The study of social learning and culture in humans and nonhuman animals has expanded exponentially. The significance of this work for many disciplines covered by Behavioral and Brain Sciences has become increasingly apparent, with implications from evolutionary biology at large (Whiten Reference Whiten2017a; Whiten et al. Reference Whiten, Hinde, Stringer and Laland2011) to human nature (Henrich Reference Henrich2015; Laland Reference Laland2017). The Précis and Heyes (Reference Heyes2018) offer a theory of deeper penetrations of human culture into our minds, shaping, perhaps even creating, major aspects of cognition. This potentially opens up exciting new landscapes for culturally focused interdisciplinary research.
The basic idea, however, appears not as shiny and new as Heyes implies. It's now 20 years since Tomasello's (Reference Tomasello1999) The Cultural Origins of Human Cognition (perhaps tellingly mis-cited as 2009 in Heyes’ book and circulated précis), whose title seems perfectly apt for Heyes’ argument. Heyes argues that her approach differs from Henrich's (Reference Henrich2015) in defining cultural learning “by ostension: by pointing at putative examples” (Heyes Reference Heyes2018, p. 88), but Tomasello et al. (Reference Tomasello, Kruger and Ratner1993a) did this already in proposing three examples, including imitation and instructed learning via mindreading, now echoed by two of Heyes’ four examplars. Still, Heyes’ overall thesis is stimulating in exploring such ideas much further.
Much in the book begs questions that I hope Heyes may answer in her Response. First, I see no explicit definition of what counts as Heyes’ central concept of a cognitive gadget. Can she offer a definition, in a sentence or two? The metaphorical contrast between grist and mills appears clear, but how this maps to cognition is not. Heyes (Reference Heyes2018) notes that cultural evolutionary theory covers “large-scale conceptual structures such as fairy tales, systems of religious belief, and scientific theories,” but to her these are “cognitive grist” rather than mills (p. 36). Yet the religious beliefs of, say, an ancient Mayan or Egyptian would surely shape “not just what they think but how they think it” (Précis, sect. 1, para. 2) in radically different ways to each other and to a contemporary atheist, so what makes these not cognitive gadgets? Saying the latter are “neurocognitive” does not seem to help, for don't we nowadays assume that everything cognitive is instantiated in the brain and hence “neurocognitive”?
A related difficulty in what “counts” as a gadget arises in relation to testing selectionist evolutionary theories. I agree with Heyes that memetics has foundered in trying to unitize and count memes (her example is “heaven” [Reference Heyes2018, p. 38]: does this count as one meme? Then what about subsidiaries like angels?). Heyes suggests that by contrast, cognitive mechanisms are “unitized” by cognitive science, making them (as gadgets) more tractable for testing Darwinian selectionist theories of cultural evolution. But don't the same problems arise? Taking the example of language, does a bilingual person have one gadget, or two? Or many? Developmental changes raise further questions paralleling those concerning angels and heaven: For example, when a child adds irregular verbs to their language competence, is that adding a gadget (mini-gadget?) to their larger linguistic skill (macro-gadget?)? If not, why not? Imitation and mindreading also change, bit by bit, throughout development, challenging unitization. None of this is to disparage Darwinian analyses of cultural phenomena, of course. Colleagues and I, noting that Darwin knew nothing of genes, set aside the difficulties of genes’ purported equivalent of memes and simply returned to the principles set out by Darwin, revealing their operation in human (Mesoudi et al. Reference Mesoudi, Whiten and Laland2004) and animal (Whiten Reference Whiten2017a; Reference Whiten2019b) cultural evolution.
