Feedback, alignment and routinization in dialogue. Martin Pickering Dialogue is successful to the extent that interlocutors align relevant aspects of their mental representations. A range of evidence suggests that this occurs by a largely automatic process involving priming at many different linguistic levels (lexical, syntactic, etc.). So what effect does feedback have on the process of production during dialogue? Following Pickering and Garrod (in press), I propose that feedback does not affect priming, but is instead used for 3 basic purposes: to indicate that the addressee has achieved alignment, that the addressee has failed to understand, and that the addressee believes that the speaker's representation of the situation is wrong. These processes affect the speaker's subsequent contributions, so that feedback and priming both contribute to the ultimate achievement of alignment. I illustrate this process in terms of referential communication, and describe how feedback and alignment contribute to the development of novel expressions or expressions with novel interpretations, which may become lexicalized as routines.