In general, we find the book highly stimulating and a pleasure to read! We endorse the effort of presenting all the compelling evidence of core knowledge in both young human beings and animals – as well as the thought-provoking suggestion regarding language as the “secret sauce,” allowing core knowledge representations to meet and to be combined mainly because of the recursively combinatorial and compositional properties of language. However, it may not be language as such, but rather the ability to abstract core knowledge features from a given domain and to combine these in an abstract medium that is key. This medium can, but does not, in our view, have to be, language-based per se. Language may thus be an important subset of this abstraction medium – not the medium as such.
Our reasoning for questioning that the hub for core knowledge information to meet and be combined should be exclusively language-based is the following: Although language is beyond any doubt a powerful abstract medium for representing and combining information from different core knowledge domains, language may not be the candidate medium for all such tasks. In fact, many intellectual tasks that most researchers would consider uniquely human, and in which core knowledge is combined, do not seem to be solved in a language-based medium. Consider the highly diverse cognitive tasks of recalling an episodic memory (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005), playing chess (e.g., Coates, Reference Coates2013), or conducting mental rotation tasks (e.g., Shepard & Mentzler, Reference Shepard and Mentzler1971). Although each of these activities involves core knowledge components of at least the domains “object” and “space,” and can be, and often is, described or reported by means of language, none of them is predominantly conducted in a language-based medium. This can be elaborated as below.
Recalling episodic memories
Recalling an episodic memory refers to the task of recalling specific, personally experienced events from the past (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005), as for instance the first author remembering the birth of his third child or recalling playing golf with his brother last Saturday. Although these recollections are very different, they share the features of concerning personally experienced unique events that took place at a certain place and at a certain point in time. Episodic memories are crucial for our sense of who we are and the history we share with other people (e.g., Harris, Rasmussen, & Berntsen, Reference Harris, Rasmussen and Berntsen2014). From an evolutionary perspective, episodic memories also have important directive functions by reminding us of successes and mistakes in the past, which helps us to fare better tomorrow than yesterday (e.g., Allen & Fortin, Reference Allen and Fortin2013).
A key feature of episodic memories is autonoetic awareness, that is, the “I-was-there-sense,” which reflects an inherent part of recall of episodic memories in our mind's eye (Tulving, Reference Tulving, Terrace and Metcalfe2005). For the sake of simplicity, we will use the “golf-memory” mentioned above as an example. When recalling playing golf with his brother last Saturday, the first author relives playing in the pine woods at Nordvestjysk Golf Club, Denmark. This recollection is filled with images (e.g., the beautiful view from the first tee), sensory input concerning the weather (e.g., it was remarkably warm and sunny for September in Denmark, although the forecast had predicted rain), and the setting (e.g., the distinct smell of pine trees in the forest, and experiencing that the greens were slower than they looked), as well as emotions concerning the quality of his play (he played terrible, with several lost balls and missed puts). Although the recollection involves language (e.g., conversations with his brother on the golf course) and can be reported by means of language as attempted here, the autonoetic awareness of recalling this (or any other) specific event is not primarily language-based. Rather the recollection is predominantly multimodal and constituted by mental images (e.g., recalling the vision of his drive on hole 2 disappearing in the woods), sensory information (e.g., the pleasant surprise of experiencing the warm sun), and emotions (e.g., being disappointed of the terrible play, but enjoying the company of his brother; e.g., Rubin, Reference Rubin2005, Reference Rubin2006; Tulving, Reference Tulving, Terrace and Metcalfe2005).