Heyes’ mission is to identify what creates distinctively human cognition, proposing that cultural learning supplies the answer. But if other animals’ lives are shaped by their cultural inheritance (Whitehead & Rendell [Reference Whitehead and Rendell2015, p. 6], for example, conclude from their comprehensive survey that “Culture is a major part of what the whales are”; also see Whiten Reference Whiten2017b; Whiten & van de Waal Reference Whiten and van de Waal2018, for primates), then presumably their cultures too are predicted to shape their cognition. Gruber et al. (Reference Gruber, Muller, Strimling, Wrangham and Zuberbuhler2009) offered the novel affordance of honey-filled holes in logs to wild chimpanzees from communities that either did (Kanyawara) or did not (Budongo) have stick-tool use in their cultural repertoires, and found that although the Kanyawara apes applied sticks to efficiently extract honey, Budongo individuals applied just their habitual leaf-sponging technology, with less success. The authors’ conclusion, encapsulated in their title, was that chimpanzees “rely on their cultural knowledge” to solve such novel problems. If this interpretation is correct, would it translate in Heyes’ terms as the Kanyawara chimpanzees having a “cognitive gadget” like imagining modifying stick-tool use appropriate to solving a novel problem – a cultural cognitive gadget the Budongo chimpanzees lack? If not, why not?
This chimpanzee study of Gruber et al., contrasting two populations, is relevant to Heyes’ proposals for dissecting the effects on cognitive competences/gadgets of “nature, nurture, culture” (Précis, sect. 2). Developmental ethologists like Hinde (Reference Hinde1970) long ago recognized that one cannot easily dissect these at the level of the individual because these factors interact in complex ways from conception to death; yet when it comes to differences – between individuals, populations, or species – such causes can be empirically distinguished. Taking Heyes’ example of literacy, one can show that culture is crucial to the emergence of literacy because literacy does not appear in societies that lack the requisite cultural background. However, this is different from claiming, as Heyes appears to do, that literacy is a cognitive gadget in human individuals, created by cultural evolution alone, because literacy arose only a few thousand years ago. We know from contemporary hunter-gatherer studies that hunters are impressively skilled in reading, in signs in the dirt, a narrative of recent events (“a kudu with a lame calf travelled north around an hour ago”) (Wannenburgh et al. Reference Wannenburgh, Johnson and Bannister1979). The signs are not images of kudu; they are abstract signs that have meaning for readers of them, as printed words do for us. Hunting is known to have an ancient ancestry (Whiten & Erdal Reference Whiten and Erdal2012) during which genetic changes could be associated with such a capacity, so this might well underlie our ability to culturally acquire literacy. A hominin lacking an evolutionary history of hunting and gathering should thus struggle to acquire literacy. This is not practicable to test, of course, but doesn't the possibility mean it is premature to conclude that literacy is any “purely culturally created” competence?
A somewhat pedantic aside: I am puzzled that the term “teleosemantic” has emerged and been adopted by Heyes. It seems to mean “teleonomic,” the term introduced by Pittendrigh (Reference Pittendrigh, Roe and Simpson1958) to contrast with “teleogical,” the latter meaning guided by purpose, the former distinguishing the mere appearance of purpose, for reasons such as evolution by natural selection. We say such biological phenomena have evolved to serve certain fitness-supporting functions, rather than to fulfil anyone's purposes. The term teleonomic has been widely used by well-known evolutionary biologists like Mayr and G. C. Williams and philosophers of biology like Nagel and Hull. In the same spirit Lorenz (Reference Lorenz1966) talked of the ways in which both genome and learning function to “acquire and store information on the environment” (p. 8). If teleosemantic is indeed only a synonym for teleonomic, good scholarship suggests settling on the latter and abandoning the other term.
Relating to such distinctions, Heyes states that “‘Cultural learning’ is a subset of social learning … specialized for cultural evolution” (Reference Heyes2018, pp. 86, 89 [my italics]). But surely cultural transmission functions only to benefit either a receiver, like an observational learner, or a transmitter engaged in some form of teaching; its function is not so futuristically oriented as to be about creating cultural evolution – except perhaps for vocational inventors. But Heyes is not focused on such inventors. Elsewhere Heyes more correctly says “specialized for cultural inheritance” (e.g., Précis, Fig. 4 [my italics]), which avoids the impression that every imitator or mindreader is out to create cultural evolutionary change; should that not always be the terminology used?
