In the book Cognitive Gadgets: The Cultural Evolution of Thinking (Reference Heyes2018), Cecilia Heyes takes a strong stance against the established idea that natural selection of genetic variants is the force that has selected and shaped human cognitive capacities. Heyes negates the existence of innate cognitive instincts. She suggests instead a fascinating and radical alternative: that cultural evolution occurring through social interactions in childhood has “built” and “assembled” the pieces of mental technology that underlie some unique human cognitive capacities. Heyes does not negate the natural selection of variants; however, she believes that these variants are not genetic but cultural.
The attempt to provide a neurobiological, mechanistic explanation of theories of cultural psychology and social anthropology (Shweder & Sullivan Reference Shweder and Sullivan1993) is admirable. As would any radical position, Heyes’ theory of uniquely human cognitive gadgets being assembled in the brain by cultural evolution requires a careful scrutiny. We note that some of the foundations on which Heyes builds her reasoning are based on a fractional selection of the available empirical evidence.
A first tenet of Heyes’ theory is the denial that cognitive mechanisms such as social attentional biases and the ability to imitate are genetically inherited. As a consequence of this negation, Heyes proposes that these cognitive capacities are physically assembled in the brain only after birth, through social interactions in childhood. Although Heyes considers these capacities as uniquely human (more on this later), an ethological perspective is fruitful in order to examine the solidity of this assumption.
An interesting example is associated with newborns’ responses to face-like stimuli, which Heyes considers at length in her book (Reference Heyes2018; pp. 60–63). She argues that “domain-general processes of associative learning are sufficient to explain why, in the first year of life, a simple preference for inverted triangles of blobs becomes a highly robust and selective preference for fellow humans ‘looking at me’” (p. 62). Quite in contrast, it seems to us that in the absence of such a “simple preference,” domain-general processes would simply have no time and opportunity to build up complete face representations on the sole basis of the exposure to real-world exemplars of human faces.
We believe that organisms are equipped with dedicated orienting and learning mechanisms that work as adaptive priors, engrained in cortical architectures that have been shaped by natural selection to deal appropriately with environmental stimuli. This approach may provide a different view of why, as also stressed by Heyes, face preference at birth is not only human-specific but also widespread among vertebrates (p. 62). Research on comparative cognition may prove useful in this regard. As stated recently by Versace et al. (Reference Versace, Martinho-Truswell, Kacelnik and Vallortigara2018):
(Cognitive) priors imply some assumptions about the external world that guide learning, but can, and must, allow errors (…). Research has shown that (e.g.) early preferences of chicks are not strictly species-specific but apply equally to hen face-like or polecat face-like features, or to the biological-motion appearance of either a hen or a cat. This is due to the fact that the orienting mechanisms cannot be too specific for the individual features of the mother hen, which are to some extent unpredictable from the genetic repertoire. A level of non-specificity is functional in avoiding excessive false negatives in the form of failed recognition caused by variability between adults within a species, and by changes in the appearance of even a single individual. (p. 963)
A noticeable example related to our species is provided by recent results suggesting that a cortical route specialized for face processing is already functional at birth. Buiatti et al. (Reference Buiatti, Di Giorgio, Piazza, Polloni, Menna, Taddei, Baldo and Vallortigara2019) used electroencephalography to record neural activity in one- to four-day-old newborns who were exposed to schematic patterns of upright and inverted face-like stimuli. Compared to inverted faces, upright faces elicited stronger responses in a partially right-lateralized network including lateral occipitotemporal and medial parietal areas that largely overlap with the adult face-processing circuit (Rossion & Jacques Reference Rossion, Jacques, Kappenman and Luck2011). Most interestingly, a negative correlation between age and the face-like pattern response was observed, in striking contrast with the idea that the face-specific cortical response increases as a function of exposure to faces. This can be explained as follows: The highly simplified face-like geometrical patterns (the inverted triangles of blobs to which Heyes Reference Heyes2018 alludes, p. 61) act for newborns as, using ethological terms, key or supernormal stimuli. The immature visual system of the newborn in the very first hours of life is genetically tuned to optimally detect such key stimuli, and exposure to real-world complex and variable faces may refine the face-like circuitry such that it rapidly gets more attuned to the real-world features and gradually loses sensitivity to artificial face-like geometrical patterns. This view is profoundly different from that proposed by Heyes because it posits that the unfolding of a genetically inherited face processing mechanism is indeed at work here, and that its lack of specificity is expected as part of such an adaptive prior to account for “variability between adults within a species, and by changes in the appearance of even a single individual” (Versace & Vallortigara Reference Versace and Vallortigara2015, p. 963). Another glaring omission to this discourse is the robust evidence that newborn humans can imitate (e.g., Meltzoff et al. Reference Meltzoff, Murray, Simpson, Heimann, Nagy, Nadel, Pedersen, Brooks, Messinger, De Pascalis, Subiaul, Paukner and Ferrari2018).
