Help in analysing pre/post stats in a repeated measures ANCOVA

Hi all,

I am currently trying to work out if it is possible to run a repeated measures ANCOVA in SPSS where the two levels of a factor are each tied to a seperate covariate.

To clarify I have a study with a 2x2 design of timing (pre/post) and valance (good/bad). In the study the participants have a pre test of face recognition for two different faces then experience a social interaction with the two faces (which is aimed to induce a valanced response to each) and then do the face recognition task again.

I've found several papers and text books that recommend analysing pre/post designs by co-varying out the pre scores from the post scores in an ANCOVA and keeping the intervention factors as IVs. This seems to work fine in SPSS for between subjects designs but my issue is that for a within subject design there's no way to specify that the pre-bad scores should be co-varied from the post-bad and the pre-good from the post-good.

One way I have considered trying to do this is to run two separate ANOCVAs whithout including any IVs (so just post as DV and pre as covariate) for each of the two levels of the valance factor and getting SPSS to save the predicted values and then running a repeated measures t-test on these two new variables to see if there was a significant difference between the two groups post manipulation one their pre manipulation scores have been taken into account.

I basically wanted to see whether this was an acceptable method of analysing the data, and if so whether I should save the unstandardised predicted variables or the standard errors. If people feel this is not a correct way to analyse the data I would appreciate any other suggestions for running a repeated measures ANOCVA when faced with a pre/post design.

Many Thanks,