2.1. Mentalistic definitions of teaching

Mentalistic approaches define teaching as behavior with the intent to facilitate learning in another (Pearson Reference Pearson1989, p. 63). This approach seeks to explain the design of mental mechanisms that make teaching possible, often in order to explain individual-level and species-level variation in teaching abilities. For example, Tomasello et al. (Reference Tomasello, Kruger and Ratner1993) argue that the establishment of joint attention and holding representations about the mind of the other are necessary prerequisites to teaching. It follows that teachers need theory of mind (ToM) to identify the need for teaching, to figure out what it is that they ought to teach, and to tailor the difficulty of the task to match the skill level of the pupil (Kruger & Tomasello Reference Kruger, Tomasello, Olson and Torrance1996). Elsewhere, researchers from this school of thought apply the same line of reasoning to social learning capacities, including imitation (for review, see Caldwell & Whiten Reference Caldwell and Whiten2002). In this view, the absence of theory of mind in nonhuman animals explains why only humans engage in “powerful forms of cultural learning” such as teaching and imitation (Tomasello Reference Tomasello1999).

Similarly, the mentalistic approach uses variation in theory of mind capacities to explain variation in human teaching capacities, at the individual level. For example, Ziv and Frye (Reference Ziv and Frye2004) argue that teaching is only possible when teachers and pupils both consciously recognize (a) intentionality and (b) knowledge differences between individuals. They cite a range of studies showing that children who have better success on theory of mind (ToM) tests also teach more or use more effective methods for teaching (Davis-Unger & Carlson Reference Davis-Unger and Carlson2008; Strauss et al. Reference Strauss, Ziv and Stein2002; Ziv & Frye Reference Ziv and Frye2004), and perform better as pupils (Wellman & Lagattuta Reference Wellman and Lagattuta2004). Olson and Bruner (Reference Olson, Bruner, Olson and Torrance1996) argue that without theory of mind there can be no ascription of ignorance, and therefore there will be no attempt to teach. According to Strauss et al. (Reference Strauss, Ziv and Stein2002), “in order to teach, one needs to know when knowledge, beliefs, skills, etc. are missing, incomplete, or distorted, as well as how people learn” (p. 1476). This goes beyond simply having the intent to teach, to having the intent to teach based upon a range of mental representations of others' mental states. This approach builds on Vygotsky's (Reference Vygotsky1978) concept of the “zone of proximal development,” the narrow range of learning just outside a child's developing abilities, within which a child can learn, building on his or her present competencies through scaffolding by knowledgeable others (Guberman & Greenfield Reference Guberman and Greenfield1991; for review, see Pelissier Reference Pelissier1991). In this view, only a teacher who identifies this proximal zone of development (through ToM) can understand that teaching is needed, and can then intend to teach.

2.2. Culture-based definitions of teaching

Culture-based definitions of teaching focus on teaching as it happens in formal classrooms in Western societies, in contrast to informal social learning in more “traditional” societies. This approach is most commonly used in sociocultural anthropology and cross-cultural psychology. The research goals include describing cross-cultural variation in the prevalence of teaching, with some work focused on direct comparison of formal schooling versus everyday learning in childhood (e.g., Maynard Reference Maynard2004; Paradise & Rogoff Reference Paradise and Rogoff2009). These definitions underlie nearly all of the cross-cultural data on teaching, which are crucial to our understanding of human variation in teaching behavior.

Researchers taking this approach contrast teaching with other forms of social and individual learning outside of the Western cultural context (e.g., Lancy Reference Lancy2010). Rather than using an explicit definition, researchers in this approach identify teaching from a shifting set of ostensive features, including: (1) the teacher intends to teach, (2) knowledge transmission is unidirectional (teacher to pupil), (3) pupils are passive recipients of knowledge who do not collaborate interactively with the teacher, (4) knowledge is communicated explicitly, often by verbal instruction, and (5) the activity is marked in some way and recognized as “teaching” by its participants. These criteria are not applied as a strict checklist, but as a set of characteristics arrived at inductively via contrasts between teaching as the ethnographer understands it, and other types of social learning. Note that in this approach, “social learning” is used in the colloquial sense to refer to learning that happens in a social context or interaction; therefore, it is not limited to the functionalist learning mechanisms that have evolved to facilitate the transfer of social information.

Types of social learning that are not considered “teaching” are described as natural, simple, informal, observational, practical learning, or guided instruction (Paradise & Rogoff Reference Paradise and Rogoff2009), in contrast with “formal” learning via teaching. In informal social learning, (a) learning takes place within an activity, the focus is on completing a task rather than on learning or teaching; (b) learners are often expected to observe rather than participate; and (c) the responsibility for attending, learning, and ending a learning period lies with the learner rather than the model (Gaskins & Paradise Reference Gaskins, Paradise, Lancy, Bock and Gaskins2010). In other words, learning is common through intent participation, defined as “listening-in” and “keen observation” (Rogoff et al. Reference Rogoff, Paradise, Arauz, Correa-Chávez and Angelillo2003), or through legitimate peripheral participation, in which learning a given activity is also inextricably linked to the learner's building a sense of shared identity with other practitioners (Lave & Wenger Reference Lave and Wenger1991). Learners must, therefore, identify with their models prior to learning about an activity.

In contrast, teaching is a “marked” event, such that a behavior is only teaching when participants label the activity as “teaching” instead of something like “line-fishing” or “weaving.” In this framework, “teaching” that is embedded in another activity is informal learning, rather than formal learning or teaching. It is so fully integrated into the learner's everyday experiences that they appear to learn through “a kind of osmosis” (Gaskins & Paradise Reference Gaskins, Paradise, Lancy, Bock and Gaskins2010, p. 87). This type of learning is thought to allow for automatic, collaborative, highly effective learning through experience, “with little dependence on coercion and explicit teaching” (Paradise & Rogoff Reference Paradise and Rogoff2009, p. 124). This fashion of learning works, researchers argue, because informal learners are always interested, such that learning happens without any need for teaching, and learning happens without fail (Paradise & Rogoff Reference Paradise and Rogoff2009; Spindler & Spindler Reference Spindler, Spindler, Trueba, Spindler and Spindler1989).

From this perspective, teaching is not a generally useful mechanism to make difficult learning easier. Instead, it is a means of forcing passive or uninterested pupils to learn. A number of researchers suggest this is unique to Western societies. Mead (Reference Mead and Middleton1970, p. 12), for example, argues that in the shift toward Western ways of learning, “the emphasis has shifted from learning to teaching, from the doing to the one who causes it to be done, from the spontaneity to coercion, from freedom to power,” and that the shift toward teaching means “the shift from the need for an individual to learn something which everyone agrees he would wish to know, to the will of some individual to teach something that it is not agreed that anyone has any desire to know” (p. 3). Rogoff et al. (Reference Rogoff, Paradise, Arauz, Correa-Chávez and Angelillo2003) contrast informal learning with the “factory-efficiency” model of teaching, where information transfer is unilateral and divorced from culturally valued activities. Rather than being collaborative social partners, in this factory-efficiency model, “[t]eachers were cast as technical workers who were supposed to insert information into the children, who were seen as receptacles of knowledge or skill” (p. 181).

