Philippe,
Besides the subjective graphical methods of looking at things like Q-Q plots, there are several tests for normality. A few that look at skewness and kurtosis are D'Agostino & Tietjen 1971; D'Agostino 1971; D'Agostino, 1972; D'Agostino & Pearson 1973; Mulholland 1977; and Lin & Mudholkar 1980. These papers all seem to be in Biometrika.
For fMRI model diagnostics, you may want to look at
Diagnosis and exploration of massively univariate neuroimaging models, NeuroImage, Volume 19, Issue 3, July 2003, Pages 1014-1032
Wen-Lin Luo and Thomas E. Nichols
Dan