Poster Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions

Acoustic and Linguistic Mechanisms of Auditory Attention Switching

Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai

Jaimy Hannah¹, Sara Carta¹, Anna Sawicka¹, Emina Aličković^2,3, Johannes Zaar^2,4, Alejandro López Valdés¹, Giovanni Di Liberto¹; ¹The University of Dublin, Trinity College, ²Eriksholm Research Centre, Oticon, Denmark, ³Linköping University, ⁴Technical University of Denmark

Humans are remarkably good at selectively listening to speech streams of interest while avoiding irrelevant background sounds, a phenomenon known as the cocktail party effect. We are also able to rapidly switch our attention between different conversations if something of interest pulls our focus elsewhere. Although the neurophysiology of sustained attention in these cocktail party scenarios has been widely studied, far less is known about how attention switching takes place. In two EEG experiments, we investigated cortical speech tracking while listeners switched their attention between two competing speech streams. Participants were cued to switch their attention every 15-30 seconds, allowing us to investigate the temporal dynamics of disengaging from one stream and engaging with another. In experiment 1, we measured the neural encoding of two competing speech streams in multitalker babble. Our results revealed an asymmetry in the switching process with the engagement of the new stream emerging before disengaging from the previous stream, producing a transient period where both streams were being tracked. This period was closely mirrored by a reduction in EEG alpha power consistent with increased cognitive effort during the switching process. To probe how linguistic representations are updated across shifts in attention, we isolated cortical activity reflecting lexical prediction and compared competing hypotheses that were constructed using Large Language Models. Our findings suggest that attention switching relies not only on reorientation toward a new speech stream, but also a rapid updating of the linguistic context. Specifically, it seems that the context is dynamically reset following each switch. In experiment 2, we isolated acoustic and linguistic contributions to attention switching by manipulating speaker identity and semantic context. In one condition, participants switched between voices while listening to one continuous narrative (acoustic switch). In a second condition, participants switched between two different stories while the voice remained constant (semantic switch). The third condition was the same as in experiment 1, requiring participants to switch both the speaker and the semantic context. The temporal dynamics of the neural tracking around the switch differed across the three conditions, indicating separate contributions of acoustic and linguistic processing when switching attention. In particular, the disengagement process was significantly longer when the speaker remained consistent, and the engagement process was significantly shorter when the semantic context remained consistent. These findings suggest that attention switching in speech relies on dissociable neural mechanisms for reorienting to a new speaker and updating the linguistic context. Together, these results provide evidence that switching attention between speakers is not a unitary process. Attentional reorientation unfolds across different levels of representation, requiring listening to dynamically select new auditory targets and rapidly update linguistic context. A better understanding of these mechanisms can help explain why listeners differ in their ability to follow conversations in these complex cocktail party listening scenarios and could ultimately inform approached targeted and supporting speech comprehension in noise.

Topic Areas: Speech Perception, Computational Approaches