Presentation

Sara Beach¹, Swathi Kiran², Caroline Niziolek¹; ¹University of Wisconsin–Madison, ²Boston University

Phonetic distortions are common in the speech of persons with aphasia (PWA), but it is unknown whether their source is primarily perceptual (an insensitivity to errors) or motoric (an inability to correct detected errors). In typical speakers, the auditory system monitors itself for acoustic deviations from intended speech sounds. The M100 evoked response is suppressed to one’s own speech in comparison to playback of the same. We previously showed that the magnitude of speaking-induced suppression (SIS) is larger for prototypical vowel productions and smaller for deviant ones; moreover, the smaller the SIS, the more participants corrected their vowel formants towards the median (Niziolek et al., 2013). Thus, the degree of SIS may be a signal that drives speakers to detect and correct deviations before they become full-blown errors. Here, we use this framework to characterize the detection-correction circuit in PWA who vary in clinical subtype, severity, and lesion. In a MEG study, 15 PWA and 15 age-matched controls spoke the words ‘eat’, ‘Ed’, and ‘add’ 200 times each and listened to playback of their own utterances. We take advantage of each participant’s natural variability in vowel acoustics to identify “center” and “peripheral” trials, in which the first and second formants (F1 and F2) in the first 50 ms are closest to and farthest from, respectively, that vowel’s median. We then calculate SIS for prototypical and deviant productions. If a PWA does not show the typical pattern of less SIS to peripheral vs. central trials, it suggests that their perceptual sensitivity to acoustic deviations is compromised. If a PWA shows the typical pattern of SIS but without the subsequent behavior of acoustic correction, it suggests that their deficit is in motor control. Preliminary results indicate that, whereas SIS is left-lateralized in controls, it is right-lateralized in many PWA. This may indicate reorganization of a critical auditory function after injury. We hypothesize that PWA with more typical SIS patterns in the right hemisphere will show more acoustic evidence of online speech correction. Planned analyses delve deeper into how vowel acoustics explain neural activity. We will perform representational similarity analysis, correlating models of vowel similarity with measures of neural similarity. The acoustic model predicts greater neural similarity for utterances that are closer in F1-F2 space. The deviance model predicts similar neural activity for utterances that are similarly distant from the median, regardless of formant values. Thanks to the spatiotemporal resolution of MEG, we will test these models across the brain as speech perception unfolds. If the acoustic model is a good fit, it suggests the presence of an accurate sensory representation – the basis of intact perception. If the deviance model is a good fit, it suggests the presence of abstract information about proximity to the target sound. We hypothesize that a deviance representation supports error detection and may also be related to online speech correction. Our results may inform the tailoring of aphasia therapy to individual profiles and advance our understanding of neural mechanisms of feedback processing during speech motor control.

Topic Areas: Speech Motor Control, Disorders: Acquired