These cognitive capacities, whose neural bases have become to be understood (Lorenzi et al. Reference Lorenzi, Mayer, Rosa-Salva and Vallortigara2017; Mayer et al. Reference Mayer, Rosa-Salva, Morbioli and Vallortigara2017; Versace & Vallortigara Reference Versace and Vallortigara2015), are, in our view, innate mechanisms. Thus, the presence at birth of the specific cognitive capacities that Heyes postulates to be exclusively “acquired through sociocultural experience” (p. 5) makes this first foundation of the cognitive gadget theory unwarranted.
A second tenet of Heyes’ cognitive gadgets theory is that cognitive mechanisms such as causal understanding, imitation, and mindreading are not only acquired through sociocultural experience, but are also “distinctively human” (p. 1). However, all these cognitive mechanisms are observed in several other species, although in different grades. For example, birds display causal understanding (e.g., Jelbert et al. Reference Jelbert, Miller, Schiestl, Boeckle, Cheke, Gray, Taylor and Clayton2019), and mindreading is present in a number of nonhuman animals. Thus, the current debate pertains only to the degree by which animal mindreading differs from that of other animals (Lurz Reference Lurz2011). Throughout her discussion of the issue, Heyes affirms that what is commonly considered to be mindreading is not actual mindreading, and she specifies that the cognitive gadget theories refer to the special case of “explicit” mindreading. This construct drift towards less tractable definitions is a consequence of the use of open concepts typical of some psychological discourse: Definitions are construed theoretically rather than being naturally defined by their inherent compositional nature or causal structure. Paul Meehl ascribed the lack of cumulative progress of psychological theories to the use of these open concepts, evoking General McArthur's description of old generals: “They never die, they just slowly fade away” (Meehl Reference Meehl1978, p. 807).
