the degree of difference along a physical continuum (such as concentration of sugar in lime drink) that observers find to make a meaningful change in their ratings of difference-from-ideal (or just-right). This is analogous to finding the size of a just-noticeable difference, but translated into the realm of hedonic scaling. Their insight was that it is not only the slope of the line that is important in determining this tolerance or liking-discrimination function, but also the variance around that function. Variability-based scales are the basis for scaling in Thurstone’s models for comparative judgment (Thurstone, 1927) and its extension into determining distances between category boundaries (Edwards, 1952). Since the scale values can be found from choice experiments as well as rating experiments, the technique is quite flexible. How this type of scaling can be applied to rating data is discussed below in the derivation of the 9-point hedonic scale words. When the scale values are derived from a choice method like the triangle test or paired comparison method, this is sometimes called “indirect scaling” (Baird and Noma, 1978; Jones, 1974). The basic operation in Thurstonian scaling of choice data is to convert the proportion correct in a choice experiment (or simply the proportions in a two-tailed test like paired pference) to Z-scores. The exact derivation depends upon the type of test (e.g., triangle versus 3-AFC) and the cognitive strategy used by the subject. Tables for Thurstonian scale values from various tests such as the triangle test were given by Frijters et al. (1980) and Bi (2006) and some tables are given in the Appendix. Mathematical details of Thurstonian scaling are discussed in Chapter 5. Deriving measures of sensory differences in such indirect ways psents several problems in applied sensory evaluation so the method has not been widely used. The first problem is one of economy in data

7.6 Comparing Methods: What is a Good Scale?

collection. Each difference score is derived from a separate discrimination experiment such as a paired comparison test. Thus many subjects must be tested to get a good estimate of the proportion of choice, and this yields just one scale value. In direct scaling, each participant gives at least one data point for each item tasted. Direct scaling allows for easy comparisons among multiple products, while the discrimination test must be done on one pair at a time. Thus the methods of indirect scaling are not cost-efficient. A second problem can occur if the products are too clearly different on the attribute in question, because then the proportion correct will approach 100%. At that point the scale value is undefined as they are some unknown number of standard deviations apart. So the method only works when there are small differences and some confusability of the items. In a study of many products, however, it is sometimes possible to compare only adjacent or similar items, e.g., products that differ in small degrees of some ingredient or process variable. This approach was taken by Yamaguchi (1967) in examining the synergistic taste combination of monosodium glutamate and disodium 5 inosinate. Many different levels of the two ingredients were tasted, but because the differences between some levels were quite apparent, an incomplete design was used in which only three adjacent levels were compared. Other applications have also used this notion of variability as a yardstick for sensory difference or sensation intensity. The original approach of Fechner in constructing a psychophysical function was to accumulate the difference thresholds or just-noticeable differences (JNDs) in order to construct the log function of psychophysical sensory intensity (Boring, 1942; Jones, 1974). McBride (1983a, b) examined whether JND-based scales might give similar results to category scales for taste intensity. Both types of scaling yielded similar results, perhaps not surprising since both tend to conform to log functions. In a study of children’s pferences for different odors Engen (1974) used a paired pference paradigm, which was well suited to the abilities of young children to respond in a judgment task. He then converted the paired pference proportions to Thurstonian scale values via Z-scores and was able to show that the hedonic range of children was smaller than that of adults. Another example of choice data that can be converted to scale values is best-worst scaling, in which

167

a consumer is asked to choose the best liked and least liked samples from a set of three or more items (Jaeger et al., 2008). With three products, it can be considered a form of a ranking task. When applied to sensory intensity, this is sometimes known as maximum-difference or “max-diff.” Best-worst scaling is also discussed in Section 13.7. Simple difference scores may be calculated based on the number of times an item is called best versus worst and these scores are supposed to have interval properties. If a multinomial logistic regression is performed on the data, they are theorized to have true ratio properties (Finn and Louviere, 1992). A practical problem with the method, however, is that so many products must be tasted and compared, rendering it difficult to perform with foods (Jaeger and Cardello, 2009). The sensory professional should bear in mind that in spite of their theoretical sophistication, the indirect methods are based on variability as the main determinant of degree of difference. Thus any influence, which increases variability, will tend to decrease the measured differences among products. In the wellcontrolled psychophysical experiment under constant standard conditions across sessions and days, this may not be important-the primary variability lies in the resolving power of the participant (and secondarily in the sample products). But in drawing conclusions across different days, batches, panels, factories, and such, one has a less pure situation to consider. Whether one considers the Thurstonian-type indirect measures comparable across different conditions depends upon the control of extraneous variation.

7.6 Comparing Methods: What is a Good Scale? A large number of empirical studies have been conducted comparing the results using different scaling methods (e.g., Birnbaum, 1982; Giovanni and Pangborn, 1983; Hein et al., 2008; Jaeger and Cardello, 2009; Lawless and Malone, 1986a, b; Lawless et al., 2010a; Marks et al., 1983; Moskowitz and Sidel, 1971; Pearce et al., 1986; Piggot and Harper, 1975; Shand et al., 1985; Vickers, 1983; Villanueva and Da Silva, 2009; Villanueva et al., 2005). Because scaling data are often used to identify differences between products, the ability to detect differences is one important

168

practical criterion for how useful a scaling method can be (Moskowitz and Sidel, 1971). A related criterion is the degree of error variance or similar measures such as size of standard deviations or coefficients of variation. Obviously, a scaling method with low interinpidual variability will result in more sensitive tests, more significant differences, and lower risk of Type II error (missing a true difference). A related issue is the reliability of the procedure. Similar results should be obtained upon repeated experimentation. Other practical considerations are important as well. The task should be user friendly and easy to understand for all participants. Ideally, the method should be applicable to a wide range of products and questions, so that the respondent is not confused by changes in response type over a long ballot or questionnaire. If panelists are familiar with one scale type and are using it effectively, there may be some liability in trying to introduce a new or unfamiliar method. Some methods, like category scales, line scales, and magnitude estimation, can be applied to both intensity and hedonic (like-dislike) responses. The amount of time required to code, tabulate and process the information may be a concern, depending upon computer-assisted data collection and other resources available to the experimenters. As in any method, validity or accuracy are also issues. Validity can only be judged by reference to some external criterion. For hedonic scaling, one might want the method to correspond to other behaviors such as choice or consumption (Lawless et al., 2010a). A related criterion is the ability of the scale to identify or uncover consumer segments with different pferences (Villanueva and Da Silva, 2009). Given these practical considerations, we may then ask how the different scaling methods fare. Most published studies have found about equal sensitivity for the different scaling methods, provided that the methods are applied in a reasonable manner. For example, Lawless and Malone (1986a, b) performed an extensive series of studies (over 20,000 judgments) with consumers in central location tests using different sensory continua including olfaction, tactile, and visual modalities. They compared line scales, magnitude estimation, and category scales. Using the degree of statistical differentiation among products as the criterion for utility of the methods, the scales performed about equally well. A similar conclusion was reached by Shand et al. (1985) for trained panelists. There

7

Scaling

was some small tendency for magnitude estimation to be marginally more variable in the hands of consumers as opposed to college students (Lawless and Malone, 1986b). Statistical differentiation increased over replicates, as would be expected as people came to understand the range of items to be judged (see Hein et al., 2008, for another example of improvement over replication in hedonic scaling). Similar findings for magnitude estimation and category scales in terms of product differentiation were found by Moskowitz and Sidel (1971), Pearce et al. (1986), Shand et al. (1985), and Vickers (1983) although the forms of the mathematical relations to underlying physical variables was often different (Piggot and Harper, 1975). In other words, as found by Stevens and Galanter (1957) there is often a curvilinear relationship between the data from the two methods. However, this has not been universally observed and sometimes simple linear relationships have been found (Vickers, 1983). Similar results for category scales and line scales were found by Mattes and Lawless (1985). Taken together, these empirical studies paint a picture of much more parity among methods than one might suppose given the number of arguments over the validity of scaling methods in the psychological literature. With reasonable spacing of the products and some familiarization with the range to be expected, respondents will distribute their judgments across the available scale range and use the scale appropriately to differentiate the products. A reasonable summary of the literature comparing scale types is that they work about equally well to differentiate products, given a few sensible pcautions.

7.7 Issues

responses (i.e., judgments that might fall in-between categories). The line scale offers a continuously graded choice of alternatives, limited only by the measurement abilities in data tabulation. Baten also noted that the line scale seemed to facilitate a relative comparison among the products. This was probably due to his placement of the scales one above the other on the ballot, so judges could see both markings at the same time. In order to minimize such contextual effects it is now more common to remove the prior ratings for products to achieve a more independent judgment of the products. However, whether that is ever achieved in practice is open to question-humans are naturally comparative when asked to evaluate items, as discussed in Chapter 9. Furthermore, there may be potential for increased discrimination in methods like the relative positioning technique. The naturally comparative nature of human judgment may be something we could benefit from rather than trying to fight this tendency by over-calibration.

7.7.2 Should Category Rating Scales Be Assigned Integer Numbers in Data Tabulation? Are They Interval Scales? There is also a strong suspicion that many numerical scaling methods may produce only ordinal data, because the spacing between alternatives is not subjectively equal. A good example is the common marketing research scale of “excellent-very good-good- fair-poor.” The subjective spacing between these adjectives is quite uneven. The difference between two products rated good and very good is a much smaller difference than that between products rated fair and poor. However, in analysis we are often tempted to assigned numbers one through five to these categories and take means and perform statistics as if the assigned numbers reflected equal spacing. This is a ptense at best. A reasonable analysis of the 5-point excellent to poor scale is simply to count the number of respondents in each category and to compare frequencies. Sensory scientists should not assume that any scale has interval properties in spite of how easy it is to tabulate data as an integer series.

169

7.7.4 What is a “Valid” Scale? An ongoing issue in psychophysics is what kind of scale is a true reflection of the subject’s actual sensations? From this perspective, a scale is valid when the numbers generated reflect a linear translation of subjective intensity (the private experience). It is well established that category scales and magnitude estimates, when given to the same stimuli, will form a curve when plotted against one another (Stevens and Galanter, 1957). Because this is not a linear relationship, one method or the other must result from a non-linear translation of the subjective intensities of the stimuli. Therefore, by this criterion, at least one scale must be “invalid.”

170

7

Anderson (1974, 1977) proposed a functional measurement theory to address this issue. In a typical experiment, he would ask subjects to do some kind of combination task, like judging the total combined intensity of two separately psented items (or the average lightness of two gray swatches). He would set up a factorial design in which every level of one stimulus was combined with every level of the other (i.e., a complete block). When plotting the response, a family of lines would be seen when the first stimulus continuum formed the X-axis and the second formed a family of lines. Anderson argued that only when the response combination rule was additive, and the response output function was linear, would a parallel plot be obtained (i.e., there would be no significant interaction term in ANOVA). This argument is illustrated in Fig. 7.6. In his studies using simple line and category scales, parallelism was obtained in a number of studies, and thus he reasoned that magnitude estimation was invalid by this criterion. If magnitude estimation is invalid, then its derivatives such as the LMS and LAM scales are similarly suspect.

S1

p1

S2

p2

P Psychophysical process

Response output process

Integration process

Response for Combination Judgment

R

(High)

(Med.)

Levels of Stimulus or Variable 2

(Low)

(Low)

(Med.)

(High)

Levels of Stimulus or Variable 1

Fig. 7.6 The functional measurement scheme of Anderson (1974).

Scaling

Others have found support for the validity of magnitude estimation in studies of binaural loudness summation (Marks, 1978). This argument continues and is difficult to resolve. A review of the matter was published by Gescheider (1988). For the purposes of sensory evaluation, the issue is not terribly important for two reasons. First, any scale that produces statistically significant differentiation of products is a useful scale. Second, the physical ranges over which category scaling and magnitude estimation produce different results is usually quite large in any psychophysical study. In most product tests, the differences are much more subtle and generally do not span such a wide dynamic range. The issue dissolves from any practical perspective.

7.8 Conclusions Much sound and fury has been generated over the years in the psychophysical literature concerning what methods yield valid scales. For the sensory practitioner, these issues are less relevant because the scale values do not generally have any absolute meaning. They are only convenient indices of the relative intensities or appeal of different products. The degree of difference may be a useful piece of information, but often we are simply interested in which product is stronger or weaker in some attribute, and whether the difference is both statistically significant and practically meaningful. Scaling provides a quick and useful way to get intensity or liking information. In the case of descriptive analysis, scaling allows collection of quantitative data on multiple attributes. The degree of variability or noise in the system is, to a large part, determined by whether the panelists have a common frame of reference. Thus reference standards for both the attribute terms and for the intensity anchors are useful. Of course, with consumer evaluations or a psychophysical study such calibration is not possible and usually not desired. The variability of consumer responses should offer a note of caution in the interptation of consumer scaling data. Students and sensory practitioners should examine their scaling methods with a critical eye. Not every task that assigns numbers will have useful scale properties like equal intervals. Bad examples abound

Appendix 1

in the commodity grading (quality scoring) literature (Pangborn and Dunkley, 1964). For example, different numbers may be assigned to different levels of oxidation, but that is scoring a physical condition based on inferences from sensory experience. It is not a report of the intensity of some experience itself. It is not tracking changes along a single perceptual continuum in the psychophysical sense. Scoring is not scaling. All hedonic scales seem to measure what they are intended to measure rather effectively, as long as no gross mistakes are made (Peryam, 1989, p. 23).

Appendix 1: Derivation of Thurstonian-Scale Values for the 9-Point Scale The choice of adjective words for the 9-point hedonic scale is a good example of how carefully a scale can be constructed. The long-standing track record of this tool demonstrates its utility and wide applicability in consumer testing. However, few sensory practitioners actually know how the adjectives were found and what criteria were brought to bear in selecting these descriptors (slightly, moderately, very much, and extremely like/dislike) from a larger pool of possible words. The goal of this section is to provide a shorthand description of the criteria and mathematical method used to select the words for this scale. One concern was the degree to which the term had consensual meaning in the population. The most serious concern was when a candidate word had an ambiguous or double meaning across the population. For example, the word “average” suggests an intermediate response to some people, but in the original study by Jones and Thurstone (1955) there were a group of people who equated it with “like moderately” perhaps since an average product in those days was one that people would like. These days, one can think of negative connotations to the word “average” as in “he was only an average student.” Other ambiguous or bimodal terms were “like not so much” and “like not so well.” Ideally, a term should have low variability in meaning, i.e., a low standard deviation, no bimodality, and little skew. Part of this concern with the normality of the distribution of psychological reactions to a word was the fact that the developers used Thurstone’s model

171

172

7

that marks the ordinate in equal standard deviations units according to the cumulative normal distribution. Either of these methods will tend to make the cumulative S-shaped curve for the item into a straight line. The X-axis value for each point is the “category z-value” for that bucket. 6. Fit a line to the data and interpolate the 50% point on the X-axis (the re-scaled category boundary estimates). These interpolated values for the median for each item now form the new scale values for the items. An example of this interpolation is shown in Fig. 7.7. Three of the phrases used in the original scaling study of Jones and Thurstone (1955) are pictured, three that were not actually chosen but for which we have approximate proportions and z-scores from their ps. The small vertical arrows on the X-axis show the scale values for the original categories of −4 to +3 (+4 has cumulative proportion of 100% and thus the z-score is infinite). Table 7.1 gives the values and

Scaling

proportions for each phrase and the original categories. The dashed vertical lines dropped from the intersection at the zero z-score (50% point) show the approximate mean values interpolated on the X-axis (i.e., about -1.1 for “do not care for it” and about +2.1 for “pferred.”). Note that “pferred” and “don’t care for it” have a linear fit and steep slope, suggesting a normal distribution and low standard deviation. In contrast, “highly unfavorable” has a lower slope and some curvilinearity, indicative of higher variability, skew, and/or pockets of disagreement about the connotation of this term. The actual scale values for the original adjectives are shown in Table 7.2, as found with a soldier population circa 1950 (Jones et al., 1955). You may note that the words are not equally spaced, and that the “slightly” values are closer to the neutral point than some of the other intervals, and the extreme points are a little farther out. This bears a good deal of similarity to the intervals found with the LAM scale as shown in the column where the LAM values are re-scaled to the same range as the 9-point Thurstonian Values.

Appendix 2: Construction of Labeled Magnitude Scales “

“

“

‘

“

“

“

Fig. 7.7 An illustration of the method used to establish spacings and scale values for the 9-point hedonic scale using Thurstonian theory. Arrows on the X-axis show the scale points for the z-scores based on the complete distribution of the original -4 to +4 ratings. The Y-axis shows the actual z-scores based on the proportion of respondents using that category for each specific term. Re-plotted from data provided in Jones et al. (1955).

There are two primary methods for constructing labeled magnitude scales and they are very similar. Both require magnitude estimates from the participants to scale the word phrases used on the lines. In one case, just the word phrases are scaled, and in the second method, the word phrases are scaled among a list of common everyday experiences or sensations that most people are familiar with. The values obtained by the simple scaling of just the words will depend upon the words that are chosen, and extremely high examples (e.g., greatest imaginable liking for any experience) will tend to compss the values of the interior phrases (Cardello et al., 2008). Whether this kind of context effect will occur for the more general method of scaling amongst common experiences is not known. But the use of a broad frame of reference could be a stabilizing factor. Here is an example of the instructions given to subjects in construction of a labeled affective magnitude scale. Note that for hedonics, which are a bipolar continuum with a neutral point, it is necessary to collect a tone or valence (plus or minus) value as well as the overall “intensity” rating.

Appendix 2

173

Table 7.1 Examples of scaled phrases used in Fig. 7.7 Original “Preferred” category Proportion Z-score Proportion 4 3 2 1 0 −1 −2 −3 −4

1.000 0.999 0.983 0.882 0.616 0.383 0.185 0.068 0.008

Table 7.2 Actual 9-point scale phrase values and comparison to the LAM values

(undef.) 3.0 2.1 1.2 0.3 −0.3 −0.9 −1.5 −2.4

0.80 0.50 0.20 0.07 0.03

Z-score 0.84 0.00 −0.84 −1.48 −1.88

Descriptor

Scale value (9-point)

Like extremely Like very much Like moderately Like slightly Neither like nor dislike Dislike slightly Dislike moderately Dislike very much Dislike extremely

4.16 2.91 1.12 0.69 0.00 −0.59 −1.20 −2.49 −4.32

Next to each word label a response area appeared similar to this: Phrase: Like extremely

Words or phrases are psented in random order. After reading a word they must decide whether the word is positive, negative or neutral and place the corresponding symbol on the first line. If the hedonic tone was not a neutral one (zero value), they are instructed to give a numerical estimate using modulus-free magnitude estimation. The following is a sample of the instructions taken from Cardello et al. (2008): After having determined whether the phrase is positive or negative or neutral and writing the appropriate symbol (+, −, 0) on the first line, you will then assess the strength or magnitude of the liking or disliking reflected by the phrase. You will do this by placing a number on the second blank line (under “How Much”). For the first phrase that you rate, you can write any number you want on the line. We suggest you do not use a small number for this word/phrase. The reason for this is that subsequent words/phrases may reflect much lower levels of liking or disliking. Aside from this restriction you can use any numbers you want. For each subsequent

“Do not care for it” Proportion Z-score

“Highly unfavorable” Proportion Z-score

0.96 0.83 0.55 0.30 0.14

1.75 0.95 0.13 −0.52 −1.08

0.96 0.93 0.92 0.90 0.86 0.84 0.82 0.46

Interval

LAM value

LAM rescaled

Interval

1.26 1.79 0.43 0.69 0.59 0.61 1.29 1.83

74.2 56.1 36.2 11.2 0.0 −10.6 −31.9 −55.5 −75.5

4.20 3.18 2.05 0.63 0.00 −0.60 −1.81 −3.14 −4.28

1.02 1.13 1.52 0.63 0.60 1.21 1.33 1.14

1.75 1.48 1.41 1.28 1.08 0.99 0.92 −0.10

word/phrase your numerical judgment should be made proportionally and in comparison to the first number. That is, if you assigned the number 800 to index the strength of the liking/disliking denoted by the first word/phrase and the strength of liking/disliking denoted by the second word/phrase were twice as great, you would assign the number 1,600. If it were three times as great you would assign the number 2,400, etc. Similarly, if the second word/phrase denoted only 1/10 the magnitude of liking as the first, you would assign it the number 80 and so forth. If any word/phrase is judged to be “neutral” (zero (0) on the first line) it should also be given a zero for its magnitude rating.

