Although I agree with Heyes’ main stance that emphasizes the power of general-purpose mechanisms in contributing to higher cognitive function, I think that she has taken a particular construal that obscures a potentially more interesting situation. Characterizing language as one of multiple mechanisms (selective social learning, imitation, mindreading, and language) resulting from cultural evolution is misleading, as it obscures its role as a principal vehicle of cultural learning and evolution.
It is likely that there are strong interactions between the four mechanisms. By laying out the structure of the book as an enumeration of how each could have come about, the potentially more interesting aspect of their interrelations, and dynamics, is significantly downplayed. This contributes to a false impression that they are independent.
This false impression is particularly the case for language. The cultural structuring role of language has been eloquently elaborated by many including Bruner, Lakoff, Riceour, and Polkingborn. Yet, from the outset the role of language is underplayed: “I argue in this book that it is the information we get from others, handled by general purpose mechanisms, that builds distinctively human ways of thinking” (Heyes Reference Heyes2018, p. 2). Where does this information come from? The problem with Heyes’ characterization is that it hides the notion that the information that we get from others via language may have a special status, so that language is not a mechanism of the same caliber as the other gadgets. Setting language alongside the other cognitive gadgets downplays the importance of language in cultural evolution, as she states “and linguistic communication is far from the only significant channel of cultural inheritance” (p. 169), yet at the same time she notes that “language comes first” (p. 166) with respect to mindreading, and that “through language children learn to read minds” (p. 166).
I will develop three characteristics and arguments for elevating language to a special status in cultural learning. I will then focus on one of these, illustrating how – based on general cognitive mechanisms – language creates a capability for analogical mapping that contributes to theory of mind.
The first special characteristic of language that makes it quite distinct from the other three elements is that language is structured around the event-based organization of human experience (Goldberg Reference Goldberg1995), and thus provides a very high density vehicle for the transfer of information that is central to construction of the distinctively human ways of thinking. Through the structure-defining dimensions of language including time, mode, and aspect, the speaker can communicate very precise spatial, temporal, causal, and other relations between an agent's actions and outcomes. Importantly, in addition to re-describing the already visible aspects of events, language allows for the enrichment of descriptions, by explaining in more detail aspects of events that are not visible to the untrained observer (Bruner Reference Bruner1991; Nelson & Fivush Reference Nelson and Fivush2004), in a process referred to as narrative enrichment (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). Of particular interest here is how language can be used to specify how unseen mental states may have causal roles in the actions of others. This provides a vehicle for structuring and organizing “the information we get from others” that renders language in a different class from the other proposed cognitive gadgets. This is crucial for the transmission of complex knowledge – instructions – of how to make and build complex cultural artifacts (Stout & Chaminade Reference Stout and Chaminade2012). Reciprocally, the structure of language prefigures how human experience is perceived (Ricoeur Reference Ricoeur1984). Ricoeur developed a framework for narrative that involves the encoding of experience, the emplotment of that experience, and then the reception and comprehension of the resulting narrative by the listener. Interestingly, already in the perception phase, language has pre-shaped our perception so as to construe things in a narrative-compatible manner. This is even more apparent in the process of emplotment whereby we construe experience into a form that is consistent with normative structures. Thus, by modeling the event structure of life, language has two unique roles – it allows the precise communication about this structure, and at the same time it filters the perception of the world so that it is consistent with this structure.
The second characteristic of language that makes it the significant channel of cultural inheritance is related to its unique role in making meaning (Bruner Reference Bruner1990), through its creative and normative dimensions. In the technical realm, the creation and labeling of notions like right angle, hypotenuse, and square root allow for the further specification and discovery of properties in a typical example of meaning making. That is, by allowing the creation of expressive forms for new realms of meaning, language is a workhorse of cultural evolution. Likewise, in the social domain, the development of normative schemas of human behavior by the same kind of labeling and enrichment provides the vehicle for the specification of cultural artifacts like creation myths and social norms that are central to the human condition.
