Dear Will, Something like the following should do it. Statistical parametric maps were initially thresholded with an uncorrected voxel threshold of p = .oo1. To account for multiple testing only clusters surpassing a corrected cluster threshold of .o5 FWE-R based on Gaussian Random Field Theory / a topological FDR of .05 were considered as significant. For FWE cite Friston, Worsley, Frackowik, et al. (1994, Hum Brain Mapp - there might be other papers) for FDR cite Chumbley & Friston (2009, Neuroimage). Best Helmut