Although an implicit sense of self is already apparent before language acquisition (Rochat, 2010), it seems clear that, during infancy, the establishment of an explicit self-concept is primarily driven by the co-occurring growth in linguistic competence. This is because the comprehension and production of grammatical forms such as personal pronouns and verb conjugations require young children to draw an explicit distinction between the roles that can be assumed by themselves or by other people (Bates, 1990). In facts, recent research suggests that the mastery of person reference is closely related to the use of language to express one’s own and others’ mental states (Markova and Smolík, 2014). The underlying assumption is that, to correctly produce personal pronouns and mental state terms, children need to acquire a full awareness of their own desires, feelings, and thoughts (Rochat, 2010); furthermore, they must also understand that other people can have different or even false beliefs about a given situation (Doherty and Perner, 1998; Spataro et al., 2017). Put in other words, they need to acquire an explicit Theory of Mind (ToM) (Flavell, 2004; Wellman and Liu, 2004; Low and Perner, 2012).

Theory of Mind can be defined as the “ability to appreciate the existence of one’s own and others’ mental states (e.g., intentions, emotions, desires), and use them to explain and predict behaviors” (Longobardi et al., 2016b, p. 424; see also Flavell, 2004; Lecce et al., 2015). In the laboratory, this complex set of skills is typically examined by testing children’s understanding of first- and/or second-order false belief (Perner and Wimmer, 1985; Wellman and Liu, 2004). However, a crucial distinction has been drawn between passing false-belief tasks and spontaneously using one’s own ToM abilities during conversations and interactions with peers and adults (Meins et al., 2006; Grazzani and Ornaghi, 2012). Although the existence of a relation between these two domains has been disputed (Tager-Flusberg and Sullivan, 1995; Charman and Shmueli-Goetz, 1998; Meins et al., 2006; Longobardi et al., 2016b), the spontaneous production of mental state language (MSL: a specific type of language including perceptual, physiological, emotional, cognitive and moral terms) has been shown to forerun and predict the acquisition of ToM and, more generally, the emergence of meta-representational awareness (Beeghly et al., 1986; Bartsch and Wellman, 1995). In agreement, significant positive associations between ToM and MSL have been reported by studies examining the behavior of preschool children in a variety of interactive contexts (e.g., book reading; Dunn et al., 1991; Hughes and Dunn, 1998; Rollo and Sulla, 2016).

Importantly for the present purposes, the acquisition of an explicit ToM and the associated production of MSL are strongly influenced by the quality and the frequency of the social interactions to which children are exposed during their first years of life. More specifically, social interactionist approaches to cognitive development posit that the use of language during mother–child conversations has a central importance in promoting the development of mind-reading skills (Ruffman et al., 2002; Symons, 2004; Taumoepeau and Ruffman, 2006, 2008), since children learn to interpret the meaning of their experiences by participating in culturally shared activities (Nelson, 1996; Fivush et al., 2006; Bamberg, 2011). Children’s participation in communicative interactions is also critical to build their ability to construct and share stories with others – a conversational skill that emerges in the early childhood years and is strongly predictive of cognitive, literacy and socioemotional development, as well as school readiness and academic achievement (Deckner et al., 2006; Fivush et al., 2006; Cristofaro and Tamis-LeMonda, 2012). Parents can indeed scaffold the acquisition of children’s narrative ability by sharing personal experiences during family reminiscing, by telling stories during pretend play, or by reading stories from picture books (Whitehurst and Lonigan, 1998; Escobar et al., 2017).

