3.1. The gadget hypothesis

The alternative, cognitive gadget view suggests that cultural evolution, rather than genetic evolution, has played the dominant role in the emergence of human teaching. It proposes that the cognitive capacity to teach is assembled from ‘old parts’ during childhood through cultural learning – by observing and interacting with older children and adults as these experts exercise their more advanced capacity for teaching. The development of cultural learning is facilitated by a ‘genetic starter kit’, a set of genetic adaptations specific to the hominin line that make us peculiarly receptive to information from other agents, including enhanced social tolerance, attention and motivation, and expanded capacities for associative learning and executive controls. However, these genetically inherited resources are not specific to the development of teaching. Rather than being a programme for the development of teaching, of the kind implied by the instinct view, the starter kit functions to promote all kinds of social and cultural learning – for example, learning about foraging, tool use and language, as well as learning to teach (Heyes, Reference Heyes2018).

The cognitive ‘parts’ assembled by cultural learning into a capacity for teaching include: perceptual and attentional processes that enable the teacher to spot errors, i.e. differences between naive and expert performance; learning processes that track the novice's progress toward expert performance and the effectiveness of the teacher's verbal and non-verbal interventions in advancing that progress; executive processes allowing the teacher to switch interventions to advance the learner's progress, and to inhibit the impulse to take over; and mindreading, through which the teacher represents learners, not only in terms of what they do, but also in terms of what they know. Many of these parts are both phylogenetically and ontogenetically ‘old’; they are evolutionarily ancient and come online early in human development, long before children begin to teach. For example, reinforcement learning of the kind that tracks a novice's progress (Apps et al., 2005) is present in a broad range of vertebrate and invertebrate species (MacPhail, Reference MacPhail1982; Shettleworth, Reference Shettleworth2010) and early in human infancy (Siqueland & DeLucia, Reference Siqueland and DeLucia1969). Even mindreading, the most human-specific and late-developing part, is ‘old’ in the sense of having evolved – genetically or culturally (Heyes & Frith, Reference Heyes and Frith2014) – to fulfil a wide range of social functions. It is sometimes used for teaching, but it was not made for teaching. Even supporters of the instinct view, who see mindreading as the defining feature of teaching, do not argue that mindreading evolved specifically for teaching (Strauss & Ziv, Reference Strauss and Ziv2012).

The gadget perspective does not cast mindreading as the central or defining feature of teaching. Instead, motivated by research indicating cultural diversity in the mature practice and development of teaching (see below), we suggest that, during ontogeny, humans assemble different cognitive packages for teaching depending on their socio-cultural environment. The packages contain components of roughly the same kinds – perception, attention, learning, executive processes and mindreading – but the relative importance of these components, and the way they are configured vary widely depending on local ecology, what needs to be taught within a society, and cross-cultural contact, the extent to which a society has been exposed to the teaching gadgets of other cultures.

Social anthropologists have long argued that teaching practices associated with formal education, such as direct active teaching and frontal instruction, are culturally transmitted and can diffuse across cultures (Lancy, Reference Lancy2015; LeVine et al., Reference LeVine, LeVine, Schnell-Anzola, Rowe and Dexter2012). The gadget hypothesis develops this view by incorporating all forms of teaching and extending the focus from behaviour to cognition, proposing that there is cultural transmission, not only of ways of acting, but also of ways of thinking. It suggests that teaching hijacks cognitive mechanisms involved in bonding, coordination and communication more generally along with domain-general abilities such as reinforcement learning and executive function. The kind of teaching children receive shapes their teaching-related behaviour and cognition in a synthetic, plastic ontogenetic process.

The gadget hypothesis does not merely acknowledge that the development of teaching is modulated by cultural factors. That is true of the vast majority of, if not all, human traits. Rather, the gadget hypothesis suggests that, without exposure to teaching by others in their society, children would not learn to teach at all, or would teach in the same simple ways observed in other animals. These sadly deprived children would have the cognitive resources provided by the genetic starter kit and could use them to learn by observing others and by direct engagement with the inanimate world. However, without models of teaching behaviour, they would develop, at best, a rudimentary capacity to teach others – inferior to that observed in any contemporary human society.

