What Babies Know (Spelke, Reference Spelke2022) is a big book – some 532 pages. As is to be expected from one of the founders of cognitive science, it is uniformly well-written, thoroughly well-researched, and chock-full of fascinating, original, and compelling arguments. It is also only half of Spelke's theory of human cognition: What Babies Know sets out the theoretical foundations for the next volume What Babies Learn, and together, the two books explain how humans come to be the kinds of thinkers that they are. Because the latter book is not part of this symposium, assessing the overall picture that emerges from Spelke's masterful synthesis of decades worth of work on the peculiarities of human cognition is not trivial. Still, the very fact of this division of her overall framework is meaningful – and raises some questions. To get at these questions, though, a brief background of the cornerstones of Spelke's theory of core knowledge is necessary.
According to this theory, human cognition is built on the foundation of a number of “core knowledge systems.” These are evolutionarily ancient, domain-specific sets of mental representations that structure how we interpret the world we live in. For example, there is a core knowledge system centering on objects: Beyond just carving the world into edges and corners (as on Marrian theories of vision), we are also born expecting the world to contain objects that have various features, such as the fact that they exclude each other (they cannot both occupy the same part of space). Further core knowledge systems concern the domains of space, number, form (roughly, what objects are plausible biological entities), agency, and sociality. At heart of Spelke's theory then is the claim that these core knowledge systems make for the basis of human cognition: They structure our thoughts, and as their outputs get further integrated through our linguistic capacities – another core knowledge system – they build up the kinds of minds we have. For present purposes, two aspects of this theory are crucial. (Spelke's theory is built on several decades' worth of empirical studies, most of which she was instrumental in conducting. Although the results of these studies can be interpreted in different ways – a point often noted explicitly in the book – I shall leave this aside here.)
First, it is an explicitly nativist picture of human cognition, and thus contrasts with, say, the account of Heyes (Reference Heyes2018), according to which human cognition is the result of a handful of general purpose learning tools that are then culturally harnessed to yield human knowledge structures. As a nativist account, Spelke's picture has to grapple with exactly what it means for something – like a knowledge system – to be innate. To address this question, Spelke employs the following characterization of innateness: a “cognitive system is innate if it is not learned: that is, if it is present and functional on the infant's first effective perceptual encounters with the entities to which it applies” (p. 71).
As an account of innateness, though, the part behind the colon here is a bit tricky. If kids culturally learn about something – for example, that big snakes are dangerous – then this knowledge can be “present and functional on the infant's first effective perceptual encounters with the entities to which it applies.” However, we would not want to see this as a case of innateness: It is learned, after all (as also pointed out by the first part of the above characterization). On the flipside of this, if an infant encounters something – such as agents acting on false beliefs – several times, but if these encounters do not result in the infant learning about the entity, then a later capacity for dealing with the entity could still be a case of innateness, even though this capacity is not “present and functional on the infant's first effective perceptual encounters with the entities to which it applies.”
It is thus really the first part of the sentence that does the work here: The issue is that core knowledge systems are not individually, culturally, or socially learned. Here it is noteworthy that Spelke of course accepts that such learning does occur – we are not born knowing how to play piano, say, or how to read music. A key tenet of her theory is just that many of our major representational expectations are not learned. This will become important again below.
This leads to the second and most important point to note here: The fact that the focus of Spelke's explanatory framework is squarely on the ancient, innate knowledge structures that are shared with many nonhuman animals. The question I want to raise in this commentary is whether this is the best way to get at distinctively human cognition. Put differently: Even if we accept that human minds contain core knowledge systems, which have evolved because they prepared many organisms (including humans) for successful interactions with the world, the question remains of what makes these systems so special that they should be taken to be fulcrum on which human cognition rests.
To make this question clearer, it is useful to note that, in the précis of the book, there is a somewhat misleading statement about the evolutionary presuppositions of core knowledge systems. So, in the précis, it is said: “An ancient system that first emerged in highly distant ancestors is likely to center on abstract content, because it had to be applicable to the diverse environments that the descendants of that last common ancestor came to inhabit.” This is misleading, as it seems to imply that processes of biological evolution have foresight about which traits will be adaptive in the future – which is not the case. A psychological system cannot be selected for conferring fitness benefits to the descendants of the bearer of the system (in whatever environment they live): Somehow, these benefits have to come back to the bearer itself. In the book itself, Spelke is clearer in noting that the issue is that psychological traits that are consistently adaptive over many generations and in different environments are more likely to go into fixation in the population than traits that are not adaptive in these ways. According to Spelke, core knowledge systems are such traits. Still, there is an important point buried here.
