The amount of material copied varies between a single C (mavit / ma-m-vit ‘lion(s)’) and CV (hodai / ho-ho-dai ‘rock(s)’). The former is preferred unless copying a single C would give rise to an illicit coda or cluster, in which case CV is copied. In contrast to previous analyses of similar patterns in Tohono O’odham and Lushootseed, I analyze the reduplicant as an infix rather than a prefix. The infixation of the reduplicant can be generated via constraints requiring the first vowel of the stem to correspond to the first vowel of the word. Furthermore, the preference for copying the initial consonant of the word can be generated by extending positional faithfulness to the base-reduplicant relationship. I argue that the infixation analysis is superior on two grounds. First, it reduces the C vs. CV variation to an instance of reduplicant size conditioned by phonotactics. Second, unlike the prefixation analyses, which must introduce a new notion of faithfulness to allow syncope in the base just in the context of reduplication (e.g. “existential faithfulness ” (Struijke 2000a)), the infixation analysis uses only independently necessary constraints of Correspondence Theory.