The claim that language is not a crucial aspect of episodic memories is further supported by two key empirical findings. First, language impairments do typically not affect the ability to recall personally experienced events from the past (Rubin, Reference Rubin2006). If language had been key in episodic memory recollections, then language impairments should have affected recall. Second, the case of KC (Rosenbaum et al., Reference Rosenbaum, Köhler, Schacter, Moscovitch, Westmacott, Black, Gao and Tulving2005), who because of a motorcycle accident had virtually no episodic memories, had no apparent language impairments. Again, if language had been the medium for episodic recollections, it becomes difficult to explain the virtually complete lack of episodic memories in KC, who at the same time had no obvious language impairments. Finally, some scholars claim that episodic memories as well as the ability to envision the future may have evolved before language, not after (Kellogg, Reference Kellogg2023). In short: Episodic memories are central in human beings' mental life, and according to some scholars, uniquely human (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005). Although episodic memories typically involve representing and combining information from core knowledge systems (e.g., “object” and “space”), the defining autonoetic awareness (the “I-was-there-sense”) is not language-based in nature.
Playing chess
Playing chess involves a range of cognitive skills, including pattern recognition, memory, imagery, and decision making (e.g., Campitelli, Reference Campitelli, Hambrick, Campitelli and Macnamara2017). It is common to distinguish between tactical and strategical aspects of the game. Tactical operations concern forced sequences of moves leading to a clear advantage for one player. Strategical operations involve more long-term planning concerning positional aspects of the position such as for instance pawn structures and piece placement and activity (e.g., Coates, Reference Coates2013). Because of space limitations, we here consider tactical aspects only. Occasionally, during a mating attack, there may for instance be a theoretical forced mate in three moves. This means that a player can win for certain, if she manages to find the correct sequence of three consecutive moves, regardless of which moves the opponent choses as responses. To find such a sequence the player will have to, in her minds' eye, imagine all possible move orders until the end – a process at times (and slightly misleading) called search in the cognitive literature on chess (Coates, Reference Coates2013). Finding a forced mate in three moves clearly draws on core knowledge information from at least the domains “object” and “space,” as well as conforming to the rules of chess, of course. This information needs to be represented, combined, and processed in a mental space. However, when solving such a task, language plays a very minor role, if any at all. Rather, the process mainly draws on imagery in a mental space, in which the moves of the chess pieces are simulated one by one in the minds' eye of the attacker (e.g., Campitelli, Reference Campitelli, Hambrick, Campitelli and Macnamara2017; Coates, Reference Coates2013). The process of finding a forced mate in three moves (or any other tactical chess maneuver) can be described to others in language, but the process as such is not language-based.
The importance of imagery (but not language) when playing chess is further supported by the results from studies on blindfold chess (i.e., an especially challenging version of chess where you play without a visible chess board and pieces, and hence, in your mind's eye only). In one illustrative experiment (Saariluoma, Reference Saariluoma1991, exp. 1), chess players were given the task of following three already played chess games in their mind's eye. The moves of the games were read out to the players by means of conventional algebraic chess annotation (i.e., a standard chess annotation system) one by one at the pace of one move (for both white and black) every fourth second. After 15 and 25 moves the players were asked about the position of the pieces. While attempting to follow the games, the players were exposed to one of the three different kinds of possible interferences in the form of additional tasks: (1) No interference (control), (2) articulatory interference (repetitive pronunciation of a syllable), and (3) imaging interference (to imagine the syllable). The results revealed that although the articulatory interference (and the control) had no diminishing effect on performance, imaging interference had (Saariluoma, Reference Saariluoma1991). Thus, even though the moves were presented in a language format, only imaging (but not articulatory) interference negatively affected performance. If language had been crucial for keeping track of the game, then articulatory interference should have affected performance, but it did not. Hence, the task of keeping track of an unfolding chess game by means of verbally presented chess moves in algebraic annotation does not seem to be conducted in a language-based mental medium, but predominantly in a medium based on visual imagination.
