Conversation and ‘Voices’ in the Narratives of 5- to- 8-Year-Old French-Speaking Children

 ABSTRACT According to Bakhtin, polyphonic dialogism, a general principle for advancing knowledge, is well instantiated in the literary novel, and more generally in narratives, where different ‘voices’ can find expression. Accordingly, in this paper we analyzed picture-based narratives constructed by thirty 5-to-8-year-old French-speaking children, produced before and after a conversation on the causes of the events. The analysis focused on children’s ability to make different voices heard, particularly the voices of the characters which children made heard not only through direct or reported discourse forms, but also through the expression of their inner states such as beliefs, intentions and feelings. Results show a development in children’s ability to make these voices heard in their narratives, as well as a facilitating role of the conversational exchange on their expression. It will be argued that these results provide support to the central place that dialogue occupies in the overall Bakhtinian


Introduction
According to Bakhtin (1986Bakhtin ( [1979), discourse genres are learnt at the same time as the lexical and structural forms of the language through everyday dialogic interactions between the self and others. For Bakhtin, the meaning of language forms, words and utterances, are shaped through their different occurrences in dialogues with others, or even within oneself (BAKHTIN, 1981(BAKHTIN, [1934). Dialogues, particularly those that bring forth different points of view, are a central means for advancing understanding and knowledge, as well as for attaining the specifically human characteristics of selfconsciousness. This general principle is consistent with the interactional and functionalist approaches to language acquisition which have shown the beneficial effects of conversational functioning on the acquisition of different aspects of language knowledge and use (CLARK, 2018;SALAZAR ORVIG, 2017;VENEZIANO, 2000).
Another central point of the Bakhtinian approach 1 was to show that different discourse genres can coexist within the same overall text, of which the literary novel was taken as a particular object of study (BAKHTIN, 1986(BAKHTIN, [1979). Within this approach, the novel is a 'complex' discourse genre in which several discourse types, and even languages, are imbricated and where different 'voices', that of the author, of the narrator and of the characters, find expression within a dynamic and unfinalized dialogue, namely, a dialogue where each intervention is seen as a reaction to a previous intervention and is projected towards a subsequent reaction, with no first nor final word being thus possible.
This notion of dialogism involves all aspects of knowledge including language, art and the process of learning itself.
The different 'voices' made heard in the novel, and more generally in narratives, create polyphonic dialogues. According to Bakhtin (e.g.,1979Bakhtin (e.g., /1986, reported or indirect discourse can occur in everyday conversations and it is thus considered a first order, or simple, type of discourse. However, it becomes a second order, more complex, discourse type when it is inserted in a new domain of exchanges such as the novel or the narration more generally. The same holds for direct discourse as well as for the expression of 1 There is an ongoing controversy concerning the authorship of some of the notions ascribed to Bakhtin, controversy that is of no interest here, but see Zenkine (2011) for a very thorough and well-informed treatment of the issue. Brazilian Editor's Note: On this theme, see also Bakhtiniana 9 (special number), 2014 https://revistas.pucsp.br/index.php/bakhtiniana/issue/view/1255/showToc All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0 thoughts and feelings. When these are expressed by the speaker in everyday conversation to convey his/her own feelings, beliefs and thoughts, it is a first order, simple discourse type. However, when feelings, intentions and thoughts are expressed to make the voices of the characters of the story heard, this is considered a second order, more complex type of discourse. Indeed, in these cases, the speaker/narrator not only attributes those feelings and thoughts to others, but also these voices become an integral part of the overall story plot that they contribute to create.
The complexity of airing this type of discourse is high in the literary novel. Though at a much lower level of sophistication, also children who are led to narrate a story on the basis of a sequence of pictures -an elicitation situation widely used to trace and better understand the development of children's narrative competences (BERMAN; SLOBIN, 1994;STRÖMQVIST;VERHOEVEN, 2004, for studies using the Frog story as the pictured basis of the story)engage in a cognitively challenging activity where they not only are called to narrate but also, at the same time, to create the plot of the story, and this within a rather short period of time. Indeed, it requires children to focus on and integrate into one structured discourse different aspects at the same time: the creation of a coherent and communicatively clear narrative content, the choice of appropriate linguistic expressions adapted in terms of lexicon, grammatical structure and discourse type, as well as the expression of the different 'voices' to be made heard in the narrative.
In earlier studies, we traced the development of children's ability to narrate a story on the basis of five pictures without text in this complex situation, and explored whether a Short Conversational Intervention (SCI) between the child and the Experimenter promoted children's abilities in terms of narrative content and linguistic expression (VENEZIANO et al, 2020;NIR, 2017). This paper, inspired by Bakhtin's notion of polyphonic dialogism of narration, proposes an original analysis focused on whether and how children of ages comprised between 5 and 8 years, make different voices heard in their stories, and whether this ability is affected by the conversational intervention they hold with the Experimenter about the causal aspects of the story (see below in the Method section).
In the elicitation and procedural setting used here, children could make several voices heard: the voice of the child who interacts with the experimenter and the voice of the child as author who may express his uncertainties or his own beliefs. These are first All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0 order, 'simple discourse' types in which the child speaks in the 'first person' and expresses his own subjectivity. There is then the baseline voice of the narrator which is heard when the child takes the viewpoint of an observer about what happens and relates it in a rather objective or descriptive way. Here the child enters the narrative discourse by talking about the characters and the events from his/her own perspective. The most interesting voices children can make heard are those of one or both of the characters, and not only when their voices are heard through direct or reported discourse forms, but also when their 'voices' are heard through the expression of their inner states such as beliefs, intentions and feelings. When children make the characters' voices heard in any of these forms, they talk about the events that make up the plot through the characters' subjectivity and from the latter's point of view.
The analysis will concentrate on whether there is a development in children's ability to make these voices heard in their stories, and on whether the SCI has an effect on this ability. The early presence of polyphony in oral narratives would provide support to the central place that dialogically-expressed points of view occupy in the overall Bakhtinian approach.