Lancy and Grove (Reference Lancy, Grove, Lancy, Bock and Gaskins2010) equate teaching with explicit or abstract verbal instruction, which they claim is rare in non-Western societies. They discuss “the near total absence of children being taught (in the explanatory, didactic sense) by adults” (p. 145), and explain in an endnote that only three examples of this kind of teaching exist in the ethnographic literature on non-Western societies. Lancy and Grove also present several examples of guided learning that they do not consider “teaching,” but which would be considered teaching by the mentalistic definition mentioned above. For example, teaching is absent in canoe-building among the Warao, where “there is not much verbal instruction… but the father does correct the hand of his son and does teach him how to overcome the pain in his wrist from working with the adze” (Wilbert Reference Wilbert1976; quoted in Lancy & Grove Reference Lancy, Grove, Lancy, Bock and Gaskins2010, p. 161). Weaving apprenticeships in which “actual instruction begins in earnest when the master sits beside the boy at the loom and begins to demonstrate some simple patterns, which the novice copies” are not classed as “teaching,” because prior to instruction, pupils offset the cost of teaching through menial labor, and again there is little explicit verbal explanation (Lancy & Grove Reference Lancy, Grove, Lancy, Bock and Gaskins2010, p. 160).

2.3. Functionalist definitions of teaching

Functionalist definitions of teaching are grounded in the observable behavioral causes and outcomes of teaching behavior, instead of the teacher's motivational state or on local context. This approach aims to explain the evolutionary roots and adaptive design of teaching behavior, in part by demonstrating that “teaching” is not a uniquely human behavior. Functionalist approaches have similarly been applied to social learning more broadly, opening up the spectrum of social learning mechanisms beyond pure imitation, to include a range of mechanisms by which nonhuman (and human) animals may gather information through social transmission (for reviews, see Heyes & Galef Reference Heyes and Galef1996; Fragaszy & Perry Reference Fragaszy and Perry2003). The functionalist approach to teaching differs from the mentalistic one in that there is less focus given to mental mechanisms – a practical innovation for studying nonverbal species (see Caro & Hauser Reference Caro and Hauser1992). To this end, functionalist definitions build on a basic theoretical definition of teaching as behavior evolved to facilitate learning in others. To enable the empirically rigorous study of teaching in nonhuman animals, Caro and Hauser (Reference Caro and Hauser1992) established three operational criteria for identifying behavior that functions as teaching: (1) behavior is contingent on the presence of a naïve learner, (2) it provides no immediate benefit (or even generates a cost) for the teacher, and (3) it can be shown to facilitate learning in others.

Subsequent functionalist approaches to teaching have sought to update Caro and Hauser's (Reference Caro and Hauser1992) operational criteria, but not their conceptual definition. Hoppitt et al. (Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008) argue that the cost criterion does not reliably distinguish between behavior that evolved for teaching versus alternative functions, so it may lead to false positives. For example, food provisioning may result in offspring learning about parental food preferences, but probably evolved because it increases offspring survival rates (Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008). However, if food provisioning leads to learning by offspring at an immediate cost to parents, the Caro and Hauser (Reference Caro and Hauser1992) definition would categorize this as teaching. Without rejecting the usefulness of the Caro and Hauser operational definition, Hoppitt et al. (Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008) offer a revision of the broader theoretical definition of teaching that separates the process of learning from the behavior of teaching, such that teaching can be viewed as an accessory to other well-established social learning processes, and teaching does not have a single behavioral profile. For example, learning may happen through teaching by tolerance of close observation – where the adaptation of tolerance is on the part of the teacher – as compared with close observation, where the adaptation is on the part of the learner, who persists in observing. In this way, teaching can evolve alongside existing social learning processes, increasing learning accuracy through small modifications of a model-turned-teacher's behavior. The typology laid out by Hoppitt and colleagues distinguishes among teaching via local enhancement, observational conditioning, imitation, opportunity provisioning, and coaching. Teaching behavior is conceptually distinguishable from other social learning behavior because non-teaching is inadvertent, such that the adaptation for teaching is not on the part of the demonstrator.

Thornton and Raihani (Reference Thornton and Raihani2008) point out that the Caro and Hauser (Reference Caro and Hauser1992) operational definition can also lead to false negatives in the study of teaching in nonhuman animals (see also Thornton & McAuliffe Reference Thornton and McAuliffe2012; cf. Byrne & Rapaport Reference Byrne and Rapaport2011). Thornton and Raihani suggest that teaching be defined instead by “key characteristics” (p. 1825): “(1) it is a form of cooperative behavior with response-dependent fitness payoffs; (2) its function is to facilitate learning in others; and (3) it involves the coordinated interaction of a donor and a receiver of information.” The key characteristics do not stipulate behavioral guidelines for identifying these key characteristics in the field, but do predict that teaching will be observed where its utility in increasing the pupil's learning efficiency is the highest (e.g., it most improves on other learning mechanisms). A major function of these key characteristics is to distinguish teaching from other forms of social learning, communication, and social interactions such as “punishment,” which do not evolve because of the learning benefits they create for the learner.

Thornton and Riahani (2008) and Hoppitt et al (Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008) do not include human-specific forms of teaching within their frameworks. However, the novel functionalist definition of Csibra and Gergely (Reference Csibra and Gergely2009; Reference Csibra and Gergely2011) focuses on teaching as a uniquely human trait. Although like the mentalistic approaches, this definition describes a number of mental characteristics of teaching, it differs from mentalistic approaches (e.g., Kruger & Tomasello Reference Kruger, Tomasello, Olson and Torrance1996), in that its primary research question is on the functional or adaptive role that teaching serves, and on how teaching evolved (in humans). Csibra and Gergely (Reference Csibra and Gergely2009; Reference Csibra and Gergely2011) argue that the adaptive value of teaching is the driving causal force in the evolution of those mental capacities, rather than the capacities acting as constraints on the evolution of teaching. Since they argue that the adaptive function of teaching is to facilitate transfer of abstract, generalizable knowledge, their definition of teaching is narrower than other functionalist definitions. In Csibra and Gergely's approach to teaching, which they term natural pedagogy, the adaptation that makes knowledge transmission possible is a motivational system. This system evolves to facilitate the sharing of generalizable knowledge with others, and its adaptive function is to speed up the rate at which naïve (human) learners gain “reliable, new, and relevant information” about the world (Gergely et al. Reference Gergely, Egyed and Király2007, p. 140).

According to Csibra and Gergely (Reference Csibra, Gergely, Munakata and Johnson2006) this function does not necessarily require theory of mind capacities or a conscious intentionality to achieve, but rather any evolved psychology that would function to facilitate learning in others, for this particular informational domain. Thus, the teaching adaptation includes communicative capacities: behavioral markers, or ostensive cues, that highlight and mark the act of teaching for teacher and pupil. By this definition, teaching behavior requires: “(1) explicit manifestation of generalizable knowledge by an individual (the ‘teacher’), and (2) interpretation of this manifestation in terms of knowledge content by another individual (‘the learner’)” (Csibra & Gergely Reference Csibra, Gergely, Munakata and Johnson2006, p. 5). It does not require that teachers or pupils are consciously aware of these processes, unlike the mentalistic or culture-based definitions. By this definition, the teaching adaptation is both a special learning mechanism (because it requires an adaptation by both parties) and a special type of communication (because it evolved specifically to convey generalizable, abstract content such as “tigers are always dangerous,” as opposed to fleeting, concrete content about the current state of the world, such as “there is a tiger behind you”). Thus, it characterizes teaching as an adaptation to facilitate learning in others, for a particular type of content (generalizable knowledge) and kind of learning (knowledge transfer) in a single species (humans).