On the one hand, many of the kinds of cognitive traits that are consistently adaptive over many generations and in different environments would seem to be extremely “abstract.” Indeed, some key examples of such traits would seem to be powerful abilities for individual and cultural learning. This is because of the fact that former can be used in many different environments and circumstances to acquire locally adaptive ways of acting: How to be socially successful, say, or which foods are edible in which ways. As a matter of fact, this is also precisely the reasoning underlying the common accounts of the evolution of cultural learning (see, e.g., Boyd & Richerson, Reference Boyd and Richerson2005; Henrich, Reference Henrich2015; Heyes, Reference Heyes2018). By its own logic, Spelke's theory thus needs to acknowledge the existence of these abilities for learning.
On the flipside of this, it is also plausible that there are psychological traits that have been specifically adaptive in the human lineage. Now, Spelke appears reluctant to admit this, but it is unclear exactly why that is. For example, in response to Gergely and Csibra's idea (see, e.g., their Csibra & Gergely, Reference Csibra and Gergely2011) that humans, specifically, have evolved to be teachers and learners, Spelke writes (p. 428): “Good teachers cannot efficiently fill the gaps in students’ knowledge unless they are aware of what their students do and do not understand” – that is, unless they rely on a core knowledge system of other agents. This, though, seems false: We could be born with certain expectations of what is important to teach; this could still be adaptive even if it is not fully accurate. For example, even if teachers unnecessarily belabor points, teaching and learning can be adaptive if there are enough morsels of insight in the teaching.
Something similar goes for a uniquely human talent for symbol cognition. In response, Spelke writes (pp. 432–433): “A further reason for doubting that a species-unique talent for learning and using spatial symbols underlies our cognitive accomplishments, including our construction of social agent concepts, comes from studies of symbol learning in nonhuman animals. Some chimpanzees, monkeys, and parrots have been trained to use pictures or objects as symbols since infancy, and they developed impressive abilities both to communicate by means of those symbols and to use the symbols as tools for thought.” However, this clearly seems to overstate the case here. Despite much training, no nonhuman animal has been shown to be able to read and write. In general, the issue here is not whether nonhuman animals have some ability for symbol cognition; the question is whether humans have a unique adaptation for symbol cognition. Overall, it is just not clear what the case against human-specific cognitive adaptations here is meant to be, or what the theoretical reasons are for thinking there are no such adaptations.
Putting all of this together, this leads to the following question. Why should we think that at the heart of human cognition are core knowledge systems, rather than a combination of our capacities for cultural or individual learning, our species-specific cognitive adaptations, and more widely shared innate representational endowments? Put differently, given that we should accept that all of (cultural and individual) learning, core knowledge, and uniquely human cognitive structures are parts of our innate psychological endowments, why give core knowledge such a prominent pride of place?
In response, the book makes a somewhat cryptic remark (p. 446): “I present this hypothesis as an alternative to theories that root our uniquely human cognitive achievements in our innate propensities for shared intentionality, for pedagogical learning, or for symbolic thought, but the hypothesis suggests there is truth to all these theories. Infants learn language from the members of their social world, and they put language to use, first and foremost, to share their experiences with the people they care about.” This is a little puzzling, though. Given that, as argued earlier, we need to accept that many organisms – including humans – have found individual and social learning to be adaptive, the latter cannot be seen as derived from systems of core knowledge. Similarly, because Spelke has not provided a compelling reason to doubt the existence of some human-specific cognitive adaptations, the latter cannot be explained as the result of the workings of core knowledge systems either.
Importantly, accepting this does not entail giving up on the importance of core knowledge systems. Rather, it just means acknowledging that cultural and individual learning, uniquely human cognitive adaptions, symbols and other forms of cognitive technology, and core knowledge all matter to human cognition. Indeed, I defend an account of distinctively human cognition of exactly this type: According to this, human cognition is the result of a variety of innate representations – some, but not all of which are integrated into sets of core knowledge systems – that are expanded and used through cultural learning and technology (see, e.g., my Schulz, Reference Schulz2020).
In a nutshell, the question at the heart of this commentary is: Why emphasize core knowledge so strongly – why make that its own book? I look forward to hearing the answer to this question; it will undoubtedly be illuminating and lead to fruitful exchange.