Like Bruner, Polkinghorne (Reference Polkinghorne1988) considered narrative as the primary form by which human experience is made meaningful. He considered that having narrative as one of our fundamental structures of comprehension shapes the character of our existence in a particular way, similar to the effects of Ricoeur's emplotment (Ricoeur Reference Ricoeur1984). The realm of meaning is an open system in which new forms of organization can emerge and new meaning systems can develop. Polkinghorne (Reference Polkinghorne1988, p. 31) summarizes the position of Merleau-Ponty that “language takes up the contingencies of existence, and the perceptual openness of life to the natural and intersubjective worlds, and molds them into a meaningfulness that is greater than the meaningfulness they originally hold.” In this context, a crucial aspect of narrative is its capacity for the disclosure or creation of possible worlds (Bruner Reference Bruner2009). Riceour noted that the adequate self-understanding of man is dependent on this dimension of language as a disclosure of possibility. “It is by an understanding of the worlds, actual and possible, opened up by language that we may arrive at a better understanding of ourselves” (Ricoeur & Kearney Reference Ricoeur and Kearney1978, p. 118).
It is worth noting that this notion of meaning, and language's role in creating and porting meaning, is essentially not addressed in Heyes’ analysis, although the notion of information is highly present. Perhaps language becomes more like the other gadgets when it is considered independent of its role of making meaning in Bruner's sense. From an historical perspective on the status of meaning, Bruner (Reference Bruner1990) noted that the initial goal of the cognitive sciences was to understand meaning and the symbolic activities humans employ to make sense of the world and themselves. Then, for various reasons (greatly related to the rise of theoretical computer science and information theory), cognitive science became a science of information processing. In Bruner's words, critics argued that cognitive science may have “gained its technical successes at the price of dehumanizing the very concept of mind it had sought to re-establish” (p. 1). This trend in cognitive science – away from meaning towards information processing – is partially reflected here in the analysis of cultural tools, wherein the notion of meaning in Bruner's sense is largely left unaddressed.
The third (related) characteristic of language is its mechanism for building up new structures from existing structures, and labeling these so that they can in turn be composed into ever more structured representations. Goucha et al. (Reference Goucha, Zaccarella and Friederici2017) postulated that the uniquely human language capability relies on something like the syntactic merge operator, and crucially, the ability to label the results of the merge, and to reuse these new labeled structures in future merge operations. This recursive capability is the foundation for the building of complex hierarchical structure that is crucial to cultural evolution, and is not present in the other cultural learning components that Heyes analyses.
If language is so different, how can it be built up from existing cognitive mechanisms? The answer might be found in the construction grammar framework, which characterizes language as a structured inventory of form-to-meaning mappings. Part of the beauty of this characterization is that the mechanism that learns and generalizes over these mappings between form (words, sentences) and meanings (e.g., event representations) can indeed be from a more general-purpose toolbox. The cue competition hypothesis (Bates & MacWhinney Reference Bates, MacWhinney, MacWhinney and Bates1987; Bates et al. Reference Bates, Wulfeck and MacWhinney1991, Li & Macwhinney Reference Li and Macwhinney2013) holds that across languages, cues including word order, grammatical marking, and intonation are used to code thematic roles and other dimensions of meaning. Using this hypothesis, we re-used a well-characterized model of the primate cortico-striatal sensorimotor sequence learning system (Dominey Reference Dominey1995) in order to learn grammatical constructions (Dominey Reference Dominey2013; Dominey et al. Reference Dominey, Hoen, Blanc and Lelekov-Boissard2003; Dominey et al. Reference Dominey, Inui and Hoen2009; Hinaut & Dominey Reference Hinaut and Dominey2013). Cues including word order and grammatical marking are used by the model to appropriately map sentence form to predicate-argument meaning. Thus, grammatical function words indicate the relations between open class words in the sentence. We then extended this notion to the narrative level, where narrative function words (such as because, before, after) indicate relations between elements of events in a narrative (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). This introduces the concept of narrative construction, which extends the notion of grammatical construction as a form-to-meaning mapping at the sentence level to the narrative level, consistent with the construction framework, where there is a continuum between lexical elements and constructions.