In the cased of Cardello et al. (2008), positive and negative word labels were analyzed separately. Raw magnitude estimates were equalized for scale range using the procedure of Lane et al. (1961). All positive and negative magnitude estimates for a given subject were multiplied by an inpidual scaling factor. This factor was equal to the ratio of the grand geometric mean (of the absolute value of all nonzero ratings) across all subjects pided by the geometric mean for that subject. The geometric mean magnitude estimates for each phrase were then calculated based on this range-equated data. These means became the distance

174

from the zero point for placement of the phrases along the scale, usually accompanied by a short cross-hatch mark at that point.

7

Scaling

Basker, D. 1988. Critical values of differences among rank sums for multiple comparisons. Food Technology, 42(2), 79, 80-84. Baten, W. D. 1946. Organoleptic tests pertaining to apples and pears. Food Research, 11, 84-94. Bendig, A. W. and Hughes, J. B. 1953. Effect of number of verbal anchoring and number of rating scale categories upon transmitted information. Journal of Experimental Psychology, 46(2), 87-90. Bi, J. 2006. Sensory Discrimination Tests and Measurement. Blackwell, Ames, IA. Birch, L. L., Zimmerman, S. I. and Hind, H. 1980. The influence of social-affective context on the formation of children’s food pferences. Child Development, 51, 865-861. Birch, L. L., Birch, D., Marlin, D. W. and Kramer, L. 1982. Effects of instrumental consumption on children’s food pferences. Appetite, 3, 125-143. Birnbaum, M. H. 1982. Problems with so-called “direct” scaling. In: J. T. Kuznicki, R. A. Johnson and A. F. Rutkiewic (eds.), Selected Sensory Methods: Problems and Approaches to Hedonics. American Society for Testing and Materials, Philadelphia, pp. 34-48. Borg, G. 1982. A category scale with ratio properties for intermodal and interinpidual comparisons. In: H.-G. Geissler and P. Pextod (Eds.), Psychophysical Judgment and the Process of Perception. VEB Deutscher Verlag der Wissenschaften, Berlin, pp. 25-34. Borg, G. 1990. Psychophysical scaling with applications in physical work and the perception of exertion. Scandinavian Journal of Work and Environmental Health, 16, 55-58. Boring, E. G. 1942. Sensation and Perception in the History of Experimental Psychology. Appleton-Century-Crofts, New York. Brandt, M. A., Skinner, E. Z. and Coleman, J. A. 1963. The texture profile method. Journal of Food Science, 28, 404-409. Butler, G., Poste, L. M., Wolynetz, M. S., Agar, V. E. and Larmond, E. 1987. Alternative analyses of magnitude estimation data. Journal of Sensory Studies, 2, 243-257. Cardello, A. V. and Schutz, H. G. 2004. Research note. Numerical scale-point locations for constructing the LAM (Labeled affective magnitude) scale. Journal of Sensory Studies, 19, 341-346. Cardello, A. V., Lawless, H. T. and Schutz, H. G. 2008. Effects of extreme anchors and interior label spacing on labeled magnitude scales. Food Quality and Preference, 21, 323-334. Cardello, A. V., Winterhaler, C. and Schutz, H. G. 2003. Predicting the handle and comfort of military clothing fabrics from sensory and instrumental data: Development and application of new psychophysical methods. Textile Research Journal, 73, 221-237. Cardello, A. V., Schutz, H. G., Lesher, L. L. and Merrill, E. 2005. Development and testing of a labeled magnitude scale of perceived satiety. Appetite, 44, 1-13. Caul, J. F. 1957. The profile method of flavor analysis. Advances in Food Research, 7, 1-40. Chambers, E. C. and Wolf, M. B. 1996. Sensory Testing Methods. ASTM Manual Series, MNL 26. ASTM International, West Conshohocken, PA. Chen, A. W., Resurreccion, A. V. A. and Paguio, L. P. 1996. Age appropriate hedonic scales to measure the food pferences of young children. Journal of Sensory Studies, 11, 141-163.

References Chung, S.-J. and Vickers, 2007a. Long-term acceptability and choice of teas differing in sweetness. Food Quality and Preference 18, 963-974. Chung, S.-J. and Vickers, 2007b. Influence of sweetness on the sensory-specific satiety and long-term acceptability of tea. Food Quality and Preference, 18, 256-267. Coetzee, H. and Taylor, J. R. N. 1996. The use and adaptation of the paired comparison method in the sensory evaluation of hamburger-type patties by illiterate/semi-literate consumers. Food Quality and Preference, 7, 81-85. Collins, A. A. and Gescheider, G. A. 1989. The measurement of loudness in inpidual children and adults by absolute magnitude estimation and cross modality matching. Journal of the Acoustical Society of America, 85, 2012-2021. Conner, M. T. and Booth, D. A. 1988. Preferred sweetness of a lime drink and pference for sweet over non-sweet foods. Related to sex and reported age and body weight. Appetite, 10, 25-35. Cordinnier, S. M. and Delwiche, J. F. 2008. An alternative method for assessing liking: Positional relative rating versus the 9-point hedonic scale. Journal of Sensory Studies, 23, 284-292. Cox, E. P. 1980. The optimal number of response alternatives for a scale: A review. Journal of Marketing Research, 18, 407-422. Curtis, D. W., Attneave, F. and Harrington, T. L. 1968. A test of a two-stage model of magnitude estimation. Perception and Psychophysics, 3, 25-31. Edwards, A. L. 1952. The scaling of stimuli by the method of successive intervals. Journal of Applied Psychology, 36, 118-122. Ekman, G. 1964. Is the power law a special case of Fechner’s law? Perceptual and Motor Skills, 19, 730. Einstein, M. A. 1976. Use of linear rating scales for the evaluation of beer flavor by consumers. Journal of Food Science, 41, 383-385. El Dine, A. N. and Olabi, A. 2009. Effect of reference foods in repeated acceptability tests: Testing familiar and novel foods using 2 acceptability scales. Journal of Food Science, 74, S97-S105. Engen, T. 1974. Method and theory in the study of odor pferences. In: A. Turk, J. W. Johnson and D. G. Moulton (Eds.), Human Responses to Environmental Odors. Academic, New York. Finn, A. and Louviere, J. J. 1992. Determining the appropriate response to evidence of public concern: The case of food safety. Journal of Public Policy and Marketing, 11, 12-25. Forde, C. G. and Delahunty, C. M. 2004. Understanding the role cross-modal sensory interactions play in food acceptability in younger and older consumers. Food Quality and Preference, 15, 715-727. Frijters, J. E. R., Kooistra, A. and Vereijken, P. F. G. 1980. Tables of d’ for the triangular method and the 3-AFC signal detection procedure. Perception and Psychophysics, 27, 176-178. Gaito, J. 1980. Measurement scales and statistics: Resurgence of an old misconception. Psychological Bulletin, 87, 564-587. Gay, C., and Mead, R. 1992 A statistical appraisal of the problem of sensory measurement. Journal of Sensory Studies, 7, 205-228.

175 Gent, J. F. and Bartoshuk, L. M. 1983. Sweetness of sucrose, neohesperidin dihydrochalcone and sacchar in is related to genetic ability to taste the bitter substance 6-npropylthiouracil. Chemical Senses, 7, 265-272. Gescheider, G. A. 1988. Psychophysical scaling. Annual Review of Psychology, 39, 169-200. Giovanni, M. E. and Pangborn, R. M. 1983. Measurement of taste intensity and degree of liking of beverages by graphic scaling and magnitude estimation. Journal of Food Science, 48, 1175-1182. Gracely, R. H., McGrath, P. and Dubner, R. 1978a. Ratio scales of sensory and affective verbal-pain descriptors. Pain, 5, 5-18. Gracely, R. H., McGrath, P. and Dubner, R. 1978b. Validity and sensitivity of ratio scales of sensory and affective verbal-pain descriptors: Manipulation of affect by Diazepam. Pain, 5, 19-29. Green, B. G., Shaffer, G. S. and Gilmore, M. M. 1993. Derivation and evaluation of a semantic scale of oral sensation magnitude with apparent ratio properties. Chemical Senses, 18, 683-702. Green, B. G., Dalton, P., Cowart, B., Shaffer, G., Rankin, K. and Higgins, J. 1996. Evaluating the “Labeled Magnitude Scale” for measuring sensations of taste and smell. Chemical Senses, 21, 323-334. Greene, J. L., Bratka, K. J., Drake, M. A. and Sanders, T. H. 2006. Effective of category and line scales to characterize consumer perception of fruity fermented flavors in peanuts. Journal of Sensory Studies, 21, 146-154. Guest, S., Essick, G., Patel, A., Prajpati, R. and McGlone, F. 2007. Labeled magnitude scales for oral sensations of wetness, dryness, pleasantness and unpleasantness. Food Quality and Preference, 18, 342-352. Hein, K. A., Jaeger, S. R., Carr, B. T. and Delahunty, C. M. 2008. Comparison of five common acceptance and pference methods. Food Quality and Preference, 19, 651-661. Huskisson, E. C. 1983. Visual analogue scales. In: R. Melzack (Ed.), Pain Measurement and Assessment. Raven, New York, pp. 34-37. Jaeger, S. R.; Jørgensen, A. S., AAslying, M. D. and Bredie, W. L. P. 2008. Best-worst scaling: An introduction and initial comparison with monadic rating for pference elicitation with food products. Food Quality and Preference, 19, 579-588. Jaeger, S. R. and Cardello, A. V. 2009. Direct and indirect hedonic scaling methods: A comparison of the labeled affective magnitude (LAM) scale and best-worst scaling. Food Quality and Preference, 20, 249-258. Jones, F. N. 1974. History of psychophysics and judgment. In: E. C. Carterette and M. P. Friedman (Eds.), Handbook of Perception. Psychophysical Judgment and Measurement, Vol. 2. Academic, New York, pp. 11-22. Jones, L. V. and Thurstone, L. L. 1955. The psychophysics of semantics: An experimental investigation. Journal of Applied Psychology, 39, 31-36. Jones, L. V., Peryam, D. R. and Thurstone, L. L. 1955. Development of a scale for measuring soldier’s food pferences. Food Research, 20, 512-520. Keskitalo, K. Knaapila, A., Kallela, M., Palotie, A., Wessman, M., Sammalisto, S., Peltonen, L., Tuorila, H. and Perola, M. 2007. Sweet taste pference are partly genetically

176 determined: Identification of a trait locus on Chromosome 161-3 . American Journal of Clinical Nutrition, 86, 55-63. Kim, K.-O. and O’Mahony, M. 1998. A new approach to category scales of intensity I: Traditional versus rank-rating. Journal of Sensory Studies, 13, 241-249. King, B. M. 1986. Odor intensity measured by an audio method. Journal of Food Science, 51, 1340-1344. Koo, T.-Y., Kim, K.-O., and O’Mahony, M. 2002. Effects of forgetting on performance on various intensity scaling protocols: Magnitude estimation and labeled magnitude scale (Green scale). Journal of Sensory Studies, 17, 177-192. Kroll, B. J. 1990. Evaluating rating scales for sensory testing with children. Food Technology, 44(11), 78-80, 82, 84, 86. Kurtz, D. B., White, T. L. and Hayes, M. 2000. The labeled dissimilarity scale: A metric of perceptual dissimilarity. Perception and Psychophysics, 62, 152-161. Land, D. G. and Shepard, R. 1984. Scaling and ranking methods. In: J. R. Piggott (ed.), Sensory Analysis of Foods. Elsevier Applied Science, London, pp. 141-177. Lane, H. L., Catania, A. C. and Stevens, S. S. 1961. Voice level: Autophonic scale, perceived loudness and effect of side tone. Journal of the Acoustical Society of America, 33, 160-167. Larson-Powers, N. and Pangborn, R. M. 1978. Descriptive analysis of the sensory properties of beverages and gelatins containing sucrose or synthetic sweeteners. Journal of Food Science, 43, 47-51. Lawless, H. T. 1977. The pleasantness of mixtures in taste and olfaction. Sensory Processes, 1, 227-237. Lawless, H. T. 1989. Logarithmic transformation of magnitude estimation data and comparisons of scaling methods. Journal of Sensory Studies, 4, 75-86. Lawless, H. T. and Clark, C. C. 1992. Psychological biases in time intensity scaling. Food Technology, 46, 81, 84-86, 90. Lawless, H. T. and Malone, J. G. 1986a. The discriminative efficiency of common scaling methods. Journal of Sensory Studies, 1, 85-96. Lawless, H. T. and Malone, G. J. 1986b. A comparison of scaling methods: Sensitivity, replicates and relative measurement. Journal of Sensory Studies, 1, 155-174. Lawless, H. T. and Skinner, E. Z. 1979. The duration and perceived intensity of sucrose taste. Perception and Psychophysics, 25, 249-258. Lawless, H. T., Popper, R. and Kroll, B. J. 2010a. Comparison of the labeled affective magnitude (LAM) scale, an 11-point category scale and the traditional nine-point hedonic scale. Food Quality and Preference, 21, 4-12. Lawless, H. T., Sinopoli, D. and Chapman, K. W. 2010b. A comparison of the labeled affective magnitude scale and the nine point hedonic scale and examination of categorical behavior. Journal of Sensory Studies, 25, S1, 54-66. Lawless, H. T., Cardello, A. V., Chapman, K. W., Lesher, L. L., Given, Z. and Schutz, H. G. 2010c. A comparison of the effectiveness of hedonic scales and end-anchor compssion effects. Journal of Sensory Studies, 28, S1, 18-34. Lee, H.-J., Kim, K.-O., and O’Mahony, M. 2001. Effects of forgetting on various protocols for category and line scales of intensity. Journal of Sensory Studies, 327-342. Likert, R. 1932. Technique for the measurement of attitudes. Archives of Psychology, 140, 1-55.

7

Scaling

Lindvall, T. and Svensson, L. T. 1974. Equal unpleasantness matching of malodourous substances in the community. Journal of Applied Psychology, 59, 264-269. Mahoney, C. H., Stier, H. L. and Crosby, E. A. 1957. Evaluating flavor differences in canned foods. II. Fundamentals of the simplified procedure. Food Technology 11, Supplemental Symposium Proceedings, 37-42. Marks, L. E. 1978. Binaural summation of the loudness of pure tones. Journal of the Acoustical Society of America, 64, 107-113. Marks, L. E., Borg, G. and Ljunggren, G. 1983. Inpidual differences in perceived exertion assessed by two new methods. Perception and Psychophysic, 34, 280-288. Marks, L. E., Borg, G. and Westerlund, J. 1992. Differences in taste perception assessed by magnitude matching and by category-ratio scaling. Chemical Senses, 17, 493-506. Mattes, R. D. and Lawless, H. T. 1985. An adjustment error in optimization of taste intensity. Appetite, 6, 103-114. McBride, R. L. 1983a. A JND-scale/category scale convergence in taste. Perception and Psychophysics, 34, 77-83. McBride, R. L. 1983b. Taste intensity and the case of exponents greater than 1. Australian Journal of Psychology, 35, 175-184. McBurney, D. H. and Shick, T. R. 1971. Taste and water taste for 26 compounds in man. Perception and Psychophysics, 10, 249-252. McBurney, D. H. and Bartoshuk, L. M. 1973. Interactions between stimuli with different taste qualities. Physiology and Behavior, 10, 1101-1106. McBurney, D. H., Smith, D. V. and Shick, T. R. 1972. Gustatory cross-adaptation: Sourness and bitterness. Perception and Psychophysics, 11, 228-232. Mead, R. and Gay, C. 1995. Sequential design of sensory trials. Food Quality and Preference, 6, 271-280. Mecredy, J. M. Sonnemann, J. C. and Lehmann, S. J. 1974. Sensory profiling of beer by a modified QDA method. Food Technology, 28, 36-41. Meilgaard, M., Civille, G. V. and Carr, B. T. 2006. Sensory Evaluation Techniques, Fourth Edition. CRC, Boca Raton, FL. Moore, L. J. and Shoemaker, C. F. 1981. Sensory textural properties of stabilized ice cream. Journal of Food Science, 46, 399-402. Moskowitz, H. R. 1971. The sweetness and pleasantness of sugars. American Journal of Psychology, 84, 387-405. Moskowitz, H. R. and Sidel, J. L. 1971. Magnitude and hedonic scales of food acceptability. Journal of Food Science, 36, 677-680. Muñoz, A. M. and Civille, G. V. 1998. Universal, product and attribute specific scaling and the development of common lexicons in descriptive analysis. Journal of Sensory Studies, 13, 57-75. Newell, G. J. and MacFarlane, J. D. 1987. Expanded tables for multiple comparison procedures in the analysis of ranked data. Journal of Food Science, 52, 1721-1725. Olabi, A. and Lawless, H. T. 2008. Persistence of context effects with training and reference standards. Journal of Food Science, 73, S185-S189. O’Mahony, M., Park, H., Park, J. Y. and Kim, K.-O. 2004. Comparison of the statistical analysis of hedonic data using

References analysis of variance and multiple comparisons versus and R-index analysis of the ranked data. Journal of Sensory Studies, 19, 519-529. Pangborn, R. M. and Dunkley, W. L. 1964. Laboratory procedures for evaluating the sensory properties of milk. Dairy Science Abstracts, 26-55-62. Parducci, A. 1965. Category judgment: A range-frequency model. Psychological Review, 72, 407-418. Park, J.-Y., Jeon, S.-Y., O’Mahony, M. and Kim, K.-O. 2004. Induction of scaling errors. Journal of Sensory Studies, 19, 261-271. Pearce, J. H., Korth, B. and Warren, C. B. 1986. Evaluation of three scaling methods for hedonics. Journal of Sensory Studies, 1, 27-46. Peryam. D. 1989. Reflections. In: Sensory Evaluation. In Celebration of our Beginnings. American Society for Testing and Materials, Philadelphia, pp. 21-30. Peryam, D. R. and Girardot, N. F. 1952. Advanced taste-test method. Food Engineering, 24, 58-61, 194. Piggot, J. R. and Harper, R. 1975. Ratio scales and category scales for odour intensity. Chemical Senses and Flavour, 1, 307-316. Pokor´ny, J., Davídek, J., Prnka, V. and Davídková, E. 1986. Nonparametric evaluation of graphical sensory profiles for the analysis of carbonated beverages. Die Nahrung, 30, 131-139. Poulton, E. C. 1989. Bias in Quantifying Judgments. Lawrence Erlbaum, Hillsdale, NJ. Richardson, L. F. and Ross, J. S. 1930. Loudness and telephone current. Journal of General Psychology, 3, 288-306. Rosenthal, R. 1987. Judgment Studies: Design, Analysis and Meta-Analysis. University Press, Cambridge. Shand, P. J., Hawrysh, Z. J., Hardin, R. T. and Jeremiah, L. E. 1985. Descriptive sensory analysis of beef steaks by category scaling, line scaling and magnitude estimation. Journal of Food Science, 50, 495-499. Schutz, H. G. and Cardello, A. V. 2001.. A labeled affective magnitude (LAM) scale for assessing food liking/disliking. Journal of Sensory Studies, 16, 117-159. Siegel, S. 1956. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New York. Sriwatanakul, K., Kelvie, W., Lasagna, L., Calimlim, J. F., Wels, O. F. and Mehta, G. 1983. Studies with different types of visual analog scales for measurement of pain. Clinical Pharmacology and Therapeutics, 34, 234-239. Stevens, J. C. and Marks, L. M. 1980. Cross-modality matching functions generated by magnitude estimation. Perception and Psychophysics, 27, 379-389. Stevens, S. S. 1951. Mathematics, measurement and psychophysics. In: S. S. Stevens (ed.), Handbook of Experimental Psychology. Wiley, New York, pp. 1-49.

