UCLA

This thesis explores two theoretical frameworks describing intervocalic lenition: a hearer-focused, perception-based framework versus a production-based framework. The perception-based framework (Kingston, 2008; Katz, 2016) posits that intervocalic lenition is primarily driven by a preference for higher intensity within prosodic constituents to aid the listener in distinguishing prosodic phrase boundaries. The production-based framework is an effort-driven theory (Kirchner, 1998, 2004) proposing that speakers wish to minimize articulatory effort when speaking. Articulatory gestures requiring movement that is further and/or faster are dispreferred. The production-based framework predicts that intervocalic lenition should be sensitive to the height of the surrounding vowels, whereas the perception-based framework does not. To distinguish these frameworks empirically, I propose a Poverty of Stimulus artificial grammar learning experiment to explore the extent to which surrounding vowel height influences how readily learners generalize the lenition pattern of spirantization. The results turn out to support a third hypothesis, perceptual saliency (Steriade, 2001a, 2001b; Fleischhacker, 2005). Learners generalize asymmetrically according to the condition they are trained on. The learner is more likely to generalize the pattern if she learns the alternation in a more perceptually salient position and is asked to generalize to a less salient position than vice versa.