The Implications of a Dialogical Approach to Language Acquisition: the Example of a Research Study on the Acquisition of Referring Expressions

ABSTRACT This paper aims to illustrate the contribution of dialogism to the field of language acquisition. According to dialogical approaches, children do not experience linguistic units and structures per se; they experience language in socially meaningful contexts. More specifically, speech genres, activities and interactional settings appear as mediators between individual discourse, social uses and a particular language. In order to illustrate the implications of a dialogical approach, this paper presents a research study on the acquisition of referring expressions (the DIAREF Project). Referring expressions are particularly relevant because their mastery involves both formal and functional aspects of language. The results show that children’s uses of nouns, personal pronouns, demonstrative pronouns and dislocations are jointly determined by discourse-pragmatic factors, such as the position in the referential chain and socio-discursive factors, such as speech genres, activities and interactional settings.


Introduction
Since antiquity, philosophers have raised questions about how children learn words, and therefore, how children acquire language, opposing nature and nurture.
Developmental psychology and linguistics have inherited this ancient debate: the nativist approaches assume the existence of pre-built structures without which knowledge and development could not be explained, and the non-nativist approaches assume that language is built through experience. Dialogism, in its epistemological dimension, falls within non-nativist approaches, and more precisely, within interactionist approaches to language acquisition, inspired by Vygotsky (1962Vygotsky ( [1934) and Bruner (1983). Dialogism contributed to the field by highlighting the social and dialogical foundation of language (BRONCKART, 1996(BRONCKART, , 2004FRANÇOIS;SABEAU-JOUANNET, 1984).
In this regard, the intrinsically dialogical nature of any utterance appears as the fundamental vehicle of language acquisition. It resonates with the Vygotskian notions of inner speech and interpsychological dimension (CHEYNE; TARULLI, 2005;WERTSCH, 1991). At the same time, dialogism extends this characteristic to the macrodiscursive level, involving socially and culturally shared uses of language. According to a dialogical approach, it is under these lenses that the acquisition of linguistic units and structures is to be examined.
All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type  In this paper, 2 we aim to illustrate the contribution of dialogism to language acquisition research through the example of a study on the acquisition of referring expressions conducted in the framework of the DIAREF project. 3 Before presenting this study, we will address some theoretical issues focusing on dialogism as a framework for research on language acquisition, and more specifically, the acquisition of referring expressions.

A Dialogical Approach to Language Acquisition
This section first focuses on the challenges in language acquisition that dialogism may contribute to respond to (1.1), and on the specific features which make that response possible (1.2). We then address the specific issue of referring expressions (1.3); the section ends with a brief summary of the DIAREF project (1.4).

The Theoretical Challenges of Language Acquisition
The field of language acquisition faces major challenges. In the last decades, usage-based and emergentist approaches (AMBRIDGE; LIEVEN, 2011;MACWHINNEY;O'GRADY, 2015) have succeeded in presenting alternative models showing that children do not develop their language from a pre-existing innate system of rules and categories, 4 as they build it from their direct experience of the input. Among the several challenges still to be met, Lieven (2016) mentions, in a synthesis paper, the need to address the relationship between meaning and form. This is certainly a crucial issue for understanding the foundations of the formal aspects of the acquisition process, such as 2 We wish to thank the two anonymous reviewers for their relevant comments and questions that helped us to clarify our text. 3 The DIAREF Project (Acquisition des Expressions Référentielles en dialogue: approche multidimensionnelle, funded by the French National Agency for Research, ANR-09-ENFT-055) gathered several teams and institutions and numerous colleagues around a multidimensional approach of the acquisition of referring expressions (see webpage: http://www.univ-paris3.fr/anr-diaref-37421.kjsp?RH=1505727285324). The main results of the project will be published in Salazar Orvig, De Weck, Hassan & Rialland (in press). 4 In this article, our goal is to highlight the pragmatic, interactional and dialogic aspects of the language acquisition process. The debate with the nativist approaches is, precisely, about the need to consider these aspects of language that are marginal to them.
All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0 frequency effects. In most of the emergentist and usage-based approaches, the sociopragmatic dimension appears as the most relevant factor to account for the acquisition of word meaning (TOMASELLO, 1999). However, it fades away when it comes to studying the formal aspects of language, such as constructions and syntax, because their main goal is to respond in their own terms to the nativist models (LIEVEN, 2016).

