Neural Speech Tracking during Selective Attention: A Spatially Realistic Audiovisual Study.

Abstract

Paying attention to a target talker in multitalker scenarios is associated with its more accurate neural tracking relative to competing non-target speech. This "neural bias" to target speech has largely been demonstrated in experimental setups where target and non-target speech are acoustically controlled and interchangeable. However, in real-life situations this is rarely the case. For example, listeners often look at the talker they are paying attention to while non-target speech is heard (but not seen) from peripheral locations. To enhance the ecological-relevance of attention research, here we studied whether neural bias toward target speech is observed in a spatially realistic audiovisual context and how this is affected by switching the identity of the target talker. Group-level results show robust neural bias toward target speech, an effect that persisted and generalized after switching the identity of the target talker. In line with previous studies, this supports the utility of the speech-tracking approach for studying speech processing and attention in spatially realistic settings. However, a more nuanced picture emerges when inspecting data of individual participants. Although reliable neural speech tracking could be established in most participants, this was not correlated with neural bias or with behavioral performance, and >50% of participants showed similarly robust neural tracking of both target and non-target speech. These results indicate that neural bias toward the target is not a ubiquitous, or necessary, marker of selective attention (at least as measured from scalp-EEG), and suggest that individuals diverge in their internal prioritization among concurrent speech, perhaps reflecting different listening strategies or capabilities under realistic conditions.