53
variance using a mixed-effects model, with stimulus type as the fixed factor and subject
as the random factor.
Most statistical tests were performed only on the average time series created
from all active voxels in each subject's STS, mitigating the need to perform corrections
for multiple comparisons. To create activation maps, a significance level of p<0.05
(single voxel, uncorrected for multiple comparisons) was used for the single modality
activation maps and p<0.01 for the conjunction analysis. The actual probability of the
observed STSms activations being due to chance fluctuations in the MR signal is
considerably lower, approximately p<Pn, where P is the single-voxel р-value and n is the
number of voxels in the STSms (Xiong, et al. 1995). For individual subjects, mean n=17;
for the group map, n=55.
Results
Experimeritl
Subjects received Vibrotactile somatosensory stimulation on their left hand and
auditory stimulation in their left ear while making behavioral responses with their right
hand. To determine brain areas responsive to sensory stimulation, we focused our
analysis on the right hemisphere, collapsing across different intensities Ofstimulation.
As shown in Fig. 2A, tactile-only trials activated a broad network of frontal, parietal and
temporal, including the post-central gyrus (the location of primary somatosensory
cortex, SI), the parietal operculum (the location of secondary somatosensory cortex,
S2), intraparietal sulcus, and the STS. Auditory-only trials activated a similar network of