Dear SPMlers,
consider the following event related fMRI experiment on semantic
priming.
A single trial, included in 2 1/2 TRs +/- 1/2 TR for jittering, consists
of the following sequence of events
Fixation cross: 750 msec
Prime word: 200 ms
Target word: Presented till subjects respond, mean time. 700 msec
Hash mark: Presented until the next fixation cross, Timing is
variable because jittering is included here.
There are four conditions:
A, B, C with n trials
D with 3xn trials
We are interested in the differences of brain activation when comparing
different semantic prime-target relations as defined by the conditions
A, B, C and D.
Our question is how best to modell the experiment. As we understand it
it is the best to modell all or most of the events that actually have
happened.
Event related (this was our first try):
We modell the onset of two events for every condition. That is: Fixation
cross and onset of prime (WE thought that it does not make much sense to
modell the target as a separate event because their onset is only 200
msecs apart). In our design matrix we can then define six event types:
A, B, C and D (defined by the onset of the prime), a event type Fix
which includes ALL onsets of the fixation cross in all conditions and a
event type FALSE which includes all trials with a false response defined
by its prime onset.
If we look at the results we see most activity in the event type FIX
including activity of language regions. In the event types A, B, C D
however we do not find very much activity in those regions. This results
is quite contrary to our expectations! So we suspect that we have done
something wrong, - perhaps of the close temporal succession between
fixation cross and the prime?
Now the question arises how to better modell our experiment. We have
thought of the following alternatives:
1. Do not modell the event FIX, i.e. only modell the onset of the prime
word in the four condition.
2. Modell epochs.
a) Modell the time from Onset of fixation cross till reaction as an
epoch.
As we understand it than we can look at two things.
If we define a contrast, say a 1 for Condition A, then this contrast
tells us what happens in the brain during condition A compared to what
is not defined as an epoch. Or we can define a contrast say a 1 for
condition A and a -1 for condition B. Then this contrast tells us what
is activated during a trial epoch in condition A more than in condition
B. Is this correct?
b) Modell the time from onset of prime word till reaction time because
we are not really interested in "fixation cross related activity".
(Definition of contrasts is as in a.) But can we do that?
We would appreciated some advice!
Best
--
Dr. Dr. Henrik Walter
Abteilung Psychiatrie III
Universitätsklinikum Ulm
Leimgrubenweg 12
89075 Ulm
fon: 0731/502 1489
fax: 0731/502 1549
e-mail: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|