In this broad context, several researchers have begun to examine the social origins of memory, by asking whether differences in maternal narrative styles can affect children’s memory for the narrated events (e.g., Engel, 1986, Unpublished; Fivush, 1994; Nelson, 1996; Nelson and Fivush, 2000; Lange and Carroll, 2003). Fivush et al. (2006), for instance, studied mother–child conversations about past events and found that the children of elaborative mothers, who engaged in richly detailed descriptions (e.g., “Memaw and Grandad came over, and Daddy cooked hamburgers out on the grill”: Haden, 1998) and open-ended questions (e.g., “What did we do at the zoo?” and “Who was there with us?”: Fivush et al., 2006), recalled significantly more information than the children of repetitive mothers (Fivush and Fromhoff, 1988; Reese et al., 1993; Haden et al., 2001). A similar distinction between narrative and paradigmatic interaction styles has been drawn by Tessler (1986, Unpublished) and Tessler and Nelson (1994), on the basis of previous work by Bruner (1986). According to these authors, narrative mothers tend to interpret events and actions in terms of intentions and mental states (e.g., “The frog is jumping on the stone because she’s afraid to drown”: Tessler and Nelson, 1994); in addition, they make frequent references to the autobiographical experiences of their children (e.g., “Do you remember when we went to the rugby match?”: Farrant and Reese, 2000); paradigmatic mothers, on the other hand, focus on labeling objects and specifying their relations with superordinate categories (e.g., “What kind of animal is that? A bird, right!”: Tessler and Nelson, 1994), or on repeating the exact content of a previous utterance (e.g., Mother asks, “We had fun, didn’t we?” and in her next turn repeats, “Yes, we had fun”: Haden, 1998). Although all parents produce both narrative and paradigmatic utterances when interacting with their children, most of them show a clear predominance of one style over the other (Lange and Carroll, 2003). Adopting this classification, Tessler and Nelson (1994) coded the spontaneous conversations occurring between mothers and their children during a museum visit: the results showed that the children of narrative mothers recalled the museum objects more accurately than the children of Paradigmatic mothers. Furthermore, in later studies, the use of a narrative style during parent–child interactions has been found to relate to other aspects of children’s cognitive development, including conscience development (Laible and Thompson, 2000), emotion recognition (Ontai and Thompson, 2002), socio-communicative competence (Tadić et al., 2013), and ToM (Welch-Ross, 1997; Ontai and Thompson, 2008).

Besides narrative styles, a large body of research point to the conclusion that the frequency with which mothers refer to MSL during conversations is positively and significantly associated with children’s performance in ToM tasks and the development of their psychological lexicon (Meins et al., 2002; Ruffman et al., 2002; de Rosnay et al., 2004; Slaughter et al., 2007; Ornaghi et al., 2011). Ruffman et al. (2002), for example, found that a composite measure of mothers’ mental state utterances (including the use of desire and think/know terms) at 3 years predicted unique variance in children’s ToM performance at 4 years. Likewise, LaBounty et al. (2008) reported that the frequency of mothers’ emotional explanations at 3.5 years predicted children’s concurrent emotion understanding, whereas the frequency of fathers’ explanations in terms of desires and emotions predicted children’s concurrent and later ToM at 5 years. As concerns the relation between mothers and children’s MSL, two classical studies by Taumoepeau and Ruffman (2006, 2008) showed that maternal references to desires at 15 months significantly predicted children’s later MSL and emotion task performance at 24 months. In addition, mothers’ talk about others’ thought and knowledge to 24-month-old children significantly predicted their MSL at 33 months (see also Furrow et al., 1992; Jenkins et al., 2003; Rudek and Haden, 2005; Symons et al., 2006; Pearson and Pillow, 2016).

Narratives provide a coherent structure for recounting the past to others, but also for representing the past to ourselves (Rollo and Benelli, 2003). Since language is the medium through which children learn the canonical narrative structures (Bruner, 1991, 1996), it is also the primary means through which adults “socialize” children (Fivush, 2014). In this context, mother–child conversations are essential instruments through which children learn the cultural norms that guide adult’s behaviors. As Fivush et al. (2006) wrote: “Children’s lives are organized such that the activities in which children are encouraged to engage are those that the culture deems important, and it is through participation in culturally mediated, socially structured activities that children learn the skills necessary to become competent members of their culture. With time, as children engage in adult-guided activities, they internalize the skills initially displayed by others and add them to their own intrapersonal repertoire” (Fivush et al., 2006, p. 1,568). It is then important to thoroughly investigate this continuous and circular process where adults narrate their own experiences and children learn to build their personal representations: how is the adults’ narrative toward children organized?

Parents and their children spend many time in joint picture-book reading. This activity conforms to the turn-taking structure of conversation and occurs in a repetitive sequence formerly described by Ninio and Bruner (1978). Furthermore, it is from the frequent exposure to this format that children acquire the capacity to create stories. Since picture-book reading is characterized by a unique three-way intersection of narration, face-to-face interaction and self-reference, it provides an optimal setting for exploring the processes of self-construction (Lonigan and Whitehurst, 1998). Indeed, numerous studies have investigated the relationship between maternal mental lexicon during picture-book reading and their children’s ToM (Ontai and Thompson, 2008). These studies examined both the variations in maternal styles and the consequences on children’s understanding of mind. Although the results have highlighted a close association between maternal mental lexicon and the development of children’s social comprehension, it is not yet clear which aspects of maternal language are essential to the establishment of this relation.