2.4. On the benefits of integrating existing approaches

The mentalistic, culture-based, and functionalist approaches are each tailor-made research programs designed to address questions about specific aspects of teaching. Mentalistic approaches investigate the psychological prerequisites for teaching, and use these factors to explain individual- and species-level variation in teaching capacities. Culture-based definitions focus on the features of Western-style classroom teaching, and highlight qualitative cross-cultural differences in the means by which children learn. Finally, functionalist definitions of teaching focus on the evolutionary costs and benefits of teaching behavior to better understand the evolutionary origins and functional design features of teaching behavior. The problem with the current division of labor is that each approach could make better use of the findings in the other two fields to address their own primary research questions. Recent attempts at dialogue across these approaches have, unfortunately, been stymied by definitional debates about what constitutes teaching (see, e.g., Csibra Reference Csibra2007; Thornton et al. Reference Thornton, Raihani and Radford2007). The systematic framework I propose subsequently aims to integrate core research questions from each approach, and synthesize extant data on teaching behaviors, as identified by these approaches.

Mentalistic definitions would benefit from integration with the functionalist and culture-based approach in several ways. First, integrating mentalistic questions about the psychological prerequisites for teaching with functionalist studies of teaching-like behavior in nonhuman animals would reshape mentalistic endeavors to ask why intentional and theory-of-mind based teaching might have evolved in humans alone. Some nonhuman animals regularly facilitate learning in conspecifics without human-like theory of mind capacities, so the answer cannot be that only humans can facilitate others' learning. Mentally representing the intent-to-teach has no adaptive value by itself; only the teaching behavior has adaptive value. It follows that as long as pupils act as if they expect to receive relevant information, and learn it effectively, an interaction can fulfill the adaptive function of teaching. What, then, is unique about intentional teaching? Mentalistic approaches integrated with a functionalist framework could make better use of cross-species comparative data by relating psychological capacities more closely to behavioral adaptations, and to the social learning problems these behaviors address. One possibility is that theory of mind helps to address only a subset of adaptive problems in which one individual facilitates learning in another, and that there are alternative psychological mechanisms that might also suffice (see sect. 5 for possibilities). This can generate new and interesting questions for the mentalistic approach to teaching. For example, are there socio-ecological conditions that are specific to human evolution, under which conscious intent-to-teach might provide adaptive benefits? Under what conditions is our species' apparently unique teaching psychology better-adapted than the simpler alternatives?

Integrating the mentalistic approach with the culture-based approach's emphasis on documenting cross-cultural variation in learning capacities could provide a broader range of human behaviors against which to test hypotheses about the psychology of intentional teaching. When humans facilitate others' learning, do they always use theory of mind, or are there other mental mechanisms that promote functionally equivalent behavior? Descriptively, what sorts of social learning problems does intentional teaching solve for humans, and how does this vary cross-culturally? Explaining the full range of variation in human teaching behavior may shed light on teaching's role in making cumulative cultural adaptation possible.

Culture-based approaches would benefit from integration with the mentalistic and functionalist approaches. While useful in generating rich descriptions of cross-cultural differences in how children learn, culture-based approaches could benefit from the integrative study of teaching in order to better explain how such cross-cultural variation emerges, why it persists, and why this type of variation is characteristic across human populations but atypical in nonhuman animals. By establishing a baseline of what is universal behavior across human societies, researchers could more precisely focus on aspects of teaching behavior that may be culture-specific, in contrast. Both the mentalistic and functionalist approaches assume that teaching behavior is driven by a species-typical psychology of teaching. Researchers adopting the culture-based approach, especially those conducting long-term ethnographic fieldwork, are in the unique position to evaluate the validity of these claims; however, this requires a common framework across the major approaches. This framework would provide a yardstick by which to compare cross-cultural variation quantitatively, strengthening the original research aims of the culture-based approach to the study of teaching.

Functionalist definitions of teaching are focused primarily on the evolutionary origins and adaptive consequences of teaching. Functionalist approaches are to some degree already the basis for productive comparative work. For example, one typology links kinds of teaching behavior to the forms of learner-driven social learning behavior from which teaching may be derived (Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008), and the earliest functionalist definition was an explicit attempt to enable cross-species comparisons on the basis of behavior rather than a teacher's mental state (Caro & Hauser Reference Caro and Hauser1992). Similar efforts have proven fruitful in the social learning literature more broadly (e.g., see Whiten [2011] on chimpanzee learning). However, the functionalist approach could benefit from integration with the culture-based and mentalistic approaches in several ways. Despite acknowledging exceptional levels of teaching in humans, existing functionalist frameworks do not explain the apparently unique qualities of human teaching within the same framework they use to study teaching in other species. This is akin to studying kin selection according to a framework that does not explain the behavior of eusocial insects. One exception is Csibra and Gergely's (Reference Csibra and Gergely2009; Reference Csibra and Gergely2011) explanation of the evolution of teaching as an adaptation; however, their approach defines teaching as a uniquely and universally human adaptation, while not explaining behaviors that function to facilitate learning in others for nonhuman animals.

If humans are abnormally prolific and sophisticated teachers in comparison to other animals, it is all the more important that any framework for understanding the evolution of teaching should be tested against the human case. This is not to say that functionalists should abandon their study species of choice, only that the frameworks used would benefit by integrating research on human teaching. The best data available on the range of human variation in teaching behavior come from ethnographers in the culture-based approach, and this evidence suggests that Western-style teaching is cross-culturally rare. Even by the broader functionalist definition, there is a lack of systematic empirical work documenting that teaching happens across all human societies. Only a few papers have recently begun to address the question at all, and none include direct cross-cultural comparisons, in part because there has so far been no comprehensive framework by which to collect comparative data (Hewlett et al. Reference Hewlett, Fouts, Boyette and Hewlett2011; Kline et al. Reference Kline, Boyd and Henrich2013). Despite this, functionalist researchers state that “[t]eaching is ubiquitous in human societies” (Thornton & McAuliffe Reference Thornton and McAuliffe2006, p. 227), and that “[i]ndeed, all human children must be taught” (Thornton & McAuliffe Reference Thornton and McAuliffe2012, p. e8). This disconnect is, again, most likely due to a difference in definitions: surely, humans in all societies facilitate learning in other humans – but only an integrative framework can allow us to evaluate human uniqueness in theoretically interesting ways, and with quantitative data.

