Dear Stephan,
> We would like to compare two different word association tasks which a
> theory predicts should activate quite different regions. The problem is
> that the two tasks have different time-courses: one (A) is short --
> give a cue, subject prepares to respond for 1.5 seconds, then silently
> generates responses for about 3 seconds and then is finished with the
> trial, vs. task (B) -- after the cue, subject needs about 13 seconds to
> prepare to give answers and generates responses for about 8 seconds.
>
> It seems that an event-related design would be called for here, with
> model elements for the cue, preparation phase, and response phase.
> However, because the duration of each of these phases is different
> between the A and B tasks (except the cue phase, which is constant),
> it's unclear how best to contrast the two task types. For example, if
> we equate the number of trials of each task type, then we end up with
> many more scans corresponding to B than to A, since B is longer.
> Perhaps someone has encountered a situation like this before and/or has
> some pointers?
Having unequal number of scans during each trial type is no problem.
The more important issue, in comparing the two trials ,is that more
neuronal activity is likely to be induced by the longer trials. One
way to deal with this is to model the trial components with box cars
that have different lengths for the different trial types. This has
been used effectively to estimate delay-period activity in motor
paradigms.
In this instance the same level of activity per unit time should give
you the same parameter estimates and faciltiate some sort of comparison
between the two trial types. However, one should note that the best
way to render different evoked responses comparable is by designing
more appropriate control tasks (as opposed to post hoc modeling of
incidental differences one is not interested in).
I hope this helps - karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|