Don't look any further: How a model's linguistic group membership limits exploration and discovery in preschool children

Abstract

From an early age, children are highly motivated to learn from others how ‘we’ do things in our cultural communities, and in a variety of domains (Csibra & Gergely, 2009; Schmidt & Tomasello, 2012; Tomasello, 2016). The human unique propensity for rapid acquisition of knowledge from others allows us to develop and maintain a vast array of belief systems, arbitrary conventions, cultural tools and complex institutions (Boyd et al., 2011; Gergely & Csibra, 2006; Legare, 2019; Schmidt & Rakoczy, 2019; Tomasello, 1996). However, when children observe others' behaviours, some actions they observe may be erroneous, purposefully deceptive or, in some cases, simply irrelevant. Think of human-made artefacts whose causal structure and intended function(s) are often opaque to naïve learners (Hernik & Csibra, 2009; Schmidt & Rakoczy, 2023). Chopsticks might be designed for eating, but they also have other potential functions: you can scratch your back with them, you can use them to tidy up your hair, you can use them to stir food or pick up out-of-reach objects. Observing all these different functions of the very same object, how do children decide which of these is the proper and culturally right way to use the artefact?

According to current theories of socio-cultural learning, human children have a propensity to engage in shared and collective forms (‘we-mode’) of cognition that allow for acquiring and constructing cultural-normative knowledge (Schmidt & Rakoczy, 2018, 2019; Tomasello, 2019). Children are also attuned to ostensive communicative signals (e.g., eye contact and infant-directed speech) displayed by others that indicate their pedagogical intention (Natural Pedagogy Theory, Csibra & Gergely, 2009, Csibra & Gergely, 2011). Ostensive communicative signals instantiate a presumption of relevance (Sperber & Wilson, 1986) and epistemic trust in their addressees (Gergely et al., 2007). Hence recognising the teaching intention of others helps young children to selectively attend to and acquire knowledge from trustworthy resources in a noisy social learning environment. Computational modelling work on teaching further indicates that a naïve learner expects a teacher to purposefully generate samples to increase their belief in the hypothesis that is to be learned: there is a basic epistemic expectation in naïve learners that a teacher would sample useful data for them (Bonawitz et al., 2011; Buchsbaum et al., 2011; Shafto & Goodman, 2008).

Whilst to date, there are several lines of empirical evidence showing that children indeed construe teachers as selectively demonstrating evidence to promote efficient and rapid information gain that allows for inductive inferences (e.g., Butler & Markman, 2012, 2014), previous research also uncovered that teaching contexts restrict children's exploratory behaviours (Bonawitz et al., 2011; Gweon et al., 2014). Yet an important question remains: Do children exercise context-sensitive epistemic vigilance against teachers taking into account their linguistic group membership and thus the social relevance of the demonstration? When it comes to the demonstration of an artefact function (i.e., a piece of normative cultural knowledge; Casler et al., 2009; Schmidt & Rakoczy, 2018), we argue that children will use explicit (linguistic) cues about the teacher's social group membership to rapidly acquire relevant cultural knowledge and thus restrict their exploratory behaviour on the artefact upon viewing a teacher from the same social group as them demonstrating the artefact function. In other words, we predict that if children observe a demonstrator speaking in their own native language showing a particular function on a novel artefact with multiple potential affordances, then they would be more likely to prioritise the demonstrated function in their own reenactment. On the other hand, when they see the same function demonstration of the artefact performed by a foreign language speaker they would be more likely to explore the artefact to realise other potential functions as they would not interpret the demonstration as indicating conventional, normative use of the object relevant to them.

