Worth a thousand words?

With taste and smell—the so-called “chemical” senses, which are more complex (humans have about 400 different types of olfactory receptors) and less well understood than the others, we don’t have the luxury of those points of reference. That’s why we so often resort to loose analogies—“tastes like chicken”—and it’s also why reproducing tastes and smells is so difficult (grape soda doesn’t taste much like grapes, and nobody’s yet synthesized a bottle of 1945 Pétrus—an activity that would surely yield tremendous profit).

To challenge this barrier, we resort to analogy. Coffee tastes like nuts and chocolate; Sauvignon Blanc smells like grapefruit and cat pee. In a Sauternes, you might sense the brine of the first green olive you tasted in Italy; in a Pedro Ximénez sherry, the viscous maple syrup that your grandmother once drizzled on your pancakes.

But how carefully are we really choosing these adjectives and analogies? How often do they correspond to real chemical commonalities? Does that matter? Do the analogies more frequently serve a more poetic (or at least suggestive) purpose, forging new neural assemblies that connect relatively arbitrary taste and smell memories with each other—connections that, reinforced over time, turn into sensory reality?

Two papers at last month’s meeting of the American Association of Wine Economists in Reims (this is my second of two articles about the conference) investigated this question with respect to the wine industry, which is, if not a microcosm of all consumer-products industries, at least an increasingly apt caricature of them. While creative adjectivism has long characterized in the wine world, the practice in other taste industries—chocolaty coffee, metallic fish, grassy honey, peaty whiskey—is now ascendant.

The canonical work in the wine-adjective field is Princeton economist Richard Quandt’s “On Wine Bullshit” (a riff on his fellow Princetonian Harry Frankfurt’s “On Bullshit”). Writes Quandt:

Two things have to be true before wine ratings can become useful for the average wine drinker. Since there are many wine writers, and there is a substantial overlap in the wines they write about (particularly Bordeaux wines), it is important that there be substantial agreement among them. And secondly, what they write must actually convey information; that is to say, it must be free of bullshit. Regrettably, wine evaluations fail on both counts.

At the AAWE meeting, Coco Krumme of M.I.T., who is also a Fearless Critic food writer, studied data from critical descriptions of more than 3,500 wines from recent vintage years, ranging from $4.99 to $137.99 in retail price, and employed a Bayesian filter to “find those words that best predict the price category of a bottle” (abstract here). She found that “about 65% of commonly occurring words are non-overlapping.” Words like “old,” “elegant,” “intense,” “supple,” “velvety,” “smoky,” “tobacco,” and “chocolate” predict expensive wines; “pleasing,” “refreshing,” “value,” “enjoy,” “bright,” “light,” “fresh, “tropical,” “pink,” “fruity,” “good,” “clean,” “tasty,” and “juicy” predict cheap wines. As for suggested pairings, “steak” and “shellfish” predict expensive wines; “chicken” predicts cheap wines.

Perhaps most amusingly, Krumme reports that “words with the same meaning are preferentially used for expensive over cheap wines: for example, ‘vintage’ is six times more likely to describe an expensive wine; ‘harvest’ is used for cheap wines.”

Economist Carlos Ramirez of George Mason University, meanwhile, ran a regression (abstract here) on a data set of 800 Wine Spectator descriptions of Napa Valley Cabernet Sauvignon wines from for wines from the 2004, 2005, and 2006 vintages, and found a length-of-review effect—that is, “longer wine descriptions are associated with higher prices—a 10 percent increase in the length of a wine description (adding about 23 characters) is associated with a statistically significant increase of 4 to 13 percent to the price of the bottle.” Like Krumme, Ramirez also found some particular wine descriptors (about 20 of the 208 he looked at) that, controlling for other variables, signal higher wine prices.

If you’re familiar with wine ratings and reviews, neither of these results might surprise you. But the interesting, unanswered question is: which way does the causality go?

Here are three potential theories:

(1) Expensive wines are generally fairly similar to each other, and their particular properties lead critics to refer more frequently to certain flavors (e.g. chocolate) and to write longer reviews of these wines. That is, there’s just a specific expensiveness to expensive wines that explains these differences. (Quandt would likely doubt this, and the empirical evidence, as described in The Wine Trials, would be against it, too.)

(2) Tasting is not done blind, and thus critics are influenced to write more and refer to certain flavors when they taste expensive wines.

(3) Tasting is done blind, but the sensory reviews of expensive wines are edited after the fact by editors who know what the wines are.

Regardless of which of these theories is correct, what’s highly likely is that the descriptors are self-fulfilling—reading an expensive wine description primes the drinker to have a more typically expensive wine experience. That is, the adjectives and analogies we read in wine reviews fuse with our experience of drinking the wine in such a complete way that the liquid’s intrinsic and extrinsic properties become inseperable.

Is this why it’s so difficult to undermine the conventional wisdom that very expensive wine is worth the money?

Maybe we just synthesize whatever we seek, creating value as we go: search for chocolate, and it will magically appear.