Children probably learn to teach in the first instance through experiencing communicative acts that are embedded in everyday life. Depending on the cultural context, this may involve eye contact, pointing and infant-directed speech, or situations in which caregivers orient the child's body towards interactions between other people. Children's attentional bias towards other people and their prosocial emotions motivate them to join these interactions. As a result, children are increasingly exposed to teaching events. In many small-scale societies, this occurs when caregivers allow them to interact with objects they are using, give them commands or assign simple chores to them. The tasks that children participate in get more complex with age, which necessitates more costly forms of teaching such as detailed demonstrations. As children experience this process, they start to emulate and imitate what they see and hear, using abstract communication, demonstrations or task assignment when they interact with their peers, depending on what they have been exposed to. Children might receive situational feedback that helps them refine their own teaching and adjust to the abilities of others – for example, when they try to engage a younger sibling in difficult activities such as nut-cracking but are told that their sibling is too small to hold the hammer. As they get older, they eventually become receptive to the ethno-theories of teaching, knowledge and authority endorsed in their particular cultural environment (rather than developing a discrete, universal notion of teaching), although they probably do not fully grasp these until adolescence or even adulthood. Teaching thus relies on a genetically inherited starter kit made up of prosocial psychological mechanisms but also needs regular teaching interactions with caregivers and peers to get off the ground. This process probably involves a combination of multiple different mechanisms of cultural transmission (such as observation, imitation and teaching by evaluative feedback); our account does not imply that teaching is primarily learnt through active instruction (i.e. that caregivers take children aside to instruct them in how to teach).

We now present four lines of evidence which together support the cultural evolutionary account of teaching. The first, based on psychological experiments in industrialised societies, indicates that domain-general cognitive processes – old parts – are important for teaching. The second and third lines, based on naturalistic and experimental research in small-scale societies, indicate marked cross-cultural variation in mature teaching practice, and in the ontogeny of teaching among children. The fourth line indicates that teaching has been subject to cumulative cultural evolution, i.e. the gradual accumulation of functional changes across generations.

These lines of evidence are convergent. By itself, the evidence of cross-cultural variation (second and third lines) could be taken to show only that the development of a cognitive instinct for teaching is sensitive to a person's cultural context. However, when evidence of cultural variation is combined with evidence that teaching depends on a range of domain-general cognitive processes (first line) and has been subject to cumulative cultural evolution (fourth line), the gadget hypothesis – postulating a minimal genetic starter kit and a dominant role for cultural selection – becomes more plausible than the nativist alternative.

3.2. Teaching from old parts

The gadget account suggests that a mature cognitive capacity to teach is a culturally variable configuration of domain-general processes, in which mindreading does not consistently play a dominant role. In contrast, the instinct view places mindreading, the most socially specific component of the configuration, at centre stage. It defines teaching in relation to the mental capacity to represent a knowledge gap between the teacher and the learner (Strauss et al., Reference Strauss, Ziv and Stein2002; Ziv & Frye, Reference Ziv and Frye2004).

Recent studies of children in industrialised societies suggest that mindreading is less important for teaching than the instinct view suggests. For example, when 3–6-year-olds in the US were told that one puppet ‘knew how to play the game’ and another ‘had never seen the game before’, they answered correctly when asked which puppet was knowledgeable, but they were no more likely to teach the puppet with whom they had a knowledge gap, than the puppet that already ‘knew how to play the game’ (Corriveau et al., Reference Corriveau, Ronfard and Cui2018; Ronfard et al., Reference Ronfard, Poutre, Minigan, Atre, Wang, Strauss and Corriveau2015). In a more extreme case, 3–6-year-old children in the US positively preferred to impart information to a more knowledgeable individual than to a less knowledgeable individual, even when the former had stated plainly that they already knew the information to be taught (Kim et al., Reference Kim, Kalish, Weisman, Johnson and Shutts2016).

