I recently ran a partial-trials event-related fMRI with two conditions (A and B, B is presented 70% of the time following A). Each condition consisted of the stimulus onset, a response (from the participant), and "feedback". Feedback was evoked by the response of the participant and simple highlighted the option selected by the subject. Jitter (simple fixation cross) was added between the presentation of A's using a geometric distribution with a max jitter of 4 TR's (each TR is 2000ms). When I run FEAT, I provide two EV's (A and B) described by the onset time (relative to the start of scanning) and the response time associated with that trial. From these two EV's, I develop 4 contrasts (A [1 0], B [0 1], A>B [1 -1] and B>A [-1 1]). The differential contrasts are masked with voxels of A and B, respectively, which Z stats pass threshold.
When running the group analysis (simple mean, one group), the first three contrasts are as expected, but the fourth (B>A) shows activation of the entire cortex and cerebellum (whole brain) minus those areas activated by B. I have ran several analyses before, blocked and event-related, and have not seen this and neither have my colleagues. Can anyone help point to the cause?
|