| Statistical learning---the process of extracting patterns from distributional properties of the input---has been proposed as a key mechanism for acquiring knowledge of the probabilistic dependencies intrinsic to linguistic structure. While such a view would predict that greater sensitivity to statistical structure should lead to better language performance, this theoretical assumption has rarely been tested empirically.;Accordingly, the work presented in this thesis is among the first to establish empirical links between statistical learning and language through the framework of studying individual differences. Contrary to assumptions that incidental learning abilities are invariant across individuals, the first small-scale individual-differences study reported systematic differences in statistical learning among normal adults, which were substantially correlated with broad cognitive measures, including language comprehension.;In two subsequent studies, a novel experimental paradigm (the AGL-SRT; Misyak, Christiansen, & Tomblin) was used to probe for within-subjects associations between individual differences in statistical learning and online sentence processing. The findings point to an overall positive relationship between individual differences in the statistical learning of adjacent or nonadjacent dependencies and learners' processing for corresponding types of structures occurring in natural language (such as for local and long-distance dependencies entailed by subject-object relatives and subject-verb agreement sentences).;However, the complexity of the pattern of interrelations observed throughout the three studies also suggests that language and statistical learning may be related in more intricate, and sometimes counterintuitive, ways than traditionally supposed. Through discussion of theoretical implications, it is claimed that future efforts to empirically bridge together differences in statistical learning with variations in language should aid in elucidating further the broad perceptual-cognitive processes upon which statistical learning and language mechanisms may commonly supervene. |