Without a common framework across which human and nonhuman behavioral data can be compared, functionalist frameworks will continue to only hint at the explanation for uniquely human teaching behavior, and as a result may only be providing part of the evolutionary explanation for why teaching evolves in any species. It may be that functionalist researchers prefer to eschew ethnographic data because of its qualitative, often subjective nature. However, the range of variation in human teaching behavior documented by ethnographic descriptions can inform the development and evaluation of integrative functionalist frameworks (see sect. 4.2) and the richness of anthropological descriptions of human behavior can shape operational criteria, for example, in the development of an ethogram of human teaching behavior for cross-cultural application (Kline Reference Kline2013). A synthesis along these lines might also help to make sense of the theoretical and empirical rift between traditional functionalist approaches as applied to nonhuman animals, and Csibra and Gergely's (Reference Csibra, Gergely, Munakata and Johnson2006; Reference Csibra and Gergely2009; Reference Csibra and Gergely2011) functionalist approach to natural pedagogy as a human-specific adaptation for teaching. The typology proposed below (see sect. 3) will suggest that these two approaches are discussing two different teaching types, evolved to solve two different adaptive problems in the domain of social learning.

The functionalist approach would also benefit from further integration with the mentalistic approach, and such a synthesis need not detract from the ease with which functionalists can study teaching behavior in nonhuman animals, as it has in the past (see Caro & Hauser Reference Caro and Hauser1992). Integrating questions from the mentalistic approach with functionalist methodologies may generate and address novel research questions. For example, functionalist frameworks excel at outlining adaptive problems and the behavioral adaptations that might solve them. Mentalistic approaches work to enumerate the psychological mechanisms that produce behavioral consequences. Translating empirical research between approaches via a cohesive framework would, therefore, shed light on the socio-ecological circumstances under which a given psychology might be selected for. These psychological adaptations need not be established as criteria for identifying teaching, but rather could be used to generate novel hypotheses about the evolution of teaching psychology and the psychological underpinnings of cumulative cultural evolution. Investigating such hypotheses can shed light on the role of psychological constraints on the evolution of teaching and other social learning capacities, across species. Given the potentially important function of teaching as a means of faithful cultural transmission, this question is relevant to broader problems in the comparative study of the evolution of culture, and cumulative cultural adaptation.

3.1. Teaching types

Teaching potentially solves a number of different adaptive problems that arise in social learning, and that cannot be addressed by learner behavior alone. In order to learn, pupils must (a) attend to, and (b) have access to, relevant information, defined as novel information which is useful and/or in some way connected to the pupil's prior knowledge (Sperber & Wilson Reference Sperber and Wilson1995). This includes information that is accessed through experience, observation, or direct communication. When attention or relevant information (or both) would otherwise be lacking, teaching may make learning possible, more efficient, or more accurate. Teaching can therefore be thought of as a set of adaptive behaviors, perhaps derived from existing social learning behaviors (see Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008).

3.2. How this framework integrates across existing approaches

Much of the present debate over the form and distribution of teaching behaviors results from different schools of researchers focusing on different teaching adaptations as if these represent the entire category of “teaching.” The more fine-grained approach outlined here has the potential to restructure these debates by integrating the full range of documented variation in teaching adaptations across human and nonhuman animals. Frameworks with similar goals have proven fruitful for asocial and social learning mechanisms other than teaching (e.g., Heyes Reference Heyes1994), especially concerning the evolutionary origins of imitation and its purported uniqueness in humans (for reviews, see Caldwell & Whiten Reference Caldwell and Whiten2002; Tennie et al. Reference Tennie, Call and Tomasello2009; Tomasello et al. Reference Tomasello, Kruger and Ratner1993), leading to a more sophisticated understanding of why humans alone have cumulative culture (e.g., Enquist & Ghirlanda Reference Enquist and Ghirlanda2007; Tennie et al. Reference Tennie, Call and Tomasello2009). Because this framework covers all of the informational contexts in which teaching might evolve, it provides a comprehensive structure by which to categorize known (and future) teaching behaviors – a necessity for future comparative and anthropological studies of teaching.

This framework categorizes teaching types according to theoretically significant learning problems, rather than by behavioral profile, or based on known mechanisms of social learning. The latter is important, because the adaptive problems solved by teaching may not be a one-to-one match to those solved by already-catalogued social or asocial learning mechanisms. These learning problems are not content-specific, but instead describe informational contexts in which teaching could facilitate learning. While some mechanisms may have non-teaching analogues (e.g., Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008), this approach allows for straightforward comparative empirical applications, since learning problems can be analyzed in terms of these contexts rather than according to domains of content to be learned. This approach does not prevent researchers motivated by the research questions of mentalistic or culture-based approaches from continuing to focus such questions. Instead, it clarifies the interpretation of their research as it is relevant to the other approaches, and facilitates comparative and anthropological work. Subsequently, I demonstrate the utility of this framework by using it to reinterpret the existing data on the prevalence of teaching in humans and other animals, and to re-examine what evolutionary theory can tell us about why humans are such prolific and intensive teachers.

4.1. Broad claims and limited evidence

The prevalence of teaching in humans and other animals is a contentious issue. Some researchers make sweeping statements about the omnipresence of teaching cross-culturally, contrasting animal societies with “human societies, where teaching is common” (Thornton & Raihani Reference Thornton and Raihani2010, p. 297) or claiming that “it is (almost) incontrovertible that teaching is ubiquitous in human beings, which means that, with few exceptions, every person in every society has taught and has been taught by others” (Strauss et al. Reference Strauss, Ziv and Stein2002, p. 1476). In contrast, others claim teaching is uniquely Western (Gaskins & Paradise Reference Gaskins, Paradise, Lancy, Bock and Gaskins2010; Lancy & Grove Reference Lancy, Grove, Lancy, Bock and Gaskins2010; Paradise & Rogoff Reference Paradise and Rogoff2009; for review, see Hewlett et al. Reference Hewlett, Fouts, Boyette and Hewlett2011). In reality, the story may be more complicated. For example, my own field work in Fiji suggests that not only is teaching common there, contrary to previous claims (Hocart Reference Hocart1929; Lancy & Grove Reference Lancy, Grove, Lancy, Bock and Gaskins2010; Ritchie & Ritchie Reference Ritchie and Ritchie1979), but also, that the prevalence of teaching varies adaptively across domains and life stage (see Kline et al. Reference Kline, Boyd and Henrich2013). This suggests that a study including only some age groups or domains might seem to document an absence of teaching, especially if teaching is defined as Western-style formal verbal instruction, as it has been in the mentalistic approach.

Based on mentalistic or culture-based definitions of teaching, it seems obvious that animals lacking in the ability to consciously represent others' mental states cannot teach. On the other hand, animal behaviorists identify a number of species that display what they call teaching, based on observable behavior. Unlike much of the anthropological field research on teaching in non-Western human populations, these studies typically adhere to strict criteria enumerated in operational definitions (e.g., Caro & Hauser Reference Caro and Hauser1992), which may cause false negatives in the study of teaching (Thornton & Raihani Reference Thornton and Raihani2010). Even so, Csibra (Reference Csibra2007, p. 96) critiques established examples of teaching in nonhuman animals on the grounds that they “do not seem to be particularly good examples of the activity that, as humans, we would call ‘teaching’” and suggests further that these behaviors might instead be labeled “scaffolding,” or “charitable information donation.” The problem with these and other claims about teaching-in-general is that each is based on a definition of teaching that is aimed at a focused research question within particular disciplines, and disagreements often result when researchers have not settled on the same approach. The present taxonomy illuminates the distribution of different teaching behaviors across species and societies, and thereby facilitates comparative studies of the psychological adaptations they may entail, their evolutionary function and origin, and their cross-cultural prevalence. In the following two sections, I use qualitative and descriptive observations of teaching behavior in humans and nonhuman animals to illustrate how the framework might be applied. The examples given below are best viewed as candidate instances of teaching behavior, rather than as clear evidence of teaching behavior, in both humans and animals.

