As exciting as it is erudite, Spelke's (Reference Spelke2022) book provides a remarkable overview of her groundbreaking research into “innate systems of core knowledge.” To her, those domain-specific, innate cognitive capacities are essential for carving out “our cognitive territory into more manageable units.” As shown by Spelke's successive attempts, identifying core knowledge systems is however a complicated task. In her book, she argues for six core systems – objects, numbers, places, forms, agents, and social beings. The phylogenetic and ontogenetic functions of those core systems, interposed between percepts and beliefs, are to support the “first informative encounters with the entities they serve to represent” (p. xix). Although Spelke's hypothesis is quite convincing, we would like to make two critical arguments. Our first argument concerns the ontological divergence between the six core systems and the “entities” they are supposed to deal with. The second argument focuses on the conceptual primitives at the heart of the “core system of social beings.”
From an ontological point of view, numbers and shapes are not individual entities in themselves, but universal properties of given entities. Without reactivating here the old metaphysical debate between the ontology of individuals and the ontology of universals, it is hard to ignore that properties such as quantity, form, or place have a higher degree of abstraction than objects or agents. Wheras first-order concrete entities are defined by their material existence and causal power in the physical world, second-order abstract properties are more complex ways of referring to and qualifying those entities. There is an ontological hierarchy between individual entities endowed with spatiotemporal coordinates and the abstract properties that these same entities instantiate. Herein lies our question. If entities and properties are not on the same ontological page, can they be situated in the same cognitive territory? Or do we need to organize so-called core systems into a multilayer cognitive architecture in which properties have not the same status as objects and agents? This question, both ontological and cognitive, also underlies the point we would like to make about social cognition.
Within Spelke's framework, newborns, who are literally infants, that is, children who are “not able to speak,” have at their disposal two independent, automatic core knowledge systems dedicated to the nonphysical world: the “agent system” and the “social system.” The agent system allows infants to represent self-moving entities as agents who act in a causally effective and perceptually guided way. Present in human newborns and also in newly hatched chicks, the agent system consists in seeing self-propelled entities as goal-oriented beings that cause changes in other objects only on contact and through their motion. As for the social system, it targets social beings, that is, people “who endow one another with experiences like their own and who share their experiences in states of engagement.” To Spelke, evidence for a core system dedicated to “shareable experiences of known, individual people” is drawn from newborn's sensitivity to mutual gazes and affective engagements with their caregivers. As core knowledge is modular, the distinction between the agent and the social systems has strange consequences for the first year of infants’ social life: People around them can be conceived either as entities sharing phenomenal experiences with them (social beings) or as intentional entities exerting their causal power over objects (agents). It is only at the end of the first year of life that these two central systems can be combined, and infants finally become able to see people as social agents, endowed with phenomenal and intentional properties. According to Spelke, this cognitive achievement is because of the progressive mastery of language and the access it provides to the plurality of others’ perspectives.
It is worth mentioning that Spelke used to have a different view of social core knowledge, which served to identify “members of one's own social group” and “to guide social interactions with in- and out-group members” (Kinzler & Spelke, Reference Kinzler and Spelke2007, p. 257). In her book, Spelke appears to have modified her perspective. Her definition of the social is primarily, if not solely, intersubjective, as evidenced by her insistence on mimicry, intimacy, proximity, imitation, and emotional sharing. Certainly, the prosocial inclination to build “We-ness” or “togetherness” is essential to individual survival, on both phylogenetic and ontogenetic scales. However, it is questionable whether the social is reducible to “like-ness” in both senses of similarity and kindness. Indeed, in Spelke's core social system, the social world is a matter of affiliation or sociability: It is made up of entities “like-me,” then “like-us.” In this egocentric model, newborns build their social world progressively, as if their knowledge followed a series of concentric circles. First exposed to intersubjective sharing and face-to-face interaction, ego bridges the gap between itself and the minds of those around it through mimicry and experience sharing, and then gradually learns to expand its social network.