Dialogism and Language Acquisition
Even though Bakhtin's and Vološinov's work did not deal either with child language or with child development, a dialogical conception of language lays the groundwork for a complementary approach to language acquisition in several ways.
The most obvious one lays on the dialogical conception of an utterance. As adults, children produce utterances in the continuity with (and in response to) the other's utterances. And as adults, children's utterances are heteroglotic, in the sense that they are also constructed from the other's discourse (BAKHTIN, 1982(BAKHTIN, [1975). This fundamental feature of utterances accounts for the role dialogue plays in the concrete process of acquisition. As Vygotsky 5 (1960, 1981 laid down, higher mental functions, and thus language, first develops in an interpsychological plane before being internalized in an intrapsychological plane. Moreover, dialogical and interactionist approaches have shown the pivotal role of actual dialogue moves, such as taking up and reformulating moves, in the process of acquiring words, structures and uses (BERNICOT; SALAZAR ORVIG; VENEZIANO, 2006;CLARK, 2014;DE WECK, 2006;VENEZIANO, 2014).
Also, according to a dialogical conception of language, speakers (adults or children) experience language through the others' discourses. This assumption entails that children do not deal with an abstract system but with the adult's discourse in dialogue.
We know our native language -its lexical composition and grammatical structure -not from dictionaries and grammars but from concrete utterances that we hear and that we ourselves reproduce in live speech communication with people around us (BAKHTIN, 1986(BAKHTIN, , p.78 [1979). Therefore, forms (lexical and/or grammatical) are acquired for their communicational, pragmatic properties. These functional properties arise from their use in meaningful communicative contexts. Bakhtin -as well as Vološinov -situate these contexts at the level of activities and speech genres.
To learn to speak means to learn to construct utterances [...]. Speech genres organize our speech in almost the same way as grammatical (syntactical) forms do. We learn to cast our speech in generic forms and, when hearing others' speech, we guess its genre from the very first words (BAKHTIN, 1986(BAKHTIN, , p.78 [1979). Therefore, we can draw from Bakhtin's and Vološinov's proposal that genres (and activities) 6 are mediators between language practices and children's construction of their language.
This stance implies, as Vološinov argued, a major shift when analyzing linguistic/language phenomena. From what has been established, it follows that the methodologically based order of study of language ought to be: (1) the forms and types of verbal interaction in connection with their concrete conditions; (2) 6 The definition of speech genres is intimately associated with the notion of activity, as we can see in the beginning of Bakhtin's chapter on speech genres: "All the diverse areas of human activity involve the use of language. Quite understandably, the nature and forms of this use are just as diverse as are the areas of human activity" (BAKHTIN, 1986(BAKHTIN, , p.60 [1979). forms of particular utterances, of particular speech performances, as elements of a closely linked interaction -i.e., the genres of speech performance in human behavior and ideological creativity as determined by verbal interaction; (3) a reexamination, on this new basis, of language forms in their usual linguistic presentation (1986, pp.95-96 [1929]). This is a crucial reversal of perspective: the usual research procedure adopted by structural and generative linguistics is to consider first the linguistic units -phonemes, words -and then look into how they are used. Vološinov contends that, on the contrary, the study of language must follow the "actual generative process of language" which corresponds first to the initial immersion in social and meaningful contexts. In terms of language acquisition, this involves, in line with interactionist perspectives (BRUNER, 1983;VYGOSTKY, 1962VYGOSTKY, [19341960;1981), considering first the immersion of children in social contexts -and thus the communicative and interactive acts of meaning -before focusing on the formal aspects of the construction of linguistic units. We shall now address how this methodological reversal can affect the study of referring expressions.