4.2. Teaching in humans

Here, I provide examples for each type of teaching distinguished in the present framework. This is not a comprehensive review of teaching in humans, but rather a collection of illustrative examples drawn from throughout the ethnographic literature on teaching in humans. In many cases, these same ethnographic examples have been used as examples of “non-teaching” under culture-based definitions. Ethnographers often describe these behaviors in terms of teachers' mental states, including teachers' expected outcomes. These mental states are sometimes included in the descriptions below. However, in this framework, mental states do not function to distinguish between teaching types, so these details should be regarded only as contextual information, not as diagnostic features of teaching behavior.

4.3. Teaching in other animals

As with the review of teaching in humans, the review of teaching in nonhuman animals is not meant to be exhaustive; therefore, some possible examples of teaching may not be included here. In addition, it should be noted that the examples offered below include quantitative studies, some of which adhere to Caro and Hauser's (Reference Caro and Hauser1992) criteria for documenting teaching, as well as qualitative observations and anecdotal evidence. I include the latter in order to be consistent across discussions of humans and nonhuman animals, and because anecdotal cases presented here may inspire excellent future quantitative studies.

Teaching by social tolerance can also be described as tolerance of information scrounging. High-tolerance behavior such as this has been documented in several nonhuman primate species, though it has not always been characterized as having evolved to facilitate learning in others. In vervets, yellow baboons, mantled howlers, and capuchins (see Rapaport & Ruiz-Miranda Reference Rapaport and Ruiz-Miranda2002), adults are highly tolerant of close observation and even intrusive touching by immatures. Unlike food scrounging among adults, social tolerance of interference by immatures cannot be explained by the potential cost of an ensuing fight (Blurton-Jones Reference Blurton Jones and Morris1967), because of the relative weakness and small size of immatures as compared to the adults they are observing – such that the costs of exclusion would predict the opposite pattern (exclude immatures, tolerate more formidable adults) from that predicted by teaching by social tolerance.

Teaching by opportunity provisioning has also been observed in the wild, especially among predators. Wild meerkats hunt and disable prey (Ewer Reference Ewer1969), then present it to meerkat pups. Careful quantitative work shows that adult meerkats disable the prey to different degrees depending on the developmental stage of the pups, as gauged by the pups' food calls. The pups then finish processing the prey and so develop the skills to manage the prey without being exposed to the risk of injury, for instance from scorpion stings (Thornton & McAuliffe Reference Thornton and McAuliffe2006). Domestic cat mothers also capture and recapture prey, allowing their offspring to interact with it (observations by Ewer Reference Ewer1969; followed by experimental work by Caro Reference Caro1980). This has also been observed in river otters (Liers Reference Liers1951), cheetahs (Kruuk & Turner Reference Kruuk and Turner1967), and tigers (Schaller Reference Schaller1967). Observational studies suggest that Orcas that are beached along with their offspring sometimes flip live prey toward their offspring, and may do the same in the ocean (Lopez & Lopez Reference Lopez and Lopez1985).

Teaching by stimulus or local enhancement has been described in animal skill-acquisition, and especially foraging, across species. In river otters, some mothers have been seen to drag their offspring into the water repeatedly until they learn to swim (Liers Reference Liers1951), and a similar behavior has been observed in the California sea lion (see Caro & Hauser Reference Caro and Hauser1992). Quantitative observational data show that long-tailed macaque mothers in one population intensify and exaggerate tool use (in this case, using human hair as dental floss) in the presence of their infants (Masataka et al. Reference Masataka, Koda, Urasopon and Watanabe2009). Golden lion tamarin adults have occasionally been observed to locate hidden prey and then emit a food-transfer call (typically used for transferring already-obtained food items), leaving the young to extract the prey on their own before eating it (Rapaport & Ruiz-Miranda Reference Rapaport and Ruiz-Miranda2002). Experimental work shows that domestic hens increase pecking intensity towards palatable food, when their chicks appear to be eating unpalatable food (Nicol & Pope Reference Nicol and Pope1996). Experiments also show that tandem-running ants that have located a food source will lead naïve ants to that food source and are responsive to feedback from the naïve ants (Franks & Richardson Reference Franks and Richardson2006; Richardson et al. Reference Richardson, Sleeman, McNamara, Houston and Franks2007). Quantitative observational data show that Atlantic spotted dolphins chase prey longer and make more referential body-orienting movements in the direction of prey when calves are present (Bender et al. Reference Bender, Herzing and Bjorklund2009). Observational work suggests that Orcas accompany and may lead offspring toward hunting grounds, and charge alongside their offspring in beaching attempts. The adult does not actually beach in these cases, but the offspring do (Lopez & Lopez Reference Lopez and Lopez1985). In some of these foraging examples, it would most likely be more efficient for adults to kill the prey and provision offspring directly – which suggests that these behaviors may function to facilitate learning in offspring.

Teaching by evaluative feedback has been observed across a number of nonhuman animal taxa. Female river otters have been seen to nip at the noses of their young if they run ahead, rather than following behind (Liers Reference Liers1951). In chimpanzees, mothers or older siblings may take unfamiliar food away from an infant or juvenile, restricting its diet, and captive macaques may prevent offspring from exploring novel stimuli, or stimuli known to be dangerous (see Caro & Hauser Reference Caro and Hauser1992). However, it is not clear how widespread these food-limiting behaviors are among nonhuman primates (Fairbanks Reference Fairbanks1975; see also Hikami et al. Reference Hikami, Hasegawa and Matsuzawa1990). Primate mothers of many species encourage their offspring to walk by setting them down and then looking/calling for them to join (e.g., chimps, gorillas, rhesus macaques, free-living yellow baboons, spider monkeys: see Caro & Hauser Reference Caro and Hauser1992; see also Maestripieri Reference Maestripieri1995; Reference Maestripieri1996). Experimental work shows that adult pied babblers emit a purr-call while provisioning nestlings, such that nestlings learn to associate the call with food (Raihani & Ridley Reference Raihani and Ridley2008); further work shows that the purr-call encourages fledglings to approach adults in both food-related and non-food-related circumstances (Raihani & Ridley Reference Raihani and Ridley2007). An alternative explanation for such examples is that giving feedback evolved to alter behavior in real time – for instance, to lower the costs of parental care by eliciting independent locomotion – rather than to facilitate learning. For this teaching type, it may often be necessary to distinguish between direct and indirect benefits to the potential teacher, in order to distinguish between teaching and other explanations for behavior. For example, while feedback from female cowbirds shapes male cowbird song development and content, there is no evidence that female fitness is dependent upon males modifying their song (Smith et al. Reference Smith, King and West2000). As a result, this kind of feedback resulting in behavioral modification is more likely to have evolved via intersexual selection than through selection on teaching behavior.

5.1. Pupil psychology

For teaching to have any adaptive consequences, the pupil's psychology must be sensitive to relevant information in the context of teaching. For teaching by social tolerance and by opportunity provisioning, pupils are already motivated to learn about relevant stimuli, and sensitive to relevant information (in both cases, the learning problem is access rather than interest). For these teaching types, teacher behaviors evolved to take advantage of pre-existing behaviors or motivations in pupils, and pupil behaviors and motivations would need little to no modification for teaching to be effective, at a minimum.

