You May Also Like by Tom Vanderbilt Page B

something in a taste test, why would they not like it in the real world? Pepsi certainly did not just chuck Crystal Pepsi into the marketplace on a whim. Some ninety people, it was said, worked fifteen months on the product, cycling through several thousand versions. And well before it even made it to a regional taste market, one can be sure it was tested in-house by any number of sensory and consumer panels, the bulk of which, one presumes, must have said that they liked it.
    As it happens, this very thing—measuring people’s liking for a product on a food company “consumer panel”—was developed and perfected at Natick.The program itself was founded in 1944 at Chicago’s Quartermaster Food and Container Institute, in response to an ongoing problem of ration quality and its impact on troop morale. A team of psychologists, many of whom would go on to do seminal work in the food industry, was assembled. “One of the first issues that cameup,” Cardello told me, “was how do you measure how much someone likes something?”
    Pioneering psychologists like Wilhelm Wundt had tried to quantify, through “psychophysics,” the inexact ways our senses responded to various stimuli (for example, when you double the sweetness of something, and it does not taste twice as sweet).
    But no one had been able, or had much tried, to quantify liking. And so the “nine-point hedonic scale” was born. First used on soldiers, it eventually found its way into the test kitchens of just about every major food manufacturer. Whatever is in your refrigerator at this moment, chances are that someone, somewhere has indicated his liking of it on a scale from one to nine. There was, according to one account, an early attempt to introduce an “11-point scale,” but it would not fit on government-regulated paper. Humans have even been trotted out and asked to write down their scale-of-one-to-nine reaction on products like cat food. Why? Felines, as an accompanying report noted, are “clearly unable to verbalize their likes and dislikes.” They may haughtily strut away from the bowl, their tail a flag of disdain, but this gesture does not easily translate into a numerical scale. “Perhaps surprisingly,” the report concluded, “the grand mean of all hedonic scores was 4.7, placing it between the ‘neither like nor dislike’ and ‘like slightly’ scale adjectives.” People thought cat food was not too bad, at least
as
cat food.
    The simplicity, relative accuracy, and value of hedonic scores as an industry standard has overshadowed the ongoing methodological issues in trying to put a number on liking.Other methods, like polygraphs, have failed dismally. But issues abound. There are semantic problems. Does “like slightly” mean the same thing to one person as another? There are issues with the math.The number eight, Cardello noted, does not mean twice as much liking as the number four. Could liking and disliking be expressed on the same scale?As work by Timothy Wilson and colleagues at the University of Virginia has made clear, asking people to analyze why they chose something can lead them to change their original choice—and usually not for the better.
    But merely asking consumers
what
they like is also not as simple as you might think. In one common tool, the “just about right” scale, people are given samples of a product. Each will have, say, a different gradation of sweetening. The consumer indicates which is “just aboutright.” Sounds fine, no?There is just one problem: The level of sweetening a subject chooses is often different from what he says he likes.
    Then there is the fact that most people do not pick the number one or nine. Those seem too artificial. People hedge. It becomes, by default, a seven-point scale. “You’re never sure that you’re not going to get a product in the next sample that’s even