"What" and "When" Predictions Jointly Modulate Speech Processing

Area of Science:

• Neuroscience
• Cognitive Science
• Auditory Processing

Background:

• Adaptive behavior relies on predicting environmental regularities, encompassing both stimulus content ('what') and timing ('when').
• Speech processing involves hierarchical predictions at multiple levels of content (syllables, words) and timing (fast, slow scales).
• The neural mapping between these content and timing hierarchies remains unclear.

Purpose of the Study:

• To investigate how hierarchical 'what' and 'when' predictions in speech interact within the brain.
• To determine if distinct neural hierarchies process interactions for different speech units (syllables vs. words).
• To explore whether 'what'-'when' integration relies on sensory-specific or generic mechanisms.

Main Methods:

• Magnetoencephalography (MEG) recorded neural activity in 22 healthy volunteers.
• Manipulated 'what' (syllable, pseudoword) and 'when' (fast, slow temporal regularity) regularities.
• Analyzed neural responses to deviant stimuli and effective connectivity.

Main Results:

• 'When' regularity modulated 'what' mismatch responses with hierarchical specificity.
• Deviant pseudoword responses were amplified by slower temporal regularity, while syllable responses were affected by faster regularity.
• Interactive effects were localized to shared frontal and parietal cortical regions, indicating distributed processing.

Conclusions:

• The brain integrates 'what' and 'when' speech predictions when they are hierarchically congruent.
• This integration is mediated by a shared, distributed cortical network rather than distinct sensory-specific hierarchies.
• Findings suggest a unified mechanism for integrating temporal and content-based predictions in auditory processing.