Here, we focus on linguistic group membership as language-based social categories reliably mark the boundaries of different social communities, unlike race or gender (Baker, 2001; Giles & Billings, 2004; Henrich & Henrich, 2007; Moya & Henrich, 2016; Pietraszewski & Schwartz, 2014a; Pietraszewski & Schwartz, 2014b). Even if infants show sensitivity to the dimension of race or gender from a very young age as reflected by their different visual attention to these social category distinctions (Bar-Haim et al., 2006; Quinn et al., 2002), we argue that linguistic group membership plays a special role in children's learning from others. Before their first birthday, infants not only pay more attention to individuals who speak their own language but also critically they pay more attention to novel stimuli when these are introduced by their own language speakers than by foreign language speakers (Marno et al., 2016; Soley & Sebastián-Gallés, 2015). They also selectively prefer the objects or food items offered by their native speakers to those offered by foreign language speakers (Kinzler et al., 2007; Shutts et al., 2009). At 11 months, infants also represent an informant who speaks the same language as they do as someone who would provide them with relevant information as indicated by the greater anticipatory theta activity upon viewing a native language speaker versus a foreign language speaker (Begus et al., 2016). In toddlerhood, they show propensity to selectively imitate sub-efficient instrumental actions from demonstrators who speak the same language as them (Altınok et al., 2022; Buttelmann et al., 2013; Howard et al., 2015) and as they grow older, they keep these sub-efficient actions in their repertoire despite viewing efficient action alternatives in achieving the same goal as shown by a linguistic outgroup demonstrator (Altınok et al., 2020).

Critically, empirical evidence also indicates that the native language survives over race: when racial group membership of the agents 5-year-olds viewed was pitted against their accent, children privileged native accent over same-race (Kinzler & Dautel, 2012, see similar findings with infants Kinzler & Spelke, 2011). All these findings corroborate the evidence showing the privileged role linguistic group membership has in guiding young children's learning from others. After all, shared-language cue signals the potential of a common cultural background between two parties—between the teacher and the naïve learner, and also the teacher's capacity to transmit culturally relevant information. Observing a teacher who speaks the same language as themselves, the naïve learners could infer that they are not only similar in the language they speak but also in several other culturally significant domains that they have yet no expertise in (see also Oláh et al., 2019; Soley & Aldan, 2020) In this vein, shared language¹ can reliably guide naïve learners to the informants who could be the optimal teachers from whom they can learn culturally relevant knowledge.

This is directly demonstrated by studies on how children learn artefact functions—given that artefacts, human-made tools surrounding our daily lives, are fundamental to our material cultural knowledge. Upon being demonstrated the use of a novel artefact, children represent the demonstrated function as the target artefact's enduring function and avoid using the artefact for any other function (Casler & Kelemen, 2005). Children also appreciate that the proper function of artefacts—that is, how to use a human-made object—is not an idiosyncratic affair, but generalizable and normatively prescribed (Rakoczy & Schmidt, 2013; Schmidt & Tomasello, 2012). Upon being taught what the novel artefact is for, they protest (as a third-party observer) against an agent who uses that target artefact differently (Casler et al., 2009). Young children also show selective protest against ingroup rather than outgroup individuals in the context of violations of conventional norms on how to act on human-made artefacts (Schmidt et al., 2012). They can reason that artefacts ought to be used in conventionally agreed-upon ways despite appearing to afford other functions, and there is a proper way in which ‘we’ ought to use them (Chaigneau et al., 2016 on normative accounts of artefact functions; see also Schmidt & Rakoczy, 2019, 2023).

Previous research has shown that even toddlers understand that the conventional use of familiar objects is part of and linked to the broader shared knowledge. Two-year-olds expect the agent to be the source of a foreign language if they have previously seen the agent using an artefact unconventionally, but not if the agent has used it conventionally (Oláh et al., 2014). Toddlers also expect those who speak their language to use the same artefact for the same function, suggesting that they understand the social consensus behind the normative use of novel artefacts (Pető et al., 2021). They form representations of artefacts and their functions in inflexible ways when these are demonstrated by their native language speakers, as revealed by their persistent endorsement of tools that are presented by members of their own language group, even though the tool does not effectively achieve the intended purpose (Oláh et al., 2016). Around the age of 5, children also evaluate a novel tool as serving only one function but only when this tool use demonstration is delivered by their linguistic group member, a person who speaks the same language as they do, leading enduring mutually exclusive tool-function mappings (Pető et al., 2018).