177 Stevens, S. S. 1956. The direct estimation of sensory magnitudes-loudness. American Journal of Psychology, 69, 1-25. Stevens, S. S. 1957. On the psychophysical law. Psychological Review, 64, 153-181. Stevens, S. S. 1969. On pdicting exponents for cross-modality matches. Perception and Psychophysics, 6, 251-256. Stevens, S. S. and Galanter, E. H. 1957. Ratio scales and category scales for a dozen perceptual continua. Journal of Experimental Psychology, 54, 377-411. Stoer, N. L. and Lawless, H. T. 1993. Comparison of single product scaling and relative-to-reference scaling in sensory evaluation of dairy products. Journal of Sensory Studies, 8, 257-270. Stone, H., Sidel, J., Oliver, S., Woolsey, A. and Singleton, R. C. 1974. Sensory Evaluation by quantitative descriptive analysis. Food Technology, 28, 24-29, 32, 34. Teghtsoonian, M. 1980. Children’s scales of length and loudness: A developmental application of cross-modal matching. Journal of Experimental Child Psychology, 30, 290-307. Thurstone, L. L. 1927. A law of comparative judgment. Psychological Review, 34, 273-286. Townsend, J. T. and Ashby, F. G. 1980. Measurement scales and statistics: The misconception misconceived. Psychological Bulletin, 96, 394-401. Vickers, Z. M. 1983. Magnitude estimation vs. category scaling of the hedonic quality of food sounds. Journal of Food Science, 48, 1183-1186. Villanueva, N. D. M. and Da Silva, M. A. A. P. 2009. Performance of the nine-point hedonic, hybrid and selfadjusting scales in the generation of internal pference maps. Food Quality and Preference, 20, 1-12. Villanueva, N. D. M., Petenate, A. J., and Da Silva, M. A. A. P. 2005. Comparative performance of the hybrid hedonic scale as compared to the traditional hedonic, self-adjusting and ranking scales. Food Quality and Preference, 16, 691-703. Ward, L. M. 1986. Mixed-modality psychophysical scaling: Double cross-modality matching for “difficult” continua. Perception and Psychophysics, 39, 407-417. Weiss, D. J. 1972. Averaging: an empirical validity criterion for magnitude estimation. Perception and Psychophysics, 12, 385-388. Winakor, G., Kim, C. J. and Wolins, L. 1980. Fabric hand: Tactile sensory assessment. Textile Research Journal, 50, 601-610. Yamaguchi, S. 1967. The synergistic effect of monosodium glutamate and disodium 5 inosinate. Journal of Food Science 32, 473-477.

Chapter 8

Time-Intensity Methods

Contents 8.1 8.2 8.3

8.4

8.5

8.6

Introduction . . . . . . . . . . . . . . . . . A Brief History . . . . . . . . . . . . . . . Variations on the Method . . . . . . . . . . 8.3.1 Discrete or Discontinuous Sampling . . . 8.3.2 “Continuous” Tracking . . . . . . . . . 8.3.3 Temporal Dominance Techniques . . . . Recommended Procedures . . . . . . . . . . 8.4.1 Steps in Conducting a Time-intensity Study . . . . . . . . . . . . . . . . . 8.4.2 Procedures . . . . . . . . . . . . . . 8.4.3 Recommended Analysis . . . . . . . . Data Analysis Options . . . . . . . . . . . . 8.5.1 General Approaches . . . . . . . . . . 8.5.2 Methods to Construct or Describe Average Curves . . . . . . . . . . . . . . . . 8.5.3 Case Study: Simple Geometric Description . . . . . . . . . . . . . . 8.5.4 Analysis by Principal Components . . . Examples and Applications . . . . . . . . . . 8.6.1 Taste and Flavor Sensation Tracking . . . 8.6.2 Trigeminal and Chemical/Tactile Sensations . . . . . . . . . . . . . . 8.6.3 Taste and Odor Adaptation . . . . . . .

179 180 182 182 183 184 185 185 186 186 187 187 188 189 192 193 193 194 194

8.6.4 Texture and Phase Change . . 8.6.5 Flavor Release . . . . . . . 8.6.6 Temporal Aspects of Hedonics 8.7 Issues . . . . . . . . . . . . . . . 8.8 Conclusions . . . . . . . . . . . . References . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

195 195 196 197 198 198

8.1 Introduction Perception of aroma, taste, flavor, and texture in foods is a dynamic not a static phenomenon. In other words, the perceived intensity of the sensory attributes change from moment to moment. The dynamic nature of food sensations arises from processes of chewing, breathing, salivation, tongue movements, and swallowing (Dijksterhuis, 1996). In the texture profile method for instance, different phases of food breakdown were recognized early on as evidenced by the separation of characteristics into first bite, mastication, and residual phases (Brandt et al., 1963). Wine tasters often discuss

H.T. Lawless, H. Heymann, Sensory Evaluation of Food, Food Science Text Series, DOI 10.1007/978-1-4419-6488-5_8, © Springer Science+Business Media, LLC 2010

179

180

how a wine “opens in the glass,” recognizing that the flavor will vary as a function of time after opening the bottle and exposing the wine to air. It is widely believed that the consumer acceptability of different intensive sweeteners depends on the similarity of their time profile to that of sucrose. Intensive sweeteners with too long a duration in the mouth may be less pleasant to consumers. Conversely, a chewing gum with longlasting flavor or a wine with a “long finish” may be desirable. These examples demonstrate how the time profile of a food or beverage can be an important aspect of its sensory appeal. The common methods of sensory scaling ask the panelists to rate the perceived intensity of the sensation by giving a single (uni-point) rating. This task requires that the panelists must “time-average” or integrate any changing sensations or to estimate only the peak intensity in order to provide the single intensity value that is required. Such a single value may miss some important information. It is possible, for example, for two products to have the same or similar time-averaged profiles or descriptive specifications, but differ in the order in which different flavors occur or when they reach their peak intensities. Time-intensity (TI) methods provide panelists with the opportunity to scale their perceived sensations over time. When multiple attributes are tracked, the profile of a complex food flavor or texture may show differences between products that change across time after a product is first tasted, smelled, or felt. For most sensations the perceived intensity increases and eventually decreases but for some, like perceived toughness of meat, the sensations may only decrease as a function of time. For perceived melting, the sensation may only increase until a completely melted state is reached. When performing a TI study the sensory specialist can obtain a wealth of detailed information such as the following for each sample: the maximum intensity perceived, the time to maximum intensity, the rate and shape of the increase in intensity to the maximum point, the rate and shape of the decrease in intensity to half-maximal intensity and to the extinction point, and the total duration of the sensation. Some of the common time-intensity parameters are illustrated in Fig. 8.1. The additional information derived from time-intensity methods is especially useful when studying sweeteners or products like chewing gums and hand lotions that have a distinctive time profile.

8

Time-Intensity Methods

INTENSITY Other Parameters: DUR, Total duration = Tend – Tstart AUC, Area under the curve

TIME

Fig. 8.1 Example of a time-intensity curve and common curve parameters extracted from the record.

The remainder of this chapter will be devoted to an overview of the history and applications of this method, as well as recommended procedures and analyses. For the student who wants only the basic information, the following sections are key: variations on the method (Section 8.3), steps and recommended procedures (Section 8.4), data analysis options (Section 8.5), and conclusions (8.8).

8.2 A Brief History Holway and Hurvich (1937) published an early report of tracking taste intensity over time. They had their subjects trace a curve to repsent the sensations from a drop of either 0.5 or 1.0 M NaCl placed on the anterior tongue surface over 10 s. They noted several general effects that were later confirmed as common trends in other studies. The higher concentration led to a higher peak intensity, but the peak occurred later, in spite of a steeper rising slope. Most importantly they noted that taste intensity was not strictly a function of concentration: “while the concentration originally placed on the tongue is ‘fixed,’ the intensity varies in a definite manner from moment to moment. Saline intensity depends on time as well as concentration.” A review of studies of temporal factors in taste is found in Halpern (1991). A review of TI studies of the 1980s and early 1990s can be found in Cliff and Heymann (1993b). Sjostrom (1954) and Jellinek (1964) also made early attempts to quantify the temporal response to perceived sensory intensities. These authors asked their panelists to indicate their perceived bitterness of beer at 1 s intervals on a ballot, using a clock to indicate time.

8.2 A Brief History

They then constructed TI curves by plotting the x-y coordinates (time on the x-axis and perceived intensity on the y-axis) on graph paper. Once panelists had some experience with the method it was possible to ask them simultaneously to rate the perceived intensities of two different attributes at 1 s intervals. Neilson (1957) in an attempt to make the production of the TI curves easier, asked panelists to indicate perceived bitterness directly on graph paper at 2 s timed intervals. The clock could be distracting to the panelists and thus Meiselman (1968), studying taste adaptation and McNulty and Moskowitz (1974), evaluating oilin-water emulsions, improved the TI methodology by eliminating the clock. These authors used audible cues to tell the panelists when to enter perceived intensities on a ballot, placing the timekeeping demands on the experimenter rather than the participant. Larson-Powers and Pangborn (1978), in another attempt to eliminate the distractions of the clock or audible cues, employed a moving strip-chart recorder equipped with a foot pedal to start and stop the movement of the chart. Panelists recorded their responses to the perceived sweetness in beverages and gelatins sweetened with sucrose or synthetic sweeteners, by moving a pen along the cutter bar of the strip-chart recorder. The cutter bar was labeled with an unstructured line scale, from none to extreme. A cardboard cover was placed over the moving chart paper to pvent the panelists from watching the evolving curves and thus pventing them from using any visual cues to bias their responses. A similar setup was independently developed and used at the General Foods Technical Center in 1977 to track sweetness intensity (Lawless and Skinner, 1979). In this apparatus, the actual pen carriage of the chart recorder was grasped by the subject, eliminating the need for them to position a pen; also the moving chart was obscured by a line scale with a pointer attached to the pen carriage. In yet another laboratory at about the same time, Birch and Munton (1981) developed the “SMURF” version of TI scaling (short for “Sensory Measuring Unit for Recording Flux”). In the SMURF apparatus, the subject turned a knob graded from 1 to 10, and this potentiometer fed a variable signal to a strip-chart recorder out of sight of the panelist. The use of strip-chart recorders provided the first continuous TI data-collection methods and freed the panelists from any distractions caused by a clock or auditory signal. However, the methods required a fair degree of mental and physical

181

coordination by the participants. For example, in the Larson-Powers setup, the strip-chart recorder required the panelists to use a foot pedal to run the chart recorder, to place the sample in the mouth, and to move the pen to indicate the perceived intensity. Not all panelists were suitably coordinated and some could not do the evaluation. Although the strip-chart recorder made continuous evaluation of perceived intensities possible, the TI curves had to be digitized manually, which was extremely time consuming. The opportunity to use computers to time sample an analog voltage signal quite naturally led to online data collection to escape the problem of manual measurement of TI curves. To the best of our knowledge, the first computerized system was developed at the US Army Natick Food Laboratories in 1979 to measure bitter taste adaptation. It employed an electric sensor in the spout just above the subject’s tongue in order to determine the actual onset of stimulus arrival. Subthreshold amounts of NaCl were added to the stimulus and thus created a conductance change as the flow changed from the pliminary water rinse to the stimulus interface in the tube. A special circuit was designed to detect the conductance change and to connect the response knob to the visual line scale. Like the SMURF apparatus developed by Birch and Munton, the subject turned a knob controlling a variable resistor. The output of this potentiometer moved a pointer on a line scale for visual feedback while a parallel analog signal was sent to an analog-to-digital converter and then to the computer. The programming was done using FORTRAN subroutines on a popular lab computer of that era. The entire system is shown in Fig. 8.2. The appearance of desktop computers led to an explosion in the use of TI methodology in the 1980s and 1990s. A number of thesis research projects from U.C. Davis served as useful demonstrations of the method (Cliff, 1987; Dacanay, 1990; Rine, 1987) and the method was championed by Pangborn and coworkers (e.g., Lee and Pangborn, 1986). Several scientists (Barylko-Pikielna et al., 1990; Cliff, 1987; Guinard et al., 1985; Janusz et al., 1991; Lawless, 1980; Lee, 1985; Rine, 1987; Yoshida, 1986) developed computerized TI systems using a variety of hardware and software products. Computerized TI systems are now commercially available as part of data-collection software ensembles, greatly enhancing the ease and availability of TI data collection and data processing.

182

8

Time-Intensity Methods

Fig. 8.2 An early computerized system for time-intensity scaling used for tracking bitterness adaptation in a flow system for stimulating the anterior tongue. Heavy lines indicate the flow of stimulus solution, solid lines the flow of information, and dashed lines the experimenter-driven process control. Stimulus arrival at the tongue was monitored by conductivity sensors fitted into the glass tube just above the subject’s tongue. The subjects’

responses change on a line scale while the experimenter could view the conductivity and response voltage outputs on the computer’s display screen, which simultaneously were output to a data file. The system was programmed in FORTRAN subroutines controlling the clock sampling rate and analog-to-digital conversion routine. From Lawless and Clark (1992), reprinted with permission.

However, despite the availability of computerized systems, some research was still conducted using the simple time cueing at discrete intervals (e.g., Lee and Lawless, 1991; Pionnier et al., 2004), and the semi-manual strip-chart recorder method (Ott et al., 1991; Robichaud and Noble, 1990). A discussion of some common applications of TI methods is given in Section 8.6.

oldest approach is simply to ask the panelists to rate the intensity of sensation during different phases of consuming a food. This is particularly applicable to texture which may be evaluated in phases such as initial bite, first chew, mastication, and residual. An example of time pision during the texture evaluation of a baked product is shown in Table 8.1. When using a descriptive panel, it may be useful to have residual flavor or mouthfeel sensations rated at a few small intervals, e.g., every 30 s for 2 min, or immediately after tasting and then again after expectoration. For an example of this approach used with hot pepper “burn” see Stevens and Lawless (1986). Each measurement is then treated like a separate descriptive attribute and analyzed as a separate variable, with little or no attempt to reconstruct a time-connected record like the TI curve shown in Fig. 8.1. For researchers

8.3 Variations on the Method 8.3.1 Discrete or Discontinuous Sampling The sensory scientist has several options for collecting time-dependent sensory data. The methods for timerelated scaling can be pided into four groups. The

8.3 Variations on the Method

183

Table 8.1 Texture attributes at different phases of descriptive analysis Phase Attributes Surface

First bite

First chew Chew down

Residual

Roughness Particles Dryness Fracturability Hardness Particle size Denseness Uniformity of chew Moisture absorption Cohesiveness of mass Toothpacking Grittiness Oiliness Particles Chalky

interested in some simple aspect like “strength of bitter aftertaste” this method may suffice. Another related approach is to ask for repeated ratings of a single or just a few attributes at repeated smaller time intervals, usually cued by the panel leader or experimenter. These ratings are then connected and graphed on time axis. This is a simple procedure that can be used to track changes in the intensity of a flavor or texture attribute and requires no special equipment other than a stopwatch or other timing device. The panel must be trained to rate their sensations upon the time cue and to move rapidly through the list of attributes. The cue may be given verbally or on a computer screen. It is not known how many attributes can be rated in this way, but with faster time cueing and shorter intervals, obviously fewer attributes may be rated. This method also requires some faith in the assumption that the attributes are being rated close to the actual time when the cue is given. The accuracy with which panelists can do this is unknown, but given that there is a reaction time delay in any perceptual judgment, there must be some inherent error or delayrelated variance built into the procedure. An example of the repeated, discrete time interval method with verbal cueing and multiple attributes can be found in studies of sweetener mixtures (e.g., Ayya and Lawless, 1992) and astringency (Lee and Lawless, 1991). The time record is treated as a connected series and time is analyzed as one factor (i.e., one independent variable) in the statistical analysis.

Word anchors Smooth-rough None-many Oily-dry Crumbly-brittle Soft-hard Small-large Airy-dense Even-uneven None-much Loose-cohesive None-much None-much Dry-oily None-much Not chalky-very chalky

184

trained to move a mouse diagonally and a visual scale included pointers on both horizontal and vertical scales to repsent the intensity of the inpidual attributes. With a slowly changing product like chewing gum, with a sampling time that is not too frequent (every 9-15 s in this case), the technique would seem to be within the capabilities of human observers to either rapidly shift their attention or to respond to the overall pattern of the combined flavors. However, as currently used, most TI tracking methods must repeat the evaluation in order to track additional attributes. Ideally, this could lead to a composite profile of all the dynamic flavor and texture attributes in a product and how they changed across time. Such an approach was proposed by DeRovira (1996), who showed how the descriptive analysis spider-web plot of multiple attributes could be extended into the time dimension to produce a set of TI curves and thus to characterize an entire profile.

8

Time-Intensity Methods

8.4 Recommended Procedures

attributes are rated in each trial, (2) that it is simple and easy to do, requiring little or no training, and (3) that it provides a picture of enhanced differences relative to the TI records. Because panelists are forced to respond to only one attribute at a time, differences in the temporal profile may be accentuated. However, at the time of this publication, no standard procedure seemed to be agreed upon. The technique requires specialized software to collect the information, but at least one of the major sensory software systems has implemented a TDS option. Attributes are assumed to have a score of zero before they become dominant, and some attributes may never be rated. This would seem to necessarily lead to an incomplete record. Different panelists are contributing at different times to different attributes, so statistical methods for comparison of differences between products using the raw data set are difficult. However, products can be statistically compared using the simplified summary scores (summed intensity by duration measures, but losing time information) or by comparison of the proportion of responders (losing intensity information) as in Labbe et al. (2009). Qualitative comparisons can be made from inspecting the curves, such as “this product is initially sweet, then becoming more astringent, compared to product X which is initially sour, then fruity.” Is information provided by TDS and by traditional TI tracking methods different? One study found a strong resemblance of the constructed time curves for both methods, providing virtually redundant information for some attributes (Le Reverend et al., 2008). In another study, correlations with TI parameters were high for intensity maxima versus dominance proportion maxima, as might be expected since a higher number of persons finding an attribute dominant should be related to the mean intensity in TI. Correlations with other time-dependent parameters such as time to Imax (Tmax ) and duration measures were low, due to the different information collected in TDS and the limited attention to one attribute at a time (Pineau et al., 2009).

8.4 Recommended Procedures 8.4.1 Steps in Conducting a Time-intensity Study The steps in conducting a time-intensity study are similar to those in setting up a descriptive analysis procedure. They are listed in Table 8.2. The first

185 Table 8.2 Steps in conducting a time-intensity study 1. Determine project objectives: Is TI the right method? 2. Determine critical attributes to be rated. 3. Establish products to be used with clients/researchers. 4. Choose system and/or TI method for data collection. a. What is the response task? b. What visual feedback is provided to the panelists? 5. Establish statistical analysis and experimental design. a. What parameters are to be compared? b. Are multivariate comparisons needed? 6. Recruit panelists. 7. Conduct training sessions. 8. Check panelist performance. 9. Conduct study. 10. Analyze data and report.

important question is to establish whether TI methods are appropriate to the experimental research objective. Is this a product with just one or a few critical attributes that are likely to vary in some important way in their time course? Is this difference likely to impact consumer acceptance of the product? What are these critical attributes? Next, the product test set should be determined with the research team, and this will influence the experimental design. The TI method itself must be chosen, e.g., discrete point, continuous tracking, or TDS. The sensory specialist should by this time have an idea of what the data set will look like and what parameters can be extracted from the TI records for statistical comparisons such as intensity maxima, time to maxima, areas under the curve, and total duration. Many of the TI curve parameters are often correlated, so there is little need to analyze more than about ten parameters. Practice is almost always essential. You cannot assume that a person sitting down in a test booth will know what to do with the TI system and feel comfortable with the mouse or other response device. A protocol for training TI panelists was outlined by Peyvieux and Dijksterhuis (2001) and this protocol or similar versions have been widely adopted. It is also wise that some kind of panel checking be done to make sure the panelists are giving reliable data (see Bloom et al., 1995; Peyvieux and Dijksterhuis, 2001) and to examine the reasonableness of their data records. At this time the researchers and statistical staff should also decide how to handle missing data or records that may have artifacts or be incomplete. As in any sensory study, extensive planning may save a lot of headaches and problems, and this is especially true for TI methods.