The problem is that “being one of us” is not the only meaning of the social. Social groups are defined by a normative structure that enables their members to predict how others are likely to behave, given their position in the social order. Besides intersubjective, horizontal relationships between acquaintances, society is also based on vertical subordination to impersonal constraints and social hierarchies. In this sense, the mark of the social is not We-ness but the impersonal relations of interdependence between status bearers or role takers. Social structures are not a matter of voluntarily engagement but a matter of enlistment. As shown by ethology and primatology, others are not only benevolent social partners but also malevolent rivals. Social agents are not only those who voluntarily engage themselves in rewarding close relationships but also those who are enrolled, whether they want it or not, in impersonal scripts, situations, and roles. In short, human and nonhuman social life cannot be reduced to experience sharing.
Our proposal is to separate the components of social cognition that Spelke tends to intertwine, namely (a) the conception of “people as individuals with mental experiences and the simulation of shareable actions and experiences of known, individual people” and (b) “the relationships that connect them.” The first component involves the ability to understand and represent the mental states of others, often referred to as “naïve psychology.” The second component of social cognition focuses on “the kind of people” we are dealing with, and on the way we use group-level identification – be it race, gender, kinship, status, or occupation – as a basis for inference, prediction, and action (Hirschfeld, Reference Hirschfeld2001). This ability, frequently referred to as “naïve” or “intuitive sociology,” might be viewed as a distinct core knowledge and could also be expanded. In fact, group membership, by definition, is a social relationship that influences and defines how group members relate to one another. Viewing group membership as a form of primitive social relationship, rather than merely a category-based perception of individuals, is theoretically heuristic: It directs attention toward other basic types of relationships, such as exchange, cooperation, competition, and dominance.
Various elements from Spelke's impressive list of studies argue in favor of a relational conception of social knowledge, a conception that could well be integrated into her account. As mentioned above, she has defended few years ago an account of the core social system based on in-group and out-group reasoning. Another discussion in What Babies Know (Spelke, Reference Spelke2022) could pave the way for a relational view of social cognition, one (too) quickly sidestepped: dominance. Indeed, 9- and 10-month-old infants expect small agents to bow and prostrate in subordination to others of more formidable physical size (Thomsen, Frankenhuis, Ingold-Smith, & Carey, Reference Thomsen, Frankenhuis, Ingold-Smith and Carey2011). Infants as young as 15-month-old demonstrate a strong sensitivity to third-party asymmetric relationships that they expect to remain stable from one conflict (e.g., when the dominant agent repeatedly pushes the subordinate to monopolize a specific area) to another conflict (e.g., when the two agents compete over a desired resource) (Mascaro & Csibra, Reference Mascaro and Csibra2012). When citing these studies, Spelke insists on the fact that these infants are a least 12 months old and are therefore “too old” for their cognitive performances to fall within the scope of core knowledge. Nevertheless, on p. 412, she does mention that 6-month-old infants, when they see a conflict between two individuals from separate social groups, expect the individual from the larger group to prevail (Pun, Birch, & Baron, Reference Pun, Birch and Baron2016). In her persistent effort to exclude social relations from her scope, Spelke even downplays in a note the remarkable social abilities of nonhuman primates by attributing their behavior to a “mix of evolutionary adaptations and by a slow, associative learning” (note 5, p. 412).
And yet, studies on dominance clearly demonstrate that the social world cannot be reduced to mind interactions. From an evolutionary perspective, it makes sense. It would be surprising if humans evolved a finely tuned core knowledge dedicated to numbers or places while only having a single-core system devoted to social relations. To navigate their social environment, people and especially children must identify relevant social relationships, adjust to the norms that govern them, and anticipate the sanctions that reinforce them (Charafeddine et al., Reference Charafeddine, Mercier, Clément, Kaufmann, Berchtold, Reboul and Van der Henst2015; Clément, Bernard, & Kaufmann, Reference Clément, Bernard and Kaufmann2011). They must assess, within a given relationship, who bears obligations toward whom, who possesses the authority to impose duties, who has the entitlement to claim certain goods, and what forms of retaliation are deemed appropriate for obligation violations (Jackendoff, Reference Jackendoff, Jackendoff, Bloom and Wynn1999). Social relationships have a normative structure that, for instance, excludes the very possibility of experience sharing in situation of dominance. Another recent study might plead for a relational model of social cognition. In their wonderful experiments on the early concept of intimacy, Spelke and colleagues show that 8- to 10-month-old infants can identify the concrete features of an interaction (saliva sharing and food licking) to infer a type of social relationship (intimacy) and then to use it to expect some apparently unrelated behaviors to occur (helping a person in distress) (Thomas, Woo, Nettle, Spelke, & Saxe, Reference Thomas, Woo, Nettle, Spelke and Saxe2022). Because infants set aside the individual traits of specific characters (such as the prosociality of the actress) to focus on the social relationship itself, this experiment supports the idea of a domain-specific relational processing.