Conducting mental rotation
In their seminal paper, Shepard and Mentzler (Reference Shepard and Mentzler1971) asked participants to assess whether possibly identical three-dimensional (3D) objects presented at different angles were the same or different. To compare the two objects from the same angle, participants had to mentally rotate one of the objects. The results revealed a linear relation between the required rotation angle for direct comparison and response time. Mental rotation involves core knowledge information from the domains “object” and “space,” but how are these bits of information represented and combined? It has been debated whether mental rotation is a modal process (e.g., Kosslyn, Ganis, & Thompson, Reference Kosslyn, Ganis and Thompson2001), or whether it may be conducted by means of abstract language-based propositions (e.g., Pylyshyn, Reference Pylyshyn2003). However, the evidence seems to favor the modal approach. First, as mentioned above, Shepard and Mentzler (Reference Shepard and Mentzler1971) found a linear relationship between rotation angle and response time, which is in accordance with the interpretation that mental rotation is a modal act. If mental rotation had been carried out by language-based propositions, there is no obvious reason why response time should be a linear function of rotation angle. Second, brain-imaging studies reveal that both the ventral and the dorsal streams are activated, when participants conduct mental rotation tasks (e.g., Milivojevic, Hamm, & Corballis, Reference Milivojevic, Hamm and Corballis2008). Again, these results are in accordance with the modal interpretation, whereas a language-based interpretation seems less straightforward.
Conclusion
In summary, the above examples represent different tasks in which core knowledge information is represented and combined in an abstract medium that does not seem to be language-based as proposed by Spelke. Although we acknowledge that many cognitive tasks in which core knowledge information is represented and combined take place in a language-based medium (as for instance writing this commentary), there are important exceptions as exemplified above. Consequently, language only seem to constitute a subset of a candidate abstract medium serving to represent and combine core knowledge components when solving problems.
You have
Access
In her book What babies know, Spelke (Reference Spelke2022) presents the most recent and comprehensive account of her core knowledge theory on infant cognition. A central claim in Spelke's earlier accounts on her core knowledge approach (e.g., Spelke, Breinlinger, Macomber, & Jacobson, Reference Spelke, Breinlinger, Macomber and Jacobson1992) was that only human beings possessed domain-specific core knowledge systems. However, several of these core knowledge systems have subsequently been found to be present in other animals. Therefore, core knowledge systems per se could no longer explain why the world of human beings is so radically different from that of other species. For instance, only human beings invent and use smart phones, spaceships, and generic symbol systems such as the alphabet and the 10-digit systems. Hence, on accepting the core knowledge approach, it raises the key question of why human beings, relative to any other species, are so much more efficient in employing and combining their core knowledge domains. In short: What is the extra “secret sauce”? In What babies know, Spelke (Reference Spelke2022) suggests that it is our language that sets human beings apart from all other creatures on Earth. First, Spelke presents the bold claim that language is the driving force when infants around their first birthday develop uniquely human concepts of other people and their mental states. Second, Spelke claims that it is language that provides the medium in which contents from different core knowledge systems can be represented and combined economically and efficiently. Here, we concentrate on the second claim.
In general, we find the book highly stimulating and a pleasure to read! We endorse the effort of presenting all the compelling evidence of core knowledge in both young human beings and animals – as well as the thought-provoking suggestion regarding language as the “secret sauce,” allowing core knowledge representations to meet and to be combined mainly because of the recursively combinatorial and compositional properties of language. However, it may not be language as such, but rather the ability to abstract core knowledge features from a given domain and to combine these in an abstract medium that is key. This medium can, but does not, in our view, have to be, language-based per se. Language may thus be an important subset of this abstraction medium – not the medium as such.
Our reasoning for questioning that the hub for core knowledge information to meet and be combined should be exclusively language-based is the following: Although language is beyond any doubt a powerful abstract medium for representing and combining information from different core knowledge domains, language may not be the candidate medium for all such tasks. In fact, many intellectual tasks that most researchers would consider uniquely human, and in which core knowledge is combined, do not seem to be solved in a language-based medium. Consider the highly diverse cognitive tasks of recalling an episodic memory (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005), playing chess (e.g., Coates, Reference Coates2013), or conducting mental rotation tasks (e.g., Shepard & Mentzler, Reference Shepard and Mentzler1971). Although each of these activities involves core knowledge components of at least the domains “object” and “space,” and can be, and often is, described or reported by means of language, none of them is predominantly conducted in a language-based medium. This can be elaborated as below.
