Hi Sarah,
You can use 3dttest++, 3dMEMA, or 3dMVM for such a 2x2 repeated-measures design plus a covariate. 3dMVM is probably the easiest approach, but if you want to try 3dMEMA or 3dttest++, you would have to run piecewise t-tests. For example, for A1B1-A1B2 at group level, you need to have effect estimate of A1B1-A1B2 from each subject plus its t-statistic. Similarly for other scenarios such as A1B1, A1B2, A1B2, A2B2, A2B1-A2B2, A1B1-A2B1, A1B2-A1B2, or (A1B1-A1B2)-(A2B1-A2B2).
Gang