We used 3dttest to examine activation from functional run1 to run 2 (identical stimuli). I assumed that using the -paired option would result in a t-value smaller than without the -paired option given that a repeated measures should result in greater power than a independent samples t-test. However, the t-values associated with a per voxel p<.01 indicated that the t-value was lower for the 3dttest without the -paired option than with the -paired option. Same order of the subjects and number of subjects. Should this not be the opposite?