Admitting social relationship within the realm of core knowledge brings us back to our initial, ontological comment about numbers and forms. What might be the place of relationship processing in our cognitive architecture? Is social relationship a quasi-perceptual entity that deserves an ontological status of its own or a second-order, language-infected property? We have argued elsewhere that social relationships might be an ontological primitive (Kaufmann & Clément, Reference Kaufmann and Clément2014). Patterns of relationships are recognizable in situ as a succession of constraining and enabling affordances, each action affording a set of possible subsequent actions. Moreover, they have the strong inductive potential that characterizes abstract concepts. Finally, they are mostly shared with nonhuman primates, one of the main conditions for core knowledge according to Spelke.
Within this perspective, the phrase “navigating in the social world” must be taken literally. During their first months, infants must position themselves not only in physical space but also in social space. They must identify the normative boundaries, obstacles, and gaps that obstruct their path. They have also to map the power dynamics they are entangled in. Indeed, dominance is not solely a concern for animals in the wild or adults in culture; it is a vital concern for infants, who are constantly confronted with their complete state of dependency (Hrdy & Burkart, Reference Hrdy and Burkart2020). They have to evaluate their own power, in terms of capacity and possibility of acting as well as in terms of permission and constraint. To put it otherwise, the task of newborns is not primarily to know their surroundings, but rather to find their place among others and assume the position that awaits them within a pre-established kinship and community. As Rochat (Reference Rochat2003) suggests it, the infant's sense of self is not “ego-logical” but “eco-logical”: It is based on the proprioceptive cartography and relational mapping of her spatial and social environment. Of course, to better understand this self-in-the-making, Spelke's brilliant odyssey of the human mind is indispensable.
You have
Access
As exciting as it is erudite, Spelke's (Reference Spelke2022) book provides a remarkable overview of her groundbreaking research into “innate systems of core knowledge.” To her, those domain-specific, innate cognitive capacities are essential for carving out “our cognitive territory into more manageable units.” As shown by Spelke's successive attempts, identifying core knowledge systems is however a complicated task. In her book, she argues for six core systems – objects, numbers, places, forms, agents, and social beings. The phylogenetic and ontogenetic functions of those core systems, interposed between percepts and beliefs, are to support the “first informative encounters with the entities they serve to represent” (p. xix). Although Spelke's hypothesis is quite convincing, we would like to make two critical arguments. Our first argument concerns the ontological divergence between the six core systems and the “entities” they are supposed to deal with. The second argument focuses on the conceptual primitives at the heart of the “core system of social beings.”
From an ontological point of view, numbers and shapes are not individual entities in themselves, but universal properties of given entities. Without reactivating here the old metaphysical debate between the ontology of individuals and the ontology of universals, it is hard to ignore that properties such as quantity, form, or place have a higher degree of abstraction than objects or agents. Wheras first-order concrete entities are defined by their material existence and causal power in the physical world, second-order abstract properties are more complex ways of referring to and qualifying those entities. There is an ontological hierarchy between individual entities endowed with spatiotemporal coordinates and the abstract properties that these same entities instantiate. Herein lies our question. If entities and properties are not on the same ontological page, can they be situated in the same cognitive territory? Or do we need to organize so-called core systems into a multilayer cognitive architecture in which properties have not the same status as objects and agents? This question, both ontological and cognitive, also underlies the point we would like to make about social cognition.