Considering the accumulating evidence on the role of native language influencing the way children map the functions to novel artefacts, and how artefact use is rooted in cultural learning as the way they are used is shaped by conventional agreement, here we are using the classic paradigm of Bonawitz et al. (2011) to investigate how children limit their exploratory behaviour on a novel artefact following a demonstration by linguistic ingroup versus outgroup. The study has two aims. Our first aim is to conceptually replicate the original study of Bonawitz et al. (2011) in a different cultural setting, in [ANONYMIZED] sample of children. Secondly, we aim to extend the interpretation of the original findings by arguing that children's spontaneous restriction of their exploration of novel artefacts may be a consequence of their strong motivation to rapidly acquire relevant cultural knowledge from their linguistic ingroups.

Different lines of research suggest that young children have a tendency to over-attribute functions and norms of proper use to novel artefacts, even if there are alternative and equally feasible ways of how to use the artefact (promiscuous teleology and normativity; Casler & Kelemen, Casler & Kelemen, 2005; Kelemen, Kelemen, 1999, 2004; Schmidt et al., Schmidt et al., 2011, Schmidt et al., 2016). Considering this, we propose using an unfamiliar apparatus with multiple functions as it was in the original study of Bonawitz et al. (2011). We would argue that such apparatus mimics everyday artefacts that have one intended function but multiple possible uses (e.g., a fork is to eat with, but it can be used to brush one's hair or to pull something out from underneath the drawer) and thus it can be a suitable tool for investigating cultural learning. Thus, as in Bonawitz and her colleagues' study, children will view a novel artefact with four different functions. In the same-language demonstration, they will observe a female adult speaking the same language as themselves demonstrating one function of the artefact (henceforth ‘linguistic ingroup’ condition). In the foreign-language demonstration, they will observe the same function demonstration by another female adult who speaks a different language than them (‘linguistic outgroup’ condition). Importantly, the demonstrations will be nonverbal but incorporate ostensive signals to reveal the demonstrators' teaching intention. When the demonstration phase is done, children will be given the artefact, and we will measure children's exploratory behaviours on it. We predict that children will be fixating more on the demonstrated function of the artefact, thus limiting their exploratory behaviour and discovery when the demonstration is delivered by their linguistic ingroup teacher. We expect children to widely explore the artefact and also discover its non-demonstrated functions when the demonstration is delivered by a linguistic outgroup teacher.

In line with our main hypothesis, we predict that children will limit their explorative behaviour on the target artefact upon viewing the function demonstration by their linguistic ingroup. We expect this to be evident by the greater proportion of time children will spend on the demonstrated function in the linguistic ingroup condition than in the linguistic outgroup condition. We will carry out an independent samples t-test following the previous work of Gweon et al. (2014) comparing the proportion of time children will spend on the demonstrated function in two conditions. Critically, we predict that children in both conditions will use the artefact realising its demonstrated function at least once during the response phase. To test this, we will compare the number of children doing so in the two conditions using Fisher's exact test.

Moreover, we expect children to discover fewer target functions after the linguistic ingroup's demonstration than after the linguistic outgroup demonstration. We also predict that children in the linguistic ingroup condition will perform fewer unique actions on the artefact than children in the linguistic outgroup condition. Both the number of target functions children will discover and the unique actions they will perform on the artefact in two conditions will be contrasted by an independent samples t-test.

Given Bonawitz et al. (2011) findings, we also predict condition differences in the overall amount of time children will spend playing. The overall time children will spend interacting with the artefact in the linguistic ingroup condition will be shorter than in the linguistic outgroup condition. This will be again contrasted with independent samples t-tests. However, we do not predict that children will be paying differential attention to two demonstrations (as these will be nonverbal and identical except for the identity of the demonstrator in two conditions), nor do we predict that there will be any differences between the experimental conditions in the latency they act on the artefact. We will contrast the overt attention children display to the demonstration and the latency they act on the artefact in two conditions using independent samples t-tests.

When it comes to the first action and first function children would realise we expect no differences between the two conditions (see Altınok et al., 2020 for a similar pattern of findings in an over-imitation paradigm in children's imitative behaviour after linguistic ingroup and outgroup models). Whilst we anticipate that children in both conditions will first act on the artefact using the button to make it squeak, as demonstrated, it is also likely that children would be performing a different action or bringing about a different function than the one demonstrated as the first action or as the first function they will realise. Depending on the number of children who do so, we might also compare the number of children who faithfully adhere to the demonstrated action and to the function as the first action and the first function they perform between the experimental conditions using Fisher's exact test.

