Learning morpho-phonological alternations like French liaison : an artificial language learning study

Morpho-phonological phenomena present challenges to learners because they interface with morpho-syntactic, semantic, and phonological grammar. A common difficulty encountered in these interface phenomena lies in navigating idiosyncrasies and exceptional cases. This thesis explores the nature of learning implications of morpho-phonological alternations, in particular focusing on the case study of patterns akin to French liaison. Liaison is a prevalent and complex phenomenon in French that occurs at word boundaries. I propose a theoretical account of the learnability of French liaison within the Optimality Theory framework which gives rise to predictions regarding various components and dimensions of learning. I evaluate these predictions in an experimental study with an artificial language learning paradigm that consists of a training phase and a testing phase with two-alternative-forced-choice tasks. The training phase has a between-subject design where learners experience different hypothesized learning trajectories. The aim of the experiment is to investigate the extent to which learners would face challenges in learning given that the artificial language includes idiosyncratic patterns. The results demonstrate that in a controlled experimental setting, adult learners do show partial differential learning outcomes depending on their learning trajectory in training. However, participants also show task-dependent performance: above-chance for noun phrase choices but near-chance for word segmentation with both novel and familiar nouns. These results, taken together, imply that one of the most challenging components of learning in patterns similar to liaison predominantly lies in mechanisms related to computing word boundaries and deriving representations.