Recalling episodic memories
Recalling an episodic memory refers to the task of recalling specific, personally experienced events from the past (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005), as for instance the first author remembering the birth of his third child or recalling playing golf with his brother last Saturday. Although these recollections are very different, they share the features of concerning personally experienced unique events that took place at a certain place and at a certain point in time. Episodic memories are crucial for our sense of who we are and the history we share with other people (e.g., Harris, Rasmussen, & Berntsen, Reference Harris, Rasmussen and Berntsen2014). From an evolutionary perspective, episodic memories also have important directive functions by reminding us of successes and mistakes in the past, which helps us to fare better tomorrow than yesterday (e.g., Allen & Fortin, Reference Allen and Fortin2013).
A key feature of episodic memories is autonoetic awareness, that is, the “I-was-there-sense,” which reflects an inherent part of recall of episodic memories in our mind's eye (Tulving, Reference Tulving, Terrace and Metcalfe2005). For the sake of simplicity, we will use the “golf-memory” mentioned above as an example. When recalling playing golf with his brother last Saturday, the first author relives playing in the pine woods at Nordvestjysk Golf Club, Denmark. This recollection is filled with images (e.g., the beautiful view from the first tee), sensory input concerning the weather (e.g., it was remarkably warm and sunny for September in Denmark, although the forecast had predicted rain), and the setting (e.g., the distinct smell of pine trees in the forest, and experiencing that the greens were slower than they looked), as well as emotions concerning the quality of his play (he played terrible, with several lost balls and missed puts). Although the recollection involves language (e.g., conversations with his brother on the golf course) and can be reported by means of language as attempted here, the autonoetic awareness of recalling this (or any other) specific event is not primarily language-based. Rather the recollection is predominantly multimodal and constituted by mental images (e.g., recalling the vision of his drive on hole 2 disappearing in the woods), sensory information (e.g., the pleasant surprise of experiencing the warm sun), and emotions (e.g., being disappointed of the terrible play, but enjoying the company of his brother; e.g., Rubin, Reference Rubin2005, Reference Rubin2006; Tulving, Reference Tulving, Terrace and Metcalfe2005).
The claim that language is not a crucial aspect of episodic memories is further supported by two key empirical findings. First, language impairments do typically not affect the ability to recall personally experienced events from the past (Rubin, Reference Rubin2006). If language had been key in episodic memory recollections, then language impairments should have affected recall. Second, the case of KC (Rosenbaum et al., Reference Rosenbaum, Köhler, Schacter, Moscovitch, Westmacott, Black, Gao and Tulving2005), who because of a motorcycle accident had virtually no episodic memories, had no apparent language impairments. Again, if language had been the medium for episodic recollections, it becomes difficult to explain the virtually complete lack of episodic memories in KC, who at the same time had no obvious language impairments. Finally, some scholars claim that episodic memories as well as the ability to envision the future may have evolved before language, not after (Kellogg, Reference Kellogg2023). In short: Episodic memories are central in human beings' mental life, and according to some scholars, uniquely human (e.g., Tulving, Reference Tulving, Terrace and Metcalfe2005). Although episodic memories typically involve representing and combining information from core knowledge systems (e.g., “object” and “space”), the defining autonoetic awareness (the “I-was-there-sense”) is not language-based in nature.