Regarding the baseline condition, our prediction is that the children in the baseline condition will explore the artefact similarly to those children in the linguistic outgroup condition, as will be evident in the overall playtime, in the number of target functions discovered, and in the number of unique actions performed on the artefact. We plan to run one-way ANOVAs for these variables between the three conditions using a Bonferroni correction for multiple comparisons. Following the original paper, we may also compare the three conditions using planned linear contrasts by applying the weights of 2, −1 and −1, respectively for the conditions of linguistic ingroup, linguistic outgroup and baseline for the outlined pattern of predictions explained above (namely for the overall play time, for the number of target functions discovered and for the number of unique actions performed).

We predict that children will be more likely to successfully bring about the target function in both experimental conditions than in the baseline condition. The number of children who successfully bring about the function (‘making the toy squeak’) will be contrasted by Fisher's exact test with a Bonferroni correction for multiple comparisons. When it comes to the first action and the first function children would realise, we also predict that children in the two experimental conditions will be prioritising the demonstration in their reenactments whilst we do not predict this for the baseline. Again, we will compare the number of children in three conditions using Fisher's exact test with a Bonferroni correction for multiple comparisons.

Since we aim to also have a fixed time-window coding (i.e., first 30 s) to match the duration of time children interact with the artefact across conditions, we will present the analyses for this time window separately in the Supplementary Materials for the number of unique actions children perform on the apparatus and the total number of target functions children will discover in three conditions.

Children show the propensity to assign functions to novel, unfamiliar objects from a very young age and rapidly form enduring tool-function mappings (Casler & Kelemen, 2005, 2007; Deborah et al., 2004). They also show sensitivity to the inherently cultural nature of artefact function, interpreting artefact use as reflecting conventional agreement behind it (Casler et al., 2009; German et al., 2007; Siegel & Callanan, 2007; see also Tomasello, 1999). Whilst pedagogical demonstrations limit their exploration of artefact functions, leading them to fixate more on the demonstrated function despite the artefact potentially serving other purposes (Bonawitz et al., 2011), it remains unclear whether relevant demonstrator characteristics, such as the language the demonstrator speaks, could also influence children's exploration behaviour on novel artefacts. Given that the same-language spoken is one direct social cue signalling that the teacher is likely to sample evidence that would be indicative of culturally relevant information worth acquiring, here with this proposed study, we predict that children will rely on their linguistic ingroup members for efficient information gain about artefact functions. When the demonstrator speaks the same language as children, they will be more likely to spend more time operating the novel artefact using the demonstrated function, showing limited exploratory behaviour on the artefact. When the demonstration is delivered by a foreign language speaker, we predict that children will be more likely to explore the artefact extensively.

If the data support our predictions, then the restrictive force the linguistic ingroup teacher demonstration exerts on children's exploration would suggest that children are making strong normative cultural inferences about the functions of artefacts (and more generally about how we should act in our cultural groups). However, if the linguistic group membership brings no additional interpretation of relevance about the demonstrated function then this planned study will still be valuable in conceptually replicating the original paradigm in a different cultural setting: given the nonverbal ostensive signals used by both demonstrators, children should be perceiving both demonstrations as relevant teaching episodes tailored for them and should limit their exploration after both teacher's demonstration independently of the language the demonstrators previously spoke. In this vein, we believe our baseline comparison to the planned experimental conditions will be critical in painting the full picture of whether or how children limit their exploratory behaviour in a cultural context outside of North America (but see Shneidman et al., 2016 for a study with Yucatec Mayan children) and expand our understanding of the ways cultural factors may influence preschool-aged children's understanding of artefact functions in general.

Nazlı Altınok: Conceptualization; data curation; formal analysis; investigation; methodology; project administration; writing – original draft. Rebeka Anna Zsoldos: Conceptualization; data curation; investigation; methodology; resources; writing – review and editing. Krisztina Andrási: Formal analysis; methodology; writing – review and editing. Ildikó Király: Conceptualization; funding acquisition; methodology; resources; supervision; writing – review and editing. Marco F. H. Schmidt: Conceptualization; funding acquisition; methodology; supervision; writing – review and editing.

The authors declare no competing interests.