Attentional Modulation of Hierarchical Speech Representations in a Multitalker Environment

Area of Science:

• Neuroscience
• Auditory Neuroscience
• Cognitive Neuroscience

Background:

• Humans excel at the cocktail-party task, selectively listening to one speaker amid noise.
• The precise neural mechanisms and extent of attentional modulation across speech features remain unclear.

Purpose of the Study:

• To characterize attentional effects on speech representations at various feature levels.
• To investigate how attention modulates neural activity in different brain areas during speech perception.

Main Methods:

• Recorded whole-brain blood-oxygen-level-dependent (BOLD) responses using functional magnetic resonance imaging (fMRI).
• Subjects passively listened to single stories or selectively attended to one speaker in overlaid stories.
• Developed spectral, articulatory, and semantic models of speech and used voxelwise modeling to analyze neural responses.

Main Results:

• Attention broadly modulated speech representations across multiple levels, with effects strengthening at later processing stages.
• Unattended speech was represented up to the semantic level in the parabelt auditory cortex.
• Intrinsic selectivity profiles were identified for individual voxels during passive listening.

Conclusions:

• Attentional mechanisms play a crucial role in filtering auditory information and enabling selective listening.
• Neural representations of speech are dynamically modulated by attention, influencing processing from basic acoustic features to complex semantics.
• The auditory cortex maintains representations of unattended speech, suggesting sophisticated neural strategies for managing complex acoustic environments.