R1. Cognition blindness

One of the overarching aims of Cognitive Gadgets is to encourage people interested in human evolution to think not only about brains, bodies, behaviour, and beliefs, but also in a computational way about how our minds work. I was trying to overcome “cognition blindness,” a tendency among evolutionists to look straight past an important resource – the kind of cognitive science, thriving in labs all over the world since the 1970s, that casts mental processes as software running on the brain (Block Reference Block, Smith and Osherson1995). Some commentators revealed, inadvertently, just how tenacious cognition blindness can be. While making otherwise valuable points, these commentators looked straight past the software and wrote about cognitive gadgets as if they are parts of the brain, chunks of behaviour, or airy bridges between brain and behaviour built out of folk psychology and pure maths (e.g., Badcock, Constant, & Ramstead [Badcock et al.]; Iannetti & Vallortigara; Jablonka, Ginsburg, & Dor [Jablonka et al.]; Smaldino & Spivey; Sperber; Tennie; Whiten). The arcuate fasciculus is part of the brain, not of the mind. “Social organisation,” “norm,” “conformity,” and (in frequent usage) “decision rule” refer to behavioural regularities rather than computational processes. And in many models, terms such as “inference” and “belief” are taken from folk psychological stock and applied so promiscuously that they lose all meaning, leaving maths to do the work.

Of course, it is vital to study the brain and behaviour, often with the help of mathematical models and folkweave characterisations of the mind, but Cognitive Gadgets recommends a major addition to the evolutionist's armoury. It suggests that we can better understand human evolution if we recognise that the brain interacts with behaviour via cognitive mechanisms; these mechanisms are among the targets of genetic and cultural selection; and folk psychology seldom provides the most precise and empirically grounded descriptions of how these mechanisms work. In many cases – such as the mechanisms involved in object recognition, speech production, and reading – folk psychology is simply silent. Mathematical models can help fill the silence, but without more abstract, software characterisations of what the mind is doing, these models struggle to make testable predictions (Coltheart Reference Coltheart and Wixted2002; Reference Coltheart2012).

Dominey and Fenici & Garofoli certainly do not suffer from cognition blindness. They see the computationalism of Cognitive Gadgets clearly and challenge it head-on. At the heart of their challenge are a lament and an historical claim. They lament that computationalist cognitive science makes little contact with lived experience and, therefore, with the humanities. It sides with the natural sciences, offering “explanation” rather than “understanding,” in the language of “information” rather than of “meaning.” The historical claim is that this could easily have been otherwise. If computer technology had not been advancing so rapidly when behaviourism ran out of steam, the cognitive revolution would have produced a more humane, meaning-based cognitive science (Bruner Reference Bruner1990).

I am sympathetic to the lament and I find the historical claim fully plausible. It is deeply regrettable that we are still a long way from knowing how to integrate explanation and understanding, information and meaning, science and the humanities. However, I doubt that the direction of cognitive science can be changed from on high by the kind of metaphysical arguments advanced by critics of computationalism (e.g., Baggs, Raja, & Anderson (Baggs et al.); Clark & Chalmers Reference Clark and Chalmers1998; Hutto & Myin Reference Hutto and Myin2017; Malafouris Reference Malafouris, Etzelmüller and Tewes2016), and, even if such inorganic change were possible, I am not sure there would be a net gain from switching sides. A cognitive science that jettisoned computationalism for “meaning” would lose most of the insights accumulated over the last 50 years and, although closer to the humanities, would be alienated from the natural sciences. As long as a meaning-based approach continues to dominate social and developmental psychology (e.g., Tomasello Reference Tomasello1999; Reference Tomasello2014; Reference Tomasello2019; Whiten), I see no danger that it will be abandoned completely by those who study the mind. Furthermore – and this may be where I differ most from Dominey and Fenici & Garofoli – I am not disturbed by the historically contingent origins of computationalism. I see both computationalism and folk or “belief-desire” psychology – the “meaning” framework – as products of cultural evolution. They each have strengths and weaknesses, and are eminently revisable. For now, as highlighted by van Bergen & Sutton, there are advantages to be gained from using both folk psychology and computationalism to understand the evolution of the human mind.