In teaching by local enhancement and by evaluative feedback, pupils must be sensitive to teacher cues and feedback that indicate relevant information. This is not a teaching-specific psychological adaptation in pupils, because the same kinds of cues are germane to other forms of social behavior. For example, social learners may already attend to others' demonstrated interest and to positive or negative social reinforcement. In all of these teaching types, the adaptation is on the part of the teacher, who displays these behaviors preferentially toward naïve learners, thereby facilitating learning.

In contrast, direct active teaching is characterized by manifestation of relevant information by the teacher to the pupil, as well as the pupil's interpretation of that information as generalizable. Because direct active teaching is a solution to the frame problem, it is expected to evolve only when pupils have very limited or very costly alternative individual or social learning mechanisms available. As a result, in direct active teaching, the pupil's only indication that information is relevant comes from the teacher.

There are a number of hypothetical pupil psychologies that would make this possible. First, in mentalistic approaches to teaching, pupils must theorize about the teacher's motivations and recognize that a teacher intends to teach. In this model of the pupil's psychology, a pupil who recognizes teaching may be receptive, or may reject a teacher's influence (see sect. 6.2 on skeptical pupils). This model also predicts that direct active teaching will be limited to species with theory of mind capacities, since a pupil's receptiveness to direct active teaching depends upon recognizing the teacher's intent.

Alternatively, teachers and pupils may have evolved a system of communication to reliably indicate the exchange of relevant, generalizable information. This suggests that direct active teaching is a result of teacher behavior creating new selective pressures on pupils. Such a process of coevolution may have created a novel teaching type in humans, derived from ancestral ones shared with nonhuman animals. Csibra and Gergely (Reference Csibra, Gergely, Munakata and Johnson2006) argue that such a special communication system is a defining feature of (human) teaching: When teachers mark their behavior using ostensive cues (e.g., eye gaze, use of pupil's name, soliciting joint attention, etc.), pupils act as if they expect to receive relevant information that should be learned. This does not necessarily require that teachers or pupils consciously theorize about the contents of the other's mind. It only requires that ostensive cues reliably result in learning receptivity on the part of the pupil. This is not a novel style of reasoning that applies only to teaching. For example, animals may respond to behavioral solicitations of play or grooming, without theorizing about the mind of the conspecific making the request. This suggests that direct active teaching need not be restricted to species with theory of mind capabilities, at least not from the perspective of the evolved psychology of the pupil.

5.2. Teacher psychology

For most teaching types, teachers can build upon existing pupil behaviors and motivations to shape pupil learning. However, teachers must still have psychological adaptations that make teaching more likely in situations in which it will be adaptive. This means that teachers must be selective in their efforts, teaching only those pupils from whose learning the teacher will also benefit (see sect. 6). All else equal, teachers must provide teaching when it is beneficial for the pupil – and that means providing information when it is relevant.

Again, theory of mind is one solution to the problem of relevance. For all teaching types, teachers could use theory of mind to assess what pupils know or do not know, and could subsequently provide information that is useful given the pupil's mental state. Theory of mind capacities may be necessary when the teacher's explicit proximate goal is to produce conceptual change in a pupil – shaping the pupil's understanding to match a predetermined concept as known by the teacher (e.g., Carey Reference Carey2000). In other cases, the need for theory of mind capacities is less clear. For example, theorizing about the pupil's mind may only be necessary when there is expected to be great heterogeneity in pupils' knowledge states, and when their knowledge status is not tied to any cues that would be reliably present (over an evolutionary time span).

In cases when pupils' learning needs are reliably tied to cues of age or maturity, teachers do not need to assess a particular pupil's mental state in order to provide relevant information. Instead, teachers may simply provide information that is typically relevant to pupils at a particular life stage or in a given situation. For instance, human adults in many societies interact with pre-verbal infants through modified infant-directed language (see review by Snow Reference Snow, Fletcher and MacWhinney1995) and gesture (Iverson et al. Reference Iverson, Capirci, Longobardi and Cristina Caselli1999). The proximate mechanism for this is not necessarily that adults theorize about the contents of infants' minds and strategically expose them to intensified emotional expression and to age-relevant vocabulary. Instead, it may be that the adult's own emotional reaction to young children, or their desire to see babies react, directly motivates behaviors such as motherese or “baby talk,” and other infant-directed behaviors. At the ultimate or evolutionary level, the motivational system matters very little, so long as children are exposed to relevant stimuli. The same sort of mechanism is evidently at work in meerkats. Adult meerkats provide pups with dangerous prey (scorpions), and depend primarily on vocal cues of immaturity to determine how much to disable the prey before provisioning (Thornton & McAuliffe Reference Thornton and McAuliffe2006).

Alternatively, teachers may provide relevant knowledge based on pupil-specific cues that depend on a metacognitive understanding of a behavior or task. Meta-representation for the purposes of teaching means that the teacher holds representations of her own knowledge, which is itself in the form of representations (Sperber Reference Sperber and Sperber2000). In common parlance, this is “knowing about” or having conscious knowledge of a task. Csibra and Gergely (Reference Csibra, Gergely, Munakata and Johnson2006) argue that teachers must meta-represent their own knowledge, analyze it, and selectively provide information to pupils that is relevant to the task at hand and adjusted to the learner's performance. As long as teachers meta-represent their own knowledge, they could use cues from pupil behavior or from the context to deduce which bits of their knowledge are relevant, without employing theory of mind (Csibra & Gergely Reference Csibra and Gergely2009; Reference Csibra and Gergely2011). When relevance is based on behavioral (as opposed to mental) assessment, the teacher does not distinguish between what the pupil knows and what the pupil does, and only provides feedback on aspects of the behavioral performance of a task.

There is some evidence for a role for metacognition in teaching. Conscious access to one's own knowledge may play a minor role in language-learning (Karmiloff-Smith Reference Karmiloff-Smith1985) and a more important role in what appears to be metacognition of metacognitive instruction – instruction on learning how to think about one's own metacognition in order to learn more effectively (Adams et al. Reference Adams, Treiman, Pressley, Damon, Sigel and Renninger1998; Gourgey Reference Gourgey1998). There is also evidence that metacognitive awareness of teaching may make children more sensitive and sophisticated teachers (Davis-Unger & Carlson Reference Davis-Unger and Carlson2008).

Intuitively, it may seem far easier to theorize about another's mind than to evaluate that person's task performance through metacognition and behavior-matching. However, it is important to realize that while theory of mind may be an “easy” task for humans, it is not necessarily the case for other animals. Both theory of mind and metacognition are hypothetically feasible solutions for the problem of providing relevant information from the perspective of the teacher. Alternatively, teaching behaviors that are produced reliably in response to a cue (of pupil age or maturity, for example) may not require either theory of mind or metacognition. This suggests that predictions about the role and prevalence of teaching across species that are based only on species differences in theory of mind or metacognition may not be robust. It may require more than considering cognitive constraints to explain why some teaching types are more common than others, some species are more avid teachers than others, or why direct active teaching seems to be so rare. Another factor to consider is the cooperative nature of teaching.

