You can tell AFNI (via to3d) that the time pattern of your slices is not spread out evenly across the TR interval. To do this in to3d, you use an input file for the 'tpattern' field after the -time:zt or -time:tz option on the command line. Or you could use the 3dTshift program to reinterpolate the dataset in time to a common time basis.

The analysis problem is harder. You probably want to do this by modeling the hemodynamic response using a basis function expansion. At present, this is only documented in one of our educational handouts -- I don't recall which number, but it is titled "Irregular Stimulus Timing" or something like that. You really have to understand the tools (waver+3dDeconvolve+3dcalc) to do this type of analysis at this time.

The basic idea would be to catenate all your data volumes into one long dataset. Then you would create a set of column vectors that represent the idealized response to the catenated stimuli at each actual time point; each column vector would be from one basis function; program waver would be used for this, with the -tstim option. The stimulus times would have to be adjusted to allow for the time offset between the catenated imaging runs. Then these column vectors would be used as the regressors to 3dDeconvolve.

That's not a very clear explanation, I'm afraid. Without thinking about it a little more and drawing some pictures, it's the best I can do at the moment. I hope it helps some. The analysis is do-able, and not complex in principle, but since the programs aren't really setup for this case, you have to use them cleverly to get what you want.

Buda Bob