Browsing by Author "Kanerva, Viivi"

Aims. Speech processing has traditionally been studied with simple paradigms that do not take the variety of natural speech perception situations into consideration. Speech processing in everyday situations becomes more difficult if, for example, the auditory quality of speech is poor, there are other distracting voices in the listening environment or if the semantic context of the conversation is not clear. Nevertheless, humans possess an exceptional capability to selectively attend to a specific speech stream even in suboptimal listening conditions. In a recent functional magnetic resonance imaging (fMRI) study, participants followed audiovisual dialogues amongst distracting speech. The dialogues were presented in various experimental conditions including two different levels of semantic predictability. The results gave support to the novel notion that high-level semantic information has an impact already at lower levels of auditory processing. The aim of the present study was to examine whether support for this notion would be demonstrated using a method that has a better temporal resolution than fMRI. I examined the effects and interaction of semantic context and selective attention in the processing of naturalistic speech using the same stimuli as the recent fMRI study but collected neurophysiological data with electroencephalography (EEG) as it has a temporal resolution of milliseconds. Methods. Thirty adults selectively attended to audiovisual two-person dialogues with distracting speech in the background during EEG recordings. Half of the dialogues had a coherent narrative and the other half consisted of mixed lines from unrelated dialogues. The participants answered questions on dialogue content. In addition, the participants carried out a visual control task during which they were instructed to ignore all speech. I analyzed behavioral performance and event-related potentials (ERPs) timed to the beginning of dialogue lines in the two tasks. Results and conclusions. Based on the behavioral results, a coherent semantic context enhanced speech intelligibility. The ERP results suggested that semantic contextual coherence modulated the processes of auditory selective attention at around 180 ms after dialogue line onset. This attentional effect was significant only when the dialogues were attended to, and the semantic context was coherent. Furthermore, the beginnings of lines in incoherent dialogues elicited a P300 effect which could be linked to memory-related operations of contextual updating. Alternative interpretations are discussed. In conclusion, the present study provides both behavioral and neural evidence that semantic contextual information enhances the processes of auditory selective attention in compromised listening conditions. A coherent semantic context seems to facilitate the processing of contextually relevant information through semantic priming already at early stages of auditory attentional processing.