TY - JOUR
T1 - Listening to two speakers
T2 - Capacity and tradeoffs in neural speech tracking during Selective and Distributed Attention
AU - Kaufman, Maya
AU - Zion Golumbic, Elana
N1 - Publisher Copyright:
© 2023 The Authors
PY - 2023/4/15
Y1 - 2023/4/15
N2 - Speech comprehension is severely compromised when several people talk at once, due to limited perceptual and cognitive resources. In such circumstances, top-down attention mechanisms can actively prioritize processing of task-relevant speech. However, behavioral and neural evidence suggest that this selection is not exclusive, and the system may have sufficient capacity to process additional speech input as well. Here we used a data-driven approach to contrast two opposing hypotheses regarding the system's capacity to co-represent competing speech: Can the brain represent two speakers equally or is the system fundamentally limited, resulting in tradeoffs between them? Neural activity was measured using magnetoencephalography (MEG) as human participants heard concurrent speech narratives and engaged in two tasks: Selective Attention, where only one speaker was task-relevant and Distributed Attention, where both speakers were equally relevant. Analysis of neural speech-tracking revealed that both tasks engaged a similar network of brain regions involved in auditory processing, attentional control and speech processing. Interestingly, during both Selective and Distributed Attention the neural representation of competing speech showed a bias towards one speaker. This is in line with proposed ‘bottlenecks’ for co-representation of concurrent speech and suggests that good performance on distributed attention tasks may be achieved by toggling attention between speakers over time.
AB - Speech comprehension is severely compromised when several people talk at once, due to limited perceptual and cognitive resources. In such circumstances, top-down attention mechanisms can actively prioritize processing of task-relevant speech. However, behavioral and neural evidence suggest that this selection is not exclusive, and the system may have sufficient capacity to process additional speech input as well. Here we used a data-driven approach to contrast two opposing hypotheses regarding the system's capacity to co-represent competing speech: Can the brain represent two speakers equally or is the system fundamentally limited, resulting in tradeoffs between them? Neural activity was measured using magnetoencephalography (MEG) as human participants heard concurrent speech narratives and engaged in two tasks: Selective Attention, where only one speaker was task-relevant and Distributed Attention, where both speakers were equally relevant. Analysis of neural speech-tracking revealed that both tasks engaged a similar network of brain regions involved in auditory processing, attentional control and speech processing. Interestingly, during both Selective and Distributed Attention the neural representation of competing speech showed a bias towards one speaker. This is in line with proposed ‘bottlenecks’ for co-representation of concurrent speech and suggests that good performance on distributed attention tasks may be achieved by toggling attention between speakers over time.
KW - Attention
KW - Cocktail party
KW - Distributed
KW - Neural speech tracking
KW - Selective
UR - http://www.scopus.com/inward/record.url?scp=85149421167&partnerID=8YFLogxK
U2 - 10.1016/j.neuroimage.2023.119984
DO - 10.1016/j.neuroimage.2023.119984
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
C2 - 36854352
AN - SCOPUS:85149421167
SN - 1053-8119
VL - 270
JO - NeuroImage
JF - NeuroImage
M1 - 119984
ER -