Some models of the morphology-phonology interface take morphology and phonology to be computed in the same component of the grammar, simultaneously, e.g., “P»M” instantiations of Optimality Theory (McCarthy and Prince 1993a,b, Kager 1996, i.a.). Other models, e.g., Distributed Morphology (Halle and Marantz 1993, 1994), separate morphology from phonology, with operations that involve morphology preceding those that involve only phonology. I undertake a study of the order of operations needed to derive the form of finite verbs in the Neo-Aramaic language Turoyo (Jastrow 1993). Turoyo furnishes several arguments in favor of a separation of morphology from phonology, including phonologically-conditioned allomorphy that is anti-optimizing and surface opaque (reaffirming the robust earlier findings of Paster 2006) and infixation that in its surface position is invisible to allomorphy, but visible to phonological processes. More specifically, what Turoyo shows is that exponent choice fully precedes, and is oblivious to, phonological considerations.