Using cognitive neuroscience to understand learning mechanisms: evidence from phonological processing

Abstract

This dissertation studies different learning mechanisms of phonological processing by conducting behavioral and neurophysiological experiments in the artificial grammar learning paradigm. The main goal is to identify the phonological computations that give rise to the complex combinatorics underlying human languages by providing new knowledge about whether linguistic constraints that are learned in laboratory situations are directly “channeled” into incremental, real-time phonological predictive processing. It also explores whether the learning outcome depends on the type of learning mechanisms (domain-specific vs. domain-general) and the computational complexity of the patterns. It achieves this by examining the nature of the neural commitment produced by language exposure via electroencephalogram (EEG). ☐ The specific patterns that were tested in this dissertation are well-motivated from the perspective of formal language theory and theoretical linguistics. Processing of these phonological patterns must ultimately be integrated into the complete cognitive neuroscience of language that links abstract mathematical grammars to real-time theories of word processing and ultimately to biologically plausible neural computations. The dissertation, therefore, contributes to a growing body of research that highlights the nature of acquired linguistic knowledge at the neural level that behavioral measures cannot address. It provides new insight into how artificial language learning causes acquisition mechanisms to incorporate linguistic rules acquired in the laboratory. ☐ This dissertation’s contribution to knowledge is threefold. Firstly, it has been shown that human phonological learning mechanisms are restricted mainly by domain-specific computational constraints. Secondly, specific phonological rules that are acquired in laboratory situations are directly channeled into incremental, real-time phonological predictive processing evidenced at the millisecond level by ERP components. Thirdly, domain-specific learning mechanisms associated with implicit learning and domain-general learning mechanisms associated with explicit learning converge on similar knowledge states, but with different underlying neural mechanisms. ☐ The hope is that these results will illustrate how formal language theory can be integrated with the cognitive neuroscience of language, and how experimental techniques from experimental psychology and cognitive neuroscience can reveal new insights into the critical properties of human language such as the complex interplay between domain-specific and domain-general components of cognition.