> I do have missing data, it's a 3x3 design with high, medium and low arousal
> pictures of positive, negative and neutral valence, however we have no low
> arousal negative condition.
Liz, if I understand you correctly, you don't have low arousal negative condition across ALL subjects. If that's the case, it is not really a 3X3 design then, and you can't run such a model with 3dLME. If your goal is to run those contrasts, paired t-test with 3dttest might be good enough for that purpose, assuming you don't other missing cells for some subjects.
Gang