A Dialogical Approach to the Acquisition of Referring Expressions
Because they present both morphological and pragmatic features, referring expressions (hereafter REs) are particularly interesting to a dialogical approach. The "choice" of a determiner (for a noun) or of a pronoun indicates the degree of the referent's accessibility for the addressee (ARIEL, 1990;GUNDEL;HEDBERG;ZACHARSKI, 1993). At the same time, the acquisition of determiners' and pronouns' paradigms is strongly linked to the acquisition of syntax.
Studies on toddlers in family interactions showed that early occurrences of pronouns or determiners are adequately used, according to the discursive accessibility of the referent (DA SILVA et al., in press;SALAZAR ORVIG et al., 2010, 2013SERRATRICE;ALLEN, 2015;YAMAGUCHI et al., in press). However, these early uses might be rather determined by the coalition of syntactic and discursive factors such as subject function and expression of topic, which are usually highly accessible. In that case, can we still consider that these early uses reflect early pragmatic skills? In the same vein, studies on older children, mostly focused on narratives, highlighted both their actual skills and their difficulties dealing with the absence of shared knowledge or with several concurrent referents (HICKMANN; SCHIMKE; COLONNA, 2015;REZZONICO et al., in press). Current discussions address these pending issues by investigating the cognitive basis of discourse-pragmatic skills (DE CAT, 2015). However, our contention is that a dialogical perspective can also account for both children's abilities and difficulties. As other linguistic expressions and all the more, since they carry pragmatic values, the use of REs is determined by the dialogical conditions and the socio-discursive contexts of children's communicative experience. The DIAREF project, which we will present now, aimed to address this issue.

The Aim of the DIAREF Project
The aim of the DIAREF research project was two-fold. Firstly, we aimed at giving an account of the various REs used by children -and also adults -taking into consideration multiple facets of their development (phonological, morphological, syntactic and discourse-pragmatic) in children aged between 1;7 and 7;5 in various interactional settings. The main focus was to assess children's sensitivity to accessibility and to the position of the RE in the referential chain. Secondly, we aimed at accounting for these uses by emphasizing the interplay of formal and functional factors and, more precisely, the role of children's communicative experience. This experience is gradually built up and diversified through the uses they are exposed to in the various interactional settings, first at home, and later at school. Those settings involve various activities and speech genres, and sometimes various participants, which influence children's productions.
Thus, describing the uses of REs in different interactional settings, considering the role of dialogue, activities and speech genres, allows us to understand the way these contexts influence the "choice" of REs and how their functionalities are built.

The Influence of Socio-Discursive Factors on the Acquisition of Referring Expressions
Drawing on data and results from the DIAREF project, 7 this section first presents some methodological aspects involved in data gathering and the analysis of REs relevant for this paper (3.1). It focuses then on the results concerning three main aspects of children's communicative experience, that is, the role of activities (3.2), speech genres (3.3), and social context (3.4) in children's uses of REs.

Methodological Aspects
Our approach to reference acquisition requires drawing on data from both adults and children in diversified sets of interactions collected in non-experimental settings. Our data gathered various corpora 8 collected in two settings (home and school) involving, on the one hand, participants with different social statuses (children, mothers, teachers), and on the other hand, various naturally occurring activities (different types of play, daily routines, and storytelling). 9 These methodological aspects are crucial to shed light to the language models children are exposed to.
For the purposes of this article, we will focus on the four main categories of REs: 10 'Dislocation' is a very frequent construction in oral French. It is formed by a strong expression (noun, demonstrative pronoun) and a clitic pronoun in a resumptive position (e.g. le chien il court, 'the dog he runs'). As shown by Lambrecht (1994), this construction allows a strong form such as a noun or a demonstrative pronoun (usually devoted to the expression of new information) to be used in the topic position.
All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0 relevant for the study of children's development of referential competence. The distribution of these four categories was analyzed according to the following factors: the influence of the activity being carried out, the speech genre and the social context (home/school) and their position in the referential chain. Among the positions in the referential chain, we will focus on first and subsequent mentions (hereafter FM and SM), as they were the predominant ones in our data.