186

8.4.2 Procedures If only a few attributes are going to be evaluated, the continuous tracking methods are appropriate and will provide a lot of information. This usually requires the use of computer-assisted data collection. Many, if not all, of the commercial software packages for sensory evaluation data collection have TI modules. The start and stop commands, sampling rate, and inter-trial intervals can usually be specified. The mouse movement will generally produce some visual feedback such as the motion of a cursor or line indicator along a simple line scale. The display often looks like a vertical or horizontal thermometer with the cursor position clearly indicated by bar or line that rises and falls. The computer record can be treated as raw data for averaging across panelists. However, statistical analysis by simple averaging raises a number of issues (discussed in Section 8.5). A simple approach is to pull characteristic curve parameters off each record for purposes of statistical comparisons such as intensity maximum (Imax ), time to maximum (Tmax ), and area under the curve (AUC). These are sometimes referred to as “scaffolding parameters” as they repsent the fundamental structure of the time records. Statistical comparison of these parameters can provide a clear understanding of how different products are perceived with regard to the onset of sensations, time course of rising and falling sensations, total duration, and total sensory impact of that flavor or textural aspect of the products. If a computer-assisted software package is not available or cannot be programmed, the research can always choose to use the cued/discontinuous method (e.g., with a stopwatch and verbal commands). This may be suitable for products in which multiple attributes must be rated in order to get the full picture. However, given the widespad availability of commercial sensory data-collection systems in major food and consumer product companies, it is likely that a sensory professional will have access to a continuous tracking option. The starting position of the cursor on the visible scale or computer screen should be considered carefully. For most intensity ratings it makes sense to start at the lower end, but for hedonics (like/dislike) the cursor should begin at the neutral point. For meat tenderness or product melting, the track is usually unidirectional, so the cursor should start at “not tender” or

8

Time-Intensity Methods

“tough” for meat and “not melted” for a product that melts. If the cursor is started at the wrong end of a unidirectional tracking situation, a falsely bi-directional record may be obtained due to the initial movement. An outline of panel training for TI studies was illustrated in a case study by Peyvieux and Dijksterhuis (2001) and the sensory specialist should consider using this or a similar method to insure good performance of the panelists. These authors had panelists evaluate flavor and texture components of a complex meat product. The prospective panelists were first introduced to the TI method, and then given practice with basic taste solutions over several sessions. The basic tastes were considered simpler than the complex product and more suitable for initial practice. Panelist consistency was checked and a panelist was considered reliable if they could produce two out of three TI records on the same taste stimulus that did not differ more than 40% of the time. A vertical line scale was used and an important specification was when to move the cursor back to zero (when there was no flavor or when the sample was swallowed for the texture attribute of juiciness). Problems were noted with (1) nontraditional curve shapes such as having no return to zero, (2) poor replication by some panelists, (3) unusable curves due to lack of a landmark such as no Imax . The authors also conducted a traditional profiling (i.e., descriptive analysis) study before the TI evaluations to make sure the attributes were correctly chosen and understood by the panelists. If TI panelists are already chosen from an existing descriptive panel, this step may not be needed. The authors conducted several statistical analyses to check for consistent use of the attributes, to look for oddities in curve shapes by some panelists, and to examine inpidual replicates. Improvements in consistency and evidence of learning and practice were noted.

8.4.3 Recommended Analysis For purposes of comparing products, the simplest approach is to extract the curve parameters such as Imax , Tmax , AUC, and total duration from each record. Some sensory software systems will generate these measures automatically. Then these curve parameters can be treated like any data points in any sensory evaluation and compared statistically. For three or

8.5 Data Analysis Options

more products, analysis would be by ANOVA and then planned comparisons of means (see Appendix C). Means and significance of differences can be reported in graphs or tables for each curve attribute and product. If a time by intensity curve is desired, the curves can be averaged in the time direction by choosing points at specific time intervals. This averaging method is not without its pitfalls; however, a number of alternative methods are given in the next section. An example of how to produce a simplified averaged curve is given below in the case study of the trapezoidal method (Lallemand et al., 1999).

8.5 Data Analysis Options 8.5.1 General Approaches Two common statistical approaches have been taken to perform hypothesis testing on TI data. Perhaps the most obvious test is simply to treat the raw data at whatever time intervals were sampled as the input data to analyses of variance (ANOVA) (e.g., Lee and Lawless, 1991). This approach results in a very large ANOVA with at least three factors-time, panelists, and the treatments of interest. Time and panelists effects may not be of primary interest but will always show large F-ratios due to the fact that people differ and the sensations change over time. This is not news. Another common pattern is a time-by-treatment interaction since all curves will tend to converge near baseline at later time intervals. This is also to be

187

Table 8.3 Parameters extracted from time-intensity curves Parameter Other names Definition Peak intensity Imax , Ipeak Height of highest point on TI record Total duration DUR, Dtotal Time from onset to return to baseline Self-explanatory Area under the curve AUC, Atotal Plateau Dpeak Time difference between reaching maximum and beginning descent Area under plateau Apeak Self-explanatory Area bounded by onset of decline and reaching baseline Area under descending phase Ptotal Rising slope Ri Rate of increase (linear fit) or slope of line from onset to peak intensity Rate of decrease (linear fit) or slope of line from initial declining point to baseline. Declining slope Rf Extinction Time at which curve terminates at baseline Time to reach peak intensity Time to peak Tmax , Tpeak Time to half peak Half-life Time to reach half maximum in decay portion Modified from Lundahl (1992) Other shape parameters are given in Lundahl (1992), based on a half circle of equivalent area under the curve and piding the half circle into rising and falling phase segments

188

8

2006), an effect that is sometimes described as an inpidual “signature.” Examples of inpidual signatures are shown in Fig. 8.3. The causes of these inpidual patterns are unknown but could be attributed to differences in anatomy, differences in physiology such as salivary factors (Fischer et al., 1994), different types of oral manipulation or chewing efficiency (Brown et al., 1994; Zimoch and Gullet, 1997), and inpidual habits of scaling. Some of this information may be lost when analyzing only the extracted parameters.

INTENSITY

Judge 1 Judge 2 Judge 3

0

20

40

60 80 TIME (sec)

100

120

140

Fig. 8.3 Examples of time-intensity records showing characteristic signatures or shapes. Judge 1 shows a record with multiple plateaus, a common occurrence. Judge 2 shows a smooth and continuous curve. Judge 3 shows a steep rise and fall.

A third approach is to fit some mathematical model or set of equations to each inpidual record and then use the constants from the model as the data for comparisons of different products (Eilers and Dijksterhuis, 2004; Garrido et al., 2001; Ledauphin et al., 2005, 2006; Wendin et al., 2003). Given the increasing activity and ingenuity in the area of sensometrics, it is likely that such models will continue to be developed. The sensory scientist needs to ask how useful they are in the product testing arena and whether the model fitting is useful in differentiating products. Various approaches to modeling and mathematical description of TI curves are discussed in the next section.

8.5.2 Methods to Construct or Describe Average Curves The analysis of TI records has produced a sustained interest and response from sensometricians, who have

Time-Intensity Methods

8.5 Data Analysis Options

189

INTENSITY

JUDGE 1 JUDGE 2 AVERAGE

TIME

Fig. 8.4 Two curves with different peak times, if averaged, can lead to a double-peaked curve that resembles neither original data record.

style” either by simple visual inspection of the curves or by a clustering analysis or other statistical methods (van Buuren, 1992; Zimoch and Gullet, 1997). Then these subgroups can be analyzed separately. The analysis may proceed using the simple fixed-time averaging of curve heights or one of the other methods described next. An alternative approach is to average in both the intensity and time directions, by setting each inpidual’s maximum of the mean time to maximum across all curves, and then finding mean times to fixed percentages of maximum in the rising and falling phases of each curve. This procedure was originally published by Overbosch et al. (1986) and subsequent modifications were proposed by Liu and MacFie (1990). The steps in the procedure are shown in Fig. 8.5. Briefly, the method proceeds as follows: In the first step, the geometric mean value for the intensity maximum is found. Inpidual curves are multiplicatively scaled to have this Imax value. In the second step, the geometric mean time to Imax is calculated. In the next steps, geometric mean times are calculated for fixed percentage “slices” of each curve, i.e., at fixed percentages of Imax . For example, the rising and falling phases are “sliced” at 95% of Imax and 90% of Imax and the geometric mean time values to reach these heights are found. This procedure avoids the kind of double-peaked curve that can arise from simple averaging of two distinctly different curve shapes as shown in Fig. 8.4. The method results in several desirable properties that do not necessarily occur with simple averaging at fixed times. First, the Imax value from the mean curve is the geometric mean of the Imax of the inpidual curves. Second, the Tmax value from the mean curve is the geometric mean of the Tmax of the inpidual

curves. Third, the endpoint is the geometric mean of all endpoint times. Fourth, all judges contribute to all segments of the curve. With simple averaging at fixed times, the tail of the curve may have many judges returned to zero and thus the mean is some small value that is a poor repsentation of the data at those points. In statistical terms, the distribution of responses at these later time intervals is positively skewed and left-censored (bound by zero). One approach to this problem is to use the simple median as the measure of central tendency (e.g., Lawless and Skinner, 1979). In this case the summary curve goes to zero when over half the judges go to zero. A second approach is to use statistical techniques designed for estimating measures of central tendency and standard deviations from leftcensored positively skewed data (Owen and DeRouen, 1980). Overbosch’s method works well if all inpidual curves are smoothly rising and falling with no plateaus or multiple peaks and valleys, and all data begin and end at zero. In practice, the data are not so regular. Some judges may begin to fall after the first maximum and then rise again to a second peak. Due to various common errors, the data may not start and end at zero, for example, the record may be truncated within the allowable time of sampling. To accommodate these problems, Liu and MacFie (1990) developed a modification of the above procedure. In their procedure, Imax and four “time landmarks” were averaged, namely starting time, time to maximum, time at which the curve starts to descend from Imax and ending time. The ascending and descending phases of each curve were then pided into about 20 time interval slices. At each time interval, the mean I value is calculated. This method then allows for curves with multiple rising and falling phases and a plateau of maximum intensity that is commonly seen in some judges’ records.

8.5.3 Case Study: Simple Geometric Description A simple and elegant method for comparing curves and extracting parameters by a geometric approximation was described by Lallemand et al. (1999). The authors used the method with a trained texture panel to evaluate different ice cream formulations. The labor-intensive nature of TI studies was illustrated in the fact that

190

8

JUDGE 1 JUDGE 2

Time-Intensity Methods

JUDGE 1 JUDGE 2

A Imax 1

B Geo. mean Imax

Imax 2

JUDGE 1 JUDGE 2

Tmax1

C

JUDGE 1 JUDGE 2

D

Geo. mean Tmax

Tmax2

JUDGE 1 JUDGE 2

Geo. mean 95% Tmax

T1 @ 95% Imax

E

F Geo. mean T @95% Imax

Geo. mean Tmax,Imax

T2 @ 95% Imax

Fig. 8.5 Steps in the data analysis procedure recommended by Overbosch et al. (1986). (a) Two hypothetical time-intensity records from two panelists showing different intensity maxima at different times. (b) The geometric mean value for the intensity maximum is found. Inpidual curves may then by multiplicatively scaled to have this Imax value. (c) The two Tmax values. (d) The geometric mean time to maximum (Tmax ) is calculated.

(e) Geometric mean times are calculated for fixed percentage “slices” of each curve, i.e., at fixed percentages of Imax . The rising phase is “sliced” at 95% of Imax and the time values determined. A similar value will be determined at 95% of maximum for the falling phase. (f) The geometric mean times at each percent of maximum are plotted to generate the composite curve.

12 products were evaluated on 8 different attributes in triplicate sessions, requiring about 300 TI curves from each panelist! Texture panelists were given over 20 sessions of training, although only a few of the final sessions were specifically devoted to practice

with the TI procedure. Obviously, this kind of extensive research program requires a significant time and resource commitment. Data were collected using a computer-assisted rating program, where mouse movement was linked to

8.5 Data Analysis Options

191

the position of a cursor on a 10 cm 10-point scale. The authors noted a number of issues with the data records due to “mouse artifacts” or other problems. These included sudden unintended movement of the mouse leading to false peaks or ratings after the end of the sensation, mouse blockage leading to unusable records, and occasional inaccurate positioning by panelists causing data that did not reflect their actual perceptions. Such ergonomic difficulties are not uncommon in TI studies although they are rarely reported or discussed. Even with this highly trained panel, from 1 to 3% of the records needed to be discarded or manually corrected due to artifacts or inaccuracies. Sensory professionals should not assume that just because they have a computer-assisted TI system, the human factors in mouse and machine interactions will always work smoothly and as planned. Examples of response artifacts are shown in Fig. 8.6. T1 T2

T3

False Plateau

INTENSITY

“Mouse artifacts”

TIME (sec)

Fig. 8.6 Response artifacts in TI records. The solid line shows some perhaps unintended mouse movement (muscle spasm?) near the peak intensity. The dashed line shows a bump in the mouse after sensation ceased and returned to zero. The dotted line illustrates an issue in determining at what point the intensity plateau has ended. The short segment between T1 and T2 may have simply been an adjustment of the mouse after the sudden rise, when the panelist felt they overshot the mark. The actual end of the plateau might more reasonably be considered to occur at T3. (see Lallemand et al. 1999).

Lallemand and coworkers noted that TI curves often took a shape in which the sensation rose to a plateau near peak intensity for a period during which intensity ratings changed very little and then fell to the baseline. They reasoned that a simple geometric approximation by a trapezoid shape might suffice for extracting curve parameters and finding the area under the curve (not unlike the trapezoidal approximation method used for integration in calculus). In principle, four points could

be defined that would describe the curve: the onset time, the time at the intensity maximum or beginning of the plateau, the time at which the plateau ended and the decreasing phase began, and the time at which sensation stopped. These landmarks are those originally proposed by Lui and MacFie (1990). In practice, these points turned out to be more difficult to estimate than expected, so some compromises were made. For example, some records would show a gradually decreasing record during the “plateau” and before the segment with a more rapidly falling slope was evident. How much of a decrease would justify the falling phase or conversely, how little of a decrease would be considered still part of the plateau (see Fig. 8.6)? Also, what should one do if the panelist did not return to zero sensation or unintentionally bumped the mouse after reaching zero? In order to solve these issues, the four points were chosen at somewhat interior sections of the curve, namely the times at 5% of the intensity maximum for the onset and endpoint of the trapezoid, and the times at 90% of the intensity maximum for the beginning and end of the plateau. This approximation worked reasonably well, and its application to a hypothetical record is shown in Fig. 8.7. Given the almost 3,000 TI curves in this one study, the trapezoid points were not mapped by hand or by eye, but a special program written to extract the points. However, for smaller experiments it should be quite feasible to do this kind of analysis “by hand” on any collection of graphed records. The establishment of the four trapezoid vertices now allows extraction of the six basic TI curve parameters for statistical analysis (5 and 90% of maximum intensity points, the four times at those points), as well as the intensity maximum from the original record, and derived (secondary) parameters such as rising and falling slopes and the total area under the curve. Note that the total area becomes simply the sum of the two triangles and the rectangle described by the plateau. These are shown in the lower section of Fig. 8.7. A composite trapezoid can be drawn from the averaged points. The utility and validity of the method was illustrated in one sample composite record, showing the fruity flavor intensity from two ice creams differing in fat content. Consistent with what might be expected from the principles of flavor release, the higher fat sample had a slower and more delayed rise to the peak (plateau) but a longer duration. This would be pdicted if the higher fat level was better able to sequester

192

8

Time-Intensity Methods

with their single-point intensity estimates. Consistent with the latter notion, the profiling scores could be better modeled by a combination of several of the TI parameters. The simplicity and validity of this analysis method suggests that it should find wider application in industrial settings.

Intensity trapezoidal approxiamtion to T-I curve

A,D = points at 5% of maximum B,C = points at 90% of maximum

8.5.4 Analysis by Principal Components Time DERIVED PARAMETERS;

B

C

Rd Ri Ai

Am

Ad

A

D Time Di

Dm

Dd

Fig. 8.7 The trapezoidal method of Lallemand et al. (1999) for assessing curve parameters on TI records. The upper panel shows the basic scheme in which four points are found when the initial 5% of the intensity maximum (Imax ) occurs, when 90% of Imax is first reached on the ascending segment, when the plateau is finished at 90% of Imax on the descending phase and the endpoint approximation at 5% of Imax on the descending phase. The lower panel shows the derived parameters, namely Ri , Ai , and Di for the rate (slope), area, and duration of the initial rising phase; Am and Dm for the area and duration of the middle plateau section; and Rd , Ad , and Dd for the rate (slope), area, and duration of the falling phase. A total duration can be found from the sum of Di , Dm , and Dd . The total area is given by the sum of the A parameters or by the formula for the area of a trapezoid: Total area = (I90 -I5 ) (2Dm + Di + Dd )/2. (Height times the sum of the two parallel segments, then pided by 2).

a lipophilic or nonpolar flavor compound and thus delay the flavor release. They also examined the correlation with a traditional texture descriptive analysis and found very low correlations of inpidual TI parameters with texture profiling mean scores. This would be expected if the TI parameters were contributing unique information or if the texture profilers were integrating a number of time-dependent events in coming up

Another analysis uses principal components analysis (PCA, discussed in Chapter 18) (van Buuren, 1992). Briefly, PCA is a statistical method that “bundles” groups of correlated measurements and substitutes a new variable (a factor or principal component) in place of the original variables, thus simplifying the picture. In studying the time-intensity curves for bitterness or different brands of lager beer, van Buuren noticed that inpiduals once again produced their own characteristic “style” of curve shape. Most people showed a classic TI curve shape, but some subjects were classified as “slow starters,” with a delayed peak and some showed a tendency to persist and not come back down to baseline within the test period. Submission of the data to PCA allowed the extraction of a “principal curve” which captured the majority trend. This showed a peaked TI curve and a gradual return to baseline. The second principal curve captured the shape of the minority trends, with slow onset, a broad peak and slow decline without reaching baseline. The principal curves were thus able to extract judge trends and provide a cleaned-up view of the primary shape of the combined data (Zimoch and Gullet, 1997). Although a PCA program may extract a number of principal components, not all may be practically meaningful (for an example, see Reinbach et al., 2009), and the user should examine each one for the story it tells. Reasonable questions are whether the component reflects something important relative to the simple TI curve parameters, and whether it shows any patterns related to inpidual differences among panelists. Dijksterhuis explored the PCA approach in greater detail (Dijksterhuis, 1993; Dijksterhuis and van den Broek, 1995; Dijksterhuis et al., 1994). Dijksterhuis (1993) noted that the PCA method as applied by van Buuren was not discriminating of different bitter stimuli. An alternative approach was “non-centered PCA” in which curve height information was retained during

8.6 Examples and Applications

data processing, rather than normalizing curves to a common scale. The non-centered approach works on the raw data matrix. Stimuli or treatments were better distinguished. The first principal curve tends to look like the simple average, while the second principal curve contains rate information such as accelerations or inflection points (Dijksterhuis et al., 1994). This could be potentially useful information in differentiating the subtle patterns of TI curves for different flavors. The PCA approach also involves the possibility of generating weights for different assessors that indicate the degree to which they contribute to the different principal curves. This could be an important tool for differentiating outliers in the data or panelists with highly unusual TI signatures (Peyvieux and Dijksterhuis, 2001).

8.6 Examples and Applications A growing number of studies have used TI methods for the evaluations of flavor, texture, flavor release, hedonics, and basic studies of the chemical senses. A short review of these studies follows in this section, although the reader is cautioned that the list is not exhaustive. We have cited a few of the older studies to give credit to the pioneers of this field as well as some of the newer applications. The examples are meant to show the range of sensory studies for which TI methods are suitable.

