The final-state interaction in multichannel decay processes is systematically studied with application to B decay in mind. Since the final-state interaction is intrinsically interwoven with the decay interaction in this case, no simple phase theorem like "Watson's theorem" holds for experimentally observed final states. We first examine in detail the two-channel problem as a toy-model to clarify the issues and to remedy common mistakes made in earlier literature. Realistic multichannel problems are too challenging for quantitative analysis. To cope with mathematical complexity, we introduce a method of approximation that is applicable to the case where one prominent inelastic channel dominates over all others. We illustrate this approximation method in the amplitude of the decay B to pi K fed by the intermediate states of a charmed meson pair. Even with our approximation we need more accurate information of strong interactions than we have now. Nonetheless we are able to obtain some insight in the issue and draw useful conclusions on general features on the strong phases.