The Logical and Empirical Foundations of Baker's Learnability Paradox
(Dissertation Proposal Talk)

The learnability paradox known as Baker's Paradox (Pinker 1989, in reference to Baker 1979) arises whenever it appears that there are arbitrary exceptions to a productive rule. For instance, although new verbs can be productively extended to the double object construction when they enter English (e.g. text me the address), certain transfer of possession verbs still cannot be used in this construction (e.g. *donate the library a book). Several authors, including Baker (1981), and more recently Culicover (1999) and Goldberg (2006), assume that such restrictions are arbitrary and conclude that the learner must be conservative and attentive. I question this conclusion on both the logical level and the empirical level. On the logical level, I attempt to clarify, and provide a modern perspective on, the consequences of the existence or non- existence of arbitrary exceptions for learning. On the empirical level, I question the claimed existence of arbitrary restrictions in several domains: the double object construction, the attributive adjective construction, and preposition pied-piping and stranding. In all of these domains, I argue that the relevant restrictions follow from general principles. I conclude that learning them does not require attentiveness to the use of particular words in (these) particular constructions.

Baker, C. L. (1979). Syntactic theory and the projection problem. Linguistic Inquiry, 10:533-581.

Culicover, P. W. (1999). Syntactic Nuts: Hard cases, Syntactic theory, and Language acquisition. Oxford University Press, Oxford.

Goldberg, A. (2006). Constructions at Work. Oxford University Press.

Pinker, S. (1989). Learnability and cognition: the acquisition of argument structure. MIT Press, Cambridge, MA.