Abstract During dialogue, interlocutors come to use the same words for referents, a phenomenon termed lexical alignment. Pickering and Garrod’s (2004) Interactive Activation model proposes that automatic priming mechanisms operate at each level of representation (e.g., conceptual, lexical, phonological) and percolation between the levels enhances alignment at the lexical level. However, from previous research it is unclear whether lexical alignment is wholly driven by alignment at the conceptual level, or whether it is partly driven by the repetition of word form. Using non-cognate translation equivalents (i.e., words that are highly similar in meaning, but do not share the same, or similar, word form) this study investigated this issue in a bilingual population. The results show that within-language lexical alignment is greater than between-language alignment. Such results suggest lexical alignment is partly driven by the repetition of word form. If alignment were wholly based on conceptual alignment, the alignment effect would have been of similar magnitude in both within- and between-languages. As this study involved the use of bilingual participants, it also offered an investigation into how alignment operates in a second language and whether it occurs cross-linguistically. Proficient bilinguals were found to align to the same extent in their dominant and non-dominant languages, which suggests that alignment mechanisms can be extended to a second language. Cross-linguistic alignment effects were also obtained. The results are discussed in relation to the implications for the Interactive Activation model, and the possible extension of this model to account for bilingual dialogue.