August 24, 2004

Hot features

On men, front is hot, back is not. But on women, it's back that's hot, while front is not. We're talking about vowels here, mind you, and Charles Darwin may be raising his eyebrows a bit as he discusses this matter with Edward Sapir and Aristotle at the University of Heaven.

Brad Pitt and Laura Dern have an advantage over John Wayne and Britney Spears, according to Amy Perfors, who did a study of the effects of phonetic symbolism on sexual attractiveness. She presented a paper entitled "What's in a Name? The effect of sound symbolism on perception of facial attractiveness" at CogSci 2004 (August 5-7 in Chicago) whose proceedings published the abstract, and a more informal description with some additional information can be found on her website.

What Perfors did was to post photos of 12 male and 12 female friends on, with names photoshopped in. She posted each picture multiple times, with different names. What she found was an interaction between the sex of the picture and the stressed vowel of the name: "Men whose names' stressed vowel is a front vowel were rated statistically MORE attractive than men with names with a stressed back vowel. However, the reverse was true for women: women with names with stressed BACK vowels were statistically more attractive."

She also found a sex-linked effect with initial sonorants vs. obstruents: obstruents were hotter on males, sonorants on females, though only the female effect was significant. All of these effects were small -- about a quarter of a point on a 10-point scale, or less, as you can see from the graph.

It's a cute piece of work, and it's not surprising that it made Nature, New Scientist, Guardian, Boston Herald, as well as Reuters/CNN.

So why do I think that Charles Darwin might be raising his eyebrows? Well, I'm sure that he's worried about the ease with which evolutionary psychologists can gin up stories about the reasons for perceptual preferences and contextual effects on behavior, no matter what these turn out to be. And he's familiar with the many pieces of research showing tendencies for female animals to make their vocal tracts seem smaller to emphasize their femininity, and for male animals to make their vocal tracts seem larger when they're acting masculine. So he knows, as Perfors does, that if the vowel hotness results had come out the other way around, there would have been an equally ready explanation.

I'm sure that's why, on her website, Perfors is very tentative in offering an explanation for the effect:

"Why do you think this happens? I want to stress here that I'm JUST GUESSING... but there has been other work suggesting that cross-linguistically, people think that front vowels are 'smaller' and back vowels are 'larger.' Now, you'd think that that would make guys with front vowels do worse than guys with back vowels (and vice versa), but there are other studies suggesting that women actually aren't most attracted to the super hyper-masculine, macho guys (and men aren't most attracted to the super feminine girls). It seems we like people who are somewhat masculine or feminine, but not too much. The reasoning is that maybe women want guys who can be kind of sensitive and gentle, and good providers - while guys want women who have a bit of spunk. What does this have to do with names? Well, maybe a guy with a front-vowel name seems subconsciously gentler or more sensitive, hence more attractive (and vice-versa for women). Again, this is a complete chain of guesswork; it could be another explanation entirely. More research needs to be done. But it makes a bit of sense, when you think about it."

I suspect that Darwin also might be worried about some other things, in this particular case. The nature of the English lexicon of names makes it impossible that Perfors' list was strictly controlled, phonologically and otherwise. You can't contrast (say) Beet and Boot, or Bit and Butt, or other "names" that differ only in the front-back dimension of their main-stressed vowel. Even if you could, the names would not be equally common (overall or in a particular age range), or equally associated with famous people, or whatever. Instead, the list of names with front vowels surely differed from the list of names with back vowels in many other ways, phonetically and otherwise. Perfors doesn't give the complete list that she used, or the raw results, so it's hard to tell whether there are any other plausible differences. And if she didn't start the study with the hypothesis that front-back was going to make the difference, but instead considered the 20 or so obvious phonological alternatives -- high vs. low vowels, labial vs. non-labial consonants, one syllable vs. two, open syllables versus closed, etc. -- then there's the statistical problem of multiple tests. And what's the distribution of sexual orientations of the "subjects"? These are the kinds of annoying, picky little questions that reviewers (are supposed to) ask for publication in refereed journals. Perfors may well have answers for such questions, and if she publishes in a well-refereed journal, she'll have a chance to bring them out.

I hope that if that journal is Nature, they somehow find some editors and referees who know more about the subject than the one they tapped to write their news article on Perfors' work (in, "The best in science journalism"). The writer, Michael Hopkin, seems to have good credentials:

"Michael became an online news reporter for Nature in January 2004, after two and a half years as a subeditor for Nature's print edition. Besides contributing to newspapers, magazines and online publications, he has given numerous interviews as a science expert on BBC radio. Michael has a BSc in biology from the University of Nottingham..."

but if he ever took a linguistics course, he wasn't paying attention. He closes the article with these two howlers:

Perfors argues that the discovery that vowel sounds can influence a person's perceived attractiveness is the more interesting finding, because it seems to be a subconscious effect. Experts, including the Swiss linguist Ferdinand de Saussure, have previously argued that vowel sounds are arbitrary building blocks with no intrinsic meaning.

Which brings us to the most pressing question of all: is my own name, Mike, a help or a hindrance when it comes to attractiveness? "Mike is a front vowel sound, so it's a good name," says Perfors. "If you do badly with the ladies you can't blame it on your name."

The first problem here is that there are hundreds if not thousands of published articles on sound symbolism, mimesis and even on whole systematic sound-symbolic subvocabularies called ideophones, so it's preposterous to cite Perfors' research as if it were the first suggestion ever that l'arbitraire du signe is not absolute.

The second problem is worse. Mike is NOT "a front vowel sound". The vowel in the name spelled "Mike", in all English dialects that I know of, has a low back nucleus.

If Hopkin quoted Perfors correctly, and if she was not just being polite, then she might have been confused by the spelling. The symbol "i" in IPA (as in most orthographies) denotes a high front vowel. But in English, as a result of the Great Vowel Shift, the nucleus of long vowels written with orthographic "i" lowered and backed, all the way to the bottom back corner of the vowel quadrilateral. In most contemporary dialects, it's a diphthong with a high front off-glide, so you might take it as mixed on the hot-or-not dimension, but "a front vowel sound" it is definitely not.

If you think about it, this is a sad state of affairs. A journalist who served for "two and a half years as a subeditor for Nature's print edition", who passes for "a science expert on BBC radio", and who was assigned to write a story about sound symbolism for a publication that advertises itself as "the best in science journalism", turns out to be completely ignorant of the most elementary phonetic terminology, as applied to the pronunciation of his own native language. Worse, the term in question was the key independent variable in the experiment under discussion, and he not only didn't know what it meant -- as applied to the kind of words studied -- he didn't bother to find out.

As before, I blame the linguists, for not insisting that anyone who claims to be an educated person needs elementary competence in describing and analyzing the sound, form and meaning of human language.

[link via Erika at Kittenishly Doomy Thoughts]

[Update: as Eric Bakovic points out over at Phonoblog, Perfors demonstrates in her webpage discussion of the experiment (which I cited above) that she knows about the feature composition of the diphthongal pronunciation of long i, so (unless she suffered a moment of distraction in a telephone interview, or something) the error is entirely Hopkin's. ]


Posted by Mark Liberman at August 24, 2004 07:20 PM