R2. Grist and mills

In one of my efforts to overcome cognition blindness, to point at what is missing from cultural evolutionary studies, I borrowed an 800-year-old metaphor of the mind from St. Thomas Aquinas. I said that cultural evolution operates not only on the grist of the mind (e.g., beliefs, ideas, behaviours, skills, artefacts) but also on the mills (cognitive mechanisms). Like most metaphors, this one is far from perfect. Mills work on grist and cognitive mechanisms work on beliefs, ideas, behaviours, skills, and artefacts (BIBSA); cognitive mechanisms take these particulars as input and transform them. So far, so good. But whereas mills turn grist into flour, cognitive mechanisms turn BIBSA into more BIBSA. Beliefs, ideas, behaviours, skills, and artefacts – the usual targets of cultural evolutionary analysis – are both inputs and outputs of cognitive processing.

As Smaldino & Spivey noticed, the grist-and-mills metaphor would have been even more imperfect if I had used it to capture not the synchronic relationship between cognitive processes and their contents, but the diachronic relationship between social interactions and neural mechanisms. I agree with them that “the social mechanisms of language use and the neural mechanisms of language processing may not be well treated as ‘a grist’ and ‘a mill’, respectively.” Fortunately, although the mutually formative relationship between social interactions and cognitive (rather than neural) mechanisms was a central theme of Cognitive Gadgets, I did not try to capture that relationship with a metaphor of any kind. Instead I characterised it as a relationship in which cognitive mechanisms undergo cultural evolution.

I am pleased to find that I have much in common with Baggs et al., but they are also unhappy about the grist-and-mills metaphor. At first blush it seems that, in their view, this metaphor misled me into thinking that “the things we do and make” are mere products of cognitive processes. On this reading, to cast behaviour and artefacts as grist is to overlook the vital role of the agent's own behaviour in determining the information to which s/he has access, and to underestimate the importance of both artefacts and the behaviour of other agents as carriers of information in their own right. But when the first blush has subsided, this is an implausible reading of the concern expressed by Baggs et al. Cognitive Gadgets does not say a lot about artefacts because it focusses on social cognition (language, mindreading, imitation) rather than instrumental cognition (e.g., causal understanding, spatial navigation), but it dwells at great length on the importance of social interaction – what we do with others – in informing and shaping the human mind. Given this emphasis, it is more likely that Baggs et al. are objecting to the metaphysics of the grist-and-mills metaphor. They are challenging the assumption – enshrined in both computationalism and contemporary Western folk psychology – that thinking, acting, and artefacts are three fundamentally different kinds of things. They see value in the idea of “the extended mind” (Clark & Chalmers Reference Clark and Chalmers1998), the view that behaviour and objects can be intrinsic parts of a thinking process.

There is something exhilarating about philosophical work on the extended mind. Consistent with the cultural evolution of mindreading, it shows that our thinking about thinking could easily have been both coherent and radically different from the way it is now. However (call me old-fashioned), I cannot see what would be gained, in everyday life or in cognitive science, by switching from the view that the mind is “in the head” to the view that the mind is (partly) in the world. The capacity of a puddle to constrain dance movements and inspire mischief can be captured not only by casting the puddle as “a component in our action control” (Baggs et al.), but also in the conventional way by casting the puddle as an environmental input to action control – grist to a mill. Similarly, in the diachronic case, when I say that the childhood development of imitation draws on experience with optical mirrors, and of being imitated by others, I struggle to see what would be gained by casting the mirrors and the actions of other agents as component parts of the child's developing mind. It is kind of cool to think of it that way, but would the extended mind perspective suggest different empirical questions, or make existing questions more empirically tractable?

Baggs et al. also chide me (gently) for neglecting niche construction, “the idea that animals reshape their environments through their actions, and this in turn structures the selection pressures exerted on current and future generations.” It is not clear whether niche construction is a bold new concept, like the extended mind, or a catchy new term for an important and pervasive phenomenon that has long been recognised by evolutionists (Feldman et al. Reference Feldman, Odling-Smee and Laland2017; Gupta et al. Reference Gupta, Prasad, Dey, Joshi and Vidya2017). Without attempting to resolve that issue, which is way above my pay grade, I can only say that I am puzzled when people suggest that niche construction – a ubiquitous phenomenon throughout the animal kingdom – is not just important in humans, but a key to understanding distinctively human characteristics. It is a bit like the problem posed by research on social learning strategies in nonhuman animals (Heyes Reference Heyes2018, ch. 5). If nearly all animals have social learning strategies, we need to find out what it is about human social learning strategies that makes us different. Similarly, if nearly all animals engage in niche construction, we need to find out what it is about human niche construction that makes us different. Most of the explanatory work is done by the difference-maker rather than the base concept – in the case of social learning strategies, by the recognition that, in humans, some social learning strategies are explicitly metacognitive.

