Optimal fmri design for RSA

Dear all

I’m in the planning stage of a new experiment where I would like to apply representational similarity analysis.
I plan to implement 2 sensory stimuli with 6 variations leading to a total of 12 different conditions. Should I apply a block design or event-related (fast/slow) for RSA. Does someone has experience with design for RSA and could give me some hints ?

many thanks!

Hi Mike,
in general for RSA it’s better to use event-related designs to minimize block effects. I recommend reading this paper by Kriegeskorte, which contains considerations on optimal parameters and designs (ISI, TR, etc.): Representational similarity analysis – connecting the branches of systems neuroscience.

It seems you have a relatively small set of conditions, so you might also consider looking at Type1-Index1 sequences for the trial order, which should be better compared to pseudo-random presentation: https://cfn.upenn.edu/aguirre/wiki/public:t1i1_sequences


1 Like

thank you very much. this is helpful.

@mvdoc Feel free to ping me if you need a hand with the counterbalanced sequences (i.e., t1i1).


1 Like

Not sure if you still are looking for input, but this paper might also be helpful with respect to trial timing: https://cpb-us-e1.wpmucdn.com/blogs.uoregon.edu/dist/1/8955/files/2017/06/TrialTiming-q6ozfd.pdf