June 06, 2004

Psycholinguistics in the logging industry

A little while ago, Sally Thomason's interest in sustainable forestry turned up this quote:

Mammal sightings: black bear, bobcat, cougar, white-tailed deer, mule deer, elk, chipmunk, ground squirrel, flying squirrel, snowshoe hare, mice, and vole

which led her to wonder why "the zero-plural-for-game-animals usage in this list" doesn't apply to the mice. Mice aren't game animals, but as Sally points out, neither are voles and chipmunks (though in Never Cry Wolf, Farley Mowatt claims to have survived for a time by hunting small rodents and eating them whole, in order to prove a point; but I digress).

Sally insightfully suggested that the mice/vole business might have something to do with the morphology of the non-head elements of English compound nouns:

My first thought was that the asymmetry between the plural form mice and the singular forms vole, etc., must have basically the same motivation as the parallel asymmetry in certain types of compounds, discussed, I think, by Peter Gordon: mice-eater but rat-eater, with *rats-eater impossible.

But maybe not, because as I recall (and I might be misremembering), Gordon's explanation for the asymmetry in compounds had to do with late vs. early plural formation, depending on whether it was the default regular plural (as in rats) or an irregular plural like mice; and that wouldn't be relevant for the list in the logging book. I also don't think the asymmetry is an idiosyncrasy of this author, because it sounds fine to me, and replacing mice with mouse doesn't. So what governs this pattern in the list of mammal sightings?

The same issue about noun compounds came up last winter, as Geoff Pullum and I exchanged some thoughts on the rigors of fieldwork and the distribution of activities centers in Las Vegas, Santa Cruz and Chapel Hill NC.

I was a bit surprised not to have gotten more email about those activities centers, because the (near) prison riot over strict transitivity that Sally described in another post is a tea party in comparison to what happens when you toss this topic into a roomful of psycholinguists. A clear presentation of one side of the controversy is Haskell, T.R., MacDonald, M.C., & Seidenberg, M.S. "Language learning and innateness: Some implications of compounds research". Cognitive Psychology, 47, 119-163. (2003):

In noun compounds in English, the modifying noun may be singular (mouse-eater) or an irregularly inflected plural (mice-eater), but regularly inflected plurals are dispreferred (*rats-eater). This phenomenon has been taken as strong evidence for dual-mechanism theories of lexical representations, which hold that regular (rule-governed) and irregular (exception) items are generated by qualitatively different and innately specified mechanisms. Using corpus analyses, behavioral studies, and computational modeling, we show that the rule-versus- exceptions approach makes a number of incorrect predictions. We propose a new account in which the acceptability of modifiers is determined by a constraint satisfaction process modulated by semantic, phonological, and other factors. The constraints are acquired by the child via general purpose learning algorithms, based on noun compounds and other constructions in the input. The account obviates the regular/irregular dichotomy while simultaneously providing a superior account of the data.

I don't recall that either side of this controversy has addressed the "game animal plural" facts that Sally has pointed out, but I'm not sure, and I can't easily check because I'm traveling, with limited time and limited internet access, and the relevant books and papers are back at Penn. However, I think that Sally is right to connect these two phenomena.

One of the results featured in the Haskell et al. paper (cited above) is that various semi-regular plurals (like wolf-wolves) are intermediate in acceptability as the first element of compounds -- so compounds like wolves-eater are rated as somewhat better than compounds like rats-eater, but somewhat worse than those with singular first elements (like rat-eater or wolf-eater), and (presumably) also a bit worse than mice-eater. Putting this together with gradient results from a corpus study, they argue that the facts are against dichotomous theories (like level-ordering or theories based on a regular/irregular dichotomy).

One could do an analogous experiment on the "game animal" zero plural phenomenon: if we were to add canis lupus to the Mammal sightings list, how do we feel about wolf vs. wolves, in comparison (say) to squirrel vs. squirrels? Haskell et al. would predict (I think) that wolves fits better than squirrels but not as well as mice.

For me, I think it depends on where in the list it comes -- near bear and cougar, I think I'd tend to say wolf, but near squirrel and vole, I'd be happy to say wolves. I think. My confidence in this judgment is somewhere between small and tiny.

Google tells us that "hunting wolves" is unexpectedly common, relative to the analogous patterns with other game-like animals. On the other hand, "hunting mice" is unexpectedly rare. Overall, it's not clear, in explaining these data, that the regularity of the plural formation is a major influence:

musk ox/musk oxen

Posted by Mark Liberman at June 6, 2004 10:10 PM