R3. Gadgets and instincts

In the movies, Frankenstein screams maniacally “It's alive! It's alive!” as his monster begins to twitch. We have no trouble understanding what Dr. F. is asserting (and denying) even though he is drawing on a distinction, between life and death, that affords many intermediates and ambiguous cases. A creature can be more or less alive, closer or further away from death; there are entities – viruses, zombies, Frankenstein's monster – that resist classification; and, as it says in The Book of Common Prayer (2007/1549), “In the midst of life we are in death.” The distinction between cognitive gadgets and cognitive instincts, although less profound, is similarly sinuous.

The first thing I should emphasise is that a cognitive gadget is not an entity “created by cultural evolution alone” (Whiten). As highlighted by Sperber, I am convinced that “The rich interactive complexity of developmental processes makes it absolutely clear that, in cognition as in other biological systems, there are no pure cases of nature or of nurture; no biological characteristic is caused only by ‘the genes’ or only by ‘the environment’” (Heyes Reference Heyes2018, p. 24). Rather, a cognitive gadget is a cognitive mechanism with distinctively human characteristics that have been shaped predominantly by selection operating on cultural variants. In contrast, a cognitive instinct is a cognitive mechanism with distinctively human characteristics that have been shaped predominantly by selection operating on genetic variants. The terms cognitive gadget and cognitive instinct mark the ends of a continuum of cases (Sperber), with, I argued in Cognitive Gadgets, imitation and mindreading close to the gadget end, and things like associative learning and the inborn face bias (Iannetti & Vallortigara) close to the instinct end of the continuum.

There are many evolutionary processes that could, in principle, send a cognitive process from one end of the continuum into a “messy middle ground” (Rathkopf & Dennett) between gadgetry and instinctiveness. For example, in principle, genetic assimilation (Del Giudice) could increase the role of genetically inherited information in shaping development, and genetic accommodation could amplify the roles of nature, nurture, and/or culture (Jablonka et al.). As Del Giudice underlined, this is not a “zero-sum competition.” All of these in-principle possibilities I happily embrace. What puzzles me is that those commentators who were critical of the gadget-instinct distinction seem to share my interest in examining how different factors (genetic, cultural, “plasticity,” etc.) combine to produce cognitive development, but do not seem to believe that, in order to do this, one must be able to get an empirical handle on what and how each factor is contributing in any given case. It is as if they want to know how different ingredients and oven settings contribute to the texture and flavour of a cake but do not believe that, to find out, one must be able to distinguish their contributions through intervention – for example, by adding more flour – and by examining patterns of covariance – for example, by comparing cakes baked at 180, 190, and 200⁰C.

Badcock et al., Jablonka et al., and Sperber say very little about empirical matters. They distinguish types of interaction between genetic and experiential influences – or genetic and specifically cultural influences – without considering how the types could be distinguished in practice. For example, they do not explain how we would know whether genetic accommodation had or had not occurred (Jablonka et al.), or how we can tell apart cases in which “A biological function [has been] fulfilled through the cultural evolution of an appropriate trait” and in which “cultural evolution [has taken] advantage of biologically evolved dispositions” (Sperber). On the other hand, Del Giudice, revisiting our disagreement about mirror neurons (Cook et al. Reference Cook, Bird, Catmur, Press and Heyes2014; Del Giudice et al. Reference Del Giudice, Manera and Keysers2009), concerns himself with empirical matters but offers a counsel of despair. He doubts that twin studies can provide positive evidence of genetically inherited contributions to development, and remarks ominously that “It may be impossible to fully make sense of the cross-cultural data on developmental trajectories without addressing the thorny issue of national differences in cognitive ability.” However, Del Giudice does not direct us to empirical methods that are, in his view, better able to trace the contributions of nature, nurture, and culture to cognitive development. It seems that he wants to consign cognitive mechanisms to the middle ground between gadgetry and instinctiveness because he despairs of our ever being able to find positive evidence of genetic, learning, and cultural contributions.