Other research of yet broader significance suggests that executive functions may be more important than mindreading for the development of teaching in industrialised societies (Davis-Unger & Carlson, Reference Davis-Unger and Carlson2008). Executive functions – including inhibition, attention, working memory and error detection – are domain-general processes that monitor and control all kinds of thought and action, from manual skills such as reaching and grasping, through emotion regulation, to higher cognitive functions such as financial reasoning. Davis-Unger and Carlson (Reference Davis-Unger and Carlson2008) used a range of standard tests to assess mindreading and executive function development in a sample of US children between 3.5 and 5.5 years of age. They then taught each child to play a ‘flower game’, a board game developed by Strauss et al. (Reference Strauss, Ziv and Stein2002), and asked them to teach an adult to play the game. Both mindreading (three tasks) and executive function (five tasks) composite scores predicted the children's teaching skill (e.g. time spent teaching, the range of strategies used, the number of rules taught and the number of errors detected), indicating that both contributed to teaching efficacy. Crucially, however, executive function was a better predictor than mindreading of the children's teaching skills. This suggests that US children depend more on domain-general executive processes than on mindreading when they are teaching. It also raises the possibility that other studies of Western children systematically underestimate the contribution of executive processes. For example, Ziv et al. (Reference Ziv, Solomon, Strauss and Frye2016) found that performance on standard tests of mindreading predicted teaching skill in 3–5-year-old Israeli children and interpreted this as evidence that mindreading is the crucial component of teaching. However, performance on tests of mindreading depends in part on executive functions, and Ziv et al. did not independently assess executive function. Therefore, it is possible that the teaching skill of the children in this study depended as much or more on executive function as on mindreading.

Turning from executive function to reinforcement leaning, a yet older and more domain-general process, brain imaging of Western adults indicates that reinforcement learning plays a significant role in teaching. This kind of associative learning, which proceeds via the detection and correction of prediction errors (the difference between an action outcome and a goal state; Rescorla & Wagner, Reference Rescorla, Wagner, Black and Prokasy1972), is a very ‘old part’. It contributes to learning all motor skills, is present in a wide range of species and comes online early in human infancy (MacPhail, Reference MacPhail1982; Siqueland & DeLucia, Reference Siqueland and DeLucia1969). Apps et al. (Reference Apps, Lesage and Ramnani2015) taught adults in the UK to press one of four keys when they saw each of four colours, and then asked them to teach this skill to another person by indicating whether each of the pupil's actions was right or wrong. They found that activity in the teacher's anterior cingulate cortex when they saw the pupil's response correlated with the prediction error of that response. Given that the anterior cingulate cortex signals prediction errors in ‘standard’ reinforcement learning, when an agent is learning from the outcomes of their own actions, this result suggests that ancient mechanisms of associative learning are recruited for teaching.

Building on the theory of ‘natural pedagogy’ (Csibra & Gergely, Reference Csibra and Gergely2006, Reference Csibra and Gergely2009, Reference Csibra and Gergely2011), some psychologists have proposed that, rather than co-opting ancient mechanisms of reinforcement learning, human teaching is an inherently communicative process that depends on mindreading by both teacher and pupil (Ho et al., Reference Ho, MacGlashan, Littman and Cushman2017, Reference Ho, Cushman, Littman and Austerweil2019; see also Heyes, Reference Heyes2019a). The theory of natural pedagogy suggests that human infants are born with a package of specific genetically inherited dispositions preparing them to be taught. These include a genetically inherited sensitivity to ostensive signals such as infant-directed speech and eye contact, a propensity to engage in turn-taking (contingent reactivity), and a tendency to follow eye movements (Csibra & Gergely, Reference Csibra and Gergely2009, Reference Csibra and Gergely2011). In contrast with the cognitive gadget account, which suggests that infants are receptive to teaching by virtue of domain-general prosocial adaptations, the theory of natural pedagogy claims that human infants are born to ‘expect to receive ostensive-referential communication from adults’ and to ‘expect to learn something generalisable in ostensive-referential contexts’ (Csibra & Gergely, Reference Csibra and Gergely2009: 151). However, close examination of the evidence for natural pedagogy suggests that it is not sound (Heyes, Reference Heyes2016a). For example, human infants are attentive to eye contact (Legare & Harris, Reference Legare and Harris2016), but they do not consistently prefer direct eye contact to averted gaze (Vecera & Johnson, Reference Vecera and Johnson1995; Farroni et al., Reference Farroni, Menon and Johnson2006), and attention to eye contact is found in many species that do not teach (Emery, Reference Emery2000). Similarly, a taste for contingent reactivity is found in many social species, including rats (Werner & Latane, Reference Werner and Latane1974) and precocial birds (Harshaw & Lickliter, Reference Harshaw and Lickliter2007; Harshaw et al., Reference Harshaw, Tourgeman and Lickliter2008), suggesting that it not a ‘new part’ built for teaching but an ‘old part’ built for social bonding (Heyes, Reference Heyes2016a).

