Dear list,
I have a fast event-related design experiment where subjects receive a visual or tactile stimulation on either side, and subsequently have to wait for the fixation dot to turn from green to red to give their response (on what side were they stimulated) by pressing a foot pedal. The intervall between the stimulus onset and the "go" signal is jittered between 1500 and 2500 ms.
In a first analysis that I run as a "sanity check", I included one regressor per type of sensory stimulation, and one per type of motor response (left, right and double), contrasted them vs. rest (20 sec periods included in the implicit baseline). I get sensible activations for each type of sensory stimulation, without activations related to the (correlated) motor response, which are well captured by the corresponding regressors and contrasts.
In a second analysis that is supposed to give me the contrasts I'm really interested in (i.e. differences across sensory stimulations depending on the behavioral responses), I included only the regressors corresponding to the different types of sensory events, not those modeling the motor responses, and contrasted my conditiosn of interest against each other rather than against rest. Here, I again get sensible activations with respect to the sensory stimuli, but also strong activations in primary and pre-motor cortices related to the motor response usually associated with the sensory stimulus (e.g. left M1 for right tactile stimuli) when I contrasts one condition vs. another that does not involve the same motor response (e.g. tactile left vs. tactile right). A critical contrast compares trials with double stimuli (e.g. tactile left/visual right) with a correct answer (double response) to those where the subject missed either one of the two stimuli (left or right response).
So I ran the same analysis again, only this time including an additional EV (not used in the contrasts) modeling all the motor responses (left, right, double), hoping that this would capture the activity related to motor responses and "clean" my contrasts of interest. Unfortunately, I get results very similar to those of my second analysis.
Can anyone think of a better approach to the problem? Or do I need to change my design (either make the jitter longer, or make so that repsonses are comparable across conditions (e.g. same hand, different fingers))?
Many thanks to anyone who can help, as always!
Best,
Stéphane
|