Semantic predictability mediates effects of phonetic reduction and social variation on word recognition

Abstract

Phonetic reduction due to lexical frequency, phonological neighborhood density, and discourse mention, as well as speaking style and social-indexical variation can constrain listeners’ ability to identify spoken single word tokens. In phrasal contexts, however, semantic predictability facilitates word recognition. The aim of the current study was to investigate how semantic predictability influences the intelligibility of words that vary in their lexical, stylistic, and socio-indexical properties. Listeners were presented with auditory English phrases extracted from read passages and were asked to identify each phrase. Phrases were mixed with speech-shaped noise and each contained a target word of interest. Linguistic and social properties of the target words were used to predict listeners’ target word recognition accuracy. These factors included semantic predictability, lexical frequency, neighborhood density, speaking style, discourse mention within the passage, talker gender, and talker dialect. The results revealed that greater semantic predictability increased word recognition accuracy, but only for the most phonetically reduced words (e.g., high-frequency words, second mentions, and words in a plain style). These results suggest that the semantic predictability benefit is enhanced primarily for words that might otherwise be difficult to recognize when removed from their semantic context.