What Babies Know (Spelke, Reference Spelke2022) is a big book – some 532 pages. As is to be expected from one of the founders of cognitive science, it is uniformly well-written, thoroughly well-researched, and chock-full of fascinating, original, and compelling arguments. It is also only half of Spelke's theory of human cognition: What Babies Know sets out the theoretical foundations for the next volume What Babies Learn, and together, the two books explain how humans come to be the kinds of thinkers that they are. Because the latter book is not part of this symposium, assessing the overall picture that emerges from Spelke's masterful synthesis of decades worth of work on the peculiarities of human cognition is not trivial. Still, the very fact of this division of her overall framework is meaningful – and raises some questions. To get at these questions, though, a brief background of the cornerstones of Spelke's theory of core knowledge is necessary.
According to this theory, human cognition is built on the foundation of a number of “core knowledge systems.” These are evolutionarily ancient, domain-specific sets of mental representations that structure how we interpret the world we live in. For example, there is a core knowledge system centering on objects: Beyond just carving the world into edges and corners (as on Marrian theories of vision), we are also born expecting the world to contain objects that have various features, such as the fact that they exclude each other (they cannot both occupy the same part of space). Further core knowledge systems concern the domains of space, number, form (roughly, what objects are plausible biological entities), agency, and sociality. At heart of Spelke's theory then is the claim that these core knowledge systems make for the basis of human cognition: They structure our thoughts, and as their outputs get further integrated through our linguistic capacities – another core knowledge system – they build up the kinds of minds we have. For present purposes, two aspects of this theory are crucial. (Spelke's theory is built on several decades' worth of empirical studies, most of which she was instrumental in conducting. Although the results of these studies can be interpreted in different ways – a point often noted explicitly in the book – I shall leave this aside here.)
First, it is an explicitly nativist picture of human cognition, and thus contrasts with, say, the account of Heyes (Reference Heyes2018), according to which human cognition is the result of a handful of general purpose learning tools that are then culturally harnessed to yield human knowledge structures. As a nativist account, Spelke's picture has to grapple with exactly what it means for something – like a knowledge system – to be innate. To address this question, Spelke employs the following characterization of innateness: a “cognitive system is innate if it is not learned: that is, if it is present and functional on the infant's first effective perceptual encounters with the entities to which it applies” (p. 71).
As an account of innateness, though, the part behind the colon here is a bit tricky. If kids culturally learn about something – for example, that big snakes are dangerous – then this knowledge can be “present and functional on the infant's first effective perceptual encounters with the entities to which it applies.” However, we would not want to see this as a case of innateness: It is learned, after all (as also pointed out by the first part of the above characterization). On the flipside of this, if an infant encounters something – such as agents acting on false beliefs – several times, but if these encounters do not result in the infant learning about the entity, then a later capacity for dealing with the entity could still be a case of innateness, even though this capacity is not “present and functional on the infant's first effective perceptual encounters with the entities to which it applies.”
It is thus really the first part of the sentence that does the work here: The issue is that core knowledge systems are not individually, culturally, or socially learned. Here it is noteworthy that Spelke of course accepts that such learning does occur – we are not born knowing how to play piano, say, or how to read music. A key tenet of her theory is just that many of our major representational expectations are not learned. This will become important again below.
This leads to the second and most important point to note here: The fact that the focus of Spelke's explanatory framework is squarely on the ancient, innate knowledge structures that are shared with many nonhuman animals. The question I want to raise in this commentary is whether this is the best way to get at distinctively human cognition. Put differently: Even if we accept that human minds contain core knowledge systems, which have evolved because they prepared many organisms (including humans) for successful interactions with the world, the question remains of what makes these systems so special that they should be taken to be fulcrum on which human cognition rests.
To make this question clearer, it is useful to note that, in the précis of the book, there is a somewhat misleading statement about the evolutionary presuppositions of core knowledge systems. So, in the précis, it is said: “An ancient system that first emerged in highly distant ancestors is likely to center on abstract content, because it had to be applicable to the diverse environments that the descendants of that last common ancestor came to inhabit.” This is misleading, as it seems to imply that processes of biological evolution have foresight about which traits will be adaptive in the future – which is not the case. A psychological system cannot be selected for conferring fitness benefits to the descendants of the bearer of the system (in whatever environment they live): Somehow, these benefits have to come back to the bearer itself. In the book itself, Spelke is clearer in noting that the issue is that psychological traits that are consistently adaptive over many generations and in different environments are more likely to go into fixation in the population than traits that are not adaptive in these ways. According to Spelke, core knowledge systems are such traits. Still, there is an important point buried here.