Participants 2
Participants were 30 French-speaking children (15 boys and 15 girls) aged between 5;6 and 8;8 years. Children attended kindergarten, first and second grades in a public school of the Paris region. There were 10 children in each of three class/age groups: 5 years old (kindergarteners), 6-7 years old (first graders), and 8 years old (second graders). Table 1 presents the age range, mean and standard deviation (SD) for each grade/age group. According to the teachers, the children were developing in a typical way.

Material
We used the Stone story, a sequence of five pictures without written text meant to represent the story of a misunderstanding between two characters. 3 The story, already used in previous studies (VENEZIANO; HUDELOT, 2006HUDELOT, , 2009, is an adaptation of a story originally published in a book for young children (FURNARI, 1980

Procedure
All the children were interviewed individually in a quiet room at their school.
The procedure applied here is very similar to that used in previous studies HUDELOT, 2006HUDELOT, , 2009VENEZIANO et al., 2020). It is a withinsubject design devised to compare the narratives produced by the same child before and after a short conversation. First, children looked at the five pictures of the Stone story, presented sequentially, one at a time, in the right order. Once the child was ready to tell the story, the pictures were removed, and the child narrated the story. 4 After this first narrative, the short conversational intervention (SCI) started. The child participated then in a conversation with the experimenter focused on the causes of each of the four main events of the story and the related characters' behaviors, while the pictures were again placed in front of the child. "How come that" or "Why": 1. P1 pushes P2 ('the first push'); 2. P2 pushes P1 ('the second push'); 3. P1 shows the stone ('pointing towards the stone'); and 4. P2 helps P1 to get up ('the reconciliation'). After the SCI, the child was asked to tell the story once again (the second narrative). As was the case for the first narrative, here also the child told the story after the pictures had been removed.
All the interviews were audio and video-recorded and were later transcribed verbatim in CHAT format and linked to the videos (MACWHINNEY, 2000).

