Nativism, Formal Learning Theory, and Universal Grammar
Determining the nature of the language faculty has always been one of the central problems in linguistic theory. Is this faculty a rich set of powerful domain specific constraints on language acquisition, or does it consist of weak initialization biases for domain general learning procedures? The key debates on this question turn on learnability considerations, where these are formulated as versions of the "argument from the poverty of the stimulus" (APS). This discussion continues to play a major role in shaping the field.
This course will comprise a series of lectures that address this question from a variety of perspectives. We will analyze the APS in its various forms, and we will consider how it supports linguistic nativism. We will focus on formal versions of the APS which derive their force from the computational hardness of the language learning task. We will see how these arguments can be made more precise within both traditional and modern learning paradigms. In the second half of the course we will pursue a constructive reply to the APS by offering solutions to the learnability challenges that it presents. We will show how some of the ideas and methods of distributional learning (which have their origin in the work of structuralist linguists) provide the basis for developing provably correct procedures that learn complex, richly structured language classes. We will also summarize current heuristic work on word segmentation, morphology, and syntax induction, which is yielding encouraging results for naturally occurring corpora.
Basic formal language theory including knowledge of the Chomsky hierarchy.
Textbook (Available at CU Bookstore):
Title: Linguistic Nativism and Poverty of the Stimulus
Author/Editor: Clark & Lappin
Publisher: Wiley Blackwell
Mon & Thu 1:30-3:15
Classroom: ATLAS 229
Areas of Linguistics:
Language Development and