For the post hoc tests in 3dMVM,

group1_Scan1_vs_Scan2
group2_Scan1_vs_Scan2

the reason 3dMVM gives different results is that it uses the residuals from both groups, leading a higher number of degrees of freedom than the simple paired t-tests.

> I reran the same dataset using LME, I don't have the discrepancy anymore. The lme post-hoc results in R
> match my 3dLME results and mixed model results from SPSS (compound symmetry). Why the difference?

Not sure what difference you're referring to here.

> With 3dMVM, my type III mainF results for 3dMVM match my Type I 3dLME mainF results. If I run the same
> analysis in SPSS, my Type III mainF SPSS results match my 3dLME mainF results. Why the difference? Are
> Type III's different for 3dMVM and 3dLME?

It should be the same. I cannot comment further without knowing the exact models you've run.

Gang