Through the frameworks of Conversational Analysis [Sacks et al., 1978] and Interactional Linguistics [Selting and Couper-Kuhlen, 2001], we describe the role of smiles in constructing thematic transitions. Conversation is a "collaborative production" [Sacks et al., 1978] during which interlocutors use several "multimodal resources" [Birdwhistell, 1968] and address various themes. These thematic sequences are linked with "conversational movements" [Riou, 2015] called thematic transitions : moments of agreement negotiation on the next theme. Furthermore, facial signs are particularly observed by the recipient [Guy, 2013] and the speaker's smile helps maintain the recipient's attention [Theonas et al., 2008]. Since it is assumed that the transition requires the participants' shared attention, it is interesting to investigate the role of smile as "facial gesture" [Bavelas and Gerwing, 2007] in thematic transitions. The objective of this study is twofold (1) describe the ways participants use their smile in thematic transition (2) explore this smile role depending on the degree of acquaintance of participants. This study has been conducted on two audio-video corpora of face-to-face French conversations. The interactions analyzed brought together 40 participants for 8h : 20 participants knew each other (corpus CHEESE! [Priego-Valverde et al., 2020]) and the 20 others met for the first time the day of the recording (corpus PACO [Amoyal et al., 2020]). The smile annotations were performed with a semi automatic protocol [Rauzy and Amoyal, 2020]. The results show two major findings (1) during the thematic proposition phase, the speaker drops his/her smile compared to a random moment in the conversation (χ 2 = 180.44, DF = 78, p < 0.01). In the corpus CHEESE!, the amplitude of this effect is 4.5 while in PACO it is 5.6 ; (2) when accepting the transition, the recipient increases his/her smile (χ 2 = 111.96, DF = 78, p < 0.01). In the corpus CHEESE!, the amplitude of this effect is 2.3 while in PACO it is 2.6. These two results on both the speaker and the recipient provides additional evidence that smile change is closely related to topic change. Moreover, when participants don't know each other these configurations of smile change are even more frequent.