While a great deal of research on reduplication has focused on deriving shape invariance or segmental identity, as yet no study has investigated whether there is a correlation between reduplicative size and segmentism. This paper fills this gap and presents evidence that there is a correlation between size and segmental content, which standard theories cannot account for. In languages with multiple reduplicative morphemes, no language was found in which the smaller reduplicant had more marked structure than the larger reduplicant. Based on proposals by McCarthy and Prince (1994a, 1999), a model is developed which precisely captures this pattern. The central assumption is that reduplicative morphemes can be specified as root
or affix. The larger size and more marked segments found in root reduplicants parallels findings in prespecified morphemes. A detailed analysis of Lushootseed reduplication illustrates the predictions of the model.