by Dan Jurafsky
In one marketing study, for example, Richard Klink created pairs of made-up product brand names that were identical except for having front vowels (detal) or back vowels (dutal) and asked participants to answer:
Which brand of laptop seems bigger, Detal or Dutal?
Which brand of vacuum cleaner seems heavier, Keffi or Kuffi?
Which brand of ketchup seems thicker, Nellen or Nullen?
Which brand of beer seems darker, Esab or Usab?
In each case, the product named with back vowels (Dutal, Nullen) was chosen as the larger, heavier, thicker product.
Since ice cream is a product whose whole purpose is to be rich, creamy, and heavy, it is not surprising that people seem to prefer ice creams that are named with back vowels. Eric Yorkston and Geeta Menon at New York University asked participants to read a press release describing a new ice cream about to be released. For half, the ice cream was called “Frish” (front vowel) while for the other half it was called “Frosh” (back vowel). Asked their opinions, the “Frosh” people rated this hypothetical ice cream as smoother, creamier, and richer than other participants rated “Frish,” and were more likely to say they would buy it.
In a final twist, Yorkston and Menon distracted some participants by having them perform another task simultaneously, so they couldn’t fully concentrate on reading about the ice cream. The distracted participants were even more influenced by the vowels, suggesting that the response to the vowels was automatic, at a subconscious level.
I wondered whether commercial ice creams make use of this subconscious association of ice cream names with back vowels as richer and creamier. To find out, I ran what University of Pennsylvania linguist Mark Liberman calls a Breakfast Experiment. Liberman—a tenacious advocate for bringing linguistics to bear on public affairs—often runs a quick experiment on a linguistic tip in the news before breakfast, posting the results on Language Log, the “blog of record” in linguistics. He is legendary for his ability to run complex linguistic statistical analyses in minutes, which he says comes from his days as a piano tuner.
My hypothesis was that we would see more back vowels in names of ice cream brands or flavors, and conversely that thin, light foods like crackers would have more front vowels.
Front versus back vowels in cracker names and ice cream flavors (normalized by dividing by the expected count of front and back vowels computed from a large dictionary of English)
I tested the hypothesis on two lists of food names from the web, the 81 ice cream flavors sold by either Haagen Dazs or Ben & Jerry’s, and a list of 592 cracker brands from a dieting website. For each list, I counted the total number of front vowels (i, I, , e, æ) and the total number of back vowels.
The result? As shown in the chart on the preceding page, I found more back vowels in ice cream names like Rocky Road, Jamoca Almond Fudge, Chocolate, Caramel, Cookie Dough, Coconut and front vowels in cracker names (note the extraordinary number of I vowels) like Cheese Nips, Cheez It, Wheat Thins, Pretzel Thins, Ritz, Krispy, Triscuit, Thin Crisps, Cheese Crisps, Chicken in a Biskit, Snack Sticks, Ritz bits.
Of course there are exceptions: vanilla (the orange blossom of our day), has an I. But most of the front vowels in ice cream flavors tend to be the names of small, thin ingredients in the ice cream (thin mint, chip, peanut brittle).
Sound symbolism is thus an important device in the toolbox of modern advertisers and designers of brand names, and in fact branding companies often get their insights from linguists.
While our ice cream and cracker connections might be subconscious, they are systematic, and linguists have theories about the underlying cause: about why front vowels are associated with small, thin, light things, and back vowels with big, solid, heavy things.
The most widely accepted theory, the frequency code, suggests that low frequencies (sounds with low pitch) and high frequencies (sounds with high pitch) are associated with particular meanings. The frequency code was developed by linguist John Ohala (my phonetics professor as an undergraduate at Berkeley) by extending work by Eugene Morton of the Smithsonian.
Morton noticed that mammals and birds tend to use low-frequency (deeper) sounds when they are aggressive or hostile, but use higher-frequency (higher-pitched) sounds when frightened, appeasing, or friendly. Because larger animals naturally make deeper sounds (the roar of lions) and smaller animals naturally make high-pitched sounds (the tweet of birds), Morton’s idea is that animals try to appear larger when they are competing or aggressive, but smaller and less threatening otherwise.
Morton and Ohala thus suggest that humans instinctively associate the pitch of sounds with size. All vowels are composed of different frequency resonances. When the tongue is high and in the front of the mouth, it creates a small cavity in front. Small cavities cause higher-pitched resonances (the smaller the space for vibration, the shorter the wavelength, hence the higher the frequency). One particular resonance (called the second formant) is much higher for front vowels and lower for back vowels.
Thus the frequency code suggests that front vowels like I and i are associated with small, thin, things, and back vowels like a and ο with big heavy things because front vowels have higher-pitched resonances, and we instinctively associate higher pitch with smaller animals, and by extension smaller things in general.