Within Spelke's framework, newborns, who are literally infants, that is, children who are “not able to speak,” have at their disposal two independent, automatic core knowledge systems dedicated to the nonphysical world: the “agent system” and the “social system.” The agent system allows infants to represent self-moving entities as agents who act in a causally effective and perceptually guided way. Present in human newborns and also in newly hatched chicks, the agent system consists in seeing self-propelled entities as goal-oriented beings that cause changes in other objects only on contact and through their motion. As for the social system, it targets social beings, that is, people “who endow one another with experiences like their own and who share their experiences in states of engagement.” To Spelke, evidence for a core system dedicated to “shareable experiences of known, individual people” is drawn from newborn's sensitivity to mutual gazes and affective engagements with their caregivers. As core knowledge is modular, the distinction between the agent and the social systems has strange consequences for the first year of infants’ social life: People around them can be conceived either as entities sharing phenomenal experiences with them (social beings) or as intentional entities exerting their causal power over objects (agents). It is only at the end of the first year of life that these two central systems can be combined, and infants finally become able to see people as social agents, endowed with phenomenal and intentional properties. According to Spelke, this cognitive achievement is because of the progressive mastery of language and the access it provides to the plurality of others’ perspectives.
It is worth mentioning that Spelke used to have a different view of social core knowledge, which served to identify “members of one's own social group” and “to guide social interactions with in- and out-group members” (Kinzler & Spelke, Reference Kinzler and Spelke2007, p. 257). In her book, Spelke appears to have modified her perspective. Her definition of the social is primarily, if not solely, intersubjective, as evidenced by her insistence on mimicry, intimacy, proximity, imitation, and emotional sharing. Certainly, the prosocial inclination to build “We-ness” or “togetherness” is essential to individual survival, on both phylogenetic and ontogenetic scales. However, it is questionable whether the social is reducible to “like-ness” in both senses of similarity and kindness. Indeed, in Spelke's core social system, the social world is a matter of affiliation or sociability: It is made up of entities “like-me,” then “like-us.” In this egocentric model, newborns build their social world progressively, as if their knowledge followed a series of concentric circles. First exposed to intersubjective sharing and face-to-face interaction, ego bridges the gap between itself and the minds of those around it through mimicry and experience sharing, and then gradually learns to expand its social network.
The problem is that “being one of us” is not the only meaning of the social. Social groups are defined by a normative structure that enables their members to predict how others are likely to behave, given their position in the social order. Besides intersubjective, horizontal relationships between acquaintances, society is also based on vertical subordination to impersonal constraints and social hierarchies. In this sense, the mark of the social is not We-ness but the impersonal relations of interdependence between status bearers or role takers. Social structures are not a matter of voluntarily engagement but a matter of enlistment. As shown by ethology and primatology, others are not only benevolent social partners but also malevolent rivals. Social agents are not only those who voluntarily engage themselves in rewarding close relationships but also those who are enrolled, whether they want it or not, in impersonal scripts, situations, and roles. In short, human and nonhuman social life cannot be reduced to experience sharing.
Our proposal is to separate the components of social cognition that Spelke tends to intertwine, namely (a) the conception of “people as individuals with mental experiences and the simulation of shareable actions and experiences of known, individual people” and (b) “the relationships that connect them.” The first component involves the ability to understand and represent the mental states of others, often referred to as “naïve psychology.” The second component of social cognition focuses on “the kind of people” we are dealing with, and on the way we use group-level identification – be it race, gender, kinship, status, or occupation – as a basis for inference, prediction, and action (Hirschfeld, Reference Hirschfeld2001). This ability, frequently referred to as “naïve” or “intuitive sociology,” might be viewed as a distinct core knowledge and could also be expanded. In fact, group membership, by definition, is a social relationship that influences and defines how group members relate to one another. Viewing group membership as a form of primitive social relationship, rather than merely a category-based perception of individuals, is theoretically heuristic: It directs attention toward other basic types of relationships, such as exchange, cooperation, competition, and dominance.