Turning to the first of the two of Heyes’ four cognitive gadget exemplars I have room to discuss, I agree that social learning biases appear widespread among animals and not only in humans (Kendal et al. Reference Kendal, Boogert, Rendell, Laland, Webster and Jones2018; Price et al. Reference Price, Wood and Whiten2017). And they can be complex; Bono et al. (Reference Bono, Whiten, van Schaik, Krutzen, Eichenberger, Schnider and van de Waal2018), for example, showed that interactions between as many as three biases may interact to predict wild vervet monkeys’ social learning: the sex of the observer, the sex of the potential model, and the relative payoff gained by the latter. I disagree with Heyes that such effects have “barely appeared on the radar of cognitive scientists” (sect. 5, para. 2). It depends on who counts as a cognitive scientist, of course, but a considerable research literature has now delineated a diversity of biases in nonhuman animals, and in children and adult humans (Kendal et al. Reference Kendal, Boogert, Rendell, Laland, Webster and Jones2018; Price et al. Reference Price, Wood and Whiten2017).
Heyes proposes that human distinctiveness in this domain is defined by “explicit, metacognitive rules” about the biases in play. But the only examples of such rules offered are verbalizations. If “explicit” translates only as “verbalized,” the terms “explicit” and “metacognitive” appear disappointingly redundant. The key question is how explicit metacognitive rules could be recognized in nonhuman animals in order to test hypotheses about their presence or absence in nonlinguistic beings.
Finally, I turn to imitation, which Heyes (Reference Heyes2018, p. 116) describes as the “Lamborghini” of social learning. But does the restricted sense of bodily copying that Heyes favours really play a key role in cultural transmission? I agree this is plausible particularly for gestures, including the distinctive human realm of rituals (Whiten Reference Whiten2019a); Clay & Tennie (Reference Clay and Tennie2018), for example, found that children tended to overimitate causally irrelevant hand gestures made while solving an object manipulation task, while bonobos ignored them. But I suggest that the majority of skills a child acquires through observational copying, such as how to make an unfamiliar tool do its job, do not rely on fidelity of bodily copying, which plays little part in the now more than 50 reports of “overimitation” (Hoehl et al. Reference Hoehl, Keupp, Schleihauf, McGuigan, Buttelmann and Whiten2019). Is there any evidence to the contrary? More generally, what is the empirical evidence for the oft-repeated assertion that cumulative culture relies on high-fidelity copying – especially the bodily imitation on which Heyes’ model focuses?
In that model, Heyes likewise seems overenthusiastic about the role that adults imitating a toddler can play in building a child's imitative capacity “from scratch” using domain-general associative learning. Caregivers may sometimes imitate infants’ facial expressions in face-to-face interactions, but is there any evidence they routinely imitate toddlers’ limb and other bodily movements? Is Heyes really suggesting that the boy copying “clasp hands behind back” developed the ability to imitate this because often in the past he did this or similar actions, and his parents copied him? And how could looking in mirrors, or synchronous activities, deliver this example? The same goes for chimpanzees and orangutans, able in “‘do-as-I-do” tests to copy novel test items like “touch back of head” (Call Reference Call2001; Custance et al. Reference Custance, Whiten and Bard1995), that they surely have not learned because others copied them doing this? And what of avian imitation of bodily actions, like using foot versus beak (Heyes & Saggerson Reference Heyes and Saggerson2002; Zentall et al. Reference Zentall, Sutton and Sherburne1996)? As I remarked in a critique following Heyes’ initial promotion of the ASL model (Whiten Reference Whiten, Hurley and Chater2005), there is a more general problem here too. Most of what a parent does cannot match what their infant is doing – they are attending to feeding, changing nappies, cooking, and so on – so for the infant to learn about matching, there would have to be some specific signal indicating “now, here is my rare perceptual match to what you just did.” I think no such signals are known. Moreover, bodily imitation is not “correlated” in the sense of being synchronous anyway; imitation follows a model's acts.
So does the underlying process of imitation, from perception to matching action, remain a black box? Well yes; we remain ignorant of how the brain does it and how it comes to do so. Similarly, a humanoid robot that can achieve the whole process, globally, is yet to be created? It would be illuminating to see if such a robot could build the ability if programmed only with ASL.
