repetition suppression study question

[beheryd]

Hello! Thanks for making this great package. I'm fairly new to the world of fMRI, and I'm starting an fMRI study that'll be using RS to make predictions.

For context, I have a block experiment in which I'd show participants pairs of videos that'll differ by particular parameters (I don't want to get too much into it). The idea is that we should expect RS when a given pair matches in those parameters within a particular ROI.

The trial design is as follows.

• ITI (jitterd ~4 sec)
• movie1 - 2 secs
• ISI (?? unsure whether to jitter ~2 secs)
• movie2 - 2 secs
• questions response - 2 secs

The thing is, I'm unsure whether I'd be able to extract video-specific responses with GLMsingle, considering that in the RS literature, people tend NOT to jitter their ISI. And by nature, RS requires a short ISI.

One possible nuance is that the short ISI between videos within a trial means that the second video will be contaminated with the hemodynamic response to the first video. So it might make sense to only analyze the first videos, but one could explore different versions (only first videos, only second videos, or both) to ascertain the robustness of the results.

With all that being said, to make use of this package to extract video-specific responses (there being 2 expected within a trial), should I jitter my ISI contrary to typical RS literature? If I were to jitter my ISI, what should my range be? Perhaps I'm overcomplicating this haha.

[kendrickkay]

Hi. I think I get the gist of what you are contemplating. In general, things may be tricky since you (as a matter of experimental goals) are deliberately placing videos next to each other in time. Here are some general thoughts:

"The thing is, I'm unsure whether I'd be able to extract video-specific responses with GLMsingle, considering that in the RS literature, people tend NOT to jitter their ISI. And by nature, RS requires a short ISI." ==> In general, having jittered ISIs should not be seen as a necessary requirement for GLMsingle-type estimation. (Having jittered ISIs does impact other issues, specifically, the quality of estimation of the baseline signal level.) For example, in NSD, there was a fixed 3-s ON / 1-s OFF design, with occasional blank trials. Thus, from this point of view, ISIs were more or less constant at 4 s (with an occasional long ISI of 8 s).

Certainly, in general, you should expect to be able to estimate responses for the two videos that come in each pair. That being said, you are right that there will be response overlap between those two videos, and so the estimation noise that will inevitably accompany each video may have specific structure (specifically, given that the single trial predictors for two quickly presented vidoes are positively correlated, we expect negative correlation between the errors associated with the betas of those predictors).