Dear Chris,
thanks for your input.
> I think if you properly design your experiment, ensuring a range of
> SOA's, you can still estimate the contribution of each condition
> (assuming linearity). See e.g. Miezin et al. 2000 NeuroImage, Fig. 1.
Well, but that's exactly the point: Due to the short SOAs/ITIs, we
can't assume linearity. Therefore, the analyses in Fig.1 of Miezin (who
demonstrated nonlinearities as well) doesn't hold.
> But, for SOAs below 2000ms linearity might not hold.
> Maybe you could do something like in Ollinger et al 2001
NeuroImage...
> they separated components of trials.
Thanks for these suggestions, but presently I can't see how their
method can help. They basically show that they can disentangle two
temporally overlapping responses, but that doesn't tackle the problem
that there may be two confounded effects, i.e. changes in cogntive
demands and physiological non-linearities. In their case, there were
only the non-linearities, but no postulated changes in cognitive demands
(they investigated V1).
> If you're interested in varying the SOA and looking at its effect on
> activity, perhaps you could do some kind of parametric modulation?
> Or maybe you can set up a factorial design, 2 x 2 (task A, task B;
short
> SOA, long SOA)...
All this doesn't "de-confound" the two factors. The only thing I
thought of inbetween was to use indeed a parametric manipulation
spanning a wider time range, e.g. 0 - 6 seconds. The idea is that
demands on cognitive processes (dual-task coordination) should start to
increase only with SOAs < 2s, while non-linear effect may already be
present with SOAs around 5-6s. But I'm not sure whether two slightly
different timecourses of the effects are sufficient to disentangle them.
An additional problem is that we don't know for sure the *exact* change
in cognitive demands with decreasing SOA, so we can't set up a perfect
regressor modelling them.
best wishes,
Andre
>
> Andre Szameitat wrote:
>> Dear SPMers,
>> I've sent this question before but didn't receive an answer. However
as
>> this was directly before the Easter holidays so that not all may
have
>> read it and this question is quite important for me, I'd like to
post it
>> again:
>>
>> I have a question about a very specific situation, in which
non-linear
>> effects may occur (due to the refractoriness of the hemodynamic
>> response, i.e. the second of two successive identical stimuli evokes
a
>> smaller and delayed BOLD response). However, in my view they cannot
be
>> corrected - but I would be happy if you prove me wrong :-)
>>
>> The situation is a dual-task situation: Suppose you have two tasks
A
>> and B, which are performed with varying temporal distance (stimulus
>> onset asynchrony, SOA). In cognitive psychology, it has been argued
that
>> the shorter the SOA, the higher the demands on executive functions,
e.g.
>> related to the coordination of overlapping task processing. Thus, as
a
>> first effect, in a dual-task related brain area X we would expect
>> increasing levels of activity with decreasing SOA.
>>
>> However, it is not uncommon that already the individual performance
of
>> each task activates such a dual-task related brain area as well.
Thus,
>> suppose task A performed alone activates area X, and task B
performed
>> alone activates area X as well. When we decrease the SOA to create
a
>> dual-task situation, then there should be non-linearities due to
the
>> hemodynamic refractoriness. This is particularly so because the
time
>> range of hemodynamic refractoriness and SOA manipulations in
dual-task
>> paradigms is identical (usuallly SOA varies between 0 and 1500/2000
ms).
>> Thus, as a second effect, in a dual-task related brain area X we
would
>> expect decreasing levels of activity with decreasing SOA.
>>
>> Taken together, it appears to me that there are two perfectly
>> confounded effects: Psychological processes predict an increase of
>> activity with decreasing SOA, while non-linear BOLD effects predict
a
>> decrease of activity with decreasing SOA. Stated more generally,
both
>> effects result in a parallel (negatively correlated) modulation of
the
>> BOLD response. Therefore, it seems to me that such non-linear BOLD
>> effects cannot be corrected. For instance, an additional regressor
like
>> the one added if Volterra kernels in SPM are used should be fully
>> colinear with a regressor encoding the SOA (the psychological
>> manipulation).
>>
>> It would be great if I am missing something and there would be a
method
>> that these two effects can be disentangled. Thus, any input is
greatly
>> appreciated.
>>
>> Kind Regards,
>> Andre
>>
--
______________________________
Dr. Andre J. Szameitat
Department Psychologie
Neuro-Cognitive Psychology
Ludwig-Maximilians Universität
Leopoldstrasse 13
80802 München, Germany
Tel. +49-(0)89-2180 6778
Fax. +49-(0)89-2180 4866
www.psy.uni-muenchen.de/ncp
Office: Martiusstr. 4, Room 6
______________________________
|