The influence of uncertainty about ear of presentation, gender, and linguistic structure in following streams of speech.

The ability of listeners to attend to speech streams was studied for conditions in which low-level acoustic cues (talker gender and ear of presentation) and high-level context cues (uncertainty and linguistic structure) were varied, to better understand how these factors interact when reinforcing or competing with one another. In experiment 1, two sentences spoken by two different-gender talkers were presented simultaneously to separate ears. Listeners were instructed to report the sentence according to talker gender or ear. As the sequence of words progressed in a trial, the talkers either switched unpredictably between ears (creating uncertainty) or remained fixed. Results showed that switches between ears degraded intelligibility, but the benefits of syntactic structure remained. In experiment 2, a single stream of speech was presented in which the talker gender and ear of presentation could unpredictably switch on each word. Listeners reported a subset of the words according to talker gender or ear, with target and non-target words together forming a syntactic sentence or with non-target words syntactically anomalous. More non-target word confusions were found for syntactic sentences, likely due to the binding effect of syntax. Overall, the findings emphasized the importance of both low-level and high-level predictability for speech identification.