Also see this paper on why that design is problematic: Zarahn E, Aguirre G, D’Esposito M. 1997. A trial-based experimental design for fMRI. NeuroImage. 6:122--138. Cheers, Michael On Fri, Mar 25, 2011 at 12:18 PM, Chris Watson < [log in to unmask]> wrote: > Are you saying that there is a trial every 300-500ms? If so, then yes, the > BOLD response will probably not be linear. However, it wouldn't necessarily > kill all the signal, which may be why the results you got look reasonable. > If there are, say 2 trial types (gamble vs. sure), then perhaps the > distribution of trials creates enough of a jitter. > > Andrew Jahn wrote: > >> Hi SPMers, >> >> I ran a subject through an experiment combining eyetracking and fMRI >> scanning simultaneously. I calculated where they were looking at, and >> depending on where the eye fixation fell, I coded that as a trial (for >> example, focusing on choosing a gamble or focusing on getting a sure thing). >> I included the duration of the fixation in the regressor and convolved each >> with the HRF. >> >> Most of these fixations are only a fraction of a second long, usually >> lasting ~300-500ms. My question is, with so many trials so close together, >> does this violate the assumption of linearity in the BOLD response as >> outlined by Dale & Buckner (1997)? My design matrix looks fine, and there >> are no instances of singularity. Furthermore, the results that I have >> gotten appear to be reasonable. I would appreciate any feedback. >> >> >> Thanks! >> >> -Andrew >> > -- Research Associate Gazzaley Lab Department of Neurology University of California, San Francisco