The Old Spanish verb commonly displayed an alternating stem segment that was later leveled in favor of one of the alternants. In some instances, the direction of leveling is toward the one with the most types (cf. venço reforming on the pattern of vençes, vençe, vençer, etc.), whereas in others it is the one displaying a unique phonological alternation (cf. fugio influencing foges, foge, fogir, etc.). This paper aims to account for the two types of leveling attested in Old Spanish in terms of their phonological, rather than semantic, markedness. The framework is McCarthy's Optimal Paradigms (OP) model. In this model, which builds within Correspondence Theory, an entire verb paradigm is evaluated for satisfaction of constraints on structure and faithfulness, and violations are incurred by the paradigm as a whole. It is shown that the OP model illuminates several issues of markedness in leveling, including identification of the attractor, Over- and Underapplication of processes, and competition between output forms. Shortcomings and limitations of the OP model are also considered.