Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Functional and structural connectivity of auditory areas that process talker variability in speech
Poster A82 in Poster Session A, Tuesday, October 24, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Tyler Perrachione1, Ja Young Choi2; 1Boston University, 2Harvard University
The acoustics of speech are highly variable across talkers, and listeners must recognize speech efficiently in the face of this variability. The computational challenge of talker variability is evident when listeners recognize words spoken by single talker (quickly and accurately) vs. spoken by multiple different talkers (slower and less accurately) – a phenomenon called “talker adaptation.” Prior neuroimaging, noninvasive neurostimulation, and invasive electrophysiology in humans have suggested that talker adaptation reflects processes in bilateral superior temporal gyrus (STG), but it is unknown how these local computations are integrated with processes supporting speech recognition throughout the rest of the brain. Here, we modelled generalized psychophysiological interactions (gPPI) using functional MRI and probabilistic tractography using diffusion MRI to investigate how auditory areas that are sensitive to talker variability are functionally and structurally connected to the rest of the brain during speech recognition. Using sparse-sampling fMRI, we obtained whole-brain BOLD data from healthy neurotypical participants (n=19) while they performed a word recognition task. Participants matched visually-presented pictures to auditory words in blocks where the words were spoken by either a single consistent talker or multiple different talkers. Univariate analysis of the fMRI data revealed talker adaptation (reduced BOLD response during single- vs. multiple-talker blocks) in bilateral STG. A whole-brain parcellation obtained via group-constrained, subject-specific (GCSS) analysis revealed that within-subject talker adaptation effects were localized in four probabilistic parcels, including left and right, anterior and posterior STG. We circumscribed subject-specific functional regions of interest (fROIs) within each of these parcels, corresponding to the decile of voxels with the greatest univariate response to talker variability (multiple > single talkers). We used these fROIs as the functional and structural seeds for the gPPI and probabilistic tractography analyses. From the gPPI analyses, we found that these auditory fROIs had greater functional connectivity to domain-general brain areas outside of the spoken language network (including left middle frontal gyrus, anterior cingulate, bilateral visual areas, thalamus, and right cerebellum) when listening to a single talker vs. multiple talkers. Of the fROIs, right posterior STG showed the most extensive task-dependent changes in functional connectivity to the rest of the brain. From the probabilistic tractography analyses, we found that all four auditory areas sensitive to talker variability had similar patterns of ipsilateral structural connectivity, including prominent connections to posterior parietal and occipital cortices, as well as contralateral projections via the splenium of the corpus callosum. Taken together, these results suggest that auditory areas that are sensitive to talker variability are functionally and structurally integrated in a broader whole-brain attentional network. These neural results are consistent with recent models of speech processing proposing that behavioral phenomena associated with talker adaptation are due in part to domain-general auditory attention and streaming of a single auditory source (the talker), whereas auditory detection of talker changes incurs stimulus-driven attentional reorientation and thereby imposes additional domain-general cognitive demands on speech recognition as listeners must reorient their attention from the prior talker to a new one.
Topic Areas: Speech Perception, Control, Selection, and Executive Processes