The Influence of the Activity on the Use of Referring Expressions
Numerous studies show that the activity in which participants are involved influence different aspects of language use: for example, the diversity of vocabulary (HOFF, 2010), the degree of verbal participation ROSAT, 2003), or the amount and the nature of speech acts (LEAPER; GLEASON, 1996). Concerning the use of REs, very few studies were conducted until now apart from our team's. Salazar Orvig et al. (2018) showed the impact of the activity on the use of demonstrative pronouns.
De Weck et al. (in press) investigated this influence in a corpus of mother-child interactions: 25 toddlers (aged between 1;7 and 2;6), and 25 older children (aged 3;9 to 7;4). Toddlers' use of REs was examined in the following activities: playing with toys (PT), daily routines (DR), and activities based on pictorial material (PM). Those activities share a common feature: being anchored in the "here and now" of the interaction, being driven by a specific goal, and involving object handling, though at various degrees.
Nevertheless, results showed diversified uses of REs according to the type of activity carried out. For instance, the proportion of third-person pronouns was higher when playing with toys (11.86%) and in activities based on pictorial material (10.47%) than in the daily routines (6.10%). In contrast, the ratio of nouns, and more specifically those appearing in FMs, was higher in daily routines (50.90%) than in the two other activities (40.0 % in PT and 35.42% in PM). In daily routines, participants mentioned several objects only once without taking them up later in the dialogue. Consequently, the proportion of third-person pronouns for SMs was low (11.25% in DR vs. 15.06 % for PM and 18.18% for PT). They were replaced by null forms (8.98% in DR). As we can see in example 1, null forms encode both self-reference (notice, in Elodie 8 and 9, the lack of first-person subject je, 'I', mandatory in French) and entities, in the object function (the reference to the apple in Elodie 8): (1) Elodie, 2;2, daily routines (snack time) 11 The participants are talking about the apple Elodie is snacking Finally, demonstrative pronouns and dislocations involving demonstrative pronouns were more often observed in the picture-based activities and in the playing with toys activities.
(2) Clément, 2:3, picture-based activity (puzzle) In example 2, the child (Clément29) introduces a new referent (a new piece of the puzzle) with a demonstrative pronoun associated with a gesture (picking a piece to place it on the board). His mother asks him to label the picture on the puzzle piece, using a demonstrative dislocation. The child responds with a clitic demonstrative-pronoun 12 c' 11 In the examples, the first line gives, for toddlers, the phonetic transcription of the utterance and the second its interpretation in adult form. In the right column a literal translation is given for information purposes. Non-verbal indications are given between (( )); / and // stand for brief pauses; { } indicate uncertain and uninterpretable segments. Unpronounced phonemes appear between (). 12 The quasi-frozen construction c'est is typical of spoken French, both in adults and children discourse. It is formed by the verb être ('to be') and the clitic demonstrative pronoun 'c(e)', which can be used either in referential or in non-referential ways. for storytelling). This difference can be accounted for as, in joint storytelling, nouns introduce the characters whose features and actions will be described by the participants.

Clément has placed a piece in the puzzle
In the symbolic play, demonstrative pronouns are used to draw the interlocutor's attention to new referents, without naming them, as we also observed in toddlers (see above). There was also an interaction between the activity and the position in the referential chain for SMs. As expected, in the storytelling activity, most SMs were encoded with third-person pronouns (example 3).
'they ran out of gas?' Mother9 ((hoche la tête)) '((she nods))' In this example, the child and his mother used the third-person pronouns il, ils, leur ('he', 'they', 'them') several times for SMs of previously introduced referents (twice with an indefinite NP, un seul petit cochon -'one little pig' -and des gens -'people'and once by the mother -Mother7 -with a third-person plural pronoun ils -'they', whose referents are inferable from the mention of the chariot). We can also notice that the switch from one occurrence of the pronoun to the other does not cause any intercomprehension problems. The presence of the book in the situation accounts for these uses and, thus, enhances the proportion of third-person pronouns for SMs in this type of activity (54.63% vs. 28.23% in symbolic play).
In the symbolic play, nouns were proportionally more frequent for SMs than in storytelling (31.47% vs. 19.41% in storytelling activity). In the symbolic play, a great part of the play was devoted to positioning the objects; thus, nouns appear often in post-verbal position (direct object, like la voiture in on met la voiture là -'we put the car there' -or prepositional phrase, like le salon in la lampe elle va dans le salon -'the lamp it goes in the living-room'). There was also a significant proportion of demonstrative pronouns (16.38% vs. 6.56% in storytelling activity) and dislocations (12.72% vs. 3.86% in storytelling activity) in SMs. 13 Example 4 illustrates the case of a dislocation involving a demonstrative pronoun (Samuel45).