Arguably, there is one ‘new part’ that is specific to human teaching: conceptual communication through language. Some have argued that the ability to communicate concepts may have been particularly beneficial for complex technical skills such as the construction of Acheulian hand axes (Chazan Reference Chazan2012; Gärdenfors & Högberg Reference Gärdenfors and Högberg2017). Conceptual teaching can overcome teleological opacity (by making the goals of knowledgeable individuals more explicit) and causal opacity (by illuminating the causal relationships between different aspects of a task) (Hernik & Gergely Reference Hernik and Gergely2015). Some have used this to advance the position that, in humans, teaching co-evolved with complex tool use and language (Laland Reference Laland2017; Lombao et al., Reference Lombao, Guardiola and Mosquera2017; van Schaik et al., Reference Van Schaik, Pradhan and Tennie2019; see also Battro Reference Battro2010). However, experiments suggest that the link between verbal teaching and stone tools may in fact be rather tenuous. While some have found that these skills are more efficiently transmitted through verbal communication (Morgan et al., Reference Morgan, Uomini, Rendell, Chouinard-Thuly, Street, Lewis and Laland2015), others observed no such benefit (Ohnuma et al., Reference Ohnuma, Aoki and Akazawa1997; Putt et al., Reference Putt, Woods and Franciscus2014; Cataldo et al., Reference Cataldo, Migliano and Vinicius2018). Accordingly, we believe that the cognitive abilities utilised in conceptual communication and language were probably selected because they are beneficial for solving a wide range of coordination problems, not specifically for teaching.

Studies conducted in industrialised societies have found that children and adults often depend on representation of their pupils’ mental states when teaching. This is established by the work of Strauss and colleagues reviewed in Section 2 and by Ho et al. (Reference Ho, Cushman, Littman and Austerweil2019). However, consistent with the gadget account of teaching, the research sampled in this section indicates that mindreading is less important, and ‘old parts’ – such as executive functions and reinforcement learning – are more important, than the cognitive instinct view implies. Furthermore, ‘new parts’ that are specific to humans – such as language – probably did not evolve specifically for teaching.

3.3. Cultural diversity in teaching

Children must work ‘under adult guidance or in collaboration with more capable peers’ (Vygotsky, Reference Vygotsky1978: 86) to develop abilities they cannot yet perform independently (the zone of proximal development, see Vygotsky, Reference Vygotsky1978). Teaching therefore occurs in all human cultures (Kline et al., Reference Kline, Boyd and Henrich2013; Kline, Reference Kline2015). However, unlike non-human animals – where teaching is often limited to key aspects of food acquisition, with a single target behaviour taught in each species (Premack, Reference Premack2007) – humans show numerous cross-cultural differences in what is taught and how the teaching is done.

In societies with formal education systems, children spend much of their time supervised and taught by professional teachers who are not related to them. Formal education emphasises frontal teaching, direct instruction and abstract verbal explanations (Scribner & Cole, Reference Scribner and Cole1973; Legare, Reference Legare2017, Reference Legare2019). Additionally, middle-class parents in Western countries often practice school-like interactions through intensive verbal communication with their children, often teaching skills (like walking) that do not need to be taught (Rogoff, Reference Rogoff2003; LeVine et al., Reference LeVine, LeVine, Dixon, Richman, Leiderman, Keefer and Brazelton1994; Morelli et al., Reference Morelli, Quinn, Chaudhary, Vicedo, Rosabal-Coto, Keller and Takada2018; Lancy, Reference Lancy2015; Legare, Reference Legare2017, Reference Legare2019). However, in some hunter–gatherer and mixed subsistence societies, adults rarely talk to young children. For example, among Tsimane forager–horticulturalists in lowland Bolivia, toddlers under 4 years of age receive less than a minute of ‘talking time’ per daylight hour (Cristia et al., Reference Cristia, Dupoux, Gurven and Stieglitz2019). In many small-scale societies, much learning is done by observation, imitation and pretend pay, which do not involve direct communication with an adult caregiver (Scribner & Cole, Reference Scribner and Cole1973; Gaskins & Paradise, Reference Gaskins, Paradise, Lancy, Bock and Gaskins2010; Legare, Reference Legare2017, Reference Legare2019). This is evident in African farmers and foragers (Hewlett et al., Reference Hewlett, Fouts, Boyette and Hewlett2011; Boyette, Reference Boyette, Terashima and Hewlett2016; Boyette & Hewlett, Reference Boyette and Hewlett2017; Lew-Levy & Boyette, Reference Lew-Levy and Boyette2018).

