The multiple cognitive processes required for speech comprehension rely on multiple cortical networks that operate in parallel. This functional organization resembles the anatomical separation of the primate auditory system into dorsal and ventral processing streams. In this work, we test whether coupling within these networks depends on the nature of the auditory input. If observed, modulations of functional connectivity would suggest rapid, shortterm changes in functional organization for speech comprehension: a form of neural plasticity.
The present study uses fMRI to identify brain regions involved in sentence comprehension. Davis & Johnsrude (2003) distinguished two response profiles in lateral temporal regions that show activity correlated with the intelligibility of acoustically distorted sentences. BOLD responses were either: (1) form-dependent: responding differentially to three acoustically different forms of distorted speech, or (2) form-independent: insensitive to acoustic differences in the form of speech. Formdependent regions border primary auditory cortex, suggesting early-stage acoustic processing of speech, whereas more distant form-independent regions may be primarily involved in higher-order, linguistic processes.
We scanned a further 15 neurologically normal, right-handed English volunteers on a passivelistening version of the previous study. Participants listened to and tried to understand sentences (1.7 to 4.3 sec long) presented with three types of distortion in three different amounts. fMRI data were acquired with a 3T Bruker MRI system using sparse imaging (TR 9s, TA 3s; sentences presented in silence between scans). As before, a significant correlation between sentence intelligibility and BOLD signal was observed in the left inferior frontal gyrus (LIFG) and along the superior and middle temporal gyri bilaterally. This correlation was form dependent around auditory cortex and form-independent in four left-hemisphere regions, which we chose as seed voxels for connectivity analysis: (1) anterior STS (- 56, 0, -18), (2) posterior STS (-60, -54, 20), (3) anteroventral LIFG (-50, 28, -16) and posterodorsal LIFG (-52, 16, 18).
Functional connectivity analysis of all distortedspeech conditions revealed significant functional coupling between anterior temporal and ventral frontal regions, and between posterior temporal and dorsal LIFG regions. This pattern of connectivity conforms to the anatomical organization observed in nonhuman primates, suggesting that speech perception recruits multiple, temporofrontal networks. Furthermore, coupling between the posterodorsal LIFG and posterior temporal regions increased in strength as sentence intelligibility increased. This finding indicates that functional connectivity among regions is plastic, changing with the nature of the current input and with task demands. Acknowledgements: Research supported by Medical Research Council, UK; National Institutes of Health, USA.