Within this broad context, the present study focused on maternal talk to 4- to 6-year-old children, to investigate: (a) whether narrative styles changed as a function of children’s age; and (b) whether different narrative styles were associated to quantitative differences in the use of mental state terms, including their referents. First, we wondered whether it was possible to discriminate between mothers who utilized Narrative versus Paradigmatic styles. Within our relatively small sample, we found mothers who, during the task of picture-book reading, were highly consistent in referring to the autobiographical experiences of their children and the mental states of the book characters, and mothers who were primarily focused on specifying the properties of objects and their relations. In this respect, the results of our study confirm the distinctions illustrated in the literature between elaborative versus repetitive styles (Fivush and Fromhoff, 1988), narrative versus paradigmatic styles (Tessler, 1986, Unpublished; Tessler and Nelson, 1994), or expressive versus referential styles (Nelson, 1973). Clearly, these categories should be considered as the extremes of a continuum, since all the mothers produced utterances of both types. More interestingly, we found that the mothers’ use of narrative styles varied as a function of children’s age, although the direction of this difference did not replicate the results reported by Reese et al. (1993). In facts, our data indicate that the mothers of 4-year-old children used primarily an expressive/narrative speech, more centered on interactions and shared experiences, whereas the mothers of 5- and 6-year-old children were more likely to adopt a referential/paradigmatic speech, more focused on objects. This modification parallels what happens during early child development. As children develop and refine their interaction skills, they move from a narrative to a paradigmatic way of systematizing the reality around them (Rollo, 2003, Unpublished); likewise, mothers change their way of conversing with their children.

The second aim of the present study was to identify the internal state words used by mothers during the picture-book reading. To this purpose, the transcripts were coded using a scheme illustrated by Camaioni et al. (1998), which differentiates between eight different MSL categories (physiological, perceptual, emotional negative, emotional positive, cognitive, volitional, communicative, and moral). Confirming previous studies that examined the MSL addressed to 24-month-old children (Longobardi et al., 2016a), we found that perceptual terms were used with the highest frequency, followed by emotional positive, cognitive, and volitional terms. Importantly, communicative terms were produced with the same frequency of perceptual terms, suggesting that their use might be particularly relevant for the creation of a narrative frame in which children begin to build mature representations of themselves and other people (Nelson and Fivush, 2004). Indeed, the terms “to say,” “to tell,” and “to call” were frequently used by the mothers of 4.05-year-old children.

Statistical analyses revealed that the mothers’ use of communicative terms varied as a function of children’s age, being more frequent in the mothers of younger than older children. No other difference reached the significance level, although the general trend was well in line with the available literature. For example, the use of perceptual terms was quite frequent with 4-year-old children but decreased sharply in the mothers of older children; in contrast, the use of cognitive and volitional terms was relatively limited during the interactions with the youngest group of children, but exhibited a strong increase in the mothers of 5- and 6-year-old children. These results nicely mirror the results observed in previous studies which focused on children’s production and comprehension of internal state terms (Bretherton and Beeghly, 1982; Shatz et al., 1983; Tardif and Wellman, 2000; Wellman and Liu, 2004; Kristen et al., 2012; Longobardi et al., 2014), and therefore confirm the causal role of maternal language (Meins et al., 2002; Ruffman et al., 2002; de Rosnay et al., 2004; Slaughter et al., 2007; Ornaghi et al., 2011). Furthermore, the mothers in our sample used emotional terms more frequently with girls than with boys – a finding which has been already observed in the literature (Lagattuta and Wellman, 2002).

A central aim of our research was to verify whether mothers classified as Paradigmatic or Narrative differed with respect to the use of MSL. In contrast with our original hypotheses, the present findings suggest that conversational styles and MSL are rather independent: indeed, Narrative and Paradigmatic mothers used the eight categories of MSL to the same degree, with one noteworthy exception: that Narrative mothers produced communicative terms more frequently than Paradigmatic mothers. This is interesting because the frequent use of verbs such as “tell” and “say” might facilitate children’s understanding of tensed complements, which in turn “is a precursor and possible prerequisite of successful false-belief performance” (e.g., de Villiers and Pyers, 2002, p. 1,037). In agreement, studies have suggested that mothers’ production of communicative (and more generally, cognitive) verbs might be closely related to children’s understanding of false belief, representational change and the appearance-reality distinction (Le Sourn-Bissaoui and Deleau, 2001; Adrián et al., 2007).