When they do teach, caregivers often employ less costly forms of teaching, for example by facilitating ‘learning by doing’: children learn skills as they participate in everyday activities, observing others and contributing as opportunities arise (Scribner & Cole, Reference Scribner and Cole1973; Rogoff, Reference Rogoff2003; Paradise & Rogoff, Reference Paradise and Rogoff2009; Legare, Reference Legare2017, Reference Legare2019). This approach emphasises ‘[l]earning “by osmosis”, picking up values, skills, and mannerisms in an incidental fashion through close involvement with a socializing agent’ (Rogoff, Reference Rogoff2003: 323). Caregivers primarily use speech to support the activities they are engaging their children in, for example by giving commands, instead of giving advance verbal instruction (Paradise & Rogoff, Reference Paradise and Rogoff2009; Scribner & Cole, Reference Scribner and Cole1973; Rogoff, Reference Rogoff2003). They also emphasise forms of non-verbal communication such as touch and gesture (Legare, Reference Legare2017, Reference Legare2019). This has been observed among rural farming populations such as the Gusii in Kenya, where mothers usually speak to young children in short commands (LeVine et al., Reference LeVine, LeVine, Dixon, Richman, Leiderman, Keefer and Brazelton1994). This preference is also evident in experimental studies. For example, in an experiment where caregivers taught a game to their child, US caregivers focused on direct active teaching and took on a leading role in interactions with the child (for example by guiding and praising them) (Clegg et al., Reference Clegg, Wen, DeBaylo, Alcott, Keltner and Legare2021). In contrast, ni-Vanuatu horticulturalists from the South Pacific relied more on shared interaction styles (dividing the task between the caregiver and the child), reflecting their expectation that children should learn from collaboration and observation (Clegg et al., Reference Clegg, Wen, DeBaylo, Alcott, Keltner and Legare2021).

Subtle forms of teaching also involve task assignment, where children are told to complete a simple chore (such as fetching a tool) (Rogoff, Reference Rogoff2003; Lew-Levy et al., Reference Lew-Levy, Crittenden, Boyette, Mabulla, Hewlett and Lamb2019). In a review of the ethnographic literature on hunting in hunter–gatherers and mixed-subsistence societies, MacDonald (Reference MacDonald2007) found that children usually start to gain experience with hunting weapons at a young age as adults and older children provide them with toy weapons to play with. Real weapons are increasingly provided as they grow older (MacDonald, Reference MacDonald2007). Adults let children accompany them on hunting trips and sometimes facilitate learning by focusing on easy prey and providing them with opportunities to make their first kills (MacDonald, Reference MacDonald2007). Among the Chabu in Ethiopia, children's fathers show them how to butcher meat; during hunting trips, adults respond to questions, show them how to perform important skills, tease them about mistakes and correct them (Dira & Hewlett, Reference Dira, Hewlett, Terashima and Hewlett2016). They also provide them with carcasses for practice (Dira & Hewlett, Reference Dira, Hewlett, Terashima and Hewlett2016). BaYaka caregivers use pointing, eye contact and child-directed speech to direct a child or infant's attention and to familiarise them with tools through negative feedback, demonstrations and opportunity scaffolding (providing the infant with an object to explore) (Hewlett & Roulette, Reference Hewlett and Roulette2016; Boyette & Hewlett, Reference Boyette and Hewlett2017). When hunter–gatherers and mixed subsistence societies use costly forms of teaching such as abstract verbal communication, this is usually done to communicate opaque knowledge such as social norms (Salali et al., Reference Salali, Chaudhary, Bouer, Thompson, Vinicius and Migliano2019) and complex skills such as spear hunting (Lew-Levy et al., Reference Lew-Levy, Bombjaková, Milks, Kiabiya Ntamboudila, Kline and Broesch2022). However, in many cases explanations of plant and animal knowledge still occur alongside hands-on practice and opportunity scaffolding (MacDonald, Reference MacDonald2007; Lew-Levy et al., Reference Lew-Levy, Bombjaková, Milks, Kiabiya Ntamboudila, Kline and Broesch2022).