Researchers have extended this idea to show that raising pitch or “fronting” vowels (moving the tongue a bit toward the front of mouth to make all vowels have a slightly higher second formant pitch) are both especially associated with babies or children. In an early paper I examined more than 60 languages around the world and proposed that the word endings used in many languages to indicate smallness or lightness come historically from a word originally meaning “child” or associated with names of children, like the y in pet names Barbie and Robby. My linguistics colleague Penny Eckert shows that front vowels are associated with positive affect, and that preadolescent girls sometimes use vowel fronting to subtly imbue their speech with sweetness or childhood innocence. Linguist Katherine Rose Geenberg found that speakers of American English move their vowels toward the front when using baby talk, and psychologist Anne Fernald shows that, across languages, talk to babies tends to have high pitch.
The frequency code isn’t the only kind of sound symbolism in food. To see why, we’ll need a brief digression. Consider these two pictures:
Suppose I told you that in the Martian language one of these two was called bouba and the other was called kiki and you had to guess which was which. Think for a second. Which picture is bouba? Which kiki? How about maluma versus takete?
If you’re like most people, you called the jagged picture on the left kiki (or takete) and the round one on the right bouba (or maluma). This test was invented by German psychologist Wolfgang Köhler, one of the founders of Gestalt psychology, in 1929. Linguists and psychologists have repeated this experiment using all sorts of made-up words with sounds like bouba and kiki, and no matter what language they study, from Swedish to Swahili to a remote nomadic population of northern Namibia, and even in toddlers two and a half years old, the results are astonishingly consistent. There seems to be something about jagged shapes that makes people call them kiki and rounder curvy shapes that is somehow naturally bouba.
The link to food comes from the lab of Oxford psychologist Charles Spence, one of the world’s foremost researchers in sensory perception. In a number of recent papers, Spence and his colleagues have studied the link between the taste of different foods, the curved and jagged pictures, and words like maluma/takete.
In one paper, for example, Spence, Mary Kim Ngo, and Reeva Misra asked people to eat a piece of chocolate and say whether the taste better matched the words maluma or takete. People eating milk chocolate (Lindt extra creamy 30 percent cocoa) said the taste fit the word maluma (and also matched the curvier figure). People eating dark chocolate (Lindt 70 percent and 90 percent cocoa) instead chose the word takete (and matched the jagged figure). In another paper the
y found similar results for carbonation; carbonated water was perceived as more “kiki” (and spiky) and still water was perceived as more “bouba” (and curvy). In other words, words with m and l sounds like maluma were associated with creamier or gentler tastes and words with t and k sounds like takete were associated with bitter or carbonated tastes.
These associations are very similar to what I also found with consonants in ice cream and cracker names. I found that l and m occurred more often in ice cream names, while t and d occurred more often in cracker names.
So what is it about bouba and maluma that people associate with visual images of round and curvy, or tastes of creamy and smooth, while kiki and takete are associated with jagged visual images and sharp, bitter, and sour tastes? Recent work by a number of linguists studied exactly which sounds seem to be causing the effects.
One reasonable proposal for what’s going on has to do with continuity and smoothness. Sounds like m, l, and r, called continuants because they are continuous and smooth acoustically (the sound is pretty consistent across its whole length), are more closely associated with smoother figures. By contrast, strident sounds that abruptly start and stop, like t and k, are associated with the spiky figures. The consonant t has the most distinct jagged burst of energy of any consonant in English.
To help you visualize this, look at the display on the following page of the sound waves from a recording that I made of myself saying “maluma” followed by “takete.” Note the relatively smooth wave for maluma, which has a relatively smooth flow of air. By contrast, the three sharp discontinuities in takete on the right occurred when I said the sounds t and k; for each of these consonants, the airflow is briefly blocked by the tongue in the mouth, and then a little burst of air explodes out.
The waveform (sound waves) of me saying “maluma” and “takete”
What I call the synesthetic hypothesis suggests that the perception of acoustic smoothness by one of our five senses, hearing, is somehow linked to the perception of smoothness by two other senses: vision (seeing a curvy figure instead of a jagged one) and taste (tasting a creamy instead of sharp taste).
Synesthesia is the general name for the phenomenon of strong associations between the different senses. Some people, like Dan Slobin, a Berkeley professor of psychology and linguistics, are very strong synesthetes. For Slobin, each musical key is associated with a color: C major is pink, C minor is dark red tinged with black. But the bouba/kiki results suggest that, to at least some extent, we are all a little bit synesthetic. Something about our senses of taste/smell, vision, and hearing are linked at least enough so that what is smooth in one is associated with being smooth in another, so that we feel the similarity between sharpness detected by smell (as in cheddar), sharpness detected by touch or vision (like acute angles), and sharpness detected by hearing (abrupt changes in sound).
We can see this link between the senses even in our daily vocabulary. The words sharp and pungent both originally meant something tactile and visual: something that feels pointy or subtends a small visual angle, but both words can be applied to tastes and smells as well.
It’s not clear to what extent these synesthetic links are innate or genetic, and to what extent they are cultural. For example, nomadic tribes in Namibia do associate takete with spiky pictures, but, unlike speakers of many other languages, they don’t associate either the word or the pictures with the bitterness of dark chocolate or with carbonation. This suggests that the fact that we perceive bitter chocolate as “sharper” than milk chocolate or carbonated water as “sharper” than flat water is a metaphor that we learn culturally to associate with these foods. But we really don’t know yet, because we are just at the beginning of understanding these aspects of perception.
