THE COGNITIVE AND NEURAL BASIS OF LANGUAGE LEARNING: INVESTIGATIONS IN TYPICAL AND AUTISTIC POPULATIONS

Abstract

This dissertation will focus on the processing and learning of abstract, phrasal argument structure constructions. Chapter 1 provides the theoretical framework for abstract constructions, and illustrates the importance of such representations in speakers' linguistic knowledge. Chapter 2 reviews the evidence for meaning being associated with abstract phrasal forms, including experiment 1, which supports the idea that constructional meaning is accessed online and automatically.

The dissertation then focuses on how these stable representations are initially learned. This will focus on two avenues of research, which explore this issue in two different ways. First, chapter 3 will discuss the neural mechanisms that contribute to construction learning. This chapter will present evidence from two neuroimaging studies (experiments 2 and 3) - the first scientific investigations that probe the neuroscientific basis of abstract construction learning. These findings will be discussed in detail; they suggest that construction learning may rely heavily on a largely domain general learning system, including a prediction based mechanism. We also find that such abstract learning may alleviate low level visual processing in previously unexplored ways.

Next, chapter 4 reports research on construction learning in children with autism. This will include behavioral work (experiment 4) suggesting that children have difficulty learning abstract constructions, and discuss the implications of this finding. The chapter closes with final thoughts on our current understanding of language delay in autism, which includes both social and cognitive factors.

Overall this dissertation report provides several novel insights into the learning and processing of abstract phrasal correspondences, and will argue that they are 1) a crucial element of our linguistic knowledge, allowing language to be used flexibly 2) are accessed quickly and implicitly during online sentence comprehension 3) are likely learned at least in part on the basis of a prediction-based, domain general learning mechanism, by generalizing over discrete pairings of form and function, and 4) require a type of learning which the (locally biased) cognitive style of autism is ill-suited for.