Hi Ioana,

Yes, each married file must have the exact same structure.

So you put AR information into the AI timing file? I am
not quite positive with what you are doing, but the 4
cases should partition all of the A/N events.

Actually, at the time of A and N, the type of instruction
is not known, while at instruction time, the preceding
scenario is known (assuming they pair like that). So it
seems to me you have 6 regressor classes listed:

scenario A
scenario N
instruction AI
instruction AR
instruction NI
instruction NR

I do not otherwise see why you would expect a difference
in the BOLD signal between AI and AR (assuming they do
not incorporate the instruction phases).

- rick