Dear SPMers,
I would really appreciate some help regarding the analysis of an fMRI experiment I am currently piloting. The design is quite unconventional, and though I have posted a message about it before, we have now changed the design a little, and I have come up with a few more issues I'm uncertain about.
Our paradigm involves two different semantic verbal fluency tasks (task A and task B) with self-initiated responses. There are about 6 sessions for each of the tasks, and in each session there are a few blocks (in each block subjects retrieve words from a different semantic category). For half of the subjects, we first give all sessions of task A and then of task B and for the other half we give all sessions of task B and then of task A.
According to our pilot data, the rate of responses is quick at the beginning of the each block (around 3-4 responses in the first 2-4 seconds) and then the response rate decreases (to a response every 3-7 seconds). I am interested in the following comparisons:
1. All of task A compared to resting baseline (and all of task B compared to baseline)
2. The first 2-4 responses of the blocks compared to a few responses from the middle/end of the blocks
3. The first 2-4 responses of blocks of task A with the first few responses of blocks of task B
I plan to analyze the data using the variable duration epoch method (Grinband et al., Neuroimage), where the duration of each epoch is the time between two consecutive self-initiated responses. The baseline trials cannot be added within the experimental blocks because we assume subjects are actively performing the task throughout the block.
My questions are:
1. I understood that I have to make sure that the duration of the baseline condition is as long as the duration of the experimental condition(s). So, for comparison #1, if each block of the task lasts 30 seconds, should I give a 30-second baseline following each experimental block?
2. For comparisons #2 and #3, I compare two conditions whose durations are shorter than the entire block (for each block, the total duration of each condition—which includes a few trials— is 2-7 seconds). Does this mean that for these comparisons I should use shorter baseline epochs? If so, I can think of two ways to do so: One, to model only a few seconds of each baseline block (say the first few seconds) and not the entire 30 seconds. This way, the duration of the baseline is, on average, as long as the duration of the experimental conditions. The second way of getting short blocks of baseline is to introduce another part to the experiment where I will present short blocks of some random cognitive task interleaved with short blocks of the baseline task.
Which (if any) of these two options makes more sense?
3. I think that the very first response of each block might involve a different process than the second-fourth responses. Does it make sense to disregard the first epoch from the analysis for comparisons #2 and #3?
I would appreciate, of-course, any other input regarding this non-trivial design (in case I'm missing anything important…).
Thanks a lot!
Talya
|