The Influence of the Speech Genre on the Use of Referring Expressions
The Bakhtinian notion of speech genre implies the intertwinement of discourse types and activities. Our data confirmed this intertwinement: for example, picture-based activities involved an important amount of labelling (see example 2, Clément2), whereas joint storytelling involved narrative discourse and action descriptions. One of the key issues of a dialogical approach is to investigate whether referential skills develop on their own or if they grow mediated by children's involvement in various discourse sequences.
Various studies have shown that speech genres impact syntactical forms and lexical choices in children's discourse (BERMAN; NIR-SAGIV, 2004;STRÖMQVIST, 2002 In the DIAREF project, Vinel et al. (in press) investigated two hypotheses: 1) that children's sensitivity to referent accessibility is built on to the use of REs in various genres; 2) that this link evolves with age. In order to address these two issues, the study considered the way genre affected the proportion of REs in children's discourse and how this factor interacted with the position of the referent in the referential chain. In order to neutralize, as much as possible, the activity factor, the study was conducted on activities involving pictures (either storytelling from a picture book or games involving pictures, such as lotto, cards, puzzles etc). The study concerned 14 toddlers (aged 1;7 to 2;6) for both types of activities, and 30 older children (aged 3;6 to 7;5) only for joint storytelling.
Speech genres were considered from two complementary angles: a) the genre at the level of discursive sequences, determined by the relation of discourse to the referential space. In that sense, narrative sequences, which build a fictional world, differed from here-and-now sequences in which utterances refer to the current situation, time and space of the interaction. The comparison between discursive sequences was conducted only with the toddlers' corpus. b) The genre at the utterance level: mainly labelling, action descriptions, state descriptions, explanations and argumentations.
The results of this investigation showed that discursive sequences had an impact on the "choice" of third-person pronouns and nouns. Toddlers used more third-person pronouns (18.73%) and more nouns (46.11%) in the narrative sequences than in the hereand-now sequences (respectively 6.69% and 31.56%). However, this impact was not statistically significant when discursive sequences were considered together with speech genres at the utterance level. Further research is necessary to better understand the strong overlap of these two levels.
At the utterance level, the speech genre, along with the position in the referential chain, can be a predominant factor in the choice of an RE. In certain cases, the speech genre overrides the "referential position" factor. For instance, third-person pronouns preferentially appeared in action descriptions, as we can see in example 5, in a here-andnow discursive sequence for a toddler, and in example 6, in a narrative sequence for an older child.
(5) Maxime, 2:3, puzzle The child is looking at a new piece of the puzzle  'oh well he saw a stick ((points to a picture))' In contrast, labelling was a dominant factor for clitic demonstrative pronoun c'est, 'it/this', and the only significant factor for strong demonstrative-pronouns ça, 'this/that'.
The strength of this factor is probably due to the influence of the activities, which were all picture-based (see 3.2). Salazar Orvig et al. (2018) showed, in a larger corpus including all kinds of activities, that labelling was the predominant factor for clitic demonstrative-pronouns, but that strong demonstrative-pronouns could be very often used in descriptive utterances. The difference of results between Vinel et al. (in press) and Salazar Orvig et al. (2018) highlights the strong interaction between these two levels of socio-discursive factors.
When considering the narrative sequences alone, the use of nouns and third person pronouns was determined by both the utterance speech genre and the preferential position in the referential chain (FM for nouns; SM for third person pronouns), as it can be seen in the very-adult like example reported below (partially seen in example 3): 'there is only one little pig who is leaving// et and then he is tired {there}// and then he falls asleep ((turns a page))// and the he asks to some ((points the picture)) people to go on the tractor but they don't want ((he shrugs)) ((turns a page)) and that's the story' Above all, the study of the impact of speech genres on the use of REs provides a path to account for early referential skills. The results in Vinel et al. (in press) demonstrate that the "choice" of third-person pronouns or nouns is not determined by one single factor, either the position in the referential chain or the speech genre, but by their strong interaction. Therefore, we can hypothesize that the uses of REs, constrained by speech genres, are potential mediators for the acquisition and development of the autonomous, more mature cognitive skill of assessing the accessibility of a referent for an interlocutor.