A third, and most fascinating, idea of Cognitive Gadgets is that “human cognitive mechanisms have been built by cultural evolution” (p. 22), and that these new “pieces of mental technology are not merely tuning but assembled in the course of childhood” (p. 22). The evidence that Heyes brings in support of this idea comes from functional magnetic resonance imaging (fMRI) in humans: In response to viewing written sentences, literate individuals produce stronger responses than illiterates in several areas of the brain, including the left mid-fusiform region (the so-called “visual word form area,” VWFA; Dehaene et al. Reference Dehaene, Pegado, Braga, Ventura, Nunes Filho, Jobert, Dehaene-Lambertz, Kolinsky, Morais and Cohen2010). Heyes considers this finding a proof of principle for the cognitive gadgets theory: A cultural product (literacy) builds a new specific piece of brain machinery (“If one did not know that reading is culturally inherited, it would be easy to mistake the […] precise localization of VWFA for signs that the capacity to read depends on a cognitive instinct”; p. 20). There are, however, two main problems in this reasoning. First, several forms of simple noncultural learning enhance fMRI activations in a large set of cortical areas (e.g., Buchel et al. Reference Buchel, Morris, Dolan and Friston1998): such changes in brain activity should not (and are not) considered as testimony that “new pieces of mental technology are […] assembled” (p. 22). Second, matters of specificity and sensitivity of fMRI responses, and the ensuing difficulties of unequivocally identifying a certain cognitive state on the basis of an fMRI response (Poldrack Reference Poldrack2006), are not considered. Indeed, the mid-fusiform gyrus (i.e., the VWFA) responds to a wide number of sensory stimuli, including visual stimuli that do not entail words and have no linguistic implications (Price & Devlin Reference Price and Devlin2003; van Turennout et al. Reference van Turennout, Ellmore and Martin2000). Thus, given that this brain region is also activated when no linguistic stimuli are presented, it is an incorrect reverse inference to conclude that its activation indicates that any language response has occurred, and should therefore not be labeled VWFA (Price & Devlin Reference Price and Devlin2003). Unwarranted conclusions based on reverse inferences and overlap theories of fMRI results are pervasive in human cognitive neuroscience (Iannetti et al. Reference Iannetti, Salomons, Moayedi, Mouraux and Davis2013). Some of Heyes’ propositions are not immune from this issue.
We wish to conclude by recalling the message that Valentino Braitenberg offers in his “Vehicles – Experiments in Synthetic Psychology” (Reference Braitenberg1984): The use of mentalistic terms to describe the behavior of artificial machines with an internal structure inspired by the nervous system reduces our chances to understand properly the mechanisms determining their behavior. These mechanisms are instead more easily understood by creating the structure that gives rise to the behavior. In contrast to Braitenberg's famous “law of uphill analysis and downhill invention,” Heyes states that “relationships between the brain, behavior and the world cannot be understood without describing those relationships at an abstract, mental level” (p. 9). A critical assessment of these diametrically opposed viewpoints has the potential of being revealing.
In the book Cognitive Gadgets: The Cultural Evolution of Thinking (Reference Heyes2018), Cecilia Heyes takes a strong stance against the established idea that natural selection of genetic variants is the force that has selected and shaped human cognitive capacities. Heyes negates the existence of innate cognitive instincts. She suggests instead a fascinating and radical alternative: that cultural evolution occurring through social interactions in childhood has “built” and “assembled” the pieces of mental technology that underlie some unique human cognitive capacities. Heyes does not negate the natural selection of variants; however, she believes that these variants are not genetic but cultural.
The attempt to provide a neurobiological, mechanistic explanation of theories of cultural psychology and social anthropology (Shweder & Sullivan Reference Shweder and Sullivan1993) is admirable. As would any radical position, Heyes’ theory of uniquely human cognitive gadgets being assembled in the brain by cultural evolution requires a careful scrutiny. We note that some of the foundations on which Heyes builds her reasoning are based on a fractional selection of the available empirical evidence.
A first tenet of Heyes’ theory is the denial that cognitive mechanisms such as social attentional biases and the ability to imitate are genetically inherited. As a consequence of this negation, Heyes proposes that these cognitive capacities are physically assembled in the brain only after birth, through social interactions in childhood. Although Heyes considers these capacities as uniquely human (more on this later), an ethological perspective is fruitful in order to examine the solidity of this assumption.
An interesting example is associated with newborns’ responses to face-like stimuli, which Heyes considers at length in her book (Reference Heyes2018; pp. 60–63). She argues that “domain-general processes of associative learning are sufficient to explain why, in the first year of life, a simple preference for inverted triangles of blobs becomes a highly robust and selective preference for fellow humans ‘looking at me’” (p. 62). Quite in contrast, it seems to us that in the absence of such a “simple preference,” domain-general processes would simply have no time and opportunity to build up complete face representations on the sole basis of the exposure to real-world exemplars of human faces.
