The definition of a lexeme’s inflectional paradigm makes essential reference to that lexeme’s stem(s). Canonically, a single stem serves as the basis for every word form in a lexeme’s paradigm. Often, however, a lexeme’s paradigm is based on multiple stems. The difference in form between the stems is independent of their difference in distribution (§5.1). Formal variation among stems may be attributable to sandhi processes, may be an inherent property of a particular inflection class or may reflect a juxtaposition of independent inflection classes (§5.2). The distribution of stems within an inflectional paradigm may be variously conditioned (§5.3): phonologically conditioned distribution may be an effect of automatic sandhi or may be nonautomatic; grammatically conditioned distribution may be sensitive to morphosyntactic or morphomic properties of a lexeme’s word forms. I exemplify these possibilities with evidence from Sanskrit. I summarize the ways in which the different kinds of alternation in stem form cross-cut the different patterns of stem distribution (§5.4). I propose a precise way of modeling a stem’s formal and distributional properties by means of two operations (§5.5): a function Stem mapping a paradigm cell’s lexical and morphosyntactic content to a stem form; and an operation SC whose value for a particular stem varies according to its syntagmatic context. I summarize in §5.6.