The Influence of the Social Setting on the Use of Referring Expressions
A third aspect of the impact of the socio-discursive dimension on the acquisition of REs is the influence of social contexts. Social contexts are considered here, in line with Vygotsky and Bakhtin, as cultural spaces. Indeed, dialogism is associated with the concept of polyphony/heteroglossia in Bakhtin's (1975Bakhtin's ( /1982 conception of language. Polyphony/heteroglossia is about the multiple voices and perspectives that make out the social world. We argue, based on Bakhtin's views, that any language use is mediated by the social ways of interacting that relate to a specific social field. Thus, as children grow up and start attending school, they experience other language uses than those experienced at home. Drawing on that premise, to show the influence of the social setting, we compared children's uses of REs during joint storytelling activities at home with their mothers, and at school with their teachers and a group of classmates. The children were aged 3;3 to 5;9. They attended one of the three different grades of the French "école All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0 maternelle" (nursery-school). The data were collected from four different classes. The total population is 10 children and mothers and 4 teachers (DE WECK et al., in press; HASSAN et al., in press) The results clearly show that children's uses vary not only according to the social setting, but they also follow the same trend of adults' use in each of the contexts observed.
In this regard, we found that teachers used twice as many nouns than mothers, who used more third-person pronouns and more demonstratives, considering the overall distribution of the REs. Teachers also hardly used any demonstrative pronouns, and used only 10 pronominal dislocations, ten times less than mothers. This is interesting to point out, as pronominal dislocations are a specific feature of spoken French. The results also showed that, when teachers introduced a referent (which they rarely did), they used interrogative pronouns and nouns and they never used a third-person or a demonstrative pronoun, whereas mothers did so in a non-negligible proportion, respectively 22% and 6%.
For SMs, as expected, although third-person pronouns were dominant for both groups, a difference appeared between mothers and teachers. Teachers used twice as many nouns (29.23%) than mothers (15.34%). Mothers used four times more third-person pronouns (62.54%) than nouns. As for teachers, the proportions of third-person pronouns (dominant with 41.86%) and nouns were more balanced.
As in the case regarding FMs, they hardly used any demonstrative or dislocated pronouns. The frequent use of nouns for SMs sets teachers apart from mothers. The younger children are, the more this specific use of nouns for SMs, characterizing teachers, is made evident. This situation can be explained by the fact that in the classroom, differently from the home context, the book cannot be handled by the interlocutors (who are many), as it is shared from a certain distance. It is thus necessary to explain and verbalize, as pointing is usually avoided. It can also be explained by the fact that teachers frequently uptake pupils' REs (which are mostly nouns as mentioned below) to add on them or validate them, as it is shown in example 8. ((holds the book in front of the children and turns the pages)) Kenzo1 un renard ! = ((pointe vers le livre)) 'a fox! = ((pointing to the book))' Teacher4 alors, on voit un renard 'so, we see a fox' In example 8, the teacher uptakes the noun, adding c'est vrai ('it's true') or on voit ('we see'), in order to confirm and approve the answer. This example also shows that children produce nouns in responding to teachers' questions.
As for children, their uses at school also differed from their uses at home (DE WECK et al., in press At home, although they mainly used nouns (62.5%), they also used third-person pronouns (14.58 %, as example 10 below shows) and other pronouns for nearly one-third of referent introductions. Children introduce referents, using a sort of "ritual" or formatted phrase with the verb "voir" ('to see') + an indefinite NP (like in Anaïs3 and Kenzo14). This introduction construction seems to be classroom-specific. We found only one occurrence of this construction at home in our corpora, when Anaïs answered a question by her mother who When it comes to SMs, children mainly used third-person pronouns in both settings, but they used more nouns at school than at home. The younger they were, the more they used nouns for SMs compared to third-person pronouns. This proportion can be explained by two complementary phenomena: children answer teachers' queries to identify the referents; they propose "competing" answers as in example 11, in which children discuss the identity of the previously mentioned main character: This example also suggests that children may have "internalized" the importance of vocabulary learning at school and of identifying and naming the referents involved in the story before describing their actions.