Playing chess
Playing chess involves a range of cognitive skills, including pattern recognition, memory, imagery, and decision making (e.g., Campitelli, Reference Campitelli, Hambrick, Campitelli and Macnamara2017). It is common to distinguish between tactical and strategical aspects of the game. Tactical operations concern forced sequences of moves leading to a clear advantage for one player. Strategical operations involve more long-term planning concerning positional aspects of the position such as for instance pawn structures and piece placement and activity (e.g., Coates, Reference Coates2013). Because of space limitations, we here consider tactical aspects only. Occasionally, during a mating attack, there may for instance be a theoretical forced mate in three moves. This means that a player can win for certain, if she manages to find the correct sequence of three consecutive moves, regardless of which moves the opponent choses as responses. To find such a sequence the player will have to, in her minds' eye, imagine all possible move orders until the end – a process at times (and slightly misleading) called search in the cognitive literature on chess (Coates, Reference Coates2013). Finding a forced mate in three moves clearly draws on core knowledge information from at least the domains “object” and “space,” as well as conforming to the rules of chess, of course. This information needs to be represented, combined, and processed in a mental space. However, when solving such a task, language plays a very minor role, if any at all. Rather, the process mainly draws on imagery in a mental space, in which the moves of the chess pieces are simulated one by one in the minds' eye of the attacker (e.g., Campitelli, Reference Campitelli, Hambrick, Campitelli and Macnamara2017; Coates, Reference Coates2013). The process of finding a forced mate in three moves (or any other tactical chess maneuver) can be described to others in language, but the process as such is not language-based.
The importance of imagery (but not language) when playing chess is further supported by the results from studies on blindfold chess (i.e., an especially challenging version of chess where you play without a visible chess board and pieces, and hence, in your mind's eye only). In one illustrative experiment (Saariluoma, Reference Saariluoma1991, exp. 1), chess players were given the task of following three already played chess games in their mind's eye. The moves of the games were read out to the players by means of conventional algebraic chess annotation (i.e., a standard chess annotation system) one by one at the pace of one move (for both white and black) every fourth second. After 15 and 25 moves the players were asked about the position of the pieces. While attempting to follow the games, the players were exposed to one of the three different kinds of possible interferences in the form of additional tasks: (1) No interference (control), (2) articulatory interference (repetitive pronunciation of a syllable), and (3) imaging interference (to imagine the syllable). The results revealed that although the articulatory interference (and the control) had no diminishing effect on performance, imaging interference had (Saariluoma, Reference Saariluoma1991). Thus, even though the moves were presented in a language format, only imaging (but not articulatory) interference negatively affected performance. If language had been crucial for keeping track of the game, then articulatory interference should have affected performance, but it did not. Hence, the task of keeping track of an unfolding chess game by means of verbally presented chess moves in algebraic annotation does not seem to be conducted in a language-based mental medium, but predominantly in a medium based on visual imagination.
Conducting mental rotation
In their seminal paper, Shepard and Mentzler (Reference Shepard and Mentzler1971) asked participants to assess whether possibly identical three-dimensional (3D) objects presented at different angles were the same or different. To compare the two objects from the same angle, participants had to mentally rotate one of the objects. The results revealed a linear relation between the required rotation angle for direct comparison and response time. Mental rotation involves core knowledge information from the domains “object” and “space,” but how are these bits of information represented and combined? It has been debated whether mental rotation is a modal process (e.g., Kosslyn, Ganis, & Thompson, Reference Kosslyn, Ganis and Thompson2001), or whether it may be conducted by means of abstract language-based propositions (e.g., Pylyshyn, Reference Pylyshyn2003). However, the evidence seems to favor the modal approach. First, as mentioned above, Shepard and Mentzler (Reference Shepard and Mentzler1971) found a linear relationship between rotation angle and response time, which is in accordance with the interpretation that mental rotation is a modal act. If mental rotation had been carried out by language-based propositions, there is no obvious reason why response time should be a linear function of rotation angle. Second, brain-imaging studies reveal that both the ventral and the dorsal streams are activated, when participants conduct mental rotation tasks (e.g., Milivojevic, Hamm, & Corballis, Reference Milivojevic, Hamm and Corballis2008). Again, these results are in accordance with the modal interpretation, whereas a language-based interpretation seems less straightforward.
Conclusion
In summary, the above examples represent different tasks in which core knowledge information is represented and combined in an abstract medium that does not seem to be language-based as proposed by Spelke. Although we acknowledge that many cognitive tasks in which core knowledge information is represented and combined take place in a language-based medium (as for instance writing this commentary), there are important exceptions as exemplified above. Consequently, language only seem to constitute a subset of a candidate abstract medium serving to represent and combine core knowledge components when solving problems.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interest
None.