Réunion LCA 14 décembre 2015
Salle 159, Pouchet
Title: Idempotency and the faithfulness triangular inequality.
Within the generative tradition, a phonological grammar is a mapping from underlying to surface forms. It has been argued that children manage to acquire some of the target phonology at an early stage when morphology is still lagging behind and little information on the underlying forms is therefore available (see Hayes 2004 for a review). How is that possible? Many authors have suggested that, throughout this initial learning stage, the child pairs any training surface form with a corresponding identical, completely faithful underlying form (see Prince and Tesar 2014 for a review). This learning strategy is computationally sound provided the target grammar is idempotent, namely effectively does not repair phonotactically licit forms but let them surface faithfully. Which conditions guarantee idempotency? This talk addresses this question within two constraint-based phonological frameworks, OT and HG. Within both frameworks, idempotence is shown to be related to the condition that the faithfulness constraints measure the phonological distance between underlying and surface forms in compliance with a crucial axiom of the definition of distance, namely that any side of a triangle is shorter than the sum of the other two sides.