Daniel -
In general, this is a difficult question to answer, since it
clearly depends on the size and consistency of the effect
(ie Stroop incongruency) that you're trying to detect. Thus
there is no good answer, except "more is better" (though see
qualification below). We have found significant visual-motor
event-related effects (including anterior cingulate for example)
with as few as 20-30 events.
My best guess as to why you're seeing strong main effects versus
baseline, but few differential effects, is the long SOA you're using.
Several people have shown recently that short SOA's are far more
sensitive, particularly for differential effects (Burock et al,
Zarahn et al, Friston et al, Josephs & Henson). Basically, shorter
SOAs (with random trial ordering) both increase your experimental
energy and move it into higher frequencies. For more details,
see some of the above papers.
Finally, I am assuming you were using a Fixed Effects analysis
(entering all scans into one big design matrix). Note that, if
you performed a Mixed Effects analysis (ie treating subjects
as a random variable, with only one parameter estimate per condition
per subject), it is the variability across subjects that tends
to dominate, rather than the variability within subjects
(ie between events/scans), and so there is less value in
increasing the number of events, and more value in increasing
the number of subjects.
Rik
> I have been running an event-related fMRI study in which a single
> stimulus is presented once every 16 seconds. In this study, there are 2
> factors, each with two levels. Thus, there are 4 conditions, with 40
> trials per condition. The study involves a Stroop analog and the main
> comparison is between congruent and incongruent Stroop-like stimuli.
>
> In looking at the data from 10 subjects, I haven't been seeing
> much activation for this main effect (for which there are 80 trials per
> condition). However, relative to baseline, individual trial types (e.g.,
> consistent and inconsistent trials) are producing lots of activation.
> But, the activation appears to be similar for consistent and inconsistent
> trials.
>
> My question is: are 40 trials per condition (80 when computing
> main effect of consistency) generally too few to see significant results
> in 10 subjects?
>
> Thanks,
> :> Daniel Weissman
---------------------------8-{)}-------------------------
DR R HENSON EMAIL [log in to unmask]
Wellcome Department of
Cognitive Neurology TEL (work1) +44 207 833 7483
12 Queen Square TEL (work2) +44 207 833 7472
London, WC1N 3BG FAX +44 207 813 1420
URL: http://www.psychol.ucl.ac.uk/rik.henson/index.html
---------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|