As I acknowledge repeatedly in Cognitive Gadgets, both explicitly and by poring over data, it is very difficult indeed to get an empirical handle on the contributions of nature, nurture, and culture to cognitive development. For example, after discussing a range of methods, I note:

In my view, it is neither legitimate nor helpful to respond to these challenges with a “messy middle default,” – that is, by assuming that all three sources of information contribute about equally in all cases, or by assuming out of tribal loyalty (e.g., to behaviourism or High Church evolutionary psychology) that one of them is dominant. Cognitive Gadgets offers and uses a methodological template for parsing cognitive development, based on the distinction between poverty and wealth of the stimulus. I would be flabbergasted if this template were exactly right. It certainly needs refinement and to be augmented by modelling, especially non-linear modelling (Smaldino & Spivey). But I shall stick to my guns on what the current evidence suggests – that many distinctively human cognitive mechanisms lie at the gadget end of the continuum (see below) – and, more generally, on the necessity for empirical evidence from cognitive science to back up claims about the roles of nature, nurture, and culture in cognitive development. I will be content if Cognitive Gadgets proves to be “a timely provocation” (Del Giudice) in this respect; if it encourages those interested in human evolution to recognise that claims about the innateness and genetic assimilation of cognitive processes are not helpful unless they are backed by specific, discriminative empirical evidence. We should not allow nativism to be a matter of taste.

R4. More about gadgets

Before turning to the evidence surveyed in Cognitive Gadgets, I would like to say a little more about what I had in mind when I coined the term “cognitive gadgets.” (Gadgets are out in the world now, so people can make of them what they will, but I still feel a bit proprietary.)

First, I have been convinced by Buskell (Reference Buskell2018) that “minority” cognitive processes – such as those specialised for chess (Del Giudice), lace making, or abacus calculation – are cognitive gadgets in good standing, and that they could prove to be a valuable resource in empirical research on the cultural evolution of typically human cognition. However, following High Church evolutionary psychology, I am especially interested in the types of cognitive mechanisms – such as mindreading, episodic memory, language, imitation – that are present in most people alive today. These human-nature–defining cognitive gadgets are, for me, the paradigmatic cases. Note, with Badcock et al., that many people who now identify as “evolutionary psychologists” are not High Church. I may even be one of them. But, of course, insofar as the departure from orthodoxy involves rejection of computationalism, I regard it as heresy.

Second, Del Giudice and Sperber take me to be yet more devout about associative learning than I really am. I see associative learning as a powerful engine, but not the only engine, in the construction of cognitive gadgets. As I tried to make clear in my discussions of metacognitive social learning strategies and mindreading (Heyes Reference Heyes2018, chapters 5 and 7), like Dominey, I regard language as another major generator.

Finally, I want to put my hands up and acknowledge that, although it suggests that cognitive gadgets are shaped by cultural group selection, the book says relatively little about evolutionary dynamics (Del Giudice; Smaldino & Spivey). It is the work of a cognitive scientist interested in evolution, not of an evolutionist interested in cognitive science. I hope researchers with complementary expertise will take up the challenge, using modelling and historical-anthropological data to assess the plausibility of the hypothesis that distinctively human cognitive mechanisms (along with grist – social organisation, norms, beliefs, etc.) have been shaped by cultural selection. In the meantime, let me reiterate baldly an argument in favour of cultural selection that did not make it from the book to the precis: We know of three sources of adaptive fit between a species-typical trait and its environment – intelligent design, genetic selection, and cultural selection (Dennett Reference Dennett2017). Intelligent design now contributes to the development of some distinctively human cognitive mechanisms (e.g., there are education programmes designed to promote literacy), but it is not a plausible candidate for most of these mechanisms (e.g., mindreading, imitation). Genetic selection is the option backed by High Church evolutionary psychology, but, I argue in Cognitive Gadgets, contemporary evidence from cognitive science is not consistent with the idea that genetic selection is the principal architect of the human mind. Therefore, to the extent that distinctively human cognitive mechanisms are adaptive – do their jobs well – it must be because they have been shaped by the third designer, cultural selection.

R5. Evidence

R6. Concluding remark

Although a part of me would have preferred 30,000 words “of closely reasoned adulation,” what the commentators have provided is much more invigorating and instructive. I am grateful to them all for reading the book, and offering critiques that will help evolutionary psychology to identify more and better cognitive gadgets.