Yue,
It seems that you have a dataset with a 2 x 2 within-subject (or repeated-measures) ANOVA structure. There are multiple ways to build a model, especially when there is no missing data.
> I am mostly interested in the interaction effect between factor a & b.
1) If you are ONLY interested in the interaction effect between the two factor, you can obtain the interaction effect for each subject, and then perform a one-sample t-test with 3dttest++.
2) Perform a 2 x 2 ANOVA with 3dANOVA3 -type 4 or with 3dMVM
3) If you want to use 3dLMEr, make sure you have the most recent version of AFNI. And try
3dLMEr ... -model "MOD+WORD+(1|Subj)" \
(there is no option -ICCb in 3dLMEr)
Gang