Dear Elisa,
this is due to the non-stationary extent correction. In a "smoother" region, clusters have to be larger to reach significance, while smaller clusters in "sharper" regions might already be significant. For details see Hayasaka et al. (2004, Neuroimage). Thus, you can't determine a threshold of k voxels corresponding to e.g. .05 FWE any longer, as it varies across the brain. This might require some manual masking or photo-editing to generate properly thresholded figures.
Best,
Helmut
|