On the one hand, many of the kinds of cognitive traits that are consistently adaptive over many generations and in different environments would seem to be extremely “abstract.” Indeed, some key examples of such traits would seem to be powerful abilities for individual and cultural learning. This is because of the fact that former can be used in many different environments and circumstances to acquire locally adaptive ways of acting: How to be socially successful, say, or which foods are edible in which ways. As a matter of fact, this is also precisely the reasoning underlying the common accounts of the evolution of cultural learning (see, e.g., Boyd & Richerson, Reference Boyd and Richerson2005; Henrich, Reference Henrich2015; Heyes, Reference Heyes2018). By its own logic, Spelke's theory thus needs to acknowledge the existence of these abilities for learning.
On the flipside of this, it is also plausible that there are psychological traits that have been specifically adaptive in the human lineage. Now, Spelke appears reluctant to admit this, but it is unclear exactly why that is. For example, in response to Gergely and Csibra's idea (see, e.g., their Csibra & Gergely, Reference Csibra and Gergely2011) that humans, specifically, have evolved to be teachers and learners, Spelke writes (p. 428): “Good teachers cannot efficiently fill the gaps in students’ knowledge unless they are aware of what their students do and do not understand” – that is, unless they rely on a core knowledge system of other agents. This, though, seems false: We could be born with certain expectations of what is important to teach; this could still be adaptive even if it is not fully accurate. For example, even if teachers unnecessarily belabor points, teaching and learning can be adaptive if there are enough morsels of insight in the teaching.
Something similar goes for a uniquely human talent for symbol cognition. In response, Spelke writes (pp. 432–433): “A further reason for doubting that a species-unique talent for learning and using spatial symbols underlies our cognitive accomplishments, including our construction of social agent concepts, comes from studies of symbol learning in nonhuman animals. Some chimpanzees, monkeys, and parrots have been trained to use pictures or objects as symbols since infancy, and they developed impressive abilities both to communicate by means of those symbols and to use the symbols as tools for thought.” However, this clearly seems to overstate the case here. Despite much training, no nonhuman animal has been shown to be able to read and write. In general, the issue here is not whether nonhuman animals have some ability for symbol cognition; the question is whether humans have a unique adaptation for symbol cognition. Overall, it is just not clear what the case against human-specific cognitive adaptations here is meant to be, or what the theoretical reasons are for thinking there are no such adaptations.
Putting all of this together, this leads to the following question. Why should we think that at the heart of human cognition are core knowledge systems, rather than a combination of our capacities for cultural or individual learning, our species-specific cognitive adaptations, and more widely shared innate representational endowments? Put differently, given that we should accept that all of (cultural and individual) learning, core knowledge, and uniquely human cognitive structures are parts of our innate psychological endowments, why give core knowledge such a prominent pride of place?
In response, the book makes a somewhat cryptic remark (p. 446): “I present this hypothesis as an alternative to theories that root our uniquely human cognitive achievements in our innate propensities for shared intentionality, for pedagogical learning, or for symbolic thought, but the hypothesis suggests there is truth to all these theories. Infants learn language from the members of their social world, and they put language to use, first and foremost, to share their experiences with the people they care about.” This is a little puzzling, though. Given that, as argued earlier, we need to accept that many organisms – including humans – have found individual and social learning to be adaptive, the latter cannot be seen as derived from systems of core knowledge. Similarly, because Spelke has not provided a compelling reason to doubt the existence of some human-specific cognitive adaptations, the latter cannot be explained as the result of the workings of core knowledge systems either.
Importantly, accepting this does not entail giving up on the importance of core knowledge systems. Rather, it just means acknowledging that cultural and individual learning, uniquely human cognitive adaptions, symbols and other forms of cognitive technology, and core knowledge all matter to human cognition. Indeed, I defend an account of distinctively human cognition of exactly this type: According to this, human cognition is the result of a variety of innate representations – some, but not all of which are integrated into sets of core knowledge systems – that are expanded and used through cultural learning and technology (see, e.g., my Schulz, Reference Schulz2020).
In a nutshell, the question at the heart of this commentary is: Why emphasize core knowledge so strongly – why make that its own book? I look forward to hearing the answer to this question; it will undoubtedly be illuminating and lead to fruitful exchange.
Competing interest
None.