This thesis investigates the phenomenon of flagging in code-switching. The term ‘flagging’ is normally used to describe a series of discourse phenomena occurring in the environment of a switch. In spite of a large literature on code-switching, not much is known about flagging, aside from the general assumption that it has signalling value and may draw attention to the switch (Poplack, 1988). The present study aims to offer a more eclectic understanding of flagging, by looking at the phenomenon from both a structural and an interpretive perspective. The analyses are based on two small corpora of naturalistic conversations collected amongst pairs of bilingual English-Italian speakers in the UK and Italy. The structural analysis looks at quantitative patterns of flagging. A relationship is observed between the frequency of an item in the data and its production with or without flagging. Higher frequency generally is related to less flagging. A similar relation holds between flagging and different grammatical categories, with nouns being less flagged than adjectives or verbs. The interpretive approach adopts the methods of Conversation Analysis (Auer, 1998) and investigates how the presence of flagging is instrumental in reconstructing participants’ own understanding of the interaction. Through flagging, participants reveal to one another their orientation to single instances of language alternation as belonging or not to the medium (Gafaranga, 2000) of the conversation. While it may be seen as a peripheral occurrence, flagging can reveal the degree and ease of integration of switches in speech; an appreciation of its role can further the understanding of the dynamics of language contact in naturalistic settings. By looking at the results from the two analyses, this study shows how flagging is a patterned phenomenon that speakers interpret as having communicative value, rather than a simple disfluency typical of spontaneous speech. ¶