Various elements from Spelke's impressive list of studies argue in favor of a relational conception of social knowledge, a conception that could well be integrated into her account. As mentioned above, she has defended few years ago an account of the core social system based on in-group and out-group reasoning. Another discussion in What Babies Know (Spelke, Reference Spelke2022) could pave the way for a relational view of social cognition, one (too) quickly sidestepped: dominance. Indeed, 9- and 10-month-old infants expect small agents to bow and prostrate in subordination to others of more formidable physical size (Thomsen, Frankenhuis, Ingold-Smith, & Carey, Reference Thomsen, Frankenhuis, Ingold-Smith and Carey2011). Infants as young as 15-month-old demonstrate a strong sensitivity to third-party asymmetric relationships that they expect to remain stable from one conflict (e.g., when the dominant agent repeatedly pushes the subordinate to monopolize a specific area) to another conflict (e.g., when the two agents compete over a desired resource) (Mascaro & Csibra, Reference Mascaro and Csibra2012). When citing these studies, Spelke insists on the fact that these infants are a least 12 months old and are therefore “too old” for their cognitive performances to fall within the scope of core knowledge. Nevertheless, on p. 412, she does mention that 6-month-old infants, when they see a conflict between two individuals from separate social groups, expect the individual from the larger group to prevail (Pun, Birch, & Baron, Reference Pun, Birch and Baron2016). In her persistent effort to exclude social relations from her scope, Spelke even downplays in a note the remarkable social abilities of nonhuman primates by attributing their behavior to a “mix of evolutionary adaptations and by a slow, associative learning” (note 5, p. 412).
And yet, studies on dominance clearly demonstrate that the social world cannot be reduced to mind interactions. From an evolutionary perspective, it makes sense. It would be surprising if humans evolved a finely tuned core knowledge dedicated to numbers or places while only having a single-core system devoted to social relations. To navigate their social environment, people and especially children must identify relevant social relationships, adjust to the norms that govern them, and anticipate the sanctions that reinforce them (Charafeddine et al., Reference Charafeddine, Mercier, Clément, Kaufmann, Berchtold, Reboul and Van der Henst2015; Clément, Bernard, & Kaufmann, Reference Clément, Bernard and Kaufmann2011). They must assess, within a given relationship, who bears obligations toward whom, who possesses the authority to impose duties, who has the entitlement to claim certain goods, and what forms of retaliation are deemed appropriate for obligation violations (Jackendoff, Reference Jackendoff, Jackendoff, Bloom and Wynn1999). Social relationships have a normative structure that, for instance, excludes the very possibility of experience sharing in situation of dominance. Another recent study might plead for a relational model of social cognition. In their wonderful experiments on the early concept of intimacy, Spelke and colleagues show that 8- to 10-month-old infants can identify the concrete features of an interaction (saliva sharing and food licking) to infer a type of social relationship (intimacy) and then to use it to expect some apparently unrelated behaviors to occur (helping a person in distress) (Thomas, Woo, Nettle, Spelke, & Saxe, Reference Thomas, Woo, Nettle, Spelke and Saxe2022). Because infants set aside the individual traits of specific characters (such as the prosociality of the actress) to focus on the social relationship itself, this experiment supports the idea of a domain-specific relational processing.
Admitting social relationship within the realm of core knowledge brings us back to our initial, ontological comment about numbers and forms. What might be the place of relationship processing in our cognitive architecture? Is social relationship a quasi-perceptual entity that deserves an ontological status of its own or a second-order, language-infected property? We have argued elsewhere that social relationships might be an ontological primitive (Kaufmann & Clément, Reference Kaufmann and Clément2014). Patterns of relationships are recognizable in situ as a succession of constraining and enabling affordances, each action affording a set of possible subsequent actions. Moreover, they have the strong inductive potential that characterizes abstract concepts. Finally, they are mostly shared with nonhuman primates, one of the main conditions for core knowledge according to Spelke.
Within this perspective, the phrase “navigating in the social world” must be taken literally. During their first months, infants must position themselves not only in physical space but also in social space. They must identify the normative boundaries, obstacles, and gaps that obstruct their path. They have also to map the power dynamics they are entangled in. Indeed, dominance is not solely a concern for animals in the wild or adults in culture; it is a vital concern for infants, who are constantly confronted with their complete state of dependency (Hrdy & Burkart, Reference Hrdy and Burkart2020). They have to evaluate their own power, in terms of capacity and possibility of acting as well as in terms of permission and constraint. To put it otherwise, the task of newborns is not primarily to know their surroundings, but rather to find their place among others and assume the position that awaits them within a pre-established kinship and community. As Rochat (Reference Rochat2003) suggests it, the infant's sense of self is not “ego-logical” but “eco-logical”: It is based on the proprioceptive cartography and relational mapping of her spatial and social environment. Of course, to better understand this self-in-the-making, Spelke's brilliant odyssey of the human mind is indispensable.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
None.