Data Analysis
In each narrative, produced before and after the SCI, we identified different types of 'voices' that children made heard. We distinguished: 1. the Author's voice; 2. the voice of the Narrator-as-Observer and 3. the voice of the characters. These were further distinguished into five subcategories reflecting the different ways in which the characters' voices were made heard: 1. Direct discourse; 2. Indirect or reported discourse; 3. the thoughts and beliefs of the characters; 4. the intentions of the characters, and 5. the emotions and the perceptions of the characters. We also identified a sixth caseto the effect of 'they say hello' -This was considered an ambiguous case, between direct discourse and the voice of the narrator and as such it was kept separate from the other five types of voices. All these voices are described and illustrated in the 'Results' section below where we also provide descriptive results concerning the number of children who make these voices heard at least once comparing the three grade/age groups as well as the narratives produced before and after the SCI. In addition, we also counted all the occurrences of the characters' voices (identified in subcategories 1-to-5) made heard by each child throughout the whole narrative. A score of 1 was assigned when the voice of a character was heard individually, and a score of 0.5 when the voices of the two characters were heard in an undifferentiated way (for example, ils veulent plus se bagarrer 'they don't want to fight anymore'). In this way, we could perform a mixed repeated measures ANOVA to determine whether there is a development in children's ability to make these voices heard throughout the story, whether the conversational interaction in which children participated before producing the second narrative has a positive effect on this ability, and/or there is an interaction between these two variables.

Results
As mentioned above, the setting allows children to make different voices heard.
As specified in the 'data analysis' section above, here we focused on the voice of the author, on the voice of the narrator-as-observer and on the characters' voices. In section 3.1, we describe and provide examples of how each of the identified voices are expressed by the children as well as provide descriptive results relating to the number of children who made each of the identified voices heard at least once in the two narratives combined and comparatively for each of the two narratives (before and after the SCI). In section 3.2. we will present quantitative results on the total number of characters' voices made heard through each narrative, across the age groups and the narratives produced before and after the SCI.

The Author's Voice
As mentioned, in the present story-telling setting, children had to construct and to narrate the story at the same time. As Bakhtin pointed out, there are important differences between these roles and in a literary novel they are well and skillfully distinguished. Another way in which some children had the author's voice heard was by spontaneously announcing that the story was finished (example 4): Example 4: Announcing the end of the story And then that's all! 8 The author's voice was made heard in one of the above ways by 7 children (23% of the sample), in either first or second narratives, only one of them from the older age 5 In the original: "Après y a un garçon je sais pas pourquoi il pousse l'autre." 6 In the original: "Et puis après le petit garçon il avait # peut-être qu'il avait dit une méchanceté sur l'autre garçon." 7 In the original: "Et puis après # je crois # qu' y en a un qui a dit quelque chose de méchant à l'autre." 8 In the original: "et après c'est tout!." group. In all the other cases, the author's and the narrator's voices were not clearly distinguished.

The Voice of the Narrator-as-Observer
All children in all age groups, and in both first and second narratives, made the narrator's voice heard, that is, they talked about the events and the characters' actions from an outside viewpoint, as it might be that of an Observer. When only the narrator's voice was heard throughout the story, children talked about the events in a descriptive mode, referring to one or both characters and linking the events mostly temporally (example 5) and sometimes also causally (example 6): Example 5: The narrator-as-observer's voice: Temporal links The boy with the overall pushes the boy with the shorts; then, the other one pushes him back; then he cries; and then, the other one helps him to get up. 9 Example 6: The narrator-as-observer's voice: Temporal and causal links There is a boy who stumbles because of a stone and then he pushes the girl. 10 Though all children made the narrator's voice heard to talk about at least one of the story's events, only 30% of the children made the narrator's voice heard throughout the narrative. They were 23% to do it in the first narrative, a percentage that decreased to 7% in the second narrative, with no overlap between the two narratives. That is, the children who did that in the first narrative were not the same as those who did it in the second narrative. The percentage of children who made only the narrator's voice heard was higher in kindergarteners (50%), followed by the first graders (20%), while none of the second graders made only the narrator's voice heard throughout the story.
Kindergarten children showed a marked decrease between the first and the second narrative: while they were 50% in the first narrative to make only the narrator's voice heard, none of them did that in the second narrative.

The Voices of the Characters
In telling their story, children can make the voice of one or both of the characters heard. As discussed above in the data analysis section, there are several ways in which children can do this. Some ways are very clear and easily identifiable, as when children make the characters talk in direct or indirect reported discourse. In other cases, the characters' voice is made heard through the expression of the characters' inner states, such as their thoughts and beliefs, or their intentions and emotions. As mentioned above, the characters' voices were distinguished in five subcategories, presented in details hereafter.