6.1. Teacher as donor

From the perspective of a teacher, teaching may be costly in terms of (a) the effort required for teaching, and (b) future competition with skilled or knowledgeable pupils. In addition, teachers can (sometimes) be thought of as information donors (see Thornton & Raihani [2008] for review), especially if that information is otherwise costly to acquire. If there are costs of teaching, then there must be some benefit to the teacher, in order for teaching to evolve. In teaching, these benefits are indirect and due to the pupils' gains in learning – in contrast to other forms of social reinforcement. Because the costs and benefits vary across different teaching types, the ways and degree to which benefits accrue to teachers should also be quite variable. For example, teachers may be related to their pupils, or pupils may be future allies or mates.

In addition, the relative costs and benefits of teaching will matter: teaching behaviors that require little teaching effort should be the most common, both across and within species. This may have greater predictive power than considering the cognitive capacities of a species, since more than one possible psychology could enable each teaching type (see sect. 5). A priori, lower-effort forms of teaching behavior include: teaching via social tolerance, teaching via evaluative feedback, and teaching via local or stimuli enhancement. Higher-effort forms of teaching include opportunity provisioning – because the teacher must modify the environment and transform a learning situation – and direct, active teaching, because teachers may completely cease their ongoing activities in order to verbally and/or gesturally communicate abstract, generalizable knowledge to pupils. Of course, these a priori assertions ought to be tested in terms of measurable costs, such as time, opportunity, or energetic costs to teachers. The brief review of teaching behavior given above does seem to support this prediction – all else equal, the less costly forms of teaching are apparently more common across species, and the more costly forms more rare. Observational data from Fiji show that relatedness predicts teaching between child–other dyads, and that high-cost teaching types are less common (Kline Reference Kline2013). Further, Fijians are more likely to report teaching as a means of learning for domains transmitted from parents to offspring, and for domains that are highly important to success in village life (Kline et al. Reference Kline, Boyd and Henrich2013).

If teachers are donors, they may recoup costs through indirect fitness benefits, by selectively teaching biological kin. Based on Hamilton's (Reference Hamilton1964) rule, teaching should evolve only when the cost to the pupil for independent learning, discounted by the teacher/pupil relatedness, is less than the cost to the teacher (Dewar Reference Dewar2002; Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008). This suggests that teaching should be common among closely related individuals, and in species that breed cooperatively (Burkart & van Schaik Reference Burkart and van Schaik2010; van Schaik & Burkart Reference van Schaik, Burkart, Kappeler and Silk2010). Alternatively, it may be common where background relatedness within groups is high, and teaching happens within groups while resource competition happens between them (Taylor Reference Taylor1992). Where resource competition is primarily within the groups, background relatedness would have no such effect (Taylor Reference Taylor1992). This is consistent with the literature review above: Relatively costly teaching via opportunity provisioning occurs in meerkats, which breed cooperatively (Thornton & McAuliffe Reference Thornton and McAuliffe2006); other forms of teaching happen in pied babblers and tandem-running ants (see sect. 4). High-cost forms of teaching, including direct active teaching, seem to be most common in humans, who have also been described as cooperative breeders (Hrdy Reference Hrdy, Salmon and Kennedy2008). In addition, humans often live in highly interdependent communities, which may shape caretaking, sharing, and cooperative behavior in the same ways as cooperative breeding does (Nettle et al. Reference Nettle, Panchanathan, Rai and Fiske2011; Roberts Reference Roberts2005; Tomasello et al. Reference Tomasello, Melis, Tennie, Wyman and Herrmann2012). Such high-effort teaching is rare or absent in closely related chimpanzees (cf. Boesch Reference Boesch1991), which are not cooperative breeders.

6.2. Pupil as skeptic

Teaching is, in one form or another, modification of one individual's behavior by another's influence. In this sense, the pupil is vulnerable to the same pitfalls that any social learner experiences. However, pupils may be especially at risk of manipulation by “cheating” teachers who deceive pupils for their own benefit, because the benefits to the pupil depend upon some measure of credulity. This necessary trade-off suggests that pupils should practice epistemic vigilance to guard against inaccurate or deceptive teaching that harms their fitness (Sperber et al. Reference Sperber, Clément, Heintz, Mascaro, Mercier, Origgi and Wilson2010). Even when teachers are the pupil's parents, pupil and teacher interests may differ – for example, parents often prefer to distribute resources across multiple offspring, while each individual offspring may prefer not to share with their siblings. This possibility should lead to counter-strategies used by skeptical pupils to resist manipulative teachers, while maintaining receptivity to credible teachers. Some of these strategies overlap with strategies for choosing good models for social learning more generally (for a review see Henrich & McElreath Reference Henrich and McElreath2003), while others are unique or especially applicable to teaching behavior. Pupil skepticism may be conscious and explicit, or it may be captured in learning heuristics. Heuristics may include biases based on social context of transmission, such as cues about the model or teacher's abilities or success (Boyd & Richerson Reference Boyd and Richerson1985; Henrich & McElreath Reference Henrich and McElreath2003). Alternatively, they may be based on the content of what is being learned, based on its plausibility or compatibility with what the pupil already knows (Boyd & Richerson Reference Boyd and Richerson1985; Henrich & McElreath Reference Henrich and McElreath2003; Sperber 1996).

The teaching types outlined above differ in their potential for manipulation of pupils by teachers, which will shape the type of skepticism expected to evolve in pupils. Teaching types that work by eliciting, directing, or limiting the learner's individual learning are generally least vulnerable to teacher abuse. This includes teaching via social tolerance, teaching by stimulus/local enhancement, and teaching by opportunity provisioning. In teaching by social tolerance, pupils are driven by internal motivations for social learning to seek access to another's behavior, and the “teaching” takes the form of tolerating the learner's interference. In order to achieve deception, a teacher would need to anticipate the learner's attention, and then perform the behavior in a way that would mislead the pupil to the advantage of the teacher. This type of deception would be cognitively demanding with limited effectiveness, especially if pupils have mental mechanisms for interpreting others' actions and goals, or for checking socially learned information against personal experience or observation. Similarly, teaching via stimulus or local enhancement is not prone to manipulation by selfish teachers, because teachers do not have direct influence on what pupils learn, but only facilitate pupils' individual learning by facilitating or directing their interest toward certain stimuli or sources of information. Teaching by opportunity provisioning functions similarly, as it depends on existing motives and nascent skill sets in the pupil. For example, when social predators such as meerkats provide their offspring with maimed prey, they have little effect on the pupil's motivations or the content of pupils' learning, but allow the learner to gain experience that they could not achieve safely on their own. Pure opportunity provisioning will always have these properties. Because humans often mix opportunity provisioning with other types of teaching (e.g., teaching via evaluative feedback, direct active teaching) in childhood chore assignments, opportunity provisioning contexts may be more prone to deception in humans – but this question is separate from the behavior of teaching via opportunity provisioning. For this category of teaching types, there is little benefit to pupils in being skeptical at all, as the risk of deception is very low, perhaps no greater than during any act of individual learning.

