I'm a bit confused with how SPM defines brain voxels. I've done a
segmentation prior to the statistical analysis, I then perform 16
different amounts of smoothing (0 - 15 mm FWHM) and calculate t-test
values for each case. If I calculate the number of t-test values that
not are zero, this should in my opinion be constant, i.e. not depend on
the amount of smoothing, but it can for example vary from 40 000 for no
smoothing to 46 000 for 15 mm smoothing...why is this?