Are we all subconscious marketers? Meg Maker poses the question in her column this month.
Making fun of wine tasting notes is not a new sport; recall Thurber’s 1937 cartoon about the “naïve domestic Burgundy.” But academic research about wine tasting notes is much newer, and it’s being done not only by linguists but also by mathematicians, economists, statisticians, and AI modelers.
Such quantitative research is made possible by the large databases of tasting notes amassed over years by publications like Wine Enthusiast and Wine Spectator. The datasets contain not only textual descriptions but also structured information about a wine’s vintage, region, style, producer, price, score, reviewer, and more. Researchers are particularly keen to discover how well a critic’s notes convey meaning to the reader, and also whether they are persuasive in a commercially impactful way.
Oxford researcher Martin Klimmek has shown through statistical textual analysis that wine notes that use consistent, precise language (citing the WSET rubric) are, from a mathematical perspective, more information rich and therefore presumably more useful than one populated with more poetic terms (Klimmek 2013).
But as we all know, those data sets are chock full of poetic terms, so it’s important to quantify them, too. Coco Krumme, at MIT, mined a database of 3,000 wines ranging from $5 to $200 to demonstrate the existence of “expensive” and “cheap” wine terms. Words like intense, supple, velvety, boysenberry, pork, steak, elegant, and cuvée are used for more expensive wines, while words like fruity, good, tasty, pizza, chicken, value, and juicy are used for less expensive wines. Further, “cheap” words are used frequently while “expensive” words are less often repeated, suggesting reviewers “are more likely to create new vocabulary for top-end wines” (Krumme 2011).
Researcher Kevin W. Capehart, of California State, Fresno, revisited Krumme’s work a decade later, applying her analysis to a Wine Enthusiast database of 120,000 wines. He not only confirmed her results but also found that less expensive wines have shorter tasting notes overall (Capehart 2021).
By the way, it’s also true for menus. AI researcher Dan Jurafsky and colleagues analyzed a dataset of 6,500 menus containing 650,000 dishes from restaurants in seven major U.S. cities. They found that dishes described with longer words were more expensive, and specifically, “every increase of one letter in the average length of words describing a dish is associated with an increase of 18 cents in the price” (Jurafsky 2014).
On one hand, these results make intuitive sense. I cannot imagine a top reviewer describing Haut Brion as “tasty, good for pizza.” On the other hand, it’s disturbing to think of reviewers modifying their vocabulary to suit the reputation and price of the wine.
The wine industry relies on wine communicators to be unbiased spokespeople, not marketers. George Orwell once said, “Journalism is printing what someone else doesn’t want printed. Everything else is public relations.” That’s something I now think about every time I write a review.
Photo by Meg Maker