The study of social learning and culture in humans and nonhuman animals has expanded exponentially. The significance of this work for many disciplines covered by Behavioral and Brain Sciences has become increasingly apparent, with implications from evolutionary biology at large (Whiten Reference Whiten2017a; Whiten et al. Reference Whiten, Hinde, Stringer and Laland2011) to human nature (Henrich Reference Henrich2015; Laland Reference Laland2017). The Précis and Heyes (Reference Heyes2018) offer a theory of deeper penetrations of human culture into our minds, shaping, perhaps even creating, major aspects of cognition. This potentially opens up exciting new landscapes for culturally focused interdisciplinary research.
The basic idea, however, appears not as shiny and new as Heyes implies. It's now 20 years since Tomasello's (Reference Tomasello1999) The Cultural Origins of Human Cognition (perhaps tellingly mis-cited as 2009 in Heyes’ book and circulated précis), whose title seems perfectly apt for Heyes’ argument. Heyes argues that her approach differs from Henrich's (Reference Henrich2015) in defining cultural learning “by ostension: by pointing at putative examples” (Heyes Reference Heyes2018, p. 88), but Tomasello et al. (Reference Tomasello, Kruger and Ratner1993a) did this already in proposing three examples, including imitation and instructed learning via mindreading, now echoed by two of Heyes’ four examplars. Still, Heyes’ overall thesis is stimulating in exploring such ideas much further.
Much in the book begs questions that I hope Heyes may answer in her Response. First, I see no explicit definition of what counts as Heyes’ central concept of a cognitive gadget. Can she offer a definition, in a sentence or two? The metaphorical contrast between grist and mills appears clear, but how this maps to cognition is not. Heyes (Reference Heyes2018) notes that cultural evolutionary theory covers “large-scale conceptual structures such as fairy tales, systems of religious belief, and scientific theories,” but to her these are “cognitive grist” rather than mills (p. 36). Yet the religious beliefs of, say, an ancient Mayan or Egyptian would surely shape “not just what they think but how they think it” (Précis, sect. 1, para. 2) in radically different ways to each other and to a contemporary atheist, so what makes these not cognitive gadgets? Saying the latter are “neurocognitive” does not seem to help, for don't we nowadays assume that everything cognitive is instantiated in the brain and hence “neurocognitive”?
A related difficulty in what “counts” as a gadget arises in relation to testing selectionist evolutionary theories. I agree with Heyes that memetics has foundered in trying to unitize and count memes (her example is “heaven” [Reference Heyes2018, p. 38]: does this count as one meme? Then what about subsidiaries like angels?). Heyes suggests that by contrast, cognitive mechanisms are “unitized” by cognitive science, making them (as gadgets) more tractable for testing Darwinian selectionist theories of cultural evolution. But don't the same problems arise? Taking the example of language, does a bilingual person have one gadget, or two? Or many? Developmental changes raise further questions paralleling those concerning angels and heaven: For example, when a child adds irregular verbs to their language competence, is that adding a gadget (mini-gadget?) to their larger linguistic skill (macro-gadget?)? If not, why not? Imitation and mindreading also change, bit by bit, throughout development, challenging unitization. None of this is to disparage Darwinian analyses of cultural phenomena, of course. Colleagues and I, noting that Darwin knew nothing of genes, set aside the difficulties of genes’ purported equivalent of memes and simply returned to the principles set out by Darwin, revealing their operation in human (Mesoudi et al. Reference Mesoudi, Whiten and Laland2004) and animal (Whiten Reference Whiten2017a; Reference Whiten2019b) cultural evolution.
