Hi,
If your regressors vary across trials, you could convert an epoched dataset to images do a GLM across trials in each subject and then take beta images to the second level. Both within- and between- subject steps are done using the 'Second level' button in SPM. There is an example of an analysis across trials in section 42.5 of the manual. Another way to do the same thing is with the 'Contrast' function where you could use the pseudoinverse of your design matrix as a contrast across trials for an epoched dataset and then treat the resulting dataset as you would an average with multiple conditions.
Best,
Vladimir