Direct Discourse
Here the character's voice is heard directly as he is made to talk in direct discourse.
In example 7, the child makes P1 address P2 directly preceding P1's utterance with the explicit marking 'he says': Example 7: The character's voice in direct discourse And then he says: "It is not because of me, it is the stone that made me fall." 11 Only 17% of the sample (5 children) used direct discourse to make the characters' voice heard, of which 4 children were second graders and 1 child was a kindergartener.
Overall, these children produced 9 occurrences of direct discourse, 6 of which concerned P1 clarifying to P2 the reason for the first push, while the remaining 3 occurrences concerned P2. One second grader made P2 say "d'accord" acknowledging the previous explanation of the first push by P1, while the kindergartener used direct discourse to have P2 say to P1 that he is cheating. No difference in the use of direct discourse was observed between the two narratives: one second grader used direct discourse only in the first narrative, another only in the second narrative and the other two second graders as well as the kindergartener used it in both narratives.

Indirect or Reported Discourse
Here it is still through their talk that the characters' voices are made heard; however, instead of putting the words in their mouth in direct discourse, it is in indirect reported discourse that their voices come through. This is illustrated in example 8 where the child explains the crying of P1 through what P1 had said: Example 8: The character's voice in indirect discourse And then he cries because he says that it is because of the stone. 12 As for direct discourse, also for indirect discourse only 17% of the sample (5 children) used this type of discourse form to make the characters' voices heard, three of which were second graders and the remaining two children were one a first grader and one a kindergartener. Overall, these children produced 6 occurrences of indirect discourse, of which 5 concerned P1 explaining the misunderstanding, and 1 concerned P2 responding to P1's explanation of the first push that he will help him to get up. Though there are too few occurrences to say anything meaningful, we note that the use of indirect discourse is greater in the second than in the first narrative (4 vs. 2 occurrences) and that the 2 occurrences in the first narrative were provided by second graders.

The Thoughts and Beliefs of the Characters
In direct and indirect discourse form the characters' voices are made clearly and explicitly heard through what they say. In the present and in the following two uses, the characters' voices are made heard more indirectly but still clearly through the expression of their inner states. The child reveals in this way the characters' inner world and their own outlook on the external world. The most compelling case is the expression of the thoughts and beliefs of the characters. These are at the core of the misunderstanding between the two characters and their expression reveals the characters' own points of view on the events, providing at the same time the explanation for their behaviors. All of the attributions found in the present sample concern the belief of P2 about the intentions 12 In the original: "Et après il pleure parce que il il dit qu'c'est à cause de la pierre." behind the first push by P1, which provides at the same time the motive for his pushing back behavior, as in example 9: Example 9: The character's belief about the intentions of the companion The girl believes he has done it on purpose, so she pushes him back. 13 Seven children (23% of the sample) made the beliefs of P2 heard. Of these children, 57% (4 children) were second graders and the three remaining were one a first grader and two kindergarteners. Only 2 children, both of them second graders, made the beliefs of P2 heard in the first narrative, while all seven did it in the second narrative.

The Intentions and Desires of the Characters
Attributing and expressing the intentions and desires of the characters is another way of making their voices heard. It is most often the intentions of P2 that are aired, as in example 10: Example 10: The character's intentions The boy he falls on a stone # and makes the girl fall but he has not done it on purpose. 14 Sometimes it is the desire of the characters that are made heard, as in example 11: Example 11: The characters' desires And then they do not want to fight anymore # because they are already in pain. 15 Eleven children (37% of the sample) made the intentions or desires of the characters heard. Of these children, 55% (6 children) were second graders, three were first graders and two kindergarteners. Intentions and desires were made heard in first and in second narratives by about the same number of children, respectively 72% and 83% of those who expressed them. 13 In the original: "La fille croit qu'il a fait exprès alors elle le repousse." 14 In the original: "le garçon il tombe sur une pierre # et il fait tomber la fille mais il a pas fait exprès." 15 In the original: "Et après ils veulent plus se bagarrer # parce qu'ils ont déjà mal."