Heyes’ mission is to identify what creates distinctively human cognition, proposing that cultural learning supplies the answer. But if other animals’ lives are shaped by their cultural inheritance (Whitehead & Rendell [Reference Whitehead and Rendell2015, p. 6], for example, conclude from their comprehensive survey that “Culture is a major part of what the whales are”; also see Whiten Reference Whiten2017b; Whiten & van de Waal Reference Whiten and van de Waal2018, for primates), then presumably their cultures too are predicted to shape their cognition. Gruber et al. (Reference Gruber, Muller, Strimling, Wrangham and Zuberbuhler2009) offered the novel affordance of honey-filled holes in logs to wild chimpanzees from communities that either did (Kanyawara) or did not (Budongo) have stick-tool use in their cultural repertoires, and found that although the Kanyawara apes applied sticks to efficiently extract honey, Budongo individuals applied just their habitual leaf-sponging technology, with less success. The authors’ conclusion, encapsulated in their title, was that chimpanzees “rely on their cultural knowledge” to solve such novel problems. If this interpretation is correct, would it translate in Heyes’ terms as the Kanyawara chimpanzees having a “cognitive gadget” like imagining modifying stick-tool use appropriate to solving a novel problem – a cultural cognitive gadget the Budongo chimpanzees lack? If not, why not?
This chimpanzee study of Gruber et al., contrasting two populations, is relevant to Heyes’ proposals for dissecting the effects on cognitive competences/gadgets of “nature, nurture, culture” (Précis, sect. 2). Developmental ethologists like Hinde (Reference Hinde1970) long ago recognized that one cannot easily dissect these at the level of the individual because these factors interact in complex ways from conception to death; yet when it comes to differences – between individuals, populations, or species – such causes can be empirically distinguished. Taking Heyes’ example of literacy, one can show that culture is crucial to the emergence of literacy because literacy does not appear in societies that lack the requisite cultural background. However, this is different from claiming, as Heyes appears to do, that literacy is a cognitive gadget in human individuals, created by cultural evolution alone, because literacy arose only a few thousand years ago. We know from contemporary hunter-gatherer studies that hunters are impressively skilled in reading, in signs in the dirt, a narrative of recent events (“a kudu with a lame calf travelled north around an hour ago”) (Wannenburgh et al. Reference Wannenburgh, Johnson and Bannister1979). The signs are not images of kudu; they are abstract signs that have meaning for readers of them, as printed words do for us. Hunting is known to have an ancient ancestry (Whiten & Erdal Reference Whiten and Erdal2012) during which genetic changes could be associated with such a capacity, so this might well underlie our ability to culturally acquire literacy. A hominin lacking an evolutionary history of hunting and gathering should thus struggle to acquire literacy. This is not practicable to test, of course, but doesn't the possibility mean it is premature to conclude that literacy is any “purely culturally created” competence?
A somewhat pedantic aside: I am puzzled that the term “teleosemantic” has emerged and been adopted by Heyes. It seems to mean “teleonomic,” the term introduced by Pittendrigh (Reference Pittendrigh, Roe and Simpson1958) to contrast with “teleogical,” the latter meaning guided by purpose, the former distinguishing the mere appearance of purpose, for reasons such as evolution by natural selection. We say such biological phenomena have evolved to serve certain fitness-supporting functions, rather than to fulfil anyone's purposes. The term teleonomic has been widely used by well-known evolutionary biologists like Mayr and G. C. Williams and philosophers of biology like Nagel and Hull. In the same spirit Lorenz (Reference Lorenz1966) talked of the ways in which both genome and learning function to “acquire and store information on the environment” (p. 8). If teleosemantic is indeed only a synonym for teleonomic, good scholarship suggests settling on the latter and abandoning the other term.
Relating to such distinctions, Heyes states that “‘Cultural learning’ is a subset of social learning … specialized for cultural evolution” (Reference Heyes2018, pp. 86, 89 [my italics]). But surely cultural transmission functions only to benefit either a receiver, like an observational learner, or a transmitter engaged in some form of teaching; its function is not so futuristically oriented as to be about creating cultural evolution – except perhaps for vocational inventors. But Heyes is not focused on such inventors. Elsewhere Heyes more correctly says “specialized for cultural inheritance” (e.g., Précis, Fig. 4 [my italics]), which avoids the impression that every imitator or mindreader is out to create cultural evolutionary change; should that not always be the terminology used?
