Ideal Theory

Abstract

The puzzle of how children are able to acquire linguistic structures and principles with seemingly little to no input to guide them – the “poverty of the stimulus” – has long motivated research in linguistics, particularly in generative syntax, where it has shaped the search for innate principles of grammar. I propose that the poverty of the stimulus is not an empirical generalization, but an artifact of model choice: the stimulus can be “enriched” simply by using a formalism that accounts for the interactions between the subsymbolic structures underlying symbolic language use. In this dissertation, I develop such a formalism using ideals, collections of subsymbolic cognitive patterns reflecting the range of variation in the cognitive representation for some concept, as the basic structural element. By exploiting the combinatorial properties of ideals, agents of a model gain access to vast stores of indirect positive evidence in the input that they can use to overcome the poverty of the stimulus. To show how, I build models of anaphoric one, Principle C, and the interaction between wh-movement and the complex noun phrase island constraint, all commonly considered evidence of an impoverished stimulus. In each case, agents can use ideals to construct linguistic representations that reflect the observed behavior by relying only on evidence freely available in the input. These examples demonstrate a set of strategies for solving poverty of the stimulus puzzles in general – without the need for innate grammatical principles.