There are also cross-cultural differences in who does the teaching. Formal education systems are guided by adult teachers. However, while adults play a prominent role in complex tasks such as spear hunting (Lew-Levy et al., Reference Lew-Levy, Bombjaková, Milks, Kiabiya Ntamboudila, Kline and Broesch2022), much hunter–gatherer teaching occurs between children. Hadza and BaYaka children spend much of their time in child-only play groups; as a result, they receive more teaching from other children than from adults (Lew-Levy et al., Reference Lew-Levy, Kissler, Boyette, Crittenden, Mabulla and Hewlett2020). There are also differences in the importance of kin. For example, siblings are more prominent teachers among the Hadza than among the BaYaka, which may reflect different settlement patterns (BaYaka children have more opportunities to interact with people outside the nuclear family; Lew-Levy et al., Reference Lew-Levy, Kissler, Boyette, Crittenden, Mabulla and Hewlett2020).

Cross-cultural variation in teaching practices may be driven by a range of factors such as exposure to formal education and differences in the kinds of skills children must learn, which are shaped by the subsistence system (Legare, Reference Legare2017, Reference Legare2019). These driving factors can be differentiated into ontogenetic and functional explanations, which complement each other (Tinbergen, Reference Tinbergen2005; Micheletti et al., Reference Micheletti, Brandl and Mace2022a, Reference Micheletti, Brandl, Zhang, Peacey and Maceb). Functionally, hands-on teaching as practised by foragers and subsistence agriculturalists may be especially effective at imparting the practical skills children need to survive in those environments. In contrast, abstract verbal communication prepares children for formal schooling, which enables them to gain resources through employment in a commercial economy. Numeracy and literacy may require training in a dedicated space separate from everyday life, which lends itself to the professionalisation of teaching. These contrasting ecologies thus favour the emergence of different teaching practices; children end up adopting the type of teaching they were exposed to during ontogeny.

This may help us understand why people from small-scale societies sometimes adopt more ‘schoolish’ communication styles after they have been exposed to formal education. Traditionally, Quechua caregivers in Peru teach mostly through non-verbal demonstrations and Guatemalan Maya have emphasised egalitarian ways of communicating with children (Chavajay & Rogoff, Reference Chavajay and Rogoff2002; Visscher, Reference Visscher2010). However, caregivers with more years of schooling use more verbal instruction and more hierarchical interaction styles, resembling practices they have encountered in school (Chavajay & Rogoff, Reference Chavajay and Rogoff2002; Visscher, Reference Visscher2010). In an origami folding task, ni-Vanuatu caregivers with more formal schooling also used a greater variety of teaching strategies (Boyette et al., Reference Boyette, Cebioglu and Broesch2022). During ontogeny, these caregivers may have adopted forms of instruction they encountered in the classroom and now apply them to their own children.

Other cross-cultural differences may come down to broader cultural values that do not necessarily map onto function. According to Corriveau et al. (Reference Corriveau, Ronfard and Cui2018), Chinese teachers often expect students to recognise if they do not understand a lesson and ask for help accordingly. In contrast, US teachers are expected to monitor the students’ understanding and explain information in multiple different ways, anticipating possible difficulties (Corriveau et al., Reference Corriveau, Ronfard and Cui2018). In the Central African Republic, verbal instruction is more common among Ngandu farmers than Aka hunter–gatherers, reflecting the more hierarchical values of Ngandu society (Boyette & Hewlett, Reference Boyette and Hewlett2017).

The evidence on teaching practices in adults suggests that there is a great deal of cross-cultural variation, especially between hunter–gatherers and subsistence societies on the one hand and industrialised societies with high levels of formal education on the other. It also suggests that exposure to formal schooling can shift teaching practices in small-scale societies, promoting the use of more verbal instruction and/or more diverse teaching strategies by caregivers. This is consistent with the idea that teaching strategies are culturally learned.

