Mixed models with SPSS

Hi everyone, soooo, anyone who can confirm that what I did makes sense...
I have N subjects who did some multisensory task, in which they saw (visual stimulus X 2) and were touched (touch X 2) concurrently. It's a very easy 2x2 design, in which the same condition was presented N times, but a different N times for each subject.
So, because I have a different number of trials for each subject, I thought that maybe mixed models may be appropriate. Comes out that the result is quite different from an initial repeated measure ANOVA I run, so was wondering what I am doing wrong...
I assigned the two factors as fixed effects (vision and touch) and the subject as random factor. Oh, and dependent variable is some physiological data. How is it that suddenly my Ps look crap? is it possible that the different number of trials made such a huge difference?