Teaching by evaluative feedback is more prone to abuse by selfish teachers, because its efficacy depends upon the teacher's modification of a pupil's behavior, in that effective teaching by evaluative feedback serves to change the likelihood of repeating a behavior that already exists in the pupil's repertoire. Detecting deception in teaching via evaluative feedback could be achieved if pupils are selectively receptive to encouragement or discouragement that is consistent with their intrinsic motivations, but only magnifies it (a content bias). Alternatively, pupils might be receptive to this type of teaching only from trustworthy teachers – for instance, close relatives who have little conflict of interest (a context bias). Teaching via evaluative feedback is unique because it is low-cost for teachers, but high-risk for pupils. As such, this teaching type should be relatively common in its use by teachers, but would be often accompanied by resistance and skepticism by pupils. Resistance may be less pronounced where content or context biases support the teacher's influence.

Direct active teaching is exceptionally vulnerable to exploitation by selfish teachers because it is the only teaching behavior that may include the direct communication of abstract knowledge, entirely independent of related behavioral observation. (Though note that gestural communication, or direct active teaching through demonstration, are both possible.) Since direct active teaching especially facilitates learning where pupils have little background knowledge to solve the frame problem, content biases will often be unhelpful. Instead, pupils may rely heavily on context biases dependent on characteristics of the teacher (Henrich & McElreath Reference Henrich and McElreath2003).

As mentioned, learners may be more likely to trust related teachers, because inclusive fitness benefits mean they are less likely to deceive pupils. Pupils can also assume that kin have especially relevant knowledge, because they share genes and are likely to experience similar environments (Boyd & Richerson Reference Boyd and Richerson1985; Reference Boyd, Richerson, Zentall and Galef1988). Some experimental evidence supports this conjecture: In the absence of information about accuracy, American 4–5-year-old children prefer their mothers' verbal testimony about object names and functions over a stranger's (Corriveau et al. Reference Corriveau, Harris, Meins, Ferneyhough, Arnott, Elliot, Liddle, Hearn, Vittorini and de Rosnay2009) – though when faced with accurate testimony from a stranger and inaccurate testimony from their mothers, only some children persist in favoring mother's testimony.

Without direct evidence, pupils may also choose to learn from teachers who are known to be accurate and reliable more broadly. Verifying teachers' claims against reality means the teacher is both knowledgeable and – at least in the past – truthful. This matters because learning from an honest-but-wrong teacher can be equally deleterious to a pupil's fitness as learning from a manipulative teacher. Studies with children in the United States suggest that 4-year-olds are capable of distinguishing teacher quality by tracking model accuracies, as well as preferring previously accurate models (Koenig et al. Reference Koenig, Clement and Harris2004) and knowledgeable models (Koenig & Harris Reference Koenig and Harris2005; Sabbagh & Baldwin Reference Sabbagh and Baldwin2001), and generally preferring accurate models when learning about novel object functions (Birch et al. Reference Birch, Vauthier and Bloom2008). Children also prefer accurate models over familiar ones when the models disagree (Corriveau & Harris Reference Corriveau and Harris2009), and will accept testimony from a presently well-informed model who was previously inaccurate due to being misinformed in the past – but they will not “trust” information from a model who was both well-informed and inaccurate in the past (Nurmsoo & Robinson Reference Nurmsoo and Robinson2009). Children also seem to be capable of learning according to statistical evidence drawn from multiple observations of similar events, and can integrate this with information about the model (Buchsbaum et al. Reference Buchsbaum, Gopnik, Griffiths and Shafto2011). (For an extensive review of the developmental psychology of children's learning in pedagogical contexts, see Skerry et al. Reference Skerry, Lambert, Powell and McAuliffe2013). These mechanisms only apply when the pupil can evaluate teacher accuracy or gauge a teacher's reputation for being trustworthy. Outside the laboratory, accuracy and knowledge levels may be difficult for naïve learners to assess, especially when pupils depend upon teachers for learning about dangerous or causally opaque domains.

When direct verification of teacher accuracy is not possible, pupils may use alternative strategies. Some of these strategies may have evolved for social learning more broadly (see Boyd & Richerson Reference Boyd and Richerson1985), but would also apply to cases of teaching. Simple content-based heuristics such as conformist bias (Henrich & Boyd Reference Henrich and Boyd1998) or prestige bias (Henrich & Gil-White Reference Henrich and Gil-White2001) could be particularly important for direct active teaching when the content of what is being taught is divorced from observable outcomes. For example, conformist bias explains the believability of social gossip, where information heard from more sources is rated as more credible (Hess and Hagen Reference Hess and Hagen2006). Similarly, children are more likely to believe testimony endorsed by two informants in competition with a single one (Corriveau et al. Reference Corriveau, Harris, Meins, Ferneyhough, Arnott, Elliot, Liddle, Hearn, Vittorini and de Rosnay2009). Learners may also be credulous of teachers who are respected by third parties (Boyd & Richerson Reference Boyd and Richerson1985). Prestige bias is thought to allow naïve learners to assess the quality of teachers based on the number of followers and extent of their deference (Henrich & Gil-White Reference Henrich and Gil-White2001). Consistent with this argument, 3- to 4-year-old Canadian children prefer to imitate models who were observed by others rather than models who were not watched (Chudek et al. Reference Chudek, Heller, Birch and Henrich2011).

Finally, in some cases of direct active teaching, neither direct verification of content to be learned, nor properties of the learning context may be available to pupils. In these conditions, when teachers profess beliefs in the abstract, a pupil may avoid manipulation by learning only from teachers who seem to act in accord with their own stated beliefs (Mascaro & Sperber Reference Mascaro and Sperber2009; Sperber et al. Reference Sperber, Clément, Heintz, Mascaro, Mercier, Origgi and Wilson2010). For example, knowledge of particularly rich fishing grounds was a closely kept secret among highly competitive fishermen in pre-contact Hawaii – so much so, that if a deep-sea fisherman discovered a new productive location, he might share this information only with his children, even at the threat of physical assault (Kamakau Reference Kamakau1976). In this situation, pupils should be skeptical of any information received; fishermen might misdirect others in order to protect their own preferred fishing grounds from becoming publicly known. Pupils should only believe fishermen who behave in ways that would be costly if their actual beliefs were in conflict with their stated beliefs – for instance, those who travel long distances to exploit particular fishing grounds. The same reasoning works for other behaviors, like not eating tasty but supposedly poisonous fruit. If such behaviors do not already exist, credibility-enhancing displays (CREDs) may culturally evolve along with beliefs, because they facilitate the spread of otherwise untestable beliefs (Henrich Reference Henrich2009). For direct active teaching, comparing a teacher's observed behavior against his or her professed belief may serve as a proxy for testing the validity of their beliefs. For example, imagine a number of teachers who claim that pleasurable behaviors like sex or alcohol consumption are spiritually harmful. They could be making these claims in earnest, or could be making false claims to manipulate others' behavior. If a teacher abstains from sex and alcohol, this can serve as evidence of a teacher's credibility, since it is only worth giving up pleasurable activities if one believes the negative consequences are real. (This is not evidence that a belief is true – only that the teacher believes it). On the other hand, if a teacher engages in sex and alcohol consumption, learners can infer that the teacher does not sincerely believe that these are harmful and against his or her best interest. The evolution of credibility-enhancing displays to highlight this kind of distinction may be especially prevalent for the teaching of beliefs that are difficult or dangerous to verify directly, including beliefs about disease transmission, poisonous foods, or the supernatural.