AbstractEvidence that languages differ in the initial, proximate units of phonological encoding presents a challenge to current theories of language production. The proximate units principle is the idea that variation in the proximate, first selectable, phonological units during word form retrieval for speaking is key to understanding the implementation of phonological encoding in diverse languages. It proposes that rather than imposing a unilingual phonemic template across languages, the variety of proximate units should be incorporated in theories. From this perspective, cross‐linguistic differences in speech error patterns, in phonological preparation, in automatic activation, in multi‐word coordination, and in metalinguistic awareness are more clearly interpretable. The proximate units perspective aims to accelerate scientific progress by incorporating linguistic diversity at the lexical‐phonological interface into theories of word and sentence production.