The Emotions, Feelings and the Perceptions of the Characters
Finally, the voices of the characters were also made heard when children talked about the characters' emotional and perceptual states.
The characters' emotions and feelings were made heard through the expression of their contentment, discontent, of their being angry or crossed over the other character, as in example 12: Example 12: The character's emotions And then he is not happy because he has pushed him. 16 Twenty-one children (70% of the sample) made the inner emotions of the characters heard. They are 80% of the first and of the second graders, while only 50% of the kindergarteners did it. Emotions were made heard in first and in second narratives by about the same number of first and second graders, while kindergarteners made emotions heard more in the second than in the first narrative (80% vs 20% of the kindergarteners who expressed them).
As for the characters' feelings and perceptions, these mainly concern their hurting sensation (as in example 13) or what the characters see (example 14): Fifteen children (50% of the sample) made the inner sensations of the characters heard. Sixty percent of those were kindergarteners and first graders, while only 30% of the second graders did it. Inner feelings and perceptions were made heard in both first and in second narratives, but overall they were expressed more in the second than in the first narrative. 16 In the original: "Et après il est pas content parce qu'il l'a poussé." 17 In the original: "Et il montre la pierre parce que ça lui a fait mal." 18 In the original: "C'est deux garçons # ils se voient." All content of Bakhtiniana. Revista de Estudos do Discurso is licensed under a Creative Commons attribution-type CC-BY 4.0

A Special Case: 'to Say Hello,' 'to Greet Each Other'
Children's narration of the initial setting often included expressing that the two initial setting as ambiguous: they could be a way to make the characters' voices heard but they could also reflect the narrator's voice describing, from an outsider standpoint, the greeting event that takes place between the two characters. Consequently, in the quantitative analysis -to be described below in section 3.2 -these occurrences were not included in the count of the total occurrences where the characters' voices were made heard (i.e., all the other cases described under 3.1.3).

Making the Characters' Voices Heard: Quantitative Results
Children who made one or both of the characters' voices heard could do it at least once during their stories, but they could also use this procedure several times, making thus the characters' voices heard several times and in different ways throughout their narrative. Table 2 presents the total number of times in which the characters' voices were made heard by the children during the entire narrative, per grade/age and per narrative.
We included here all occurrences of the cases described under 3.  As it can be seen in Table 2, second graders made most of the voices heard in the narratives (56%), followed by first graders (27%) and then by kindergarteners (17% of the overall occurrences). Moreover, it can also be seen that the characters' voices are made heard much more in the second than in the first narrative (63% vs. 37%).
To corroborate these descriptive results, we performed repeated measures mixed ANOVA with one within-subject factorthe narrative, with two modalities: first and second narrativeand one between-subjects factorthe grade/age level, with three modalities: kindergarteners, first graders and second graders. The ANOVA shows a main effect for grade/age (F (2.27) = 6.52, p = 0.005), and for narratives (F (1.28) = 5.30, p = 0.029), with no significant interaction between the two factors (F (2.27) = 1.677, p = 0.20). The post-hoc comparisons show that second graders made the characters' voices heard more often than first graders (t = 2.62, p <.05) and kindergarteners (t = 3.46, p <.01), but that there was no significant difference in this respect between these two latter groups (t = 0.85, ns). Moreover, the post-hoc comparison shows that children made the characters' voices heard more in their second than in their first narrative (t=2.25, p <.05).