3.4. Cultural diversity in the development of teaching

We have seen that active facilitation of learning occurs in one form or another across cultures, although the way skills are taught differs between cultures. This supports a cultural evolutionary account of teaching, but it is also important to examine evidence that children adopt the teaching methods and communicative styles to which they have been exposed. If the ontogeny of teaching differs across cultures, and the input that children receive shapes how they themselves teach, the idea that teaching practices are culturally transmitted is strengthened.

Indeed, tentative findings, primarily from cross-cultural experimental research, suggest that the diverse cultural inputs that children receive, and the teaching interactions they are exposed to, shape children's own teaching and the way they think about the transmission of information. For example, children's early social experiences appear to shape their perspective-taking in teaching situations. A screen-based study in the Netherlands showed that 5-year-olds can adapt their teaching behaviour to the (perceived) age and competence of the learner, spending more time at relevant game locations when they believed that the learner was a toddler rather than a same-aged peer (Stolk et al., Reference Stolk, Hunnius, Bekkering and Toni2013). Intriguingly, children who had spent more time in nursery were better able to take the learner's age and ability into account (Stolk et al., Reference Stolk, Hunnius, Bekkering and Toni2013). This suggests that children learn to refine their teaching through regular social interactions with other children.

Additionally, cross-cultural differences in the way caregivers interact with and instruct children may shape the developmental trajectory of children's teaching (see Corriveau et al., Reference Corriveau, Ronfard and Cui2018). When teaching a game to a naive peer, children from rural Vanuatu, where early participation in productive activities is common in everyday life, have been found to use a participatory approach to teaching that emphasised learning-by-doing and short commands up until age 8 (Brandl et al., Reference Brandl, Emmott and Macein press). Abstract teaching with verbal explanations only became common from age 9 onwards (Brandl et al., Reference Brandl, Emmott and Macein press). These patterns differ from trajectories identified in industrialised societies, where abstract verbal instruction becomes the dominant strategy by age 5 (see Section 2). However, the pattern from Vanuatu resembles observations from Maya children. By age 4, Maya children looking after their younger siblings start to initiate teaching, but 3–5-year-olds mostly use commands instead of explanations or verbal feedback (Maynard, Reference Maynard2002). The latter only increase in 6–7-year-olds and their use expands further in 8–11-year-olds (Maynard, Reference Maynard2002).

Cross-cultural differences in the conceptualisation of teaching may also shape how children reason about teaching. When reflecting about how they taught their peer, nearly half of the 4–11-year-old ni-Vanuatu children who took part in Brandl et al. (Reference Brandl, Emmott and Macein press) focused on the content (what they taught); reflection about the transmission process and their communication strategies only increased from age 9 onwards. When reflecting about whether the learner mastered the game, most children at all ages treated the fact that they taught as evidence that learning occurred, with only a minority taking the learner's actual behaviour into account (Brandl et al., Reference Brandl, Emmott and Macein press). Again, this departs from patterns observed in industrialised societies with high levels of formal education, where children tend to emphasise their communication strategies and the learner's behaviour by age 5 (see Section 2). This may reflect local ethno-theories of knowledge production. The latter treat knowledge as an external reality that people can possess and exchange, and that exists independently of internal mental processes or specific communicative acts (Lindstrom, Reference Lindstrom1990). This ethno-theory also posits that learners acquire knowledge not through personal reflection, but through osmosis from legitimate authorities (Lindstrom, Reference Lindstrom1990).

Different conceptualisations of teaching have also been found in Chinese samples. Unlike Western children, who tend to state that a teacher should preferentially teach a naive learner (as opposed to an already knowledgeable one), Chinese children state that the already knowledgeable learner should be taught, presumably to improve their existing skills (unpublished study discussed in Wang et al., Reference Wang, Wang and Chui2017). This may be due to differences between Kantian and Confucian philosophy and their cultural downstream effects: while the former defines learning as the process of acquiring knowledge, the latter also views learning as the process of perfecting oneself (Wang et al., Reference Wang, Wang and Chui2017), which has influenced Chinese folk models of learning (Li, Reference Li2002). On the other hand, a separate study found that, compared with Germans, Japanese children showed a stronger tendency to preferentially teach a naive learner as opposed to a knowledgeable one, which may reflect an emphasis on interdependence and tending to others’ needs in Japanese children's family environments (Kim et al., Reference Kim, Paulus, Sodian, Itakura, Ueno, Senju and Proust2018).

