> Should I center at 0 instead of mean-centering at time1 and time2? If you could explain this a little more I would appreciate it.

The crucial question is: if the behavior measure differs substantially between the two time points, does the difference reflect some fundamental dissimilarity between the two time points, or is it simply caused by uneven sampling? If the answer is the latter, you can center the behavior data around the overall mean. If your answer is the former, and if you are interested in comparing the two time points, it makes sense to center the behavior data around the mean for time point separately before you feed the behavior data into the 3dLME script. And set the center to 0 in the script. The reason is that you 1) want to control the behavior variability within each time point (not across time points), and 2) at the same time maintain the integrity of the difference between the two time points. If you center the behavior data around the overall mean, the difference between the two time points would be interpreted as fixing the behavior measure at the *overall* mean, which would contradict to the reality (the two time points differ in average behavior value).

Gang