> I ran the model you mentioned above using the trial-by-trial probability correct
> and a model derived trial-by-trial prediction error as modulating variables.
> ...
> However, can I use the output of this model to assess how the functional
> connectivity is changing as a function of how well the participants know the
> material? I wasn't sure how to look at that without binning.

No, I meant to suggest running a modulated context-dependent correlation analysis by creating a linear or/and a quadratic interaction term.

> For now I'll run 6 separate models in my PPI analysis. I still think I'm a little
> lost on how to run one big model.

In the one big model you simply include into the analysis all those 12 (linear) or 18 (quadratic) regressors from those 6 separate models.

HTH,
Gang