There are, however, some evolutionary implications of the synesthetic smoothness hypothesis and of the frequency code.
John Ohala suggests that the link of high pitch with deference or friendliness may explain the origin of the smile, which is similarly associated with appeasing or friendly behavior. The way we make a smile is by retracting the corners of the mouth. Animals like monkeys also retract the corners of their mouths to express submission, and use the opposite facial expression (Ohala calls it the “o-face”), in which the corners of the mouth are drawn forward with the lips possibly protruding, to indicate aggression.
Retracting the corners of the mouth shrinks the size of the front cavity in the mouth, just like the vowels I or i. In fact, the similarity in mouth position between smiling and the vowel i explains why we say “cheese” when we take pictures; i is the smiling vowel.
Ohala’s theory is thus that smiling was originally an appeasement gesture, meaning something like “don’t hurt little old me.” It evolved when mammals were in competitive situations as a way to make the voice sound more high pitched and the smiler appear smaller and less aggressive, and hence friendlier.
Both the frequency code and the synesthetic smoothness hypothesis may also be related to the origin of language. If some kinds of meaning are iconically related to sounds in the way that these hypotheses suggest, it might have been a way for speakers to get across concepts to hearers early on in the evolution of language. The origins of language remain a deep mystery. We do, however, have some hypotheses, like the “bow-wow” theory of language evolution, the idea that language emerged at least partly by copying nature, naming dogs after their bark and cats after their meow and so on. The frequency code suggests that perhaps one of the earliest words created by some cavewoman had high pitched sounds that meant “baby,” or low pitched α sounds that meant “big,” or perhaps was an acoustically abrupt kikiki meaning “sharp.” Such iconic concepts are only a small part of the vast number of things we talk about using language, but iconicity still may help us understand some of these crucial early bootstrappings of human language.
Whatever their early origins, vowels and consonants have become part of a rich and beautiful system for expressing complex meanings by combining sounds into words, just as smiling has evolved into a means of expressing many shades of happiness, love, and much else.
Whatever hidden meanings words and smiles may have, in the end there is always ice cream, as a much later bard, Wallace Stevens, told us:
Let be be finale of seem.
The only emperor is the emperor of ice-cream.
Thirteen
Why the Chinese Don’t Have Dessert
IN TRENDY SAN FRANCISCO, the latest hip dessert is ice cream made with unusual flavors, like the banana-bacon ice cream served at Humphry Slocombe down on Twenty-Fourth Street. Bacon has been appearing on all sorts of desserts for some time now: bacon brownies, candied bacon, bacon peanut brittle. The donut shop down the street even has bacon-maple donuts. Now I’m sympathetic to the argument, made most persuasively by Janet, that bacon makes everything better. But I think that’s not the only reason for bacon’s dessert popularity: there’s something else going on that’s causing the hipster trend for all these bacon desserts. To solve the mystery, let’s begin by tracing dessert back to its early origins.
First of all, dessert doesn’t just mean “sweet food.” A donut on the way to the gym is not dessert; it’s just a lack of willpower. A dessert is a sweet course that’s part of a meal, and eaten at the end.
The placement at the end of the meal is expressed in the etymology: the word dessert comes from French, where it is the participle of desservir, “to de-serve,” that is, “to remove what has been served.” The word was first used in France in 1539 and meant what you ate after the meal had been cleared away; spiced wine called hippocras accompanied by fresh or dried fruit, crisp thin wafers, or candied spices or nuts called comfits or dragées. Such dessert courses under various names have a long medieval tradition, occurring in the very first menu we still have for an English feast, from around 1285, where after “the table is taken away” the guests are served dragées and “plenty of wafers.” (Dragée is still the technical term for a confection with a hard can
died shell like Jordan almonds or M&Ms.)
But the word dragée and the tradition of eating light sweets with wine after dinner dates back far earlier than the Middle Ages. Dragée comes from the Greek word tragemata, which historian Andrew Dalby tells us was the name for the snacks eaten with wine in classical Greece after the table was cleared after a meal. A “second table” was set with wine and tragemata: cakes, fresh and dried fruits, nuts, sweets, and chickpeas and other beans.
Like comfits or dragées, their medieval descendants, however, tragemata were really snacks rather than what we now call dessert. In fact, Herodotus remarked in the fifth century BCE in his Histories that the dessert-loving Persians mocked the Greeks for not really having any proper desserts at all:
[The Persians] have few solid dishes, but many served up after as dessert [“epiphoremata”], and these not in a single course; and for this reason the Persians say that the Hellenes leave off dinner hungry, because after dinner they have nothing worth mentioning served up as dessert, whereas if any good dessert were served up they would not stop eating so soon.
It was the Persians whose love of desserts was the impetus that eventually changed these simple wafers and nuts into our modern desserts. Baghdad was established in a formerly Persian part of Mesopotamia, and a culinary new wave developed, as the great chefs of the Caliphs borrowed and enriched Persian desserts like the sweet almond pastry lauznaj and starch candy f ldhaj, sour dishes like sikbj, and the many sweet stews.
