Morphoprosodic Alignment (MPA) is a nontemplatic model of reduplication designed to account for languages with multiple reduplicative subpatterns. The premise of MPA is that reduplicative morphemes can be stem-internal or stem-external and that this distinction is visible to the phonological component through general constraints on the association of stem-internal and stem-external morphemes to prosodic categories. I illustrate the model with Moronene, Klamath, and Gooniyandi, each of which has several reduplicative morphemes. MPA meets the challenge for an optimality-theoretic model to account for such systems without resorting to morpheme-specific indexed constraints or cophonological constraint hierarchies.One of the most fruitful aspects of the study of reduplication has been the analysis of phonological alternation in reduplicative forms-that is, systems in which the reduplicative substring may alternate in size or segmental content as a function of the phonological form of the root. Where alternation in reduplicative systems is strictly phonological, all reduplicative substrings are subject to the same set of phonological generalizations.Alternation in reduplication can also be morphological: the reduplicative morpheme may alternate in size or segmental content as a reflex of something other than the phonological form of the root. Such morphological alternation displays two types of form-function mapping: isomorphic, in which multiple reduplicative morphemes map to distinct sets of phonological generalizations, and allomorphic, in which a single reduplicative morpheme maps to multiple phonological generalizations, and in which the choice of generalization is determined by the root. For example, Hawaiian has three sets of generalizations over reduplicative substrings: a syllable-sized prefix, a foot-sized prefix, and a foot-sized suffix. There is no systematic functional contrast among these formational patterns, and the choice of affix is root-controlled.Both isomorphic and allomorphic alternation pose a challenge for modeling reduplicative systems. A second challenge is one of overgeneration: languages that have multiple reduplicative generalizations nevertheless respect an upper limit on the number of phonological patterns available at either side of a root. Despite the range of prosodic categories plausible as targets of reduplication, languages appear not to use more than two at either side of a word, except where fixed segments or multiple exponence is used.