52
To identify areas responding to auditory, visual and somatosensory stimulation,
a modified conjunction analysis was used (Nichols, et al. 2005). In each subject, the t-
Statistic of the contrast between stimulation vs. rest was independently calculated for
each sensory modality in every voxel. This contrast revealed voxels that showed either a
positive or negative BOLD response to sensory stimulation. Because a task-independent
network of brain areas is deactivated (negative BOLD response) during any kind of
sensory stimulation (Raichle, et al. 2001) we selected only voxels showing a positive
BOLD response to each sensory modality. This criterion was instantiated with the
thresholding operation (Visuakt-Statistiox) AND (Auditory-t-statistic>x) AND (Tactile-t-
Statistiox) where × is the Unisensory threshold (Beauchamp 2005b). All voxels passing
this test were classified as "mu∣tisensory", mapped to the cortical surface and classified
as inside or outside the STS using an automated surface parcellation algorithm (Fischl, et
al. 2004). The time series from all multisensory STS voxels were converted to percent
signal change and averaged to create an average time series for each subject. These
time series were then averaged across subjects to create a grand mean.
A conjunction analysis was also used to create the mixed effects group map.
Individual subject brains were converted to standard space (Brett, et al. 2002), and the
percent signal change for each condition was entered into a voxel-wise ANOVAwith
subject as the random factor and condition as the fixed factor. A conjunction analysis
was performed on the output of the ANOVA to find voxels showing a significant effect to
each modality in isolation. All statistical inferences are based on between-subjects