Congruent and Incongruent Semantic Context Influence Vowel Recognition

Abstract

The influence of sentence context on the recognition of naturally spoken vowels degraded by reverberation and Gaussian noise was investigated. Target words were paired to have similar consonant sounds but different vowels (e.g., map/mop) and were embedded early in sentences which provided three types of semantic context. Fifty-eight normal-hearing, young adults were presented with sentences in which acoustic and semantic cues agreed either weakly (neutral) or strongly (congruent) or the cues strongly disagreed (incongruent). One vowel pair (/epsilon/-/ae/) was selected to be easier to recognize than the other (/a/-/ae/). Changes induced in the spectra of the vowels by degradation showed that the impact of reverberation combined with noise was quite different from either condition alone. The recognition performance of participants (n=26) for isolated word stimuli matched the predictions of the frequency analysis. In sentences the recognition of the vowel was strongly influenced by the subsequent context; performance was best with congruent context and worst with incongruent context. The deleterious impact of incongruent context was larger than the helpful impact of congruent context. Incongruent context effects were greatest in noise but were also found in quiet and in reverberation.