July 24, 2004

Caring less all the time: A variant of the etymological fallacy, and some cautions about the pragmatics-phonetics connection

Mark Liberman (most recently, here and here) has been looking, with increasing skepticism, at Steve Pinker's claim that the idiom could care less is sarcastic in intent and sarcastic in prosody. For current usage, I'm as dubious of these claims as Liberman is.

But if we're right, then Pinker's perceptions of both the pragmatics and the phonetics of could care less are mistaken. How could this happen? I suggest that Pinker is subject to a variant of the etymological fallacy, the idea that the "true" meaning of a word is its historically "original" meaning -- applying here to pragmatics and phonetics, rather than semantics and lexicon.

I also share with Liberman a deep distrust of Pinker's apparent assumption that purposes and bits of phonetics are connected to one another in very simple ways. Instead, I maintain that the usual situation is a many-to-many association, much like what we see in the connection of social meanings (or aspects of a persona) to bits of linguistic form (phonetics included), or the connection of semantics to aspects of syntactic form.

Let's start with the core of Liberman's objections: Pinker maintains that "the expression I could care less 'is not illogical, it's sarcastic.' I agree that the phrase is not a mistake in logic, but I think that Pinker is wrong about the sarcasm. And I'm pretty sure that he was wrong to argue that the melody and stress of the phrase convey -- to those who don't have a 'tin ear' -- that it's being used sarcastically."

In a small defense of Pinker, I note that I have heard occurences of I could care less with a marked prosody, occurrences that I took to be sarcastic in intent -- in particular, using a positive to convey a negative, and drawing attention to this reversal by phonetic means. On the other hand, I don't hear this often, and in fact don't recall having heard it for some time now. What I hear now, all the time, is phonetically unexceptional productions of an idiom with negative import that happens to contain no standard negative marker.

Now, one plausible hypothesis about the origin of this idiom -- hang on, I'll eventually get to Liberman's alternative hypothesis -- is that it began in sarcasm, both in intent and in prosody. The intent would be easy to miss; sarcasm is notoriously risky, likely to misfire. The prosody, too, would be easy to miss; prosodic contours do lots of things, so that even a remarkable one might be interpreted as just emphasis or foregrounding. The situation is ripe for reanalysis, after which later generations of speakers -- most current speakers -- will lack both the intent and the prosody.

Pinker is maintaining that things haven't really changed. He's fallen into a variant of the etymological fallacy. On the contrary: where something comes from isn't necessarily its description now.

In its textbook manifestation, the etymological fallacy has to do with semantics. People maintain that "decimate" can't mean 'almost entirely wipe out' because it really means 'wipe out one-tenth of'. Or that "since" and "while" can only be used as temporal connectives, not as logical ones (meaning, roughly, 'because' and 'although'), because that was their original meaning. ("Original" is, of course, a moving target here.) What's going on here is a reluctance to recognize change, and that idea can be applied to all sorts of innovations: a [t] in "often" or an [l] in "walk" (note that these things can come around in cycles); the use of past forms in counterfactual conditionals ("if I was your father"); plural subject-verb agreement with "none" ("None of the students were prepared"), because "none" is really "not one" and therefore singular; and so on.

Another manifestation of the etymological fallacy shows up in the way many ordinary people (PITS, People In The Street) think of non-standard, innovative, regional, informal, etc. usages. Many of these usages have their origins in what could broadly be labeled as "mistakes" or "errors" -- via regularization, reanalysis, generalization, hypercorrection, and the like -- and PITS are inclined to see them as still errors still, as (inadvertent) failures to attain the correct usage. This attitude towards variation leads to what I think of as the Repetition Annoyance Syndrome, or RAS: PITS are mightily annoyed when speakers or writers keep producing the manifestly incorrect usages, time after time. "There he goes again", they cry out in exasperation, as "She talked with Tom and I about it" is succeeded by "That really pleased Tom and I" and so on, one nominative coordinate object pronoun after another, in what strikes many PITS who abhor this construction as a perverse indulgence in error.