While cross-cultural research on the ontogeny of teaching is still emerging, the work published so far suggests that there is cross-cultural variation in the ways children teach and reason about teaching. Differences have been found between industrialised and subsistence societies, but also between different industrialised societies with equally high levels of formal education but different value systems and folk models of learning and knowledge. This constitutes direct evidence that teaching is culturally learned.

3.5. Cumulative cultural evolution of teaching

We have reason to believe that teaching is culturally learnt not only because teaching practices are diverse and developmentally plastic, but also because they are subject to cumulative cultural evolution (where the functionality or efficacy of cultural traits improves across cultural ‘generations’). In other words, teaching practices are gradually refined across generations. This is aided by a process in which social groups deliberately organise cultural transmission in a dedicated social context that is separate from everyday life (Scribner & Cole, Reference Scribner and Cole1973). We might call this ‘intensive teaching’ (when it occurs in small-scale societies, Scribner and Cole (Reference Scribner and Cole1973) call it ‘noninstitutional formal education’). In small-scale societies, intensive teaching is mostly used for ceremonial and spiritual expertise transmitted during initiation rituals. When it was time for male initiations in Malakula (an island in Vanuatu), the fathers of the boys planted yam gardens for their sons and appointed some trusted men to guide them through the rituals (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 250). They constructed a ceremonial house where the boys went into seclusion before being circumcised (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 250). They were then ‘taught how to make and play the panpipes’ (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 253). Furthermore, ‘the novices are made to witness at night a number of remarkable performances, innumerable and ingenious “hoaxes”’ (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 253). Men dressed up as ghosts and spirits and frightened the boys (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 254). Afterwards, ‘however, the secret of the deception is revealed to them; they are taught exactly how the “hoax” was carried out, […] and they learn how to perform it themselves against the time when, as grown men, they will have to play their part in the incision ceremonies’ (Deacon & Wedgwood, Reference Deacon and Wedgwood1934: 255).

Intense, high-arousal rituals such as this one emphasise the experiential dimension of rituals and thus episodic memory over semantic knowledge, but they are still vehicles of cultural transmission (Whitehouse, Reference Whitehouse1992). The music sessions and the reveals after the hoaxes fit the criteria for teaching, but by removing the boys from their family environment, these initiations also carve out a social sphere dedicated to experiencing a ritual which the boys will in turn transmit to the next generation. This practice requires active coordination between multiple teachers and learners at the same time. It further requires that the novices’ families plan their resource use ahead of time, which is evident in the planting of dedicated yam gardens.

In large-scale societies, these dedicated spaces are found in modern education systems, but also in workshops run by medieval craftsmen and scholarly institutions such as monasteries. In economies characterised by increasing professional specialisation, these spaces allowed people not only to accumulate specialist skills, but also to engage in highly specialised methods of instruction that were transmitted along with these skills. For example, in some societies students learned how to paint by drawing live models under the guidance of a master instructor, or they learned about poetry by repeating a teacher's recital in unison. In this way, students learned not only how to paint or recite poetry, but also how to teach painting and poetry, by interacting with expert teachers from their culture. This was only possible because resource transfers, brokered by wealthy benefactors and political leaders, freed them from subsistence production. As the professions diversified in increasingly market-based economies, some aspects of cultural transmission became themselves subject to professional specialisation, resulting in the installation of full-time educators. Once this was accomplished, expertise about teaching could itself accumulate, which we can see in the field of didactics – where teaching is taught through direct instruction at tertiary institutions of formal education. Teaching therefore not only contributes to cumulative cultural evolution in the sense that it improves the transmission of complex skills (Caldwell et al., Reference Caldwell, Renner and Atkinson2017; Lucas et al., Reference Lucas, Kings, Whittle, Davey, Happé, Caldwell and Thornton2020); it is also subject to cumulative cultural evolution.