Functional distinctions between auditory cortex and frontoparietal network in the encoding of speech features

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Neural representations of acoustic features that are key to speech perception have been studied extensively within the auditory cortex (AC), but their representations in higher-order cortical regions remain poorly understood. This study investigates the cortical encoding of three acoustic features in spoken vowels to delineate the functional distinctions across different cortical regions. We used ultra-high-field functional magnetic resonance imaging (fMRI) while human participants of both sexes listened to vowels and were asked to identify three aspects: vowel identity (probing spectral content), speaker identity (probing primarily fundamental frequency), and pitch glide direction (probing spectrotemporal modulation). Three brain regions exhibited selectivity to these features: the AC, premotor cortex (PMC), and the intraparietal sulcus (IPS). The AC demonstrated an encoding hierarchy across features, with significantly stronger encoding robustness for vowel identity than for speaker identity or pitch glide direction. In contrast, both the PMC and IPS showed uniform encoding robustness across all three features, likely reflecting their equal relevance for the identification task. Furthermore, during silent trials when participants were queried about the identity of the prior sound, AC showed no activity, whereas the PMC and IPS exhibited robust activation that was modulated by behavioral performance. These findings indicate functionally distinct roles between sensory and higher-order cortical regions: the stimulus-driven auditory cortex encodes feature-specific acoustic information, whereas the domain-general frontoparietal network performs task-relevant cognitive processing that supports goal-directed behavior. Significance Statement Understanding how the brain transforms complex acoustic signals into meaningful percepts during real-world listening is a fundamental challenge in auditory neuroscience. By combining ultra-high-field functional magnetic resonance imaging with a naturalistic multi-feature identification paradigm using spoken vowels, we reveal a functional divergence between sensory and higher-order cortical regions: the auditory cortex exhibits a feature encoding hierarchy prioritizing spectral information, whereas the frontoparietal network (premotor cortex and intraparietal sulcus) encodes speech features uniformly and remains active even in the absence of external stimuli. Our results provide empirical evidence for the functional distinction between two cortical networks, contrasting the stimulus-dependent encoding in the auditory cortex with the domain-general, task-oriented processing in the frontoparietal network, collectively supporting real-world communication.