We believe that organisms are equipped with dedicated orienting and learning mechanisms that work as adaptive priors, engrained in cortical architectures that have been shaped by natural selection to deal appropriately with environmental stimuli. This approach may provide a different view of why, as also stressed by Heyes, face preference at birth is not only human-specific but also widespread among vertebrates (p. 62). Research on comparative cognition may prove useful in this regard. As stated recently by Versace et al. (Reference Versace, Martinho-Truswell, Kacelnik and Vallortigara2018):
(Cognitive) priors imply some assumptions about the external world that guide learning, but can, and must, allow errors (…). Research has shown that (e.g.) early preferences of chicks are not strictly species-specific but apply equally to hen face-like or polecat face-like features, or to the biological-motion appearance of either a hen or a cat. This is due to the fact that the orienting mechanisms cannot be too specific for the individual features of the mother hen, which are to some extent unpredictable from the genetic repertoire. A level of non-specificity is functional in avoiding excessive false negatives in the form of failed recognition caused by variability between adults within a species, and by changes in the appearance of even a single individual. (p. 963)
A noticeable example related to our species is provided by recent results suggesting that a cortical route specialized for face processing is already functional at birth. Buiatti et al. (Reference Buiatti, Di Giorgio, Piazza, Polloni, Menna, Taddei, Baldo and Vallortigara2019) used electroencephalography to record neural activity in one- to four-day-old newborns who were exposed to schematic patterns of upright and inverted face-like stimuli. Compared to inverted faces, upright faces elicited stronger responses in a partially right-lateralized network including lateral occipitotemporal and medial parietal areas that largely overlap with the adult face-processing circuit (Rossion & Jacques Reference Rossion, Jacques, Kappenman and Luck2011). Most interestingly, a negative correlation between age and the face-like pattern response was observed, in striking contrast with the idea that the face-specific cortical response increases as a function of exposure to faces. This can be explained as follows: The highly simplified face-like geometrical patterns (the inverted triangles of blobs to which Heyes Reference Heyes2018 alludes, p. 61) act for newborns as, using ethological terms, key or supernormal stimuli. The immature visual system of the newborn in the very first hours of life is genetically tuned to optimally detect such key stimuli, and exposure to real-world complex and variable faces may refine the face-like circuitry such that it rapidly gets more attuned to the real-world features and gradually loses sensitivity to artificial face-like geometrical patterns. This view is profoundly different from that proposed by Heyes because it posits that the unfolding of a genetically inherited face processing mechanism is indeed at work here, and that its lack of specificity is expected as part of such an adaptive prior to account for “variability between adults within a species, and by changes in the appearance of even a single individual” (Versace & Vallortigara Reference Versace and Vallortigara2015, p. 963). Another glaring omission to this discourse is the robust evidence that newborn humans can imitate (e.g., Meltzoff et al. Reference Meltzoff, Murray, Simpson, Heimann, Nagy, Nadel, Pedersen, Brooks, Messinger, De Pascalis, Subiaul, Paukner and Ferrari2018).
These cognitive capacities, whose neural bases have become to be understood (Lorenzi et al. Reference Lorenzi, Mayer, Rosa-Salva and Vallortigara2017; Mayer et al. Reference Mayer, Rosa-Salva, Morbioli and Vallortigara2017; Versace & Vallortigara Reference Versace and Vallortigara2015), are, in our view, innate mechanisms. Thus, the presence at birth of the specific cognitive capacities that Heyes postulates to be exclusively “acquired through sociocultural experience” (p. 5) makes this first foundation of the cognitive gadget theory unwarranted.
