Dear Edward, > As this is my first use of spm99b, I have some basic questions > regarding model specification. Our experiment is 210sec in total, > consisting of 2 stimulus presentation blocks, each lasting > approximately 69sec. In summary the experiment consists of > rest(24sec),stim(69sec),rest(24sec),stim(69sec),rest(24sec). Since the > exact timing is variable depending on reaction time, the scans per > epoch are variable between patients, and never quite follow the above > model explicitly. How do I account for this in specifying the fMRI > model for this experiment? Each session is modelled separately in SPM99. Just enter the session-specific design parameters. In SPM99b the epoch lengths are assumed to be equal. The simplest way around this constraint, in your instance, is to treat the two periods of stimulation as two trials with different epoch lengths. You can 'recombine' them later using appropriate contrasts. SPM99 proper will allow one to specify trains of event-related responses to model variable-length epochs (but the analysis should proceed as an event-related one). > If rest periods are implicitly modeled, then do I just have 2 > sessions to specify, or 16 if I have 8 subjects? You have 8 sessions (one for each subject). Each session will have two trial types (first stim period and second period). The rest condition will be modelled implicitly and does not need to be specified (otherwise you would have three trial types). I hope this helps - Karl