The aim of the present study is to examine the interactional functions of code-switching in plurilingual conversation, with a particular focus on those related to the elaboration of the social identities of the interlocutors, i. e. their identity work. The adopted analytical model was initiated by Auer, and further developed by Gafaranga, who combined it with the Membership Categorization Analysis elaborated by Sacks. The study is based on a corpus consisting of the audio recorded dinner talk of an Italian-Swedish family whose members use Swedish, Italian and a Ligurian dialect, and focuses on the sequential progression of the talk in interaction, thus based on an approach proper of conversation analysis. The participants create and display their different roles (identities), which are related to both group membership and individuality, a subsequent distinction made by Fant. Code-switching is used as a communicative strategy among the speakers and is used to make relevant identities, as well as a tool for the management of the interaction. The analysis is divided in two distinctive parts: one related to the code-switching connected to the organisational management of the conversation, the other connected to the identity work of the interlocutors. The linguistic codes used within the group are either inclusive (Swedish, Italian) or selective (Ligurian) and the analysis indicates a general inclusion rather than exclusion of interlocutors, as the speakers accommodating to the linguistic preferences and competences of the interlocutor. The speakers, when selecting a code, thus make use of recipient design, a concept worked out by Sacks, Schegloff & Jefferson. They choose code in order to include a particular interlocutor, adjusting to the language preferences of the other. There are some instances of asymmetrical plurilingual conversation, in which the interlocutors maintain different codes, but since the code choice appears to be anticipated by the addressee the interaction runs smoothly between them. As for the direction of the code switch, two different tendencies emerge: one in the switches connected to the identity work of the participants in which the direction of the switches seems to be highly significant. The other in the switches connected to the management of the conversation in which the direction proves to be reversible.