Modeling the listener? What resets acoustic durations of repeated English words

Abstract

Listener-based accounts of speech production claim that speakers modify their speech based on their evaluation of the listener’s state of knowledge (Lindblom, 1990). In line with this, repeated words shorten when they have been previously said to the same listener (Fowler, 1988); however, repetition across an episode boundary in a narrative does not lead to decreased acoustic duration (Fowler et al., 1997). We replicate Fowler et al.’s story boundary effect and extend the study by testing whether a switch in listener has an additional effect on word duration. Speakers were asked to tell and retell the same story in the sequence of (A) listener 1/(B) listener2/(C) listener1 again (Galati and Brennan, 2010). We expect word durations to reset when the speaker starts a new narrative, especially when there is a switch in listener. In other words, word durations should be comparable in conditions (A) and (B), but shorter in (C), since the listener in condition (C) has heard the story before. Acoustic data from 20 American English native speakers have been collected and transcribed; data analysis is ongoing. This study is intended to shed light on the interplay between production economy and the need to transmit information.