Effects of semantic predictability and speaking style on phonetic variation

Abstract

Many linguistic factors contribute to phonetic reduction in «easy» linguistic contexts and phonetic enhancement in «hard» linguistic contexts. However, previous findings are mixed as to whether interactions among these factors are simply additive and as to whether their effects are differentially exhibited across acoustic domains. To better understand phonetic variation processes, we explored the interaction between two factors, semantic predictability and listener-driven speaking style, on phonetic variation in duration, f0, and vowel formant frequencies. Talkers read aloud sentences differing in the semantic predictability of the sentence-final word (high versus low predictability) to purported listeners with different linguistic competence (a native listener versus a non-native listener with high proficiency versus a non-native listener with low proficiency). Talkers produced phonetic enhancement for less predictable words and for the low-proficiency non-native listener relative to the native and high-proficiency non-native listeners across measures, but the effects of predictability and speaking style did not interact. Together, these results suggest that phonetic variation due to semantic predictability and speaking style may be additive, independent processes, rather than the result of a single listener-oriented process.