Distinct Brain Systems for Sensory and Decision Processes in Speech Perception

Abstract:
Abstract: Computational accounts of sensory perception distinguish analysis of sensory data from application of decision criteria following sensory analysis. Such models predict that during a discrimination task, the total neural activity in sensory networks should depend on stimulus quality, while activity in decision networks should correlate with response latency. We searched for the brain correlates of these systems using fMRI in 16 subjects during phoneme discrimination in noise. Stimuli were synthesized CV syllables /ba/ and /da/ differing only in F2 transition. Subjects heard syllable pairs and indicated which interval contained a designated target syllable. The level of a simultaneously presented white noise mask was systematically manipulated to effect changes in discriminability and response latency. Images were obtained at 3 Tesla using a clustered acquisition method to avoid contamination by scanner noise. Using regression analysis, responses in bilateral auditory association areas in lateral Heschl's gyrus and anterolateral planum temporale were shown to increase linearly as a function of discrimination accuracy. Bilateral regions in buried frontal operculum and anterior insula showed responses that increased linearly as a function of response latency. These data implicate auditory cortex anterolateral to AI in the analysis of spectral-temporal acoustic cues and suggest that frontal opercular activation during phoneme perception reflects post-sensory response selection processes.