Both automated priming (Pickering & Garrod, 2004) and partner-specific adaptation (Brennan & Hanna, 2009) have been proposed to underlie lexical entrainment (the repetition of words across interlocutors). Since activation levels of infrequently used words are relatively low, the effect of automated priming is predicted to be weaker in L2- than in L1-conversations, leaving more room for deliberate partner-specificity (Costa, Pickering, & Sorace, 2008). We tested this prediction by means of a production experiment, in which we varied whether participants interacted in their L1 or L2, and whether they addressed the confederate who had introduced a certain reference or another addressee. We found that in their L2, participants repeated references more frequently when addressing the person who had introduced the reference. Yet we did not find this effect of partner-specificity in the L1 conditions. Therefore, our results support the proposed combination of the two accounts.