The phonological cycle was proposed within rule-based phonology as a derivational theory of phonological phenomena sometimes referred to as "cyclic effects" (e.g., Chomsky & Halle 1968, Brame 1974). An alternative theory of these effects within parallel Optimality Theory (Prince & Smolensky 1993/2004) relies on transderivational constraints that directly compare the outputs of different derivations (Benua 1997 et seq.). In this paper, a new empirical difference between the two theories is identified: the transderivational theory but not the cycle can generate certain phonological interactions that are "counter-cyclic", where a process applying in a larger domain feeds or bleeds a process applying in a smaller domain. The unattestedness of such interactions in natural language provides new support for the cycle.