In my previous post, I quoted some work by George Lakoff about the category mother, and extrapolated it to the case of gender categories. I have a scientific caveat to make. Lakoff was trained by Noam Chomsky, and although he broke publicly with Chomsky in the 1970s, he still uses Chomsky’s methods of introspection and grammaticality judgments. When Chomsky wants to prove a grammatical point, he invents sentences in English and classifies them as “grammatical” or “ungrammatical,” and builds his arguments on those judgments. When Chomsky’s students study languages that they’re not native speakers of, they invent sentences in those languages, find a native speaker and ask that person for grammaticality judgments. This method assumes that grammaticality judgments (a) are valid and unbiased, and (b) hold for every other speaker of the language, assumptions that are not justified.
Lakoff does something similar with his list of “but tests” that are judged as “normal” or “strange,” and I repeated this in my discussion. These informal judgments are useful for speculation, but they have the same problems as Chomsky’s grammaticality judgments. However, the notions of radial categories and prototype effects are based on more than this. Eleanor Rosch herself did reproducible psycholinguistic laboratory tests, and most of the Lakoff work that I’ve described can be accounted for with tests like these.
The laboratory version of the “but tests” are reaction-time tests. In her 1973 paper “On the internal structure of perceptual and semantic categories,” Rosch presented subjects with a number of sentences and asked them to press different keys for “true” or “false.” She measured the time it took the subjects to hit a key after the sentence appeared. There was a significant difference between sentences like “A robin is a bird,” where robins are central members of the category bird, and sentences like “An ostrich is a bird,” where ostriches are peripheral members of the category. The average adult reaction time for central members was 1012 milliseconds, while the time for peripheral members was 1061 milliseconds.
Lakoff’s “but tests” examine whether a particular subcategory is central to the category; they can also test whether a particular criterion marks a member of the category as central. It seems like Lakoff’s subjective notion of “strangeness” is exactly what a reaction test is designed to measure; if it is possible to present these “but sentences” in an experiment and ask subjects reasonable questions about them, their reaction time could be a reproducible measure of “strangeness.”
Unfortunately, as far as I know nobody has done this kind of measurement, so the case for radial structure of gender categories rests on introspection.