But enough of the etymological fallacy. Let's get back to the origins of could care less. Mark Liberman has suggested an alternative history, via "negation by association", that involves no sarcasm at all. This is an attractive story, but it's not necessarily the whole story; there's no need to claim that idioms and constructions each have a single historical source, and in this case a number of people (Pinker among them, but me too) have reported hearing utterances of could care less that seemed clearly sarcastic in intent (accompanied, perhaps, by a raised eyebrow or a wry smile) and were prosodically marked, so that models were available for the scenario based on the bleaching of sarcasm. We don't have to choose between the negation-by-association and the bleaching-of-sarcasm scenarios. In fact, it seems likely to me that both effects contributed to getting us to where we are now.

A parallel. Consider the Isis construction ("The problem is is that we have to leave now"), last treated on Language Log, I think, here. There are two obvious sources for this construction: (1) the pseudocleft construction of "What the problem is is that have to leave now" (with restructuring) and (2) a pause and re-start in production, as in "The problem is, um, is that we have to leave now" (with grammaticalization of a production strategy). Opinion on the historical source of Isis is increasingly tending towards the view that both phenomena contributed to its development. Opinion on the synchronic description of Isis, meanwhile, has pretty much crystalized in the view that it doesn't involve a simple encapsulization of either source. For a recent proposal exhibiting both of these views, see Jason Brenier and Laura Michaelis's "Optimization via syntactic amalgam: Syntax-prosody mismatch and copula doubling".

Enough of diachrony vs. synchrony for this posting. What about the claim that there's a "sarcastic prosody"? Liberman is dubious about this, and I agree, whether we interpret the claim as being (a) that there's a prosodic contour devoted to conveying sarcasm (and nothing else, though sarcasm might be conveyed by other means), or as being (b) that sarcasm is (always) associated with a particular prosodic contour (which might have other functions as well). Claim (b) is straightforwardly false, since sarcasm often goes unmarked in any way, sometimes is marked only by gesture or facial expression, and can be marked linguistically by a variety of means. I myself often use creaky voice to flag that an utterance is to be taken in some special way (of which sarcasm is one possibility), and I believe that other speakers sometimes use drawl, a stretching out in time, to flag utterances this way, and that some use falsetto voice. No doubt there are other phonetic resources that could be pressed into service for this purpose.

Claim (a) is subtler. It could, conceivably, be true. There could be a phonetic resource that was used for only one purpose. It's just massively unlikely. The world of phonetic resources is really big, but the world of purposes is really big, including as it does all sorts of stuff: conveying conversational intents (like sarcasm), marking discourse functions (like foregrounding), and displaying social relationships (like intimacy), social identifications (like gender), and aspects of a persona (like flirtatiousness, dependability, authority, ฿or flamboyance). Multifunctionality is pretty much guaranteed.

Much the same is true of other choices from inventories of formal resources: lexical choices, choices of alternative inflectional forms (like past "shrank" or "shrunk"), choices of a particular inflection form (like the present participle or the past participle), choices of one syntactic construction over another. In my current mantra for such things: It's all just stuff. That is, it's all just material that can be invested with some sort of content -- pragmatic, discoursal, social, personal, semantic. Nothing is intrinsically associated with some particular content (using the higher end of your pitch range can convey femininity -- or any number of other things), and even where the associations are conventionalized, as in the associations between semantics and syntax, the usual situation is a many-to-many mapping.

So talk of a "sarcastic prosody" is pretty much sure to be misleading. Though there's nothing wrong with saying that sarcasm can, on occasion, maybe even conventionally, be conveyed though the choice of a particular prosody. Which would be a generous way of interpreting Pinker's original claim. Though then, as Liberman points out, the claim becomes devilishly hard to test.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at July 24, 2004 04:18 PM