Effect of phonological neighborhood density on word duration in spontaneous speech.

Abstract

It has been widely shown that words with many phonological neighbors are harder to recognize than words with few neighbors (L. & Pisoni 1998, among others). However, the effect of phonological neighborhood density on word production is less clear. On one hand, high‐density words are easier to produce (Vitevitch 1997, 2002), which suggests that they should be reduced in speech. Meanwhile, these words are often hyperarticulated, probably to make it easier for the listener (Wright 1997, Munson & Solomon 2004, Scarborough 2002). The present work aims to distinguish the two accounts by examining the effect of neighborhood density on word duration. We use data from the Buckeye corpus, which has 40 American English speakers, each recorded for about an hour. A mixed‐effect model is built to predict the durations of all CVC monomorphemic content words in the corpus, with speaker and word as random effects. The results show that, after controlling for other factors, neighborhood density has a small but robust negative effect on word duration, i.e., the more neighbors a word has, the shorter it becomes. The effect of average neighbor frequency is in the same direction. Present results provide unambiguous evidence for the talker‐oriented account concerning word duration.