November 13, 2006

7 - 38 - 55!

I'm not calling a football play; those are the famous Mehrabian numbers, giving -- in the usual citations of this research -- the percentages that verbal content, paralinguistic features (vocal quality, prosody, etc.), and kinesic features ("body language", broadly construed) contribute, respectively, to the total impact of a message.  When I last mentioned this research, I noted that the great avalanche of bizlore -- the lore of corporate trainers, motivational speakers, advertising advisers, and the like -- using the Mehrabian numbers went drastically far beyond Mehrabian's own claims, which were that these figures applied only to the communication of attitudes and emotions.  As it turns out, the actual results of Mehrabian's 1967 studies are much more modest than even this, as Ed Keer noted in his blog back in February (building on a longer discussion by Richard Sproat on Linguist List in 2001).

Check out Keer and Sproat for details.  The fact is that the 1967 studies weren't about the communication of attitudes and emotions in general, but about the communication of one specific set of attitudes and emotions, liking and disliking.  Ok, you say, I had no earthly idea how anyone could study the relative contributions of verbal content, paralinguistics, and kinesics to the total impact of a message (whatever that means), but I still don't see how this much more modest question could be investigated experimentally: what do you measure, and how?

Good question.  What Mehrabian did was pit features (expressing liking/disliking) in the three modes against one another to see which mode prevailed, and how often, when they were in conflict.  Even if we accept his results at face value -- and there are many details of the experimental design and the interpretation of the data that a reasonable person could fret about -- all that Mehrabian did in 1967 was, in Keer's words, to discover sarcasm, in this case the conveying by extralinguistic devices of a meaning opposite to the plain meaning of the words.

That's stage one.  In stage two, these results morph into a global generalization about language use, which then spreads into all sorts of places outside the academic world.  The details of this transformation and diffusion would be worth looking at.  (With luck, Mehrabian himself has relevant materials from the 60's and 70's.)  No, no, don't look to me to do this research; I'm the guy with over a hundred postings in his queue for Language Log, and I'm not a cultural historian.

In any case, I'd imagine that science writers for the general press, and their editors, had a hand in the spread of the Mehrabian numbers to a wider world.  (I mention editors, because many a science writer has had an original text altered, in small or large ways, to make it conform to the beliefs of editors -- whatever the content of the original.  And then, famously, headlines are often attached that seriously distort that content.)

That's stage two.  In stage three, the Mehrabian numbers become part of bizlore, indeed part of a larger set of folk beliefs.  Most people are no longer aware of the source of the numbers, and most people who cite Mehrabian haven't looked at the original studies or any careful summary of them;  it's "common knowledge" now.

Bizlore is just one part of an enormous enterprise of popular advice literature -- on education, child-rearing, exercise, diet, relationships, gardening, and more, including grammar, usage, and style.  Bizlore focuses on persuasion, power, and the fostering of positive emotions, with the aim of helping people achieve success in business dealings of all sorts. 

All sorts of popular advice literature, not just bizlore, appeal to "common sense" and folk beliefs; rely heavily on personal opinions and impressions (of the advisers and their audiences); and get points across largely via particular examples, often by telling exemplary stories of personal experience (we all love stories).  Notice that this is not at all the way scientific inquiry proceeds -- but it IS the way ordinary people reason about their world and their lives.  "Science" appears in popular advice literature mostly for its value as dressing: there are numbers, real numbers; and actual researchers or institutions, of some prominence (or apparent prominence), can be appealed to, however spuriously.  "Science" is just one more element in the rhetoric of popular advice literature, rarely an actual contributor to it.  (There are some honorable exceptions, of course.)

In any case, you can see why bizlore loves the Mehrabian numbers.  They're wonderfully impressive.  So exact, and from a real scientist!

