Hi,
In both cases, a threshold is applied to the statistical map (for example, t = 3). For each surviving cluster of voxels, a new test statistic is calculated. This test statistic can be:
a) The number of voxels in the cluster. This is the "cluster extent" or "cluster size".
b) The sum of the t-statistics of all voxels inside the cluster. This is the "cluster mass".
In both cases, the p-values are computed using permutations in randomise.
All the best,
Anderson