Part of the novelty of this system is that it easily allows a form of narrative enrichment, whereby new narrative relations, such as causal relations between a mental state and action can be created. For example, I want to grasp a glass but it is out of reach, so I ask you to give it to me. In your subsequent narration, you say “I gave you the glass because you wanted it,” and establish a causal link with an unseen mental state (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). Once the construction was learned, it could then be instantiated with different arguments in a form of analogical mapping, thus allowing expression of a narrative where I am trying to learn math but I don't get it, so I ask you to teach it to me, and you teach me, because I wanted you to. The use of such patterns can provide a basis for powerful mechanisms for enrichment including metaphor (Lakoff & Johnson Reference Lakoff and Johnson2008).
It can also provide a simple but powerful tool in the development of theory of mind. In this context, Gallagher and Hutto (Reference Gallagher, Hutto, Zlatev, Racine, Sinha and Itkonen2008) and Hutto (Reference Hutto2007) have developed a narrative practice theory of folk psychology. They noted that humans appear to have powerful mechanisms for intersubjectivity, like the detection of intentions in expressive movement and eye direction, and from an early age we are actively involved in dyadic relations with others. These interactions form the meaning component of a form to meaning mapping in our narrative construction model. Gallagher and Hutto (Reference Gallagher, Hutto, Zlatev, Racine, Sinha and Itkonen2008) then held that in addition, narrative patterns accompany this intersubjective behavior. In our narrative construction model, these narrative patterns contribute to the form component of the form-to-meaning mapping. Then, children understand why people act as they do by being exposed to psychological narratives of intersubjective relations. Through narrative practice they come to understand the norms of behavior in their society, as the narrative forms a structure in which the behavior can be interpreted, from experience (practice). Thus, as similarly noted by Heyes, in the development of theory of mind, language comes first.
In summary, although I applaud Heyes’ analysis which argues for the use of generalized cognitive mechanisms in the development of higher cognitive functions, here in the service of cultural evolution, I fear that the status of language as a unique mechanism for making meaning has been ignored, which gives a (perhaps unintentional) misconstrual. Characterizing language as one of multiple mechanisms resulting from cultural evolution is misleading, as it obscures its role as a principal vehicle of cultural learning and evolution.
Although I agree with Heyes’ main stance that emphasizes the power of general-purpose mechanisms in contributing to higher cognitive function, I think that she has taken a particular construal that obscures a potentially more interesting situation. Characterizing language as one of multiple mechanisms (selective social learning, imitation, mindreading, and language) resulting from cultural evolution is misleading, as it obscures its role as a principal vehicle of cultural learning and evolution.
It is likely that there are strong interactions between the four mechanisms. By laying out the structure of the book as an enumeration of how each could have come about, the potentially more interesting aspect of their interrelations, and dynamics, is significantly downplayed. This contributes to a false impression that they are independent.
This false impression is particularly the case for language. The cultural structuring role of language has been eloquently elaborated by many including Bruner, Lakoff, Riceour, and Polkingborn. Yet, from the outset the role of language is underplayed: “I argue in this book that it is the information we get from others, handled by general purpose mechanisms, that builds distinctively human ways of thinking” (Heyes Reference Heyes2018, p. 2). Where does this information come from? The problem with Heyes’ characterization is that it hides the notion that the information that we get from others via language may have a special status, so that language is not a mechanism of the same caliber as the other gadgets. Setting language alongside the other cognitive gadgets downplays the importance of language in cultural evolution, as she states “and linguistic communication is far from the only significant channel of cultural inheritance” (p. 169), yet at the same time she notes that “language comes first” (p. 166) with respect to mindreading, and that “through language children learn to read minds” (p. 166).
I will develop three characteristics and arguments for elevating language to a special status in cultural learning. I will then focus on one of these, illustrating how – based on general cognitive mechanisms – language creates a capability for analogical mapping that contributes to theory of mind.
