A problem with covariates

Hello everyone.

I have a problem with covariates. I have a database with a mixed design, with a manipulation as a within-subject variable. Each subject went through multiple trials both before and after the manipulation. On each trial a random stimulus was presented to each participant.

The problem is that the stimuli presented to the participants wasn't the same before and after the manipulation. That is - a random stimulus 'A' was presented before the manipulation and a different random stimulus 'B' was presented after the manipulation.

My question is - can I simply enter both stimuli into the analysis, even though each one corresponds only to one of the two measurement? The alternative I thought of was to treat the database as a between subject design and to unite the two covariates into one. But obivously, this is not the preferable option.

Does anyone have any idea as to how to treat the database?

Thanks a lot for any help!