8.6.1 Taste and Flavor Sensation Tracking A common application of continuous time-intensity scaling is tracking the sensation rise and decay from important flavor ingredients, such as sweeteners (Swartz, 1980). An early study of Jellinek and Pangborn reported that addition of salt to sucrose extended the time-intensity curve and made the taste “more rounded” in their words (Jellinek, 1964). One of the salient characteristics of many intensive or non-carbohydrate sweeteners is their lingering taste that is different from that of sucrose. Time-intensity tracking of sweet tastes was an active area of study (Dubois and Lee, 1983; Larson-Powers and Pangborn, 1978; Lawless and Skinner, 1979; Yoshida, 1986) and

193

194

8

Cliff and Noble, 1990; Matysiak and Noble, 1991). Using these techniques, enhancement of sweetness by fruity volatiles has been observed, and differences in the interactions were seen for different sweetening agents.

8.6.2 Trigeminal and Chemical/Tactile Sensations Reactions to other chemical stimuli affecting irritation or tactile effects in the mouth have been a fertile area for time-related sensory measurements. Compounds such as menthol produce extended flavor sensations and the time course is concentration dependent (Dacanay, 1990; Gwartney and Heymann, 1995). A large number of studies of the burning sensation from hot pepper compounds have used time-intensity scaling, using repeated ratings at discrete time intervals as well as continuous tracking (Cliff and Heymann, 1993a; Green, 1989; Green and Lawless, 1991; Lawless and Stevens, 1988; Stevens and Lawless, 1986). Given the slow onset and extended time course of the sensations induced by even a single sample of a food containing hot pepper, this is a highly appropriate application for time-related judgments. The repeated ingestion paradigm has also been used to study the short- and long-term desensitization to irritants such as capsaicin and zingerone (Prescott and Stevenson, 1996). The temporal profile of different irritative spice compounds is an important point of qualitative differentiation (Cliff and Heymann, 1992). Reinbach and colleagues used TI tracking to study the oral heat from capsaicin in various meat products (Reinbach et al., 2007, 2009) as well as the interactions of oral burn with temperature (see also Baron and Penfield, 1996). When examining the decay curves following different pepper compounds, different time courses can be fit by different decay constants in a simple exponential curve of the form R = Ro e−kt

Time-Intensity Methods

from peak during the decay portion of the time curve (Lawless, 1984). Another chemically induced tactile or feeling factor in the mouth that has been studied by TI methods is astringency. Continuous tracking has been applied to the astringency sensation over repeated ingestions (Guinard et al., 1986). Continuous tracking can provide a clear record of how sensations change during multiple ingestions, and how flavors may build as subsequent sips or tastings add greater sensations on an existing background. An example of this is shown in Fig. 8.8 where the saw-tooth curve shows an increase and builds astringency over repeated ingestions (Guinard et al., 1986). Astringency has also been studied using repeated discrete-point scaling. For example, Lawless and coworkers were able to show differences in the time profiles of some sensory sub-qualities related to astringency, namely dryness, roughness in the mouth, and puckery tightening sensations (Lawless et al., 1994; Lee and Lawless, 1991) depending on the astringent materials being evaluated.

Fig. 8.8 Continuous tracking record with multiple ingestions producing a “sawtooth” curve record. The abscissa shows the time axis in seconds and the ordinate the mean astringent intensity from 24 judges. The dashed curve is from a 15 ml sample of a base wine with 500 gm/l added tannin and the solid curve from the base wine with no added tannin. Sample intake and expectoration are indicated by stars and arrows, respectively. From Guinard et al. (1986), reprinted with permission of the American Society of Enology and Viticulture.

(8.1)

or

8.6.3 Taste and Odor Adaptation ln R = ln Ro − kt

(8.2)

where Ro is the peak intensity and t is time, k is the value determining how rapid the sensation falls off

The measurement of a flavor sensation by tracking over time has a close parallel in studies of taste and odor adaptation (Cain, 1974; Lawless and Skinner,

8.6 Examples and Applications

1979; O’Mahony and Wong, 1989). Adaptation may be defined as the decrease in responsiveness of a sensory system to conditions of constant stimulation providing a changing “zero point” (O’Mahony, 1986). Adaptation studies have sometimes used discrete, single bursts of stimulation (e.g., Meiselman and Halpern, 1973). An early study of adaptation used a flowing system through the subject’s entire mouth using pipes inserted through a dental impssion material that was held in the teeth (Abrahams et al., 1937). Disappearance of salt taste was achieved in under 30 s for a 5% salt solution, although higher concentrations induced painful sensations over time that were confusing to the subjects and sometimes masked the taste sensation. By flowing a continuous stream over a section of the subject’s tongue or stabilizing the stimulus with wet filter paper, taste often disappears in under a few minutes (Gent, 1979; Gent and McBurney, 1978; Kroeze, 1979; McBurney, 1966). Concentrations above the adapting level are perceived as having the characteristic taste of that substance, e.g., salty for NaCl. Concentrations below that level, to which pure water is the limiting case, take on other tastes so that water after NaCl, for example, is sour-bitter (McBurney, and Shick, 1971). Under other conditions adaptation may be incomplete (Dubose et al., 1977; Lawless and Skinner, 1979; Meiselman and Dubose, 1976). Pulsed flow or intermittent stimulation causes adaptation to be much less complete or absent entirely (Meiselman and Halpern, 1973).

8.6.4 Texture and Phase Change Tactile features of foods and consumer products have been evaluated using time-related measurements. Phase change is an important feature of many products that undergo melting when eaten. These include both frozen products like ice cream and other dairy desserts and fatty products with melting points near body temperature such as chocolate. Using the chartrecording method for time-intensity tracking, Moore and Shoemaker (1981) evaluated the degree of coldness, iciness, and sensory viscosity of ice cream with different degrees of added carboxymethyl cellulose. The added carbohydrate shifted the time for peak intensity of iciness and extended the sensations of coldness on the tongue. Moore and Shoemaker also

195

8.6.5 Flavor Release A potentially fertile area for the application of timerelated sensory measurements is in flavor release from foods during eating. Not only are texture changes

196

8

Time-Intensity Methods

cheese systems (Pionnier et al., 2004) using a discrete point TI method.

8.6.6 Temporal Aspects of Hedonics Since the pleasantness or appeal of a sensory characteristic is largely dependent on its intensity level, it is not surprising that one’s hedonic reaction to a product might shift over time as the strength of a flavor waxes and wanes. Time-related shifts in food likes and dislikes are well known. In the phenomenon known as alliesthesia, our liking for a food depends a lot on whether we are hungry or replete (Cabanac, 1971). The delightful lobster dinner we enjoyed last night may not look quite so appealing as leftovers at lunchtime the next day. Wine tasters may speak of a wine that is “closed in the glass, open on the palate, and having a long finish.” Accompanying such a description is the implicit message that this particular wine got better over the course of the sensory experience. Given the shorter time span of flavor and texture sensations in the mouth, we can ask whether there are shifts in liking and disliking. This has been measured in several studies. Taylor and Pangborn (1990) examined liking for chocolate milk with varying degrees of milk fat. Different inpidual trends were observed in liking for different concentrations, and this affected the degree of liking expssed over time. Another example of hedonic TI scaling was in a study of the liking/disliking for the burning oral sensation from chili (hot ) pepper (Rozin et al., 1982). They found different patterns of temporal shifting in liking as the burn rose and declined. Some subjects liked the burn at all time intervals, some disliked the burn at all time intervals, and some shifted across neutrality as strong burns became more tolerable. This method was revisited by Veldhuizen et al. (2006) who used a simple bipolar line scale for pleasantness and had subjects evaluated both intensity and hedonic reactions to a citrus beverage flowed over the tongue from a computer-controlled delivery system (see Fig. 8.2 for an early example of this kind of device). Note that with a bipolar hedonic scale, the mouse and cursor positions must begin at the center of the scale and not the lower end as with intensity scaling. The authors found a delayed pleasantness response compared to the intensity tracking, a similar time to maximum, but an unexpectedly quicker offset of response for pleasantness tracking. Some panelists

8.7 Issues

produced a double-peaked pleasantness response, as the sensation could rise in pleasantness, but then become too intense, but become more pleasant again as adaptation set in and the perceived strength decreased.

8.7 Issues Sensory scientists who wish to use time-intensity methods for any particular study need to weigh the potential for obtaining actionable information against the cost and time involved in collecting these data. Some orientation or training is required (Peyvieux and Dijksterhuis, 2001) and in some published studies, the training and practice is quite extensive. For example, Zimoch and Gullet (1997) trained their meat texture panel for 12 h. Panelists must be trained to use the response device and sufficiently practiced to feel comfortable with the requirements of the task in terms of maintaining focused attention to momentary sensation changes. With the use of online data collection the tabulation and processing of information is generally not very labor intensive; but without computer-assisted collection the time involved can be enormous. Even with computerized systems, the data collection is not foolproof. Responses may be truncated or fail to start at zero in some records (Liu and MacFie, 1990; McGowan and Lee, 2006) making automatic averaging of records infeasible. In one study of melting behavior (Lawless et al., 1996), some subjects mistakenly returned the indicating cursor to zero as soon as the product was completely melted, instead of leaving the cursor on maximum, producing truncated records. Such unexpected events remind us to never assume that panelists are doing what you think they should be doing. A fundamental issue is information gain. In the case where changes in duration are observed at equal maximum intensities, it can be argued that the traditional scaling might have missed important sensory differences. For example, TI can capture information such as when the TI curves cross over, e.g., the interesting case when a product with a lower peak intensity has a longer duration (e.g., Lallemand et al., 1999; Lawless et al., 1996). However, this pattern is not often seen. Usually products with stronger peak height have longer durations. In general there is a lot of redundant information in TI parameters. Lundahl (1992) studied the correlation of 15 TI parameters associated with TI

197

curves’ shapes, sizes, and rates of change. Curve size parameters were highly correlated and usually loaded on the first principal component of a PCA, capturing most of the variance (see also Cliff and Noble, 1990). Curve size parameters, including peak height, were correlated with simple category ratings of the same beverages. So an open question for the sensory scientist is whether there is any unique information in the TI parameters extracted from the records and whether there is information gain over what would be provided by more simple direct scaling using a single intensity rating. A potential problem in time-intensity methods is that factors affecting response behavior are not well understood. In TI measurements, there are dynamic physical processes (chewing, salivary dilution) leading to changes in the stimulus and resulting sensations (Fischer et al., 1994). A second group of processes concerns how the participant translates the conscious experience into an overt response, including a decision mechanism and motoric activation (Dijksterhuis, 1996). The notion that TI methods provide a direct link from the tongue of the subject to the hand moving the mouse is a fantasy. Even in continuous tracking, there must be some decision process involved. There is no information as to how often a panelist in the continuous procedure reflects upon the sensation and decides to change the position of the response device. Decisions are probably not continuous even though some records from some subjects may look like smooth curves. An indication that response tendencies are important is when the conditions of stimulation are held constant, but the response task changes. For example, using the graphic chart-recorder method, Lawless and Skinner (1979) found median durations for sucrose intensity that were 15-35% shorter than the same stimuli rated using repeated category ratings. Why would the different rating methods produce apparently different durations? Very different patterns may be observed when taste quality and intensity are tracked. Halpern (1991) found that tracked taste quality of 2 mM sodium saccharin had a delayed onset (by 400 ms) compared with tracked intensity. This might be understandable from the point of requiring a more complex decision process in the case of tracking intensity. However, it still alerts us to the fact that the behavior probably trails the actual experience by some unknown amount. What is more surprising in Halpern’s data is that tracked quality also stopped well before tracked intensity (by

198

600 ms). Can it be possible that subjects are still experiencing a taste of some definable intensity and yet the quality has disappeared? Or is there another response generation process at work in this task? A third area where potential response biases can operate is in contextual effects. Clark and Lawless (1994) showed that common contextual affects like successive contrast also operated with TI methods, as they do with other scaling tasks. Also, some ratings could be enhanced when a limited number of scales were used by subjects. As observed in singlepoint scaling, enhancement of sweetness by fruity flavors tends to occur when only sweetness is rated. When the fruity flavor is also rated, the sweetness enhancement often disappears (Frank et al., 1989), an effect sometimes referred to as halo dumping or simply “dumping.” Using the discrete-point version of TI scaling, so that multiple attributes could be rated, Clark and Lawless showed a similar effect. This is potentially troublesome for the continuous tracking methods, since they often limit subjects to responding to only one attribute at a time. This may explain in part why sweetness enhancement by flavors can occur so readily in TI studies (e.g., Matysiak and Noble, 1991). A final concern is the question of whether the bounded response scales often used in TI measurement produces any compssion of the differences among products. In analog tracking tasks, there is a limit as to how far the joystick, mouse, lever, dial, or other response device can be moved. With some practice, judges learn not to bump into the top. Yet the very nature of the tracking response encourages judges to sweep a wide range of the response scale. If this were done on every trial, it would tend to attenuate the differences in maximum tracked intensity between products. As an example, Overbosch et al. (1986, see Fig. 2) showed curves for pentanone where doubling the concentration changed peak heights by only about 8%. A similar sort of compssion is visible in Lawless and Skinner’s (1979) data for sucrose, compared to the psychophysical data in the literature.

8.8 Conclusions In most cases TI parameters show similar statistical differentiation as compared to traditional scales, but this is not universally the case (e.g., Moore and Shoemaker, 1981).

8

Time-Intensity Methods

Many sensory evaluation researchers have supported increased application of time-intensity measurements for characterization of flavor and texture sensations. In particular, the method was championed by Lee and Pangborn, who argued that the methods provide detailed information not available from single estimates of sensation intensity (Lee, 1989; Lee and Pangborn, 1986). TI methods can provide rate-related, duration, and intensity information not available from traditional scaling. However, the utility of the methods must be weighed against the enhanced cost and complexity in data collection and analysis. In deciding whether to apply TI methods over conventional scaling, the sensory scientist should consider the following criteria: (1) Is the attribute or system being studied known to change over time? Simply eating the food can often settle this issue; in many cases it is obvious. (2) Will the products differ in sensory time course as a function of ingredients, processing, packaging, or other variables of interest? (3) Will the time variation occur in such a way that it will probably not be captured by direct single ratings? (4) Is some aspect of the temporal profile likely to be related to consumer acceptability? (5) Does the added information provided by the technique outweigh any additional costs or time delays in panel training, data acquisition, and data analysis? Obviously, when more answers are positive on these criteria, a stronger case can be made for choosing a TI method from the available set of sensory evaluation tools.

References Abrahams, H., Krakauer, D. and Dallenbach, K. M. 1937. Gustatory adaptation to salt. American Journal of Psychology, 49, 462-469. Ayya, N. and Lawless, H. T. 1992. Qualitative and quantitative evaluation of high-intensity sweeteners and sweetener mixtures. Chemical Senses, 17, 245-259. Baron, R. F. and Penfield, M. P. 1996. Capsaicin heat intensity – concentration, carrier, fat level and serving temperature effects. Journal of Sensory Studies, 11, 295-316. Barylko-Pikielna, N., Mateszewska, I. and Helleman, U. 1990. Effect of salt on time-intensity characteristics of bread. Lebensmittel Wissenschaft und Technologie, 23, 422-426.

References Birch, G. G. and Munton, S. L. 1981. Use of the “SMURF” in taste analysis. Chemical Senses, 6, 45-52. Bloom, K., Duizer, L. M. and Findlay, C. J. 1995. An objective numerical method of assessing the reliability of time- intensity panelists. Journal of Sensory Studies, 10, 285-294. Bonnans, S. and Noble, A. C. 1993. Effect of sweetener type and of sweetener and acid levels on temporal perception of sweetness, sourness and fruitiness. Chemical Senses, 18, 273-283. Brandt, M. A., Skinner, E. Z. and Coleman, J. A. 1963. Texture profile method. Journal of Food Science, 28, 404-409. Brown, W. E., Landgley, K. R., Martin, A. and MacFie, H. J. 1994. Characterisation of patterns of chewing behavior in human subjects and their influence on texture perception. Journal of Texture Studies, 15, 33-48. Butler, G., Poste, L. M., Mackie, D. A., and Jones, A. 1996. Time-intensity as a tool for the measurement of meat tenderness. Food Quality and Preference, 7, 193-204. Cabanac, M. 1971. Physiological role of pleasure. Science, 173, 1103-1107. Cain, W. S. 1974. Perception of odor intensity and time-course of olfactory adaptation. ASHRAE transactions, 80, 53-75. Clark, C. C. and Lawless, H. T. 1994. Limiting response alternatives in time-intensity scaling: An examination of the Halo-Dumping effect. Chemical Senses, 19, 583-594. Cliff, M. 1987. Temporal perception of sweetness and fruitiness and their interaction in a model system. MS Thesis, University of California, Davis, USA. Cliff, M. and Heymann, H. 1992. Descriptive analysis of oral pungency. Journal of Sensory Studies, 7, 279-290. Cliff, M. and Heymann, H. 1993a. Time-intensity evaluation of oral burn. Journal of Sensory Studies, 8, 201-211. Cliff, M. and Heymann, 1993b. Development and use of time- intensity methodology for sensory evaluation: A review. Food Research International, 26, 375-385. Cliff, M. and Noble, A. C. 1990. Time-intensity evaluation of sweetness and fruitiness in a model solution. Journal of Food Science, 55, 450-454. Dacanay, L. 1990. Thermal and concentration effects on temporal sensory attributes of L – menthol. M.S. Thesis, University of California, Davis, USA. de Roos, K. B. 1990. Flavor release from chewing gums. In: Y. Bessiere and A. F. Thomas (eds.), Flavour Science and Technology. Wiley, Chichester, pp. 355-362. DeRovira, D. 1996. The dynamic flavor profile method. Food Technology, 50, 55-60. Dijksterhuis, G. 1993. Principal component analysis of time- intensity bitterness curves. Journal of Sensory Studies, 8, 317-328. Dijksterhuis, G. 1996. Time-intensity methodology: Review and pview. Proceedings, COST96 Meeting: Interaction of Food Matrix with Small Ligands Influencing Flavour and Texture, Dijon, France, November 20, 1995. Dijksterhuis, G. and van den Broek, E. 1995. Matching the shape of time-intensity curves. Journal of Sensory Studies, 10, 149-161. Dijksterhuis, G., Flipsen, M. and Punter, P. H. 1994. Principal component analysis of time-intensity data. Food Quality and Preference, 5, 121-127. DuBois, G. E. and Lee, J. F. 1983. A simple technique for the evaluation of temporal taste properties. Chemical Senses, 7, 237-247.

200 DuBois (eds.), Sweeteners: Discovery, Molecular Design and Chemoreception. ACS Symposium Series #450. American Chemical Society, Washington, DC, pp. 277-289. Jellinek, G. 1964. Introduction to and critical review of modern methods of sensory analysis (odor taste and flavor evaluation) with special emphasis on descriptive analysis. Journal of Nutrition and Dietetics, 1, 219-260. Jellinek, G. 1985. Sensory Evaluation of Food, Theory and Practice. Ellis Horwood, Chichester, England. Kroeze, J. H. A. 1979. Masking and adaptation of sugar sweetness intensity. Physiology and Behavior, 22, 347-351. Kuo, Y.-L., Pangborn, R. M. and Noble, A. C. 1993. Temporal patterns of nasal, oral and retronasal perception of citral and vanillin and interactions of these odorants with selected tastants. International Journal of Food Science and Technology, 28, 127-137. Labbe, D., Schlich, P., Pineau, N., Gilbert, F. and Martin, N. 2009. Temporal dominance of sensations and sensory profiling: A comparative study. Food Quality and Preference, 20, 216-221. Lallemand, M., Giboreau, A., Rytz, A. and Colas, B. 1999. Extracting parameters from time-intensity curves using a trapezoid model: The example of some sensory attributes of ice cream. Journal of Sensory Studies, 14, 387-399. Larson-Powers, N. and Pangborn, R. M. 1978. Paired comparison and time-intensity measurements of the sensory properties of beverages and gelatins containing sucrose or synthetic sweeteners. Journal of Food Science, 43, 41-46. Lawless, H. T. 1980. A computerized system for assessing taste intensity over time. Paper psented at the Chemical Senses and Intake Society, Hartford, CT, April 9, 1980. Lawless, H. T. 1984. Oral chemical irritation: Psychophysical properties. Chemical Senses, 9, 143-155. Lawless, H. T. and Clark, C. C. 1992. Psychological biases in time-intensity scaling. Food Technology, 46(11), 81, 84-86, 90. Lawless, H. T. and Skinner, E. Z. 1979. The duration and perceived intensity of sucrose taste. Perception and Psychophysics, 25, 249-258. Lawless, H. T. and Stevens, D. A. 1988. Responses by humans to oral chemical irritants as a function of locus of stimulation. Perception and Psychophysics, 43, 72-78. Lawless, H. T., Corrigan, C. L. and Lee, C. L. 1994. Interactions of astringent substances. Chemical Senses, 19, 141-154. Lawless, H. T., Tuorila, H., Jouppila, K., Virtanen, P. and Horne, J. 1996. Effects of guar gum and microcrystalline cellulose on sensory and thermal properties of a high fat model food system. Journal of Texture Studies 27, 493-516. Le Reverend, F. M., Hidrio, C., Fernandes, A. and Aubry, V. 2008. Comparison between temporal dominance of sensation and time intensity results. Food Quality and Preference, 19, 174-178. Leach, E. J. and Noble, A. C. 1986. Comparison of bitterness of caffeine and quinine by a time-intensity procedure. Chemical Senses, 11, 339-345. Ledauphin, S., Vigneau, E. and Causeur, D. 2005. Functional approach for the analysis of time intensity curves using B-splines. Journal of Sensory Studies, 20, 285-300. Ledauphin, S., Vigneau, E. and Qannari, E. M. 2006. A procedure for analysis of time intensity curves. Food Quality and Preference, 17, 290-295.