Discussion and Conclusion
The main goal of this paper was to show how a dialogical perspective can contribute to the interactionist and socio-pragmatic approaches of language acquisition.
Among the possible contributions, we followed the methodological reversal implied by Vološinov's proposal to examine how communicative situations determine linguistic units and structures. In order to achieve this goal, we presented and discussed the main results of a team research program on REs (the DIAREF project) focusing on the influence of socio-discursive factors on the "choice" of nouns, personal pronouns, demonstrative pronouns and dislocations. REs appeared as a particularly relevant category for this kind of study because using and mastering them involve formal (grammatical and syntactic) and functional (discourse-pragmatic) dimensions. Moreover, an adult-like mastery of REs also involves textual skills (DE WECK, 1991) and cognitive abilities (DE CAT, 2015;SERRATRICE;DE CAT, 2020). Therefore, one of the major issues in the field is resolving the apparent paradox between children's referential skills (observable both in toddlers' interactions and older children's discourse) and their actual difficulties in dealing with their interlocutors' shared knowledge and perceptual availability of referents (HICKMAN; COLONNA; SCHIMKE, 2015). One of the aspects of this paradox is that children do exhibit the capacity to use adequate forms for discourse given referents (or SMs). This means that more mature cognitive or textual abilities cannot account for these referential skills. How is it then that children exhibit this overwhelming performance? Some studies have turned to a narrow conception of dialogue influence, considering that priming (SERRATRICE, 2006) -that is formal influence -might explain children's performance. However, priming alone cannot explain young children's uses (MARCOS et al., in press) and is not relevant when considering monological productions. In contrast, taking into account the dialogical and sociodiscursive dimensions might shed light on this issue. Our results showed that children display a range of uses that vary according to the contexts, the activities and the speech genres at study. They showed that their "mastery" of the functional features of the grammatical units is anchored in specific interactional settings. If we consider nouns, for instance, their proportion varied according to activities, speech genres, and social contexts; and their expected prevalence in FMs was inflected by the specific dialogue dynamics of each interactional setting. This also applies to pronouns and dislocations, and to SMs. These results suggest that, first, the models of discourse production offered by adults (DE WECK et al., 2019;DE WECK et al., in press) have an influence on children's uses, as we have seen above, when comparing children's uses to their mothers' and teachers'. It is likely that children grasp the uses of linguistic forms in their contexts before acquiring the full mastery of their semantic and pragmatic values (NELSON, 2007). 14 However, the impact of these models does not pertain to an imitational process. If these models have an impact, it is because children learn to use REs in recurrent, meaningful sequences of dialogue, which Bruner (1983) calls 'formats'. As we showed above, these contexts, whether considered at the activity level, the speech genre level or the social context level, strongly influence the "choice" of REs. We can therefore argue that children grasp the discourse-pragmatic properties of linguistic devices at these socio-discursive levels, before growing into a more specific knowledge (if any) of the semantic and pragmatic features of the linguistic forms. Our results thus confirm the Bakhtinian idea that children build their language out of their communicative and discursive experience.