Turning to the first of the two of Heyes’ four cognitive gadget exemplars I have room to discuss, I agree that social learning biases appear widespread among animals and not only in humans (Kendal et al. Reference Kendal, Boogert, Rendell, Laland, Webster and Jones2018; Price et al. Reference Price, Wood and Whiten2017). And they can be complex; Bono et al. (Reference Bono, Whiten, van Schaik, Krutzen, Eichenberger, Schnider and van de Waal2018), for example, showed that interactions between as many as three biases may interact to predict wild vervet monkeys’ social learning: the sex of the observer, the sex of the potential model, and the relative payoff gained by the latter. I disagree with Heyes that such effects have “barely appeared on the radar of cognitive scientists” (sect. 5, para. 2). It depends on who counts as a cognitive scientist, of course, but a considerable research literature has now delineated a diversity of biases in nonhuman animals, and in children and adult humans (Kendal et al. Reference Kendal, Boogert, Rendell, Laland, Webster and Jones2018; Price et al. Reference Price, Wood and Whiten2017).
Heyes proposes that human distinctiveness in this domain is defined by “explicit, metacognitive rules” about the biases in play. But the only examples of such rules offered are verbalizations. If “explicit” translates only as “verbalized,” the terms “explicit” and “metacognitive” appear disappointingly redundant. The key question is how explicit metacognitive rules could be recognized in nonhuman animals in order to test hypotheses about their presence or absence in nonlinguistic beings.
Finally, I turn to imitation, which Heyes (Reference Heyes2018, p. 116) describes as the “Lamborghini” of social learning. But does the restricted sense of bodily copying that Heyes favours really play a key role in cultural transmission? I agree this is plausible particularly for gestures, including the distinctive human realm of rituals (Whiten Reference Whiten2019a); Clay & Tennie (Reference Clay and Tennie2018), for example, found that children tended to overimitate causally irrelevant hand gestures made while solving an object manipulation task, while bonobos ignored them. But I suggest that the majority of skills a child acquires through observational copying, such as how to make an unfamiliar tool do its job, do not rely on fidelity of bodily copying, which plays little part in the now more than 50 reports of “overimitation” (Hoehl et al. Reference Hoehl, Keupp, Schleihauf, McGuigan, Buttelmann and Whiten2019). Is there any evidence to the contrary? More generally, what is the empirical evidence for the oft-repeated assertion that cumulative culture relies on high-fidelity copying – especially the bodily imitation on which Heyes’ model focuses?
In that model, Heyes likewise seems overenthusiastic about the role that adults imitating a toddler can play in building a child's imitative capacity “from scratch” using domain-general associative learning. Caregivers may sometimes imitate infants’ facial expressions in face-to-face interactions, but is there any evidence they routinely imitate toddlers’ limb and other bodily movements? Is Heyes really suggesting that the boy copying “clasp hands behind back” developed the ability to imitate this because often in the past he did this or similar actions, and his parents copied him? And how could looking in mirrors, or synchronous activities, deliver this example? The same goes for chimpanzees and orangutans, able in “‘do-as-I-do” tests to copy novel test items like “touch back of head” (Call Reference Call2001; Custance et al. Reference Custance, Whiten and Bard1995), that they surely have not learned because others copied them doing this? And what of avian imitation of bodily actions, like using foot versus beak (Heyes & Saggerson Reference Heyes and Saggerson2002; Zentall et al. Reference Zentall, Sutton and Sherburne1996)? As I remarked in a critique following Heyes’ initial promotion of the ASL model (Whiten Reference Whiten, Hurley and Chater2005), there is a more general problem here too. Most of what a parent does cannot match what their infant is doing – they are attending to feeding, changing nappies, cooking, and so on – so for the infant to learn about matching, there would have to be some specific signal indicating “now, here is my rare perceptual match to what you just did.” I think no such signals are known. Moreover, bodily imitation is not “correlated” in the sense of being synchronous anyway; imitation follows a model's acts.
So does the underlying process of imitation, from perception to matching action, remain a black box? Well yes; we remain ignorant of how the brain does it and how it comes to do so. Similarly, a humanoid robot that can achieve the whole process, globally, is yet to be created? It would be illuminating to see if such a robot could build the ability if programmed only with ASL.