|Abstract:||This paper investigates Marantz' Generalization in the context of Optimality Theory (OT). The locality generalization implicit in Marantz' Generalization, that reduplicants tend to be adjacent to their corresponding base segments, is drawn out and focused on. A family of AdjacencyBR constraints are proposed to motivate this tendency. These constraints relate a reduplicant's placement to its anchoring, motivated by the idea that each segment in the reduplicant wants to be as close as possible to its correspondent in the base. It is shown how these constraints motivate less than full reduplication, exemplified by Dobel, Agta and Yidin, as they are at odds with MaxBR, which requires full reduplication. They are also shown to predict discontiguous reduplicants (a rare pattern, found in Marshallese and Mandarin) which are otherwise mysterious. Claims that the locality generalization is an absolute rather than a tendency are examined, and reduplication patterns from Chukchee, Indonesian, Madurese and others are analyzed. It is shown that a full typology of reduplication must be made available by the grammar, including 'opposite edge' reduplication, where the reduplicant is anchored to one edge but aligned to the opposite edge as patterns of this sort cannot be otherwise explained.