Activity in regions sensitive to auditory speech is modified during speech production: fMRI evidence for an efference copy
MetadataShow full item record
Models of speech production postulate that, in order to facilitate rapid and precise control of articulation, the predicted auditory feedback is sent to the auditory system to be compared with incoming sensory data. If this is so, an 'error' signal may be observed when the predicted auditory feedback and the sensory consequences of vocalization do not match. I used event-related fMRI to look for the neural concomitants of such an error signal. In two conditions volunteers whispered 'ted'. In one of these, voice-gated noise implemented in our real-time processing system was used to mask the auditory feedback, which should result in an error signal. Two other conditions were yoked to the production conditions (either clearly heard or masked), but were listen-only and therefore no error signal would be expected. I acquired whole-brain EPI data from 21 subjects using a fast-sparse design. Activity in the superior temporal gyrus bilaterally was significantly greater for clear than masked speech during the listen-only trials (F(1,20)≥12.84, p<0.002), and significantly higher for masked than for clear speech in the production trials (F(1,20)≥6.68, p<0.02). This crossover interaction indicates that speech production results in corollary discharge in the auditory system and furthermore suggests that this corollary discharge reflects expectations about the sensory concomitants of speech acts.