Eric,
Given that you have at least 3 conditions and probably an implicit baseline, you should use the gPPI toolbox. gPPI has been shown to be more accurate than the traditional PPI for experiments with more than 1 condition (plus baseline).
The difference is that the PPI regressor is split by condition. If you have 3 tasks, you will have 3 PPI regressors. The concept behind this is that if you would use 3 task regressors to explore the task effects, then you should use 3 regressors to explore the PPI effects. Attempts at using a single regressor assume a parametric structure to the connectivity relative to baseline.
In my opinion, studies of multiple conditions using PPI might be reporting false positives and might have also false negatives in them.
The two papers showing the difference between gPPI and PPI are:
Let me know if you have any questions.