Hi Jim,

Doing a deconvolution when the curves overlap so much is often
problematic.

But modelling only 6 seconds of the response is sure to create
curves that are contaminated. In some cases those 6 seconds will
be only the start of a response, in most it will contain that later
part of the previous response, too. And the later half of the
final curve in a trial will end up contaminating the first curve
of the next trial (since they seem close together, too).

If you have many such trials for each event type, you should come
out with a reasonable result. Otherwise, I expect your result to
be noisy.

In any case, it is important to model the entire response that you
expect. Otherwise your curve means different things at different
times. If you expect the full response to last 14 seconds, that
should be your stim_maxlag. That the resonse curves overlap is
by design: since the stimuli are so close together, the responses
will indeed overlap.

A larger jitter range might be more helpful of course, since as
you seem to be TR-locked, there are only 5 possibilities for the
ISI. But maybe that isn't an option...

- rick