Three sets of assumptions about elite and constituent relations prevail in approaches to conflict termination. The 'Interest-based', 'Identity-based', and 'Timing' approaches offer important insights into conflict termination processes but differ in how they conceive of the locus and direction of domestic political power during conflict termination. Despite such important differences, these contending approaches have not been tested comparatively. In this paper, I introduce the Dynamic Group approach and test it against these current approaches. The Dynamic Group approach differs from these others in that it is grounded in the two-level perspective on negotiation. The Dynamic Group approach treats the domestic locus and direction of power as variable (1) across cases and (2) across conflict stages within a given case. Accommodating such variation introduces the concept of domestic ratification as an important determinant of the success or failure of conflict settlement attempts. The Dynamic Group approach offers greater generalizability than the current approaches. Specifically, I demonstrate (1) that whereas these three contending approaches apply only to cases in which elite's and constituent's interests are consonant, the Dynamic Group approach covers cases of elite and constituent interest dissonance, and (2) that whereas the three approaches apply to no more than two stages of conflict termination, the Dynamic Group approach comprehends differential effects of four distinct stages. I demonstrate that the Dynamic Group approach best accounts for conflict-stage effects in the 1973 and 1985 attempts to settle the Northern Ireland conflict.