8

Time-Intensity Methods

Lee, C. B. and Lawless, H. T. 1991. Time-course of astringent materials. Chemical Senses, 16, 225-238. Lee, W. E. 1985. Evaluation of time-intensity sensory responses using a personal computer. Journal of Food Science, 50, 1750-1751. Lee, W. E. 1986. A suggested instrumental technique for studying dynamic flavor release from food products. Journal of Food Science, 51, 249-250. Lee, W. E. 1989. Single-point vs. time-intensity sensory measurements: An informational entropy analysis. Journal of Sensory Studies, 4, 19-30. Lee, W. E. and Pangborn, R. M. 1986. Time-intensity: The temporal aspects of sensory perception. Food Technology, 40, 71-78, 82. Liu, Y. H. and MacFie, H. J. H. 1990. Methods for averaging time-intensity curves. Chemical Senses, 15, 471-484. Lundahl, D. S. 1992. Comparing time-intensity to category scales in sensory evaluation. Food Technology, 46(11), 98-103. Lynch, J., Liu, Y.-H., Mela, D. J. and MacFie, H. J. H. 1993. A time-intensity study of the effect of oil mouthcoatings on taste perception. Chemical Senses, 18, 121-129. Matysiak, N. L. and Noble, A. C. 1991. Comparison of temporal perception of fruitiness in model systems sweetened with aspartame, aspartame + acesulfame K blend or sucrose. Journal of Food Science, 65, 823-826. McBurney, D. H. 1966. Magnitude estimation of the taste of sodium chloride after adaptation to sodium chloride. Journal of Experimental Psychology, 72, 869-873. McBurney, D. H. and Shick, T. R. 1971. Taste and water taste of 26 compounds for man. Perception and Psychophysics, 11, 228-232. McGowan, B. A. and Lee, S.-Y. 2006. Comparison of methods to analyze time-intensity curves in a corn zein chewing gum study. Food Quality and Preference 17, 296-306. McNulty, P. B. 1987. Flavour release-elusive and dynamic. In: J. M. V. Blanshard and P. Lillford (eds.), Food Structure and Behavior. Academic, London, pp. 245-258. McNulty, P. B. and Moskowitz, H. R. 1974. Intensity -time curves for flavored oil-in-water emulsions. Journal of Food Science, 39, 55-57. Meiselman, H. L. 1968. Magnitude estimation of the time course of gustatory adaptation. Perception and Psychophysics, 4, 193-196. Meiselman, H. L. and Dubose, C. N. 1976. Failure of instructional set to affect completeness of taste adaptation. Perception and Psychophysics, 19, 226-230. Meiselman, H. L. and Halpern, B. P. 1973. Enhancement of taste intensity through pulsatile stimulation. Physiology and Behavior, 11, 713-716. Moore, L. J. and Shoemaker, C. F. 1981. Sensory textural properties of stabilized ice cream. Journal of Food Science, 46, 399-402, 409. Neilson, A. J. 1957. Time-intensity studies. Drug and Cosmetic Industry, 80, 452-453, 534. O’Keefe, S. F., Resurreccion, A. P., Wilson, L. A. and Murphy, P. A. 1991. Temperature effect on binding of volatile flavor compounds to soy protein in aqueous model systems. Journal of Food Science, 56, 802-806. O’Mahony, M. 1986. Sensory adaptation. Journal of Sensory Studies, 1, 237-257.

References O’Mahony, M. and Wong, S.-Y. 1989. Time-intensity scaling with judges trained to use a calibrated scale: Adaptation, salty and umami tastes. Journal of Sensory Studies, 3, 217-236. Ott, D. B., Edwards, C. L. and Palmer, S. J. 1991. Perceived taste intensity and duration of nutritive and non-nutritive sweeteners in water using time-intensity (T-I) evaluations. Journal of Food Science, 56, 535-542. Overbosch, P. 1987. Flavour release and perception. In: M. Martens, G. A. Dalen and H. Russwurm (eds.), Flavour Science and Technology. Wiley, New York, pp. 291-300. Overbosch, P., Van den Enden, J. C., and Keur, B. M. 1986. An improved method for measuring perceived intensity/time relationships in human taste and smell. Chemical Senses, 11, 315-338. Owen, W. J. and DeRouen, T. A. 1980. Estimation of the mean for lognormal data containing zeroes and left-censored values, with application to the measurement of worker exposure to air contaminants. Biometrics, 36, 707-719. Pangborn, R. M. and Koyasako, A. 1981. Time-course of viscosity, sweetness and flavor in chocolate desserts. Journal of Texture Studies, 12, 141-150. Pangborn, R. M., Lewis, M. J. and Yamashita, J. F. 1983. Comparison of time-intensity with category scaling of bitterness of iso-alpha-acids in model systems and in beer. Journal of the Institute of Brewing, 89, 349-355. Peyvieux, C. and Dijksterhuis, G. 2001. Training a sensory panel for TI: A case study. Food Quality and Preference, 12, 19-28. Pineau, N., Schlich, P., Cordelle, S., Mathonniere, C., Issanchou, S., Imbert, A., Rogeaux, M., Eteviant, P. and Köster, E. 2009. Temporal dominance of sensations: Construction of the TDS curves and comparison with time- intensity. Food Quality and Preference, 20, 450-455. Pionnier, E., Nicklaus, S., Chabanet, C., Mioche, L., Taylor, A. J., LeQuere, J. L. and Salles, C. 2004. Flavor perception of a model cheese: relationships with oral and physico-chemical parameters. Food Quality and Preference, 15, 843-852. Prescott, J. and Stevenson, R. J. 1996. Psychophysical responses to single and multiple psentations of the oral irritant zingerone: Relationship to frequency of chili consumption. Physiology and Behavior, 60-617-624. Reinbach, H. C., Toft, M. and Møller, P. 2009. Relationship between oral burn and temperature in chili spiced pork patties evaluated by time-intensity. Food Quality and Preference, 20, 42-49. Reinbach, H. C., Meinert, L., Ballabio, D., Aayslyng, M. D., Bredie, W. L. P., Olsen, K. and Møller, P. 2007. Interactions between oral burn, meat flavor and texture in chili spiced pork patties evaluated by time-intensity. Food Quality and Preference, 18, 909-919. Rine, S. D. 1987. Computerized analysis of the sensory properties of peanut butter. M. S. Thesis, University of California, Davis, USA.

Chapter 9

Context Effects and Biases in Sensory Judgment

Abstract Human judgments about a sensation or a product are strongly influenced by items that surround the item of interest, either in space or in time. This chapter shows how judgments can change as a function of the context within which a product is evaluated. Various contextual effects and biases are described and categorized. Some solutions and courses of action to minimize these biases are psented. By such general principles of action as these everything looked at, felt, smelt or heard comes to be located in a more or less definite position relatively to other collateral things either actually psented or only imagined as possibly there. – James (1913, p. 342)

Contents

9.6.3 Positional or Order Bias . . . . . . Antidotes . . . . . . . . . . . . . . . . 9.7.1 Avoid or Minimize . . . . . . . . . 9.7.2 Randomization and Counterbalancing 9.7.3 Stabilization and Calibration . . . . 9.7.4 Interptation . . . . . . . . . . . 9.8 Conclusions . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . 9.7

9.1 9.2

9.3

9.4

9.5

9.6

Introduction: The Relative Nature of Human Judgment . . . . . . . . . . . . . . . . . . Simple Contrast Effects . . . . . . . . . . . 9.2.1 A Little Theory: Adaptation Level . . . . 9.2.2 Intensity Shifts . . . . . . . . . . . . 9.2.3 Quality Shifts . . . . . . . . . . . . . 9.2.4 Hedonic Shifts . . . . . . . . . . . . 9.2.5 Explanations for Contrast . . . . . . . . Range and Frequency Effects . . . . . . . . . 9.3.1 A Little More Theory: Parducci’s Range and Frequency Principles . . . . . . . . 9.3.2 Range Effects . . . . . . . . . . . . . 9.3.3 Frequency Effects . . . . . . . . . . . Biases . . . . . . . . . . . . . . . . . . . . 9.4.1 Idiosyncratic Scale Usage and Number Bias . . . . . . . . . . . 9.4.2 Poulton’s Classifications . . . . . . . . 9.4.3 Response Range Effects . . . . . . . . 9.4.4 The Centering Bias . . . . . . . . . . Response Correlation and Response Restriction . . . . . . . . . . . . . . . . . 9.5.1 Response Correlation . . . . . . . . . 9.5.2 “Dumping” Effects: Inflation Due to Response Restriction in Profiling . . . 9.5.3 Over-Partitioning . . . . . . . . . . . Classical Psychological Errors and Other Biases . . . . . . . . . . . . . . . . . . . . 9.6.1 Errors in Structured Sequences: Anticipation and Habituation . . . . . . . . . . . . 9.6.2 The Stimulus Error . . . . . . . . . .

203 206 206 207 207 208 209 210 210 210 211 212 212 213 214 215 216 216 217 218 218 218 219

. . . . . . . .

. . . . . . . .

219 219 219 220 221 222 222 223

9.1 Introduction: The Relative Nature of Human Judgment This chapter will discuss context effects and common biases that can affect sensory judgments. Context effects are conditions in which the judgment about a product, usually a scaled rating, will shift depending upon factors such as the other products that are evaluated in the same tasting session. A mediocre product evaluated in the context of some poor-quality items may seem very good in comparison. Biases refer to tendencies in judgment in which the response is influenced in some way to be an inaccurate reflection of the actual sensory experience. In magnitude estimation ratings, for example, people have a tendency to use numbers that are multiples of 2, 5, and 10, even though they can use any number or fraction they wish. At the

H.T. Lawless, H. Heymann, Sensory Evaluation of Food, Food Science Text Series, DOI 10.1007/978-1-4419-6488-5_9, © Springer Science+Business Media, LLC 2010

203

204

end of the chapter, some solutions to these problems are offered, although a sensory scientist should realize that we can never totally eliminate these factors. In fact, they are of interest and deserve study on their own for what they can tell us about human sensory and cognitive processes. An axiom of perceptual psychology has it that humans are very poor absolute measuring instruments but are very good at comparing things. For example, we may have difficulty estimating the exact sweetness level of our coffee, but we have little trouble in telling whether more sugar has been added to make it sweeter. The question arises, if people are prone to making comparisons, how can they give ratings when no comparison is requested or specified? For example, when asked to rate the perceived firmness of a food sample, how do they judge what is firm versus what is soft? Obviously, they must either choose a frame of reference for the range of firmness to be judged or be trained with explicit reference standards to understand what is high and low on the response scale. In other words, they must relate this sensory judgment to other products they have tried. For many items encountered in everyday life, we have established frames of reference based on our experiences. We have no trouble forming an image of a “large mouse running up the trunk of a small elephant” because we have established frames of reference for what constitutes the average mouse and the average elephant. In this case the judgment of large and small is context dependent. Some people would argue that all judgments are relative. This dependence upon a frame of reference in making sensory judgments demonstrates the influence of contextual factors in biasing or changing how products are evaluated. We are always prone to see things against a background or pvious experience and evaluate them accordingly. A 40◦ (Fahrenheit) day in Ithaca, New York, in January seems quite mild against the background of the northeastern American winter. However, the same 40◦ C temperature will feel quite cool on an evening in August in the same location. This principle of frame of reference is the source of many visual illusions, where the same physical stimulus causes very different perceptual impssions, due to the context within which it is embedded. Examples are shown in Fig. 9.1. A simple demonstration of context is the visual afterimage effect that gave rise to Emmert’s law (Boring, 1942). In 1881, Emmert formalized a

9

Context Effects and Biases in Sensory Judgment

A

B

C Fig. 9.1 Examples of contextual effects from simple visual illusions. (a) The dumbell version of the Muller-Lyer illusion. (b) The Ebbinghaus illusion. (c) Illusory contours. In this case the contexts induce the perceptions of shapes.

principle of size constancy based on the following effect: Stare for about 30 s at a brightly illuminated colored paper rectangle (it helps to have a small dot to aid in fixation in the center) about a meter away. Then shift your gaze to a white sheet on the table in front of you. You should see the rectangle afterimage in a complementary color and somewhat smaller in size as compared to the original colored rectangle. Next, shift your gaze to a white wall some distance off. The afterimage will now appear much larger, as the brain finds a fixed visual angle at greater distance to repsent larger physical objects. Since the mind does not immediately recognize that the afterimage is just a creation of the visual sensory system, it projects it at the distance of the surface upon which it is “seen.” The more distant frame of reference, then, demands a larger size perception. The close link between sensory judgments and context psents problems for anyone who wants to view ratings as absolute or comparable across different times, sessions, or settings. Even when the actual sensory impssion of two items is the same, we can shift the frame of reference and change the overt behavior of the person to produce a different response. This problem (or principle of sensory function) was glossed

9.1 Introduction: The Relative Nature of Human Judgment

over by early psychophysical scientists. In psychological terms, they used a simple stimulus-response (S-R) model, in which response was considered a direct and unbiased repsentation of sensory experiences. Certain biases were observed, but it was felt that suitable experimental controls could minimize or eliminate them (Birnbaum, 1982; Poulton, 1989). A more modern view is that there are two or three distinct processes contributing to ratings. The first is a psychophysical process by which stimulus energy is translated into physiological events that result in a subjective experience of some sensory intensity. The second, equally important process is the function by which the subjective experience is translated into the observed response, i.e., how the percept is translated onto the rating scale (see Fig. 9.2). Many psychophysical researchers now consider a “judgment function” to

A)

Response

Stimulus

B) Stimulus

Response

Sensation Judgment Process

Psychophysical Process C) Stimulus

Sensation

Psychophysical Process

ENCODED SENSATION, “PERCEPT”

Perceptual Process

Response

Judgment Process

Fig. 9.2 Models for sensory-response processes. (a) The simple stimulus-response model of twentieth-century behavioral psychology. (b) Two processes are involved in sensation and response, a psychophysical process and then a response output or a judgment process in which the participant decides what response to give for that sensation. (c) A more complex model in which the sensation may be transformed before the response is generated. It may exist in short-term memory as an encoded percept, different from the sensation. Contextual effects of simultaneous or sequential stimuli can influence the stimulus-response sequence in several ways. Peripheral physiological effects such as adaptation or mixture inhibition may change the transduction process or other early stages of neural processing. Other stimuli may give rise to separate percepts that are integrated into the final response. Contextual factors may also influence the frame of reference that determines how the response output function will be applied. In some models, an additional step allows transformation of the percept into covert responses that are then translated as a separate step into the overt response R.

205

be an important part of the sequence from stimulus to response (Anderson, 1974, Birnbaum, 1982, McBride and Anderson, 1990; Schifferstein and Frijters, 1992; Ward, 1987). This process is also sometimes referred to as a response output function. A third intermediate step is the conversion of the raw sensory experience into some kind of encoded percept, one that is available to memory for a short time, before the judgment is made (Fig. 9.2c). Given this framework, there are several points at which stimulus context may influence the sensory process. First, of course, the actual sensation itself may change. Many sensory processes involve interaction effects of simultaneous or sequential influences of multiple items. An item may be perceived differently due to the direct influence of one stimulus upon another that is nearby in time or space. Simultaneous color contrast is an example in color vision and some types of inhibitory mixture interactions and masking in taste and smell are similarly hard wired. Quinine with added salt is less bitter than quinine tasted alone, due to the ways that sodium ions inhibit bitterness transduction. Sensory adaptation weakens the perception of a stimulus because of what has pceded it. So the psychophysical process itself is altered by the milieu in which the stimulus is observed, sometimes because of physical effects (e.g., simple buffering of an acid) or physiological effects (e.g., neural inhibition causing mixture suppssion) in the peripheral sensory mechanisms. A second point of influence is when the context shifts the frame of reference for the response output function. That is, two sensations may have the same subjective intensity under two conditions, but because of the way the observer places them along the response continuum (due to different contexts), they are rated differently. A number of studies have shown that contextual factors such as the distribution of stimuli along the physical continuum affect primarily (although not exclusively) the response output function (Mellers and Birnbaum, 1982, 1983). A third process is sometimes added in which the sensation itself is translated into an implicit response or encoded image that may also be affected by context (Fig. 9.2c). This would provide another opportunity to influence the process if contextual factors affect this encoding step. Contextual change can be viewed as a form of bias. Bias, in this sense, is a process that causes a shift or a change in response to a constant sensation. If one situation is viewed as producing a true

206

and accurate response, then the contextual conditions that cause shifts away from this accurate response are “biased.” However, bias need only have a negative connotation if there is a reason to psume that one condition of judgment is more accurate than all others. A broader view is to accept the idea that all judgments are a function of observing conditions and therefore all judgments are biased from one another in different ways. Fortunately, many of these biases and the conditions that cause them are pdictable and well understood, and so they can be eliminated or minimized. At the very least, the sensory practitioner needs to understand how these influences operate so as to know when to expect changes in judgments and ratings. An important endpoint is the realization that few, if any, ratings have any absolute meaning. You cannot say that because a product received a hedonic rating of 7.0 today, it is better than the product that received a rating of 6.5 last week. The context may have changed.

9.2 Simple Contrast Effects By far the most common effect of sensory context is simple contrast. Any stimulus will be judged as more intense in the psence of a weaker stimulus and as less intense in the psence of a stronger stimulus, all other conditions being equal. This effect is much easier to find and to demonstrate than its opposite, convergence or assimilation. For example, an early sensory worker at the Quartermaster Corp., Kamenetsky (1957), noticed that the acceptability ratings for foods seemed to depend upon what other foods were psented during an evaluation session. Poor foods seemed even worse when pceded by a good sample. Convergence is more difficult to demonstrate, although under some conditions a group of items may seem more similar to each other when they are in the psence of an item that is very different from that group (Zellner et al., 2006).