Finally, we wanted to verify whether mothers referred internal state words to “others” (including their children and the book characters) more than to themselves. In this respect, correlational analyses showed that the more the mothers referred mental state lexicon to the book characters, the less they referred it to the children or to themselves, indicating a dialog between two different entities: the fictional characters of the story and real persons – the mothers and their children. Assuming that during the book-reading interaction there is a “match” among these three components (Reese et al., 2003; Fletcher and Reese, 2005), the social interaction established between these elements seems to move from the book and its “concrete characters” to the children and the mothers. This was confirmed by the results of the MANOVA, which showed that the attributions of MSL to the book characters were predominant in 4.05-year-old children; this suggests that mothers can tailor their mentalistic input to the abilities of younger children: in facts, the book characters provide a concrete anchoring to verbally expressed mentalistic content. In other words, the mothers of younger children use the book to objectify an internal reality, which later will be directly attributed to the child. Moreover, they adapt the language produced in shared narrative contexts to the age of their children, in order to guide them toward a gradual awareness of their inner world (Rollo, 2003, Unpublished).

In summary, the present study examined the characteristics of maternal talk during narrative interactions with their children, with a particular focus on the different use of MSL by Narrative and Paradigmatic mothers. We learned that: (a) maternal narrative styles changed as a function of children’s age, such that the mothers of 4-year-old children were more Narrative than Paradigmatic, whereas the mothers of 5- and 6-year-old children exhibited the opposite pattern (i.e., they were more Paradigmatic than Narrative); (b) the use of communicative terms decreased with children’s age, and was more frequent in Narrative than in Paradigmatic mothers; and (c) mothers adapted the referents of their MSL to the children’s age, such that internal state terms were first referred to the book characters, then to the child themselves, and lastly to the dyad.

These conclusions should be evaluated with caution, because the present study has several limitations. First, the sample size was relatively small, and this factor limited the possibility to observe significant differences between the three groups of children. A second limitation was that we focused on mothers’ speech, and therefore no data about children’s language was collected. This means that we could not assess whether changes in maternal MSL were partially due to simultaneous changes in the internal and non-internal language produced by children. Finally, no attempt was made to investigate children’s performance in false belief tasks, although previous research indicates a close link between children’s ToM and mothers’ MSL (Furrow et al., 1992; Jenkins et al., 2003; Rudek and Haden, 2005; Symons et al., 2006; Pearson and Pillow, 2016). Even with these limitations in mind, we believe that our results may represent a starting point for future research.

Longitudinal studies should be conducted to delineate any parallel change in children’s social and emotional development. Moreover, interventions can be planned in cases in which the narrative interactions between mothers and children are limited in frequency or dysfunctional (Tompkins and Farrar, 2011). In this respect, the use of language intervention practices is now well established in a variety of conditions, including children with language impairments (Cirrin and Gillam, 2008), children coming from economically disadvantaged families (Cain et al., 2005), and bilingual/immigrant children (Pirchio et al., 2015, 2017). This growing body of research has begun to document the fact that training mothers to talk elaboratively about past events can have positive effects on children’s understanding of mind (Peterson et al., 1999; Reese and Newcombe, 2007; Reese et al., 2010a,b; Taumoepeau and Reese, 2013). More generally, interventions that induced caregivers, teachers, and educators to increase the frequency of book-reading interactions have reported significant benefits to children’s oral language and their emergent psychological lexicon (Whitehurst et al., 1999; Aram and Biron, 2004; Aram et al., 2013; Grazzani et al., 2016). The present study adds to this literature by suggesting that, in addition to enhancing the overall quantity of elaborations produced by caregivers, training interventions should also focus on fostering the use of specific categories of mental state terms (communicative, volitional, and cognitive) and increasing the references to agents other than the book characters.

This study was carried out in accordance with the recommendations of AIP guidelines, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the local ethical committee.

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer SP declared a shared affiliation, with no collaboration, with the authors PS and EL to the handling Editor.