The Mehrabian numbers also plug into a powerful folk belief about how human interaction works -- that we are "communicating" (passing back and forth) "messages" to one another.  Ordinary people (and some social scientists) conceptualize interaction in terms of the "conduit metaphor" discussed in several places by Michael J. Reddy (most recently, I think, in the 2nd edition (1979) of Metaphor and Thought, edited by Andrew Ortony) and made famous by George Lakoff in many of his writings.  Now, everyone, including social scientists as a group, recognizes that paralinguistics and kinesics contribute a lot to the texture of interaction, so it's natural for ordinary people to think that linguistic expressions, paralinguistic features, and kinesic features are just three different modes of communicating the same messages, and it then makes sense to ask what their relative contributions are.

Two problems, one of substance, one of method.

The first is that there's no reason to think that the three modes are ways of conveying the SAME "meanings", or even that CONVEYING meanings is what's going on.  I would maintain, with many others, that there are many different kinds of "meaning" at issue here, and that it would be more accurate to say that the features of behavior in question (depending on the occasion) express, reflect, perform, or construct these meanings than to say that they simply convey them.

The second problem is that the question being posed -- what are the relative contributions of (strictly) linguistic content, paralinguistics, and kinesics? -- is, as I suggested above, one of those impossibly over-global questions that almost surely can't be answered.  The methodological difficulty here is that what happens in each of the three modes is exquisitely context-dependent.  I can't see any way to sample behavior, from the whole world of human interactions, while controlling for these differences in context; without such controls, what we see might well just follow from differences in the frequencies of the various contexts, rather than from some intrinsic difference between the modes.  (There's also the problem of individuating contexts.  Where do we get an inventory of the relevant types of context, even in one culture?)

What I'm saying here is that there are some questions about language and behavior that are easy to formulate but so global that they are probably unanswerable in principle.  (At least some of the questions about differences between the sexes, such as Louann Brizendine's claim that women use many more words per day than men -- now discussed here in a long series of postings by Mark Liberman -- are almost surely unanswerably over-global.  I hope to post on that eventually.)

A semi-final remark: linguists will probably be struck by what counts as (strictly) linguistic (vs. paralinguistic or kinesic) in Mehrabian's research and everything that cites it or alludes to it: apparently, only aspects of utterances that contribute to literal meaning.  To a linguist, this is desperately impoverished view of language use, excluding most of the subject matter of entire subfields of linguistics.  Almost all variation in the linguistic system is ignored, thus neglecting the ways in which, through their use of particular linguistc variants, people express, reflect, perform, and construct social group affiliations and personas; the ways they express or reflect attitudes and opinions towards their audiences (including liking/disliking!), about the nature of the interaction, etc.; and the ways in which they use linguistic choices to structure their discourses (via discourse particles, for example).  These are "social meanings" and "discourse meanings", if you want to put everything under the umbrella of "meaning".

Also missing is everything to do with non-literal meaning: for instance, implicatures of all sorts, fresh figures (especially metaphors and metonyms), and other rhetorical devices.  Emotions and attitudes can be expressed or revealed through all these means, too.

On a more constructive note, I can remind you that linguists, psychologists, sociologists, and anthropologists have long concerned themselves with the ways in which linguistic content, choices of variants, discourse organization, paralinguistic features, and kinesic features are coordinated with one another and can combine into suites of behaviors associated with "meanings" of all sorts.  For a beautiful recent example of research along (some of) these lines I recommend Rob Podesva's Stanford Ph.D. dissertation, Phonetic Detail in Sociolinguistic Variation: Its Linguistic Significance and Role in the Construction of Social Meaning, completed this summer (it will be available eventually, in chapter-sized chunks, on his website).  Podesva looks at the way three speakers' uses of one segmental variable (realization of word-final coronal stops) and two paralinguistic variables (prosody and voice quality, in particular falsetto) are associated with different personas in different contexts.  These associations are very much local, in that they are tied to particular social groups and to particular contexts, as well as to individual speakers (the three speakers -- all friends -- use the variables in different ways).

(Full disclosure: I was a member of Rob's dissertation committee.)

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at November 13, 2006 03:38 PM