9.2.1 A Little Theory: Adaptation Level As we noted above, a 40◦ day in January (in New York) seems a lot warmer than the same temperature in August. These kinds of effects are pdicted Helson’s

9

Context Effects and Biases in Sensory Judgment

theory of adaptation level. Helson (1964) proposed that we take as a frame of reference the average level of stimulation that has pceded the item to be evaluated. The mild temperature in the middle of a hot and humid summer seems a lot more cool and refreshing than is the mild temperature after a period of cold and icy weather. So we refer to our most recent experiences in evaluating the sensory properties of an item. Helson went on to elaborate the theory to include both immediate and distant pdecessors. That is, he appciated the fact that more recent items tend to have a stronger effect on the adaptation level. Of course, mere reference to the mean value of experience is not always sufficient to induce a contrast effect-it is more influential if the mean value comes to be centered near the middle of the response scale, an example of a centering bias, discussed below (Poulton, 1989). The notion of adaptation, a decrease in responsiveness under conditions of constant stimulation, is a major theme in the literature on sensory processes. Physiological adaptation or an adjustment to the ambient level of stimulation is obvious in light/dark adaptation in vision. The thermal and tactile senses also show profound adaptation effects-we become easily adjusted to the ambient room temperature (as long as it is not too extreme) and we become unaware of tactile stimulation from our clothing. So this mean reference level often passes from consciousness or becomes a new baseline from which deviations in the environment become noticeable. Some workers have even suggested that this improves discrimination-that the difference threshold is smallest right around the adaptation level or physiological zero, in keeping with Weber’s law (McBurney, 1966). Examples of adaptation effects are discussed in Chapter 2 for the senses of taste and smell. In the chemical, thermal, and tactile senses, adaptation is quite profound. However, we need not invoke the concept of neural adaptation to a pceding item or a physiological effect to explain all contrast effects. It may be simply that more or less extreme stimuli change our frame of reference or the way in which the stimulus range and response scales are to be mapped onto one another. The general principle of context is that human observers act like measuring instruments that constantly recalibrate themselves to the experienced frame of reference. What we think of as a small horse may depend upon whether the frame of reference includes Clydesdales, Shetland ponies, or tiny phistoric equine species. The

9.2 Simple Contrast Effects

207

following examples show simple effects of context on intensity, sensory quality, and hedonics or acceptability. Most of these examples are cases of perceptual contrast or a shift in judgment away from other stimuli psented in the same session.

9.2.2 Intensity Shifts Figure 9.3 shows a simple contrast effect of soups with varying salt levels psented in different contexts (Lawless, 1983). The central stimulus in the series was psented either with two lower or two higher concentrations of salt added to a low sodium soup. Ratings of saltiness intensity were made on a simple nine-point category scale. In the lower context, the central soup received a higher rating, and in the higher context, it received a lower rating, analogous to our perception of a mild day in winter (seemingly warmer) versus a mild day in summer (seemingly cooler). Note that the shift is quite dramatic, about two points on the nine-point scale or close to 25% of scale range. A simple classroom demonstration can show a similar shift for the tactile roughness of sandpapers varying in grit size. In the context of a rougher sample, a medium sample will be rated lower than it is in the context of a smoother sample. The effects of simple contrast are not limited to taste and smell.

Mean Category Rating

7 6 5

Low Context High Context

4

3

2

.12

.18

.25

.35

.50

NaCl Concentration (log scale)

Fig. 9.3 Saltiness ratings of soups with added NaCl. The sample at 0.25 M was evaluated in two contexts, one with higher concentrations and one with lower concentrations. The shift is typical of a simple contrast effect of contrast. Replotted from Lawless (1983). Copyright ASTM, used with permission.

Contrast effects are not always observed. In some psychophysical work with long series of stimuli, some item-to-item correlations have been observed. The effects of immediately pceding versus remotely pceding stimuli have been measured and a positive correlation among adjacent responses in the series was found. This can be taken as evidence for a type of assimilation, or underestimation of differences (Ward, 1979, 1987; but see also Schifferstein and Frijters, 1992).

9.2.3 Quality Shifts Visual examples such as color contrast were well known to early psychologists like William James: “Meanwhile it is an undoubted general fact that the psychical effect of incoming currents does depend on what other currents may be simultaneously pouring in. Not only the perceptibility of the object which the current brings before the mind, but the quality of it is changed by the other currents.” (1913, p. 25). A gray line against a yellow background may appear somewhat bluish, and the same line against a blue background may seem more yellowish. Paintings of the renowned artist Josef Albers made excellent use of color contrast. Similar effects can be observed for the chemical senses. During a descriptive panel training period for fragrance evaluation, the terpene aroma compound dihydromyrcenol was psented among a set of woody or pine-like reference materials. The panelists complained that the aroma was too citrus-like to be included among the woody reference materials. However, when the same odor was placed in the context of citrus reference materials, the same panelists claimed that it was far too woody and pine-like to be included among the citrus examples. This contextual shift is shown in Fig. 9.4. In a citrus context, the item is rated as more woody in character than when placed in a woody context. Conversely, ratings for citrus intensity decrease in the citrus context and increase in the woody context. The effect is quite robust and is seen whether or not a rest period is included to undo the potential effects of sensory adaptation. It even occurs when the contextual odor follows the target item and judgments are made after both are experienced (Lawless et al., 1991)! This striking effect is discussed further in Section 9.2.5.

208

9

Context Effects and Biases in Sensory Judgment

CITRUS CONTEXT

MEAN INTENSITY RATING, DIHYDROMYRCENOL

WOODY CONTEXT Citrus/woody scales

Intensity/pleasantness

used during exposure

scales during exposure

9 8 7 6 5 4 3 2 1 CITRUS RATING WOODY RATING

CITRUS RATING WOODY RATING

ODOR CHARACTERISTIC Fig. 9.4 Odor quality contrast noted for the ambiguous terpene aroma compound dihydromyrcenol. In a citrus context, woody ratings increase and citrus character decreases. In a woody context, the woody ratings decrease. The group using different

scales did not rate citrus and woody character during the contextual exposure phase, only overall intensity and pleasantness were rated. From Lawless et al. (1991) by permission of Oxford University Press.

Context effects can also alter how items are identified and characterized. When people categorize speech sounds, repeated exposure to one type of simple phoneme changes the category boundary for other speech sounds. Repeated exposure to the sound of the phoneme “bah,” which has an early voice onset time, can shift the phoneme boundary so that speech sounds near the boundary are more likely classified as “pah” sounds (a later voice onset) (Eimas and Corbit, 1973). Boundary-level examples are shifted across the boundary and into the next category. This shift resembles a kind of contrast effect.

an item of poor quality and less appealing if it followed something of better quality. The effect was known to Beebe-Center (1932), who also attributed it to Fechner in 1898. This kind of contrast has been observed for tastes (Riskey et al., 1979; Schifferstein, 1995), odors (Sandusky and Parducci, 1965), and art (Dolese et al., 2005). Another effect observed in these kinds of experiments is that a contrasting item causes other, generally lower rated stimuli to become more similar or less discriminable, an effect termed condensation (Parker et al., 2002; Zellner et al., 2006;). In the study by Zellner et al. (2006), p-exposure to a good-tasting juice reduced the magnitude of pference ratings among less appealing juices. Mediocre items were both worse and more similar. An example of hedonic shifting was found in a study on the optimization of the saltiness of tomato juice and also the sweetness of a fruit beverage using the method of adjustment (Mattes and Lawless, 1985). When trying to optimize the level of sweetness or saltiness in this study, subjects worked in two directions. In an ascending series, they would concentrate a dilute solution by mixing the beverage with a more concentrated

9.2.4 Hedonic Shifts Changes in the pference or acceptance of foods can be seen as a function of context. Hedonic contrast was a well-known effect to early workers in food acceptance testing (Hanson et al., 1955; Kamenetzky, 1959). An item seems more appealing if it followed

9.2 Simple Contrast Effects

version having the same color, aroma, and other flavor materials (i.e., only sweetness or saltiness was different). In a second descending series, they would be given a very intense sample as the starting point and then dilute down to their pferred level. This effect is shown in Fig. 9.5. The adjustment stops too soon and the discrepancy is remarkable, nearly a concentration range of 2:1. The effect was also robust-it could not be attributed to sensory adaptation or lack of discrimination and persisted even when subjects were financially motivated to try to achieve the same endpoints in both trials. This is a case of affective contrast. When compared to a very sweet or salty starting point, a somewhat lower item seems just about right, but when starting with a relatively sour fruit beverage

Fig. 9.5 Optimized concentrations of salt in tomato juice and sucrose in a fruit beverage. In the trials labeled D, the concentration was diluted from a concentrated version of the test sample. In the trials marked A, the concentration was increased from a dilute version of the test sample. Concentrations of other ingredients were held constant. The contextual shift is consistent with reaching the apparent optimum too soon as if the apparent optimum was shifted in contrast to the starting point. From Mattes and Lawless (1985) with permission.

209

or bland tomato juice, just a little bit of sugar or salt helps quite a bit. The stopping point contrasts with the starting material and seems to be better than it would be perceived in isolation. In an ascending or descending sequence of products, a change in responses that happens too soon is called an “error of anticipation.”

9.2.5 Explanations for Contrast At first glance, one is tempted to seek a physiological explanation for contrast effects, rather than a psychological or a judgmental one. Certainly sensory adaptation to a series of intense stimuli would cause any subsequent test item to be rated much lower. The importance of sensory adaptation in the chemical senses of taste and smell lends some credence to this explanation. However, a number of studies have shown that pcautions against sensory adaptation may be taken, such as sufficient rinsing or time delays between stimuli and yet the context effects persist (Lawless et al., 1991; Mattes and Lawless, 1985; Riskey, 1982). Furthermore, it is difficult to see how sensory adaptation to low-intensity stimuli would cause an increase in the ratings for a stronger item, as adaptation necessarily causes a decrement in physiological responsiveness compared to a no-stimulation baseline. Perhaps the best evidence against a simple adaptation explanation for contrast effects is from the reversed-pair experiments in which the contextual item follows the to-be-rated target item and therefore can have no physiologically adapting effect on it. This paradigm calls for a judgment of the target item from memory after the psentation of the contextual item, in what has been termed a reversed-pair procedure (Diehl et al., 1978). Due to the reversed order, the context effects cannot be blamed on physiological adaptation of receptors, since the contextual item follows rather than pcedes the item to be rated. Reversed-pair effects are seen for shifts in odor quality of aroma compounds like dihydromyrcenol and are only slightly smaller in magnitude than the contextual shift caused when the contextual item comes first (Lawless et al., 1991). The reversed-pair situation is also quite capable of causing simple contrast effects in sensory intensity. A sweetness shift was observed when a higher or a lower sweetness item was interpolated between the tasting and rating (from memory)

210

of a normal-strength fruit beverage (Lawless, 1994). Looking back at Fig. 9.2, it seems more likely that the effect changes the response function. However, not all workers in the field agree. In particular, Marks (1994) has argued that the contextual shifts are much like an adaptation process and that for auditory stimuli this is a peripheral event. It is possible that what changes is not the sensation/experience, but to some encoded version of the sensation, or to some kind of implicit response, not yet verbalized. If a person, when rating, is evaluating some memory trace of the experience, it is possible that this memory, for example, could be altered.

9.3 Range and Frequency Effects Two of the most common factors that can affect ratings are the sensory range of the products to be evaluated and the frequency with which people use the available response options. These factors were nicely integrated into a theory that helped to explain shifts in category ratings. They are also general tendencies that can affect just about any ratings or responses.

9.3.1 A Little More Theory: Parducci’s Range and Frequency Principles Parducci (1965, 1974) sought to go beyond Helson’s (1964) simple idea that people respond to the mean or the average of their sensory experiences in determining the frame of reference for judgment. Instead, they asserted that the entire distribution of items in a psychophysical experiment would influence the judgments of a particular stimulus. If this distribution was denser (bunched up) at the low ends and a lot of weak items were psented, product ratings would shift up. Parducci (1965, 1974) proposed that behavior in a rating task was a compromise between two principles. The first was the range principle. Subjects use the categories to sub-pide the available scale range and will tend to pide the scale into equal perceptual segments. The second was the frequency principle. Over many judgments, people like to use the categories an equal number of times (Parducci, 1974). Thus it is not only

9

Context Effects and Biases in Sensory Judgment

the average level that is important but also how stimuli may be grouped or spaced along the continuum that would determine how the response scale was used. Category scaling behavior could be pdicted as a compromise between the effects of the range and frequency principles (Parducci and Perrett, 1971).

9.3.2 Range Effects The range effect has been known for some time, both in category ratings and other judgments including ratio scaling (Engen and Levy, 1958; Teghtsoonian and Teghtsoonian, 1978). When expanding or shrinking the overall range of products, subjects will map their experiences onto the available categories (Poulton, 1989). Thus short ranges produce steep psychophysical functions and wide ranges produce flatter functions. An example of this can be seen in two published experiments on rating scales (Lawless and Malone, 1986a, b). In these studies, four types of response scales and a number of visual, tactile, and olfactory continua were used to compare the abilities of consumers to use the different scales to differentiate products. In the first study, the consumers had no trouble in differentiating the products and so in the second study, the stimuli were spaced more closely on the physical continua so that the task would be more challenging. However, when the experimenters closed the stimuli in, the range principle took over, and participants used more of the rating scale than expected. In Fig. 9.6, ratings for perceived thickness of (stirred) silicone samples are shown in the wide and narrow stimulus ranges. Note the steepening of the response function. For the same one log unit change in physical viscosity, the range of responses actually doubled from the wide range to the narrower range. Another kind of stimulus range effect occurs with anchor stimuli. Sarris (1967) pviously showed a strong effect of anchor stimuli on the use of rating scales, unless the anchors were very extreme, at which point their influence would tend to diminish, as if they had become irrelevant to the judgmental frame of reference. Sarris and Parducci (1978) found similar effects of both single and multiple end anchors that generally take the form of a contrast effect. For example, a low anchor stimulus, whether rated or unrated, will cause stronger stimuli to receive higher ratings than

9.3 Range and Frequency Effects

211

Fig. 9.6 A simple range effect. When products are psented over a wide range, a shallow psychophysical function is found. Over a narrow range, a steeper psychophysical function will be observed. This is in part due to the tendency of subjects to map the products (once known) onto the available scale range. From Lawless and Malone (1986b), with permission.

they would if no anchor were psented, unless the “anchor” is so extreme as to seem irrelevant. Sarris and Parducci (1978) provided the following analogy: A salesman will consider his commissions to be larger if coworkers receive less than he does. However, he is unlikely to extend or shift his scale of judgment by hearing about others who are in an entirely different bracket of income (1978, p. 39). Whether an outlying product is similar enough to have an influence (or whether it is a “horse of a different color”) should be of concern in product comparisons of perse items.

9.3.3 Frequency Effects The frequency effect is the tendency of people to try to use the available response options about the same number of times across a series of products or stimuli to be rated. The frequency effect can cause shifts that look like simple contrast and also a local steepening of the psychophysical function around points where stimuli were closely spaced or very numerous (compared with less “dense” portions of the stimulus range). The frequency principle dictates that when judging many samples, products that are numerous or bunched at the low or high ends of the distributions tend to be spad out into neighboring categories. This is illustrated in the two panels of Fig. 9.7. The upper panel shows four hypothetical experiments and how products might be bunched in different parts of the range. In the upper left panel, we see how a normal replicated psychophysical experiment would be conducted with

equal psentations of each stimulus level. The common outcome of such a study using category ratings would be a simple linear function of the log of stimulus intensity. However, if the stimulus psentations were more frequent at the high end of the distribution, i.e., negative skew, the upper categories would be overused, and subjects would begin to distribute their judgments into lower categories. If the samples were bunched at the lower end, the lower response categories would be overused and subjects would begin to move into higher categories. If the stimuli were bunched in the midrange, the adjacent categories would be used to take on some of the middle stimuli, pushing extreme stimuli into the ends of the response range, as shown in the panel for a quasi-normal distribution. Such behavior is relevant to applied testing situations. For example, in rating the psence of off-flavors or taints, there may be very few examples of items with high values on the scale and lots of weak (or zero) sensations. The frequency effect may explain why the low end of the scale is used less often than anticipated, and higher mean values are obtained than one would deem appropriate. Another example is screening a number of flavor or fragrance candidates for a new product. A large number of good candidates are sent for testing by suppliers or a flavor development group. Presumably these have been p-tested or at least have received a favorable opinion from a flavorist or a perfumer. Why do they then get only mediocre ratings from the test panel? The high end of the distribution is overrepsented (justifiably so and perhaps on purpose), so the tendency for the panel is to drop into lower categories. This may partly explain why in-house testing

212

9

Number of Stimuli Presented

Rectangular Distribution

Quasinormal

Positive Skew

Negative Skew

Predicted Rating

number of zones or categories. The effect of grouping or spacing products also intensifies as the exposure to the distributions increases. Lawless (1983) showed that the shift that occurred with a negative skew (bunching at the upper end) into lower response categories would intensify as the exposure to the skewed distribution went from none to a single exposure to three exposures. Thus the contextual effects do not suddenly appear but will take hold of the subjects’ behavior as they gain experience with the sample set.

9.4 Biases

Rectangular Distribution

Positive Skew

Quasi-normal Distribution

Context Effects and Biases in Sensory Judgment

Negative Skew

Stimulus Intensity (log scale)

Fig. 9.7 Predictions from the Parducci range-frequency theory. Distributions of stimuli that are concentrated at one part of the perceptual range (upper quartet) will show local steepening of the psychophysical functions (lower quartet). This is due to subjects’ tendencies to use categories with equal frequency, the resulting shifting into adjacent categories from those that are overused.

panels are sometimes more critical or negative than consumers when evaluating the same items. Although the great majority of experiments on the range and frequency effects have been performed with simple visual stimuli, there are also examples from taste evaluation (Lee et al., 2001; Riskey et al., 1979; Riskey, 1982; Schifferstein and Frijters, 1992; Vollmecke, 1987). Schifferstein and Frijters found similar effects of skewed distributions with line-marking responses as seen in pvious studies with category ratings. Perhaps line marking is not a response scale with infinite pisions, but panelists sub-pide the line into discrete sub-ranges as if they were using a limited

9.4.1 Idiosyncratic Scale Usage and Number Bias People appear to have pferred ranges or numbers on the response scale that they feel comfortable using. Giovanni and Pangborn (1983) noted that people using magnitude estimation very often used numbers that were multiples of 2 and 5 (or obviously 10), an effect that is well known in the psychophysical literature (Baird and Noma, 1978). With magnitude estimation, the idiosyncratic usage of a favorite range of numbers causes a correlation of the power function exponents across different sensory continua for an inpidual (Jones and Marcus, 1961; Jones and Woskow, 1966). This correlation can be explained if people are more or less expansive (versus restrictive) in their response output functions, i.e., in how they apply numbers to their sensations in magnitude estimation studies. Another version of such personal idiosyncrasy is the common observation in time-intensity scaling that people show a kind of personal “signature” or a characteristic curve shape (Dijksterhuis, 1993; McGowan and Lee, 2006). Another version of self-induced response restriction can be seen when people use only selected portions of the scale in a line-marking rating task. On a line scale with verbal labels, people may choose to make markings only near the verbal labels, rather than distributing them across the response scale. This was first observed by Eng (1948) with a simple hedonic line scale labeled Like Very Highly at one end, Dislike Very Highly at the other, and Neither Like nor Dislike at the center. In a group of 40 consumers, 24 used only the three labeled parts of the scale, and Eng deleted them

9.4 Biases

from the data analysis! This kind of behavior was also noted with the labeled affective magnitude scale (LAM scale) by Cardello et al. (2008) with both Army laboratory and student groups. Lawless et al. (2010a) found a very high frequency (sometimes above 80%) of people making marks within ±2 mm of a phrase mark on the LAM scale in a multi-city consumer central location test. Lawless et al. (2010b) found that instructions did not seem to change this behavior much but that expanding the physical size of the scale on the ballot (from about 120 to 200 mm) decreased the “categorical” behavior somewhat. Categorical rating behavior can also be seen as a step function in time-intensity records (rather than a smooth continuous curve). Finding product differences against the background of these inpidual response tendencies can be facilitated by within-subject experimental designs. Each participant is used as his or her own baseline in comparisons of products, as in dependent t-tests or repeated measure analysis of variance in complete block designs. Another approach is to compute a difference score for a comparison of products in each inpidual’s data, rather than merely averaging across people and looking at differences between mean values.

