Hi cephalopod -

> However, a more general
> question, does AFNI need the duration information?
> If I use a regular "GAM" -- will AFNI be able to
> infer how long my stimulus is, or will it assume
> that it is the length of the TR? Or, will it not
> matter?

If you use GAM with no additional parameters, it will receive default parameters for peak and dispersion. The documentation for 3dDeconvolve talks about what these default params are, and how they interact; I don't have much of an intuition for it so I'll avoid putting my foot in my mouth. However, the important thing to know is that AFNI is not estimating anything -- these are fixed params that often make sense. The more non-standard your approach, the more you'll have to experiment and see what happens.

> I looked at the x-matrix and it appeared
> that the GAM and BLOCK stim durations appear to be
> identical (ie., in width).

BLOCK is generally for block presentation (you will probably not be surprised to hear) but can also be used if you want to nail down exactly how long your presentation is for. As you've pointed out, for shortish stimulus intervals it often looks pretty close to GAM. In my own limited experience screwing around and seeing what happens, the choice of HRF seems to make surprisingly little difference on the results you get at the end of the day, provided you're not doing something really odd. As Bob is wont to say, there are a lot of features that are great if you know what you're doing. If you don't know what you're doing then doing the standard thing works pretty well.

> One last question (I hope!): In the past, I've
> been able to use stim times (in AFNI) with a colon
> that indicates on and off times, for example,
> 3.5:7 for a stimulus that starts at 3.5 seconds
> and ends at 7 seconds. I'm having difficulty do
> that now... is this still possible?

I believe the colon is for STIM_TIMES_AM1 and STIM_TIMES_AM2. Those options are still there, if you want them.

> I know there
> isn't a standard formalism for any of this, but
> what is the recommended function for dealing with
> basic event-related experiments? Is it GAM, or
> would you still suggest the BLOCK function with
> BLOCK(3,1)?

My sense of it, as above, is that for 'typical' durations people often just use GAM; and in my experience, it doesn't make that much difference. My experience is limited, but I have tried probably 4 different HRF schemes on the same data, and differences at the end were minimal. I think the AFNI ethos is to just play around and see what happens -- it will pay off with greater understanding later.