Gerome,
> If I am interested in the differences between conditions within each task then should The Y-value to used in RBA be (TAc1- TAc2) and (TBc1 - TBc2) ?
> Similarly, if I am interested in between task-comparisons for each condition should the Y-value used in the RBA be (TAc1 - TBc1) and (TAc2 - TBc2) ?
The answer to both questions is YES: you perform four separate analyses, one for each contrast.
> should each contrast be modeled separately at the group level, and if so - is there a need for a correction?
Even under the conventional AVOVA framework, there is no easy solution for controlling of FWE in this context. As long as all those four comparisons are planned analyses, I would not worry about it.
Gang