9.4.2 Poulton’s Classifications Poulton (1989) published extensively on biases in ratings and classified them. Biases in Poulton’s system go beyond Parducci’s theory but are documented in the psychophysical literature. These include centering biases, contraction biases, logarithmic response bias with numerical ratings, and a general transfer bias that is seen when subjects carry the context from a pvious session or study into a new experiment. The centering bias is especially relevant to just-right scales and is discussed in a later section. The response range bias is also a special case and follows this section. The contraction biases are all forms of assimilation, the opposite of contrast. According to Poulton, people may rate a stimulus relative to a reference or a mean value that they hold in memory for similar types of sensory events. They tend to judge new items as being close (perhaps too close) to this reference value, causing underestimation of high values and overestimation of low values. There may also be

213

overestimation of an item when it follows a stronger standard stimulus or underestimation when it follows a weaker standard stimulus, a sort of local contraction effect. Poulton also classifies the tendency to gravitate toward the middle of the response range as a type of contraction effect, called a response contraction bias. While all of these effects undoubtedly occur, the question arises as to whether contrast or assimilation is a more common and potent process in human sensory judgment. While some evidence for response assimilation has been found in psychophysical experiments through sequential analysis of response correlations (Ward, 1979), contrast seems much more to be the rule with taste stimuli (Schifferstein and Frijters, 1992) and foods (Kamenetzky, 1959). In our experience, assimilation effects are not as pvalent as contrast effects, although assimilation has certainly been observed in experiments on consumer expectation (e.g., Cardello and Sawyer, 1992). In that case, the assimilation is not toward other actual stimuli but toward expected levels. The logarithmic response bias can be observed with open-ended response scales that use numbers, such as magnitude estimation. There are several ways to view this type of bias. Suppose that a series of stimuli have been arranged in increasing magnitude and they are spaced in subjectively equal steps. As the intensity increases, subjects change their strategy as they cross into ranges of numerical responses where there are more digits. For example, they might be rating the series using numbers like 2, 4, 6, and 8, but then when they get to 10, they will continue by larger steps, perhaps 20, 40, 60, 80. In Poulton’s view they proceed through the larger numerical responses “too rapidly.” A converse way of looking at this problem is that the perceived magnitude of the higher numbers is in smaller arithmetic steps as numbers get larger. For example, the difference between one and two seems much larger compared to the difference between 91 and 92. Poulton also points out that in addition to contraction of stimulus magnitude at very high levels, the converse is also operating and that people seem to illogically expand their subjective number range when using responses smaller than the number 3. One obvious way to avoid the problems in number bias is to avoid numbers altogether or to substitute line scaling or cross-modality matching to line length as a response instead of numerical rating techniques like magnitude estimation (Poulton, 1989).

214

Transfer bias refers to the general tendency to use pvious experimental situations and remembered judgments to calibrate oneself for later tasks. It may involve any of the biases in Poulton’s or Parducci’s theories. The situation is common when subjects are used in multiple experiments or when sensory panelists are used repeatedly in evaluations (Ward, 1987). People have memories and a desire to be internally consistent. Thus the ratings given to a product on one occasion may be influenced by ratings given to similar products on pvious occasions. There are two ways to view this tendency. One is that the judgments may not shift appropriately when the panelists’ sensory experience, perception, or opinion of the product has in fact changed. On the other hand, one of the primary functions of panelist training and calibration in descriptive analysis is to build up exactly those sorts of memory references that may stabilize sensory judgments. So there is a positive light to this tendency as well. An open question for sensory evaluation is whether exposure to one continuum of sensory intensities or one type of product will transfer contextual effects to another sensory attribute or a related set of products (Murphy, 1982; Parducci et al., 1976; Rankin and Marks, 1991). And if so, how far does the transfer extend?

9.4.3 Response Range Effects One of Poulton’s biases was called the “response range equalizing bias” in which the stimulus range is held constant but the response range changes and so do the ratings. Ratings expand or contract so that the entire range is used (minus any end-category avoidance). This is consistent with the “mapping” idea mentioned for stimulus range effects (stimuli are mapped onto the available response range). Range stabilization is implicit in the way some scaling studies have been set up and in the instructions given to subjects. This is similar to the use of physical reference standards in some descriptive analysis training (Muñoz and Civille, 1998) and is related to Sarris’s work on anchor stimuli (Sarris and Parducci, 1978). In Anderson’s work with 20-point category scales and line marking, high and low examples or end anchors are given to subjects to show them the likely range of the stimuli to be encountered. The range of responses is known since

9

Context Effects and Biases in Sensory Judgment

it is visible upon the page of the response sheet or has been p-familiarized in a practice session (Anderson, 1974). Thus it is not surprising that subjects distribute their responses across the range in a nicely graded fashion, giving the appearance that there is a reasonably linear use of the scale. Anderson noted that there are end effects that work against the use of the entire range (i.e., people tend to avoid using the endpoints) but that these can be avoided by indenting the response marks for the stimulus end anchors, for example, at points 4 and 16 on the 20-point category scale. This will provide psychological insulation against the end effects by providing a comfort zone for unexpected or extreme stimuli at the ends of the scale while leaving sufficient gradations and room to move within the interior points. The “comfort zone” idea is one reason why early workers in descriptive analysis used line scales with indented vertical marks under the anchor phrases. An exception to the response range mapping rule is seen when anchor phrases or words on a scale are noted and taken seriously by participants. An example is in Green’s work on the labeled magnitude scale, which showed a smaller response range when it was anchored to “greatest imaginable sensation” that included all oral sensations including pain, as opposed to a wider range when the greatest imaginable referred only to taste (Green et al., 1996). This also looks like an example of contrast in which the high-end anchor can evoke a kind of stimulus context, at least in the participant’s mind. If the image evoked by the high-end phrase is very extreme, it acts like a kind of stimulus that compsses ratings into a smaller range of the scale. A similar kind of response compssion was seen with the LAM scale when it was anchored to greatest imaginable liking for “sensations of any kind” as opposed to a more delimited frame such as “foods and beverages” (Cardello et al., 2008). A sensory scientist should consider how the high anchor phrase is interpted, especially if he or she wants to avoid any compssion of ratings along the response range. As Muñoz and Civille (1998) pointed out, the use of a descriptive analysis scale also depends a lot on the conceptualization of the high extreme. Does “extremely strong” refer to the strongest possible taste among all sensations and products, the strongest sensation in this product type, or just how strong this particular attribute can become in this particular product? The strongest sweetness in this product might be more intense than the strongest saltiness. The definition needs to be a

9.4 Biases

215

deliberate choice of the panel leader and an explicit instruction to the panelist to give them a uniform frame of reference.

9.4.4 The Centering Bias The centering bias arises when subjects become aware of the general level of stimulus intensity they are likely to encounter in an experiment and tend to match the center or midpoint of the stimulus range with the midpoint of the response scale. Poulton (1989) distinguished a stimulus centering bias from a response centering bias, but this distinction is primarily a function of how experiments are set up. In both cases, people tend to map the middle of the stimulus range onto the middle of the response range and otherwise ignore the anchoring implications of the verbal labels on the response scale. Note that the centering bias works against the notion that respondents can use unbalanced scales with any effectiveness. For example, the “Excellent-very good-good-fair-poor” scale commonly used in marketing research with consumers is unbalanced. The problem with unbalanced scales is that over many trials, the respondents will come to center their responses on the middle category, regardless of its verbal label. The centering bias is an important problem when there is a need to interpolate some value on a psychophysical function or to find an optimal product in just-right scaling. Poulton gives the example of McBride’s method for considering bias in the justabout-right (JAR) scale (McBride, 1982; see also Johnson and Vickers, 1987). In any series of products to be tested, say for just-right level of sweetness, there is a tendency to center the series so that the middle product will come out closest to the just-right point. The function shifts depending upon the range that is tested. One way to find the true just-right point would be to actually have the experimental series centered on that value, but then of course you would not need to do the experiment. McBride gives a method for interpolation across several experiments with different ranges. The point at which the just-right function and the median of the stimulus series will cross shows the unbiased or true just-right level. This method of interpolation is shown in Fig. 9.8. In this method, you

Fig. 9.8 Adjusting for the centering bias in just-right ratings. Three series of sucrose concentrations in lemonade were tested, a low series (2-8%), a middle series (6-14%), and a high range (10-22%). In the upper panel, the method of Poulton is used to interpolate the unbiased just-right point from the series where the midpoint concentration would correspond to the just-right item. In the lower panel, the method of McBride is used to interpolate the just-right point from a series in which the average response would correspond to the just-right point. When the average response would be just right (zero on this scale), the hypothetical stimulus range would have been centered on the just-right level. Replotted from Johnson and Vickers (1987), with permission.

psent several ranges of the products in separate sessions and plot how the judgments of the JAR point shift up and down. You can then interpolate to find the range in which the just-right point would have been from the center product in the series. This obviously takes more work to do the test a couple of times, but it could avoid a mistaken estimate of the JAR level.

216

9.5 Response Correlation and Response Restriction

9

9.5.1 Response Correlation A simple example of a halo effect is shown in Fig. 9.9. In this case, a small amount of vanilla extract was added to low-fat milk, near the threshold of perceptibility. Ratings were then collected from 19 milk consumers for sweetness, thickness, creaminess and liking for the spiked sample, and for a control milk. In spite of the lack of relationship between vanilla aroma and sweet taste and between vanilla and texture characteristics, the introduction of this one positive aspect was sufficient to cause apparent enhancement in sweetness, creaminess, and thickness ratings. Apparent enhancement of sweetness is an effect long known for ethyl maltol, a caramelization product that has an odor similar to heated sugar (Bingham

ADDED VANILLA CONTROL MILK

13

MEAN RATING + 1 S.E.M.

11

Early experimental psychologists like Thorndike (1920) noted that one very positive attribute of a person could influence judgments on other, seemingly unrelated characteristics of that inpidual. In personnel evaluations of military officers, Thorndike noted a moderate positive correlation among the inpidual rated factors. People evaluate others like this in real life. If achievement in sports is influential in our assessment of a person, we might suppose a gifted athlete to also be kind to children, generous to charities, etc., even though there is no logical relationship between these characteristics. People like to have cognitive structures that form consistent wholes and are without conflicts or contradictions (called cognitive dissonance) that can make us uncomfortable. The halo effect has also been described as a carry-over from one positive product to another (Amerine et al., 1965), but its common usage is in reference to a positive correlation of unrelated attributes (Clark and Lawless, 1994). Of course, there can also be negative or horns effects, in which one salient negative attribute causes other, unrelated attributes to be viewed or rated negatively. If a product makes a mess in the microwave, it might be rated negatively for flavor, appearance, and texture as well.

Context Effects and Biases in Sensory Judgment

*Different from

* *

9

control, paired t-test N = 19.

*

7

5

3

1

SWEET

CREAMY THICK SCALED ATTRIBUTE

LIKE

Fig. 9.9 Adding a just perceivable level of vanilla extract to low-fat milk causes increases in rated sweetness, thickness, creaminess, and liking, an example of the Halo effect. From Lawless and Clark (1994), with permission.

et al., 1990). When maltol is added to various products, sweetness ratings may rise compared to products lacking this flavor. However, the effect seems to be a case of the misattribution of olfactory stimulation to the taste sense. Murphy and Cain (1980) showed that citral (a lemon odor) could enhance taste ratings, but only when the nostrils were open, which allows diffusion of the odor into the nose and stimulation of the olfactory receptors (i.e., retronasal smell). When the nostrils are pinched shut, the diffusion is effectively eliminated and the enhancement disappears. Murphy and Cain interpted this as convincing evidence that there was no true enhancement of taste intensity by citral, but only olfactory referral, a kind of confusion between taste and smell. Studies with other odors have also shown that the taste enhancement effect from volatile flavors can be eliminated by nose pinching (Frank and Byram, 1988) even for maltol (Bingham et al., 1990). The maltol effect is also minimized by training subjects who then learn to more effectively separate or localize their odor experiences from taste (Bingham et al., 1990). The sweetness enhancement may arise as a function of conditioning or experience with the pairing of sweet tastes with some odors in foods (Stevenson et al., 1995). Several lessons can be learned from the vanilla halo effect shown in Fig. 9.9. First, untrained consumers

9.5 Response Correlation and Response Restriction

9.5.2 “Dumping” Effects: Inflation Due to Response Restriction in Profiling It is part of the folklore of consumer testing that if there is one very negative and salient attribute of a product, it will influence other attributes in a negative direction, an example of a horns effect. The effect is even worse when the salient negative attribute is omitted from the questionnaire. Omission could be due to some oversight or failure to anticipate the outcome in a consumer test or simply that it was not observed in the laboratory conditions of pliminary phases of testing. In this case, consumers will find a way to dump their frustration from not being able to report their dissatisfaction by giving negative ratings on other scales or reporting negative opinions of other even unrelated attributes. In other words, restricting responses or failure to ask a relevant question may change ratings on a number of other scales. A common version of this restriction effect can be seen in sweetness enhancement. Frank et al. (1993) found that the enhancement of sweet ratings in the psence of a fruity odor was stronger when ratings were restricted to sweetness only. When

217

both sweetness and fruitiness ratings were allowed, no enhancement of sweetness was observed. Exactly the same effect was seen for sweetness and fruitiness ratings and for sweetness and vanilla ratings (Clark and Lawless, 1994). So allowing the appropriate number of attributes can address the problem of illusory enhancement. Schifferstein (1996) gave the example of hexenol, a fresh green aroma, which when added to a strawberry flavor mixture caused mean ratings in several other scales to increase. The enhancement of the other ratings occurred only when the “green” attribute was omitted from the ballot. When the “green” attribute was included in the ballot, the response was correctly assigned to that scale, and there was no apparent enhancement in the other attributes in the aroma profile. There is good news and bad news in these observations. From a marketing perspective, ratings can be easily obtained from consumers that will show apparent sweetness enhancements if the questionnaires cleverly omit the opportunity to report on sensations other than sweetness. However, the nose pinch conditions and the use of complete sets of attributes show us that these volatile odorants such as maltol are not sweet taste enhancers but they are sweet rating enhancers. That is, they are not affecting the actual perception of sweet taste intensity but are changing the response output function or perhaps broadening the concept of sweetness to go beyond taste and include pleasant aromas as well. It would not be wise to try to use maltol to sweeten your coffee. Are there other situations in sensory testing where the dumping effect can show up? One area in which responses are usually restricted to one attribute at a time is in time-intensity scaling (Chapter 8). In a common version of this technique, the subject moves a pointer, a mouse, or some other response device to provide a continuous record of sensory intensity for a specified attribute. Usually, just one attribute at a time is rated since it is very difficult to attend continuously or even by rapid shifting of attention to more than one attribute. This would seem to be a perfect opportunity for the dumping tendency to produce illusory enhancements (e.g., Bonnans and Noble, 1993). This idea was tested in experiments with repeated category ratings across time, a time-intensity procedure that allows for ratings of multiple attributes. These studies showed sweetness enhancement in sweet-fruity mixtures when sweetness alone was rated, but little or

218

no enhancement when both sweetness and fruit intensity were rated over time (Clark and Lawless, 1994). This is exactly parallel to the sweetness enhancement results seen by Frank and colleagues. Workers using single-attribute time-intensity ratings should be wary of apparent enhancements due to response restriction.

9.5.3 Over-Partitioning In the data of van der Klaauw and Frank (1996), one can also see cases in which having too many attributes causes a deflation in ratings. As in the dumping examples, their usual paradigm was to compare the sweetness ratings of a simple sucrose solution to the same concentration with a fruity odor added. When rating sweetness only, the rating is higher than when rating sweetness and fruitiness, the common dumping effect. But when total intensity and six additional attributes were rated, the sweetness rating was significantly lower than either of the other two conditions. In another example, including a bitterness rating (in addition to the sweetness and fruitiness ratings) lowered the sweetness rating compared to rating sweetness (the highest condition) and also compared to rating sweetness and fruitiness (an intermediate sweetness rating was obtained). This effect appears to be a deflation due to people over-partitioning their sensations into too many categories. The specific choices may be important in this effect. Adding only a bitter or a bitter and a floral rating had little or no effect and dumping inflation was still observed probably because there was no fruity rating. In the study by Clark and Lawless (1994), the control condition (sweetener only) showed some evidence of a decrement when the attributes for volatiles were also available. Even more dramatic was the complete reversal of sweet enhancement to inhibition when a large number of response categories were provided for simple mixtures (Frank et al., 1993). Although this effect has not been thoroughly studied, it serves to warn the sensory scientist that the number of choices given to untrained consumers may affect the outcome and that too many choices may be as dangerous as too few. Whether this effect might be seen with trained panels remains an open question. It is sometimes difficult to pdetermine the correct number of attributes to rate in order to guard against

9

Context Effects and Biases in Sensory Judgment

the dumping effect. Careful p-testing and discussion phases in descriptive training may help. It is obviously important to be inclusive and exhaustive, but also not to waste the panelists’ time with irrelevant attributes.

9.6 Classical Psychological Errors and Other Biases A number of psychological errors in judgment have been described in the literature and are commonly listed in references in sensory evaluation (e.g., Amerine et al., 1965; Meilgaard et al., 2006). They are only briefly listed here, as they serve primarily as empirical descriptions of behavior, without much reference to cause or any theoretical bases. It is important to distinguish between description and explanation, and not to confuse naming something with trying to explain why it occurred in terms of mechanism or larger theory. The sensory evaluation practitioner needs to be aware of these errors and the conditions under which they may occur.

9.6.1 Errors in Structured Sequences: Anticipation and Habituation Two errors may be seen when a non-random sequence of products is psented for evaluation, and the observer is aware that a sequence or a particular order of items is going to be psented. The error of anticipation is said to occur when the subject shifts responses in the sequence before the sensory information would indicate that it is appropriate to do so (Mattes and Lawless, 1985). An example is found in the method of limits for thresholds, where an ascending sequence is psented and the observer expects a sensation to occur at some point and “jumps the gun.” The opposite effect is said to be the error of habituation, in which the panelist stays put too long with one pvious response, when the sensory information would indicate that a change is overdue. Obviously, the psentation of samples in random order will help to undo the expectations involved in causing the error of anticipation. Perseveration is a little bit harder to understand but may have to do with lack of attention or motivation

9.7 Antidotes

on the part of the observer or having an unusually strict criterion for changing responses. Attention and motivation can be addressed by sufficient incentives and keeping the test session from being too long.

219

The stimulus error is another classical problem in sensory measurement. This occurs when the observer knows or psumes to know the identity of the stimulus and thus draws some inference about what it should taste, smell, or look like. The judgment is biased due to expectations about stimulus identity. In the old parlor game of trying to identify the origin and vintage of a wine, stimulus error is actually a big help. It is much easier to guess the wine if you know what the host is prone to drink or you have taken a peek at the bottles in the kitchen beforehand. In sensory evaluation, the principle of blind testing and the practice using random three-digit coding of samples mitigate against the stimulus error. However, panelists are not always completely in the dark about the origin or the identity of samples. Employee panels may have a fair amount of knowledge about what is being tested and they may make inferences, correctly or incorrectly. For example, in small-plant quality assurance, workers may be aware of what types of products are being manufactured that day and these same workers may serve as sensory panelists. In the worst possible scenario, the persons drawing the samples from production are actually doing the tasting. As much as possible, these situations should be avoided. In quality control panels, insertion of blind control samples (both positive controls and flawed samples) will tend to minimize the guesswork by panelists.

the sequential effects may be counterbalanced or averaged out in the group data. The second approach is to consider the order effects of interest. In this case, different orders are analyzed as a purpose of the experiment and, if order effects are observed, they are duly noted and discussed. Whether order effects are of interest will depend upon the circumstances of the product evaluation and its goals. If counterbalanced orders or randomization cannot be worked into the experimental design, the experimenter must consider whether product differences are true sensory differences or are artifacts of stimulus order. Purposeful experimentation and analysis of order effects in at least some of the routine evaluations may give one an appciation for where and when these effects occur in the product category of interest. Another well-known order effect in acceptance testing is the reception of a higher score for the first sample in a series (Kofes et al., 2009). Counterbalancing orders is of course appropriate, but one can also give a “dummy” product first to absorb the first product’s s

【#5】Estimating Crop Nutritional Status Using Smart Apps To Support Nitrogen Fertilization. A Case Study On Paddy Rice

by ^1,*, ^{2 1,*},

Department of Environmental Science and Policy, Università degli Studi di Milano, Cassandra lab, via Celoria 2, 20133 Milan, Italy

Italian National Research Council, Institute on Remote Sensing of Environment (CNR-IREA), via Bassini 15, 20133 Milan, Italy

Authors to whom correspondence should be addressed.

Received: 17 January 2021 / Revised: 5 February 2021 / Accepted: 20 February 2021 / Published: 25 February 2021