Stacia,
The problem is the way you run the paired t test. For this 1st order contrast, you are collapsing the other three factors: Distractor Salience (No,
Low, High), Visual Field (LVF, RVF), Task Difficulty. Before you can run such a paired t test with 3dttest, you need to run 3dMean for each of those Switch input files for each subject. Do the same for NoSwitch files. So in the end you should only have 26 input files for 3dttest. This is a little tedious, but that is what it should be.
Well you don't really need to run such a test as it is supposedly equivalent to the result from GroupAna. You can verify if this is the case.
Gang