Presented as part of the ANU Linguistics Seminar Series

Generative Grammar is based on the idea that there is a `Poverty of the Stimulus' problem, in that the data to which children are exposed and thereby learn their native language is relatively limited, probably not more than several tens of millions of words in some cases, if the oft-cited figures in Hart and Risley (1994} are correct. But the grammars that they then come up with give them ability to understand and produce many more than what they can have heard. This gives us the problem of 'projecting' (Peters 1972) the grammar they acquire from the limited and in a certain sense insufficent stimulus that they are getting. But I claim that we are missing some numbers that are needed to assess how serious the PoS problem really is, and whether it requires anything like an

'interesting' theory of 'universal grammar' to solve it. Namely, how rare are the complex consstructions that supposedly require UG to be acquired, versus the more basic ones that are supposedly learned as such, as basic proporties of the language.

In the talk, I try to start looking into this question by considering some aspects of English prenominal possessives in the CHILDES database for English, noting expecially that 'doubly recursive' full NP possessives such as "Zack's mother's boyfriend's sister" don't appear to occur at all in the 13.5 million words of adult & some sibling speech that I investigated, and so are plausibly projected from things that occur more often. I propose that the main bases are a) singly recursive possessives such as "daddy's friend's house", occurring at a rate of a bit more than 6 per million words, and adjectivally possessed possessors ("that poor cat's tail"), which occur at a rate of about 30-40 per million words. To get from these observations to the prediction of the doubly recursive possessors, we also seem to need a bit of theory, in partucular, something with similar effects to the X-bar theory of classic generative phrase structure, and another principle (or conseuqence of deeper principles) that I call 'Guessing Rule I'. The conclusion is that certain amount of 'modest' UG might be motived in this case.

Hart and Risley (1994) _Meaningful Differences in the Everyday Experiences of Young American Children

Peters (1972) 'The projection problem: how is a grammar to be selected', in Peters (ed) _Goals of Linguistic Theory