Dear SPMers,
I was wondering whether you might be able to advise me on how to
correctly model a rather novel event-related fMRI paradigm. There are seven
randomly intermixed trial types (the order is first order counterbalanced to
allow selective averaging), and each event lasts 3 seconds. Some events
contain just a single stimulus that appears at the onset of the 3 second
trial period. Others, compound events, contain two stimuli, one that onsets
at the beginning of the 3 second trial period and another that onsets 1.5
seconds into the trial.
Some of the compound events share the same first stimulus and differ
only in that the second stimulus is incongruent in some cases and congruent
in others (this is a study of Stroo-like interference effects). I was
expecting to see lots of prefrontal, cingulate, parietal, and extrastriate
activation when I compared incongruent and congruent compound events (i.e.,
areas typically activated by Stroop-like interference). However, I see only
the barest nonsignificant hints of such activation when analyzing the data
with SPM, which makes me think that I'm incorrectly modeling the compund
events or that there truly is no difference in activation between them.
So, let me tell you how I've already tried to model the trial types in
this paradigm. First, I've tried modelling all 7 event types as discrete
hrf events with temporal and dispersion derivatives. Second, I've tried
modelling all 7 events as 3-second blocks (i.e., I convolved an hrf with a
half-sine and used a temporal derivative). While contrasts between
non-compound events look pretty good with both approaches, contrasts between
compound events (i.e., incongruent vs congruent) show relatively little
activation.
To investigate whether there truly are differences between incongruent
and congruent compound events, I've also analyzed my data with ANOVA.
Basically, I used selective averaging to compute the average time course of
activation in every voxel, for each trial type, and for each subject. Then,
I contrasted the timecourses of incongruent and congruent compound events by
assessing the interaction between trial type and TR in a random effects
repeated measures ANOVA. As expected, I now see highly significant
prefrontal, anterior cingulate, extrastriate, and parietal activation which
is greater for incongruent than for congruent compound event types. When I
plot the time courses of the compound events, they look pretty canonical
(just slightly wider and without an undershoot at the end). So, it looks as
if the effects I'm investigating are present, but that I might not be
modelling them correctly in SPM (or, perhaps, that SPM isn't particularly
sensitive to compound events for some reason?).
I would greatly appreciate any suggestions for how I might better model
these compound events in SPM. Thanks in advance!
:> Daniel
Daniel Weissman, PhD
Center for Cognitive Neuroscience
Duke University
Durham, NC 27705
phone: (919)-681-1029
fax: (919)-681-0815
e-mail: [log in to unmask]
|