Discussion
Our results show that 5-to 8-year-old children can make different voices heard in their narratives using different discourse types such as direct and indirect discourse and the expression of the characters' internal states. Thus, early on in development, children can transpose and insert discourse types that they use in everyday conversations (according to Bakhtin, first order discourse) into the oral narrative of their constructed plots (whereby, according to Bakhtin, the first order discourse so becomes a second order complex discourse). The 5-to 6-year-old children made the characters' voices heard infrequently through their narratives and only few children could display them. The ability and necessity to use these discourse devices develop with age and with the supposed increasing cognitive and linguistic competences associated with them. Descriptive results, confirmed by statistical tests, show that 8-year-old second graders are better able than 5-to 7-year-old children to make the characters' voices heard in their narratives: Indeed, they use this device more frequently throughout their narratives and a greater number of children do that. Moreover, the older group of children makes, more often than the younger children, the thoughts and beliefs of the characters heard, which highlights their different points of view on the state of the fictitious reality.
Results also show that voices are made heard much more frequently in the second than in the first narrative. We may suppose that the conversational interaction in which children participated between the first and the second narrative facilitated the expression of this aspect. This result is in line with previous studies aiming at promoting children's explanatory and mind-related content using the same (VENEZIANO, 2010(VENEZIANO, , 2016VENEZIANO et al., , 2020 or other types of conversational procedures (SILVA & CAIN, 2019;SILVA et al., 2014). However, on the basis of the present study alone, this relation cannot be firmly established since here the effect of the conversational interaction on the polyphonic expression of the voices was not controlled for the effect that narrating the story a second time might have had independently. Nonetheless, earlier studies in which this control was performed, showed that the narratives of the control group did not present any significant differences in several measures of narrative content, while the narratives of the children who participated in the conversational interaction revealed the causal structure and the mind-related aspects of the story plot significantly more often in the post-SCI narratives than in their initially-produced ones VENEZIANO, 2016;VENEZIANO et al., 2020). It is interesting to note that the Bakhtinian notion of a chained dialogism -for Bakhtin ideally 'unfinalized' (BAKHTIN, 1984[1963) -is also at the heart of more recent conversational approaches and has been applied to describe the dynamics of early mother-child interaction and the implicit conversational rules that young children come to learn to become a competent speaker. These retroactive and proactive movements are so important in conversational analysis that they have been assumed to underlie even a simple adjacency pair such as the question-answer pair (a basic unit of conversation, according to SCHEGLOFF;SACKS, 1973). Within this framework, a question is most often asked on the basis of past experience and with a view on what the interlocutor is able or likely to answer and is itself oriented proactively towards the answer. In the same way, the answer relates retroactively to the question but is also oriented proactively to what the questioner might reply in turn. The double function of turns, which often remains implicit, is well illustrated by the explicitly marked turns referred to as 'turnabouts' (KAYE;CHARNEY, 1980) of which the clarification question is an example: the I would also like to suggest that interesting links can be found between the Bakhtinian notions about first and second order discourse (BAKHTIN, 1986(BAKHTIN, [1979) and Piaget's notion of 'vertical décalage' -or 'levels of functioning' -in developmental psychology (PIAGET, 1937(PIAGET, /1971. Vertical décalage concerns the idea that a given cognitive notion acquired at a certain level of functioning (for example, at the level of action, as it may be the case for the sensorimotor organization of space through the child's own displacements), is not transposed directly, when needs to use it arise, at a higher level of functioning such as the conceptual or metacognitive levels, but it needs to be reconstructed at these higher levels of thought and awareness (KARMILOFF-SMITH, 1992;PIAGET, 1974). With regard to discourse and the difference highlighted by Bakhtin, first order type of discourse can be considered to be readily available to children when they talk through their own subjectivity and often about their own experiences, intentions, feelings and beliefs, particularly for attaining their own goals. This first order 'know how' would however need to be reconstructed and reworked when higher levels of thinking and awareness are required. This can be the case when children, making the characters' voices heard in narratives, are required to deal with the subjectivity of others, talk about inner states not experienced by them at the moment and disconnected from their own aims, and integrate these and other related aspects into the construction of an overall narrative plot to be expressed through the specific formal and pragmatic constraints of the narrative genre (BRUNER, 1986;STEIN, 1988). The higher level of functioning required by the expression of the characters' voices in child-constructed narratives may account for the results obtained here showing that characters' voices are expressed significantly less by 5-to-7-year-old children than by second graders, a developmental trend that supports the hypothesis that a reconstruction of discourse types may indeed take place at a new level of conceptual awareness.
The results concerning the supposed facilitating role of the conversational exchange on children's expression of the characters' voices, and its ascertained facilitating role on the expression of the causal structure and the mind-related aspects of the narrative plot, lend support to the Bakhtinian notion that dialogues -particularly in the form of dialectical dialogism where different points of view across and within