Hello Dave!

> DP>> Do you mean the voxelwise average over the time window? Or the spatial average over the ROI during that window?

Sorry for the imprecision. The latter.

I'm probably misunderstanding the terminology. In the context of 3dDeconvolve, "impulse response" refers to the convolved HRF, right? Is it being used in a different context here?

> DP>> From this part, it sounds like you're effectively trying to get the impulse response for that regressor averaged over the ROI.

Does IR here refer to the actual activation (since otherwise it doesn't differ per voxel, so the ROI is irrelevant)?

> It might be easier to extract the impulse response directly with -iresp option to 3dDeconvolve then use 3dMaskAve on the iresp output to get the ROI average impulse response...

You've lost me. If this means that I can extract the activation (averaged over the ROI) for only those times during which the regressor's IR is expected to be nonzero, then we're getting close.

So let me try to explain what I'm doing, a little more clearly this time:

There's a stimulus that occurs some 20 times in the experiment, and a regressor that captures this. Each time the stimulus occurs, I want to record the average activation in the ROI, for the next few seconds (until the IR is expected to return to zero). Now I have 20 time courses, each 10 seconds long. I want to average them into one 10-second time course.

In case this was obvious, and you've already explained how to do this, I apologize :)

Cheers,
A