Keywords: fMRI, stop signal, SPM, SnPM, TFCE, cluster
Statistical parametric maps put a stringent constraint over the t- threshold for obtaining reliable clusters from 2nd level group results for fMRI. Using statistical non-parametric mapping (SnPM) as a tool, we show that the conventional statistical parametric maps generated by SPM software may be over-correcting for several small regions as it sets a t- threshold for all voxels in the brain. We use an emotional stop signal task to show that non-parametric measures uncover more information than SPM recovers for "Go" and "Stop" contrasts from cortical and subcortical brain regions. In addition, we use threshold free clustering estimation (TFCE) to demonstrate further variability in the data. SnPM and TFCE measures provide better alternatives to SPM in terms of correcting for the false postive rate (p < 0.05). Our results encourage use of permutation (500, 1000, 5000 times) instead of assuming fMRI data to follow random field theory assumption of SPM.