A second tenet of Heyes’ cognitive gadgets theory is that cognitive mechanisms such as causal understanding, imitation, and mindreading are not only acquired through sociocultural experience, but are also “distinctively human” (p. 1). However, all these cognitive mechanisms are observed in several other species, although in different grades. For example, birds display causal understanding (e.g., Jelbert et al. Reference Jelbert, Miller, Schiestl, Boeckle, Cheke, Gray, Taylor and Clayton2019), and mindreading is present in a number of nonhuman animals. Thus, the current debate pertains only to the degree by which animal mindreading differs from that of other animals (Lurz Reference Lurz2011). Throughout her discussion of the issue, Heyes affirms that what is commonly considered to be mindreading is not actual mindreading, and she specifies that the cognitive gadget theories refer to the special case of “explicit” mindreading. This construct drift towards less tractable definitions is a consequence of the use of open concepts typical of some psychological discourse: Definitions are construed theoretically rather than being naturally defined by their inherent compositional nature or causal structure. Paul Meehl ascribed the lack of cumulative progress of psychological theories to the use of these open concepts, evoking General McArthur's description of old generals: “They never die, they just slowly fade away” (Meehl Reference Meehl1978, p. 807).
A third, and most fascinating, idea of Cognitive Gadgets is that “human cognitive mechanisms have been built by cultural evolution” (p. 22), and that these new “pieces of mental technology are not merely tuning but assembled in the course of childhood” (p. 22). The evidence that Heyes brings in support of this idea comes from functional magnetic resonance imaging (fMRI) in humans: In response to viewing written sentences, literate individuals produce stronger responses than illiterates in several areas of the brain, including the left mid-fusiform region (the so-called “visual word form area,” VWFA; Dehaene et al. Reference Dehaene, Pegado, Braga, Ventura, Nunes Filho, Jobert, Dehaene-Lambertz, Kolinsky, Morais and Cohen2010). Heyes considers this finding a proof of principle for the cognitive gadgets theory: A cultural product (literacy) builds a new specific piece of brain machinery (“If one did not know that reading is culturally inherited, it would be easy to mistake the […] precise localization of VWFA for signs that the capacity to read depends on a cognitive instinct”; p. 20). There are, however, two main problems in this reasoning. First, several forms of simple noncultural learning enhance fMRI activations in a large set of cortical areas (e.g., Buchel et al. Reference Buchel, Morris, Dolan and Friston1998): such changes in brain activity should not (and are not) considered as testimony that “new pieces of mental technology are […] assembled” (p. 22). Second, matters of specificity and sensitivity of fMRI responses, and the ensuing difficulties of unequivocally identifying a certain cognitive state on the basis of an fMRI response (Poldrack Reference Poldrack2006), are not considered. Indeed, the mid-fusiform gyrus (i.e., the VWFA) responds to a wide number of sensory stimuli, including visual stimuli that do not entail words and have no linguistic implications (Price & Devlin Reference Price and Devlin2003; van Turennout et al. Reference van Turennout, Ellmore and Martin2000). Thus, given that this brain region is also activated when no linguistic stimuli are presented, it is an incorrect reverse inference to conclude that its activation indicates that any language response has occurred, and should therefore not be labeled VWFA (Price & Devlin Reference Price and Devlin2003). Unwarranted conclusions based on reverse inferences and overlap theories of fMRI results are pervasive in human cognitive neuroscience (Iannetti et al. Reference Iannetti, Salomons, Moayedi, Mouraux and Davis2013). Some of Heyes’ propositions are not immune from this issue.
We wish to conclude by recalling the message that Valentino Braitenberg offers in his “Vehicles – Experiments in Synthetic Psychology” (Reference Braitenberg1984): The use of mentalistic terms to describe the behavior of artificial machines with an internal structure inspired by the nervous system reduces our chances to understand properly the mechanisms determining their behavior. These mechanisms are instead more easily understood by creating the structure that gives rise to the behavior. In contrast to Braitenberg's famous “law of uphill analysis and downhill invention,” Heyes states that “relationships between the brain, behavior and the world cannot be understood without describing those relationships at an abstract, mental level” (p. 9). A critical assessment of these diametrically opposed viewpoints has the potential of being revealing.