Lexical other-repetition is a process that consists of repeating words that have been previously produced by another interactant. This leads to a lexical similarity of the participants' discourse. According to Tannen (2007), participants use lexical other-repetition to show their involvement in the interaction. She argues that repetition is useful at several levels: production (repetition facilitates encoding), understanding (repetition facilitates decoding), connection (it maintains cohesion in discourse), and interaction (repetition maintains the link between participants). We focus here on a subcategory of lexical other-repetitions: the phenomenon that we call "pinning" is a form of repair (Schegloff, 2007: 100-101), in so far as one element that has been uttered by one of the participants is afterward treated as a “trouble source”. It develops like a repair, and specifically like an other-repair: A turn that contains an element that will be taken as the source, "the repairable" B picks up this element in A's turn, repeats it and comments on it A/B various ways of responding to B's repair Two features characterize "repair": on the one hand, the fact that it consists in a process, and on the other hand that it is not inherently related to a real, an actual or an obvious problem. In the following definition, Schegloff insists on this last aspect: “Not only are "obvious" problems unaddressed; anything in the talk may be treated as in need of repair. Everything is, in that sense, a possible repairable or a possible trouble-source. It is overt efforts to deal with trouble-sources or repairables – marked off as distinct within the ongoing talk – that we are terming repair” (Schegloff, 2007: 100-101) For the cases we are dealing with, the repeat in the second position is neither uttered in order to clarify what has been said, nor in order to ask for an explanation, and even if it does, it is above all oriented towards underlining a more or less strange, unexpected or surprising feature of A's turn. Example 1: CLAPI. Corpus Grillage (Bert et al., 2010) A euh:::: un truc euh: bon y a personnage/ enfin tu vois y a trois: (0.6) trois trois symboles (.) et:: tu vois: (0.8) et puis sinon/ ben y a l` réglage des a- des asas euh: P des a[sas/ ((rire)) L [ouais A des asas et puis voilà In this case, P understands the element pronounced by A. Nevertheless he repeats it, with the same form as in a repair, but orienting it toward mockery, with a “savoring” function (Tannen, 2007). “Des asas” is extracted from the current discourse because it is “worthy” to be commented on. We could say that the repeated element is treated as a "commentable" (cf. "repairable") element and insofar initiates a specific form of repair. We propose to investigate whether the formal alignment (lexical similarity) due to repetition, and possibly leading to a humorous sequence, reveals convergence in interaction. During a previous collaboration in the SPIM project (Imitation in speech: From sensori-motor integration to the dynamics of conversational interaction), we gathered a large collection of examples of other-repetition, from a set of corpora of semi-spontaneous and spontaneous interactions, with various contexts and activities (conversation, storytelling) (Bertrand et al., 2008; Bert et al., 2010). Various aspects of "pinning" have been studied: - the process of extracting an element from the previous turn - the devices used to show that the repair is oriented towards mockery or humor, including the link between the first occurrence and the repeat, on the syntactic, phonetic, and interactional level - the different types of following turns (i.e., turns responding to the repeat) - the consequences for the degree of convergence within the sequence. In the cases we work on, a participant produces the first occurrence of a word or expression in the speech stream, and the other participant notes its incongruity by repeating it, eventually in a humorous perspective. The classical humorous schema consists of the presence of a connector (Greimas, 1966) and a disjunctor (Morin, 1966). In a “pinning”, A produces the connector, which allows two different interpretations (a logical one and an unexpected one), and B repeats it, adding a disjunctor, that actualizes the absurd interpretation. The gap between what A said, and what B interprets causes the incongruity, thereby creating humor. Example 2: Corpus of Interactional Data (Bertrand et al., 2008) LJ j'ai senti qu(e) ça s'adou-#cissait et bon après on a eu des rapports normaux AP mh mh LJ bon euh LJ euh mais au début putain j'é- j'étais mal quoi je euh AP mh mh AP normaux c'(es)t-à-dire euh hum LJ ((rires)) AP avec préservatif ou sans euh LJ oh putain ((rires)) ça y est t'es dedans là ((rires)) In example 2, LJ describes professional relations, and AP produces back-channel signals. Then AP repeats one word “normaux” (normal), treating it as ambiguous, which appears as a repair sequence. “Rapports” constitutes the connector, since it may be interpreted in different senses, and the repetition is already an attempt of humor. This makes LJ laugh. AP then explicitly actualizes a humorous sexual meaning, for which “préservatif” (condom) constitutes the disjunctor. In this case, there is no modification of the repeated element. In other cases from the corpora, we observe a slight phonetic modification, a syntactic restructuring, or a specific prosodic device, such as a prosodic matching (a type of prosodic orientation defined by Szcezpeck-Reed, 2006), that we will precisely analyze, in order to understand the cues given by the “repeater” to show the humorous or mockery dimension of the repetition (Bertrand & Priego-Valverde, 2011) If we consider our data, the process of “pinning” with a humorous purpose leads to three possible situations: 1/ Basic ratification: the humorous repetition receives indeed a basic ratification from A (laughter, “yes”…) but A quickly goes back to his narration. This humorous intervention of B constitutes a short digression from the current narration. 2/ Failure or non-continuation of the humorous mode of communication: the author of the first occurrence (the serious one) refuses to switch into a humorous theme, and continues his activity, whereas B develops, alone, a humorous continuation. The interactants develop parallel sequences, a humorous one and a serious one. They diverge on the kind of their activity at this point of the interaction, until B finally concedes and switches back to the serious modality. 3/ Joint fantasizing (Kotthoff, 2006): contrary to the previous situation, the humorous repetition is used as a starting point for a co-elaborated humorous sequence. In this case, A produces the first occurrence, and B repeats it (“pinning”), eventually with slight phonetic modifications, in order to switch to humor. Each participant overbids, and they co-elaborate a humoristic sequence. We consider these sequences as highly convergent moments in the interaction. Therefore, the same pattern: < S1 produces “serious” discourse, S2 produces humorous one with “pinning” of a part of the discourse > can be the starting point of a highly convergent sequence, or conversely, result in an interactional divergence. This is an evidence of the lack of equivalence between formal similarity (lexical repetition) on the one hand, and interactional convergence on the other.