Objectives: We measured language network activation using an auditory phrase recognition fMRI task in a sample of 45 high-functioning autism and 20 typically developing (TD) subjects, group matched for age and IQ. The task consisted of subjects listening to spoken phrases, with instructions to think of a word being described by each phrase. To evaluate for differences in neural timing, we used a novel method of intersubject phase correlation wherein normalized time series data were compared across individuals to obtain voxel-level maps of synchronous activation between subjects.
Methods: For each pair of 65 subjects, fMRI time series were compared at each voxel for significant correlation in time between the pair of subjects. For each pair of subjects at each voxel, a cross-correlation curve was computed and the optimal phase recorded. For TD vs. TD, TD vs. autism, and autism vs. autism pairs of subjects, the average phase difference for each sample was recorded at each voxel in the brain.
Results: Classical language areas showed significant intersubject correlation in a spatial distribution similar to the activation map obtained from conventional general linear model analysis for the language task. In addition to language areas, areas of the default mode network also showed synchronous timing across subjects to the language task. Within TD subjects, intersubject phase correlation was able to accurately measure the phase delay of left Broca’s area activation relative to left primary auditory cortex as 2.3 +/- 0.08 seconds, compared to average duration of spoken phrases of 2.2 +/- 0.2 seconds, as expected in a language task where subjects listen to the sentences and then respond covertly. Phase delays of 1.3 seconds in left Broca’s area, 1.8 seconds in right Broca’s area, 1.1 seconds in precuneus, 0.8 seconds in supplementary motor area, and 0.5 seconds in left Wernicke’s area were observed in autism group relative to the TD group. Within the autism sample, intersubject correlation was slightly but significantly less than for TD population in primary auditory cortex, left Broca’s area, left Wernicke’s area, and supplementary motor area.
Conclusions: Intersubject phase correlation was able to measure significant delays in activity in autism relative to TD subjects, including both language regions and the default mode network. This finding may represent propagation of small phase delays in early sensory processing areas to cause widespread dyssynchrony in distributed cortical networks in autism responsible for allocation of attentional resources and processing complex cognitive tasks. Intersubject phase correlation may represent a useful technique for assessing spatial distribution of neural timing delays in autism and other neurological disorders.