The first special characteristic of language that makes it quite distinct from the other three elements is that language is structured around the event-based organization of human experience (Goldberg Reference Goldberg1995), and thus provides a very high density vehicle for the transfer of information that is central to construction of the distinctively human ways of thinking. Through the structure-defining dimensions of language including time, mode, and aspect, the speaker can communicate very precise spatial, temporal, causal, and other relations between an agent's actions and outcomes. Importantly, in addition to re-describing the already visible aspects of events, language allows for the enrichment of descriptions, by explaining in more detail aspects of events that are not visible to the untrained observer (Bruner Reference Bruner1991; Nelson & Fivush Reference Nelson and Fivush2004), in a process referred to as narrative enrichment (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). Of particular interest here is how language can be used to specify how unseen mental states may have causal roles in the actions of others. This provides a vehicle for structuring and organizing “the information we get from others” that renders language in a different class from the other proposed cognitive gadgets. This is crucial for the transmission of complex knowledge – instructions – of how to make and build complex cultural artifacts (Stout & Chaminade Reference Stout and Chaminade2012). Reciprocally, the structure of language prefigures how human experience is perceived (Ricoeur Reference Ricoeur1984). Ricoeur developed a framework for narrative that involves the encoding of experience, the emplotment of that experience, and then the reception and comprehension of the resulting narrative by the listener. Interestingly, already in the perception phase, language has pre-shaped our perception so as to construe things in a narrative-compatible manner. This is even more apparent in the process of emplotment whereby we construe experience into a form that is consistent with normative structures. Thus, by modeling the event structure of life, language has two unique roles – it allows the precise communication about this structure, and at the same time it filters the perception of the world so that it is consistent with this structure.
The second characteristic of language that makes it the significant channel of cultural inheritance is related to its unique role in making meaning (Bruner Reference Bruner1990), through its creative and normative dimensions. In the technical realm, the creation and labeling of notions like right angle, hypotenuse, and square root allow for the further specification and discovery of properties in a typical example of meaning making. That is, by allowing the creation of expressive forms for new realms of meaning, language is a workhorse of cultural evolution. Likewise, in the social domain, the development of normative schemas of human behavior by the same kind of labeling and enrichment provides the vehicle for the specification of cultural artifacts like creation myths and social norms that are central to the human condition.
Like Bruner, Polkinghorne (Reference Polkinghorne1988) considered narrative as the primary form by which human experience is made meaningful. He considered that having narrative as one of our fundamental structures of comprehension shapes the character of our existence in a particular way, similar to the effects of Ricoeur's emplotment (Ricoeur Reference Ricoeur1984). The realm of meaning is an open system in which new forms of organization can emerge and new meaning systems can develop. Polkinghorne (Reference Polkinghorne1988, p. 31) summarizes the position of Merleau-Ponty that “language takes up the contingencies of existence, and the perceptual openness of life to the natural and intersubjective worlds, and molds them into a meaningfulness that is greater than the meaningfulness they originally hold.” In this context, a crucial aspect of narrative is its capacity for the disclosure or creation of possible worlds (Bruner Reference Bruner2009). Riceour noted that the adequate self-understanding of man is dependent on this dimension of language as a disclosure of possibility. “It is by an understanding of the worlds, actual and possible, opened up by language that we may arrive at a better understanding of ourselves” (Ricoeur & Kearney Reference Ricoeur and Kearney1978, p. 118).
It is worth noting that this notion of meaning, and language's role in creating and porting meaning, is essentially not addressed in Heyes’ analysis, although the notion of information is highly present. Perhaps language becomes more like the other gadgets when it is considered independent of its role of making meaning in Bruner's sense. From an historical perspective on the status of meaning, Bruner (Reference Bruner1990) noted that the initial goal of the cognitive sciences was to understand meaning and the symbolic activities humans employ to make sense of the world and themselves. Then, for various reasons (greatly related to the rise of theoretical computer science and information theory), cognitive science became a science of information processing. In Bruner's words, critics argued that cognitive science may have “gained its technical successes at the price of dehumanizing the very concept of mind it had sought to re-establish” (p. 1). This trend in cognitive science – away from meaning towards information processing – is partially reflected here in the analysis of cultural tools, wherein the notion of meaning in Bruner's sense is largely left unaddressed.
The third (related) characteristic of language is its mechanism for building up new structures from existing structures, and labeling these so that they can in turn be composed into ever more structured representations. Goucha et al. (Reference Goucha, Zaccarella and Friederici2017) postulated that the uniquely human language capability relies on something like the syntactic merge operator, and crucially, the ability to label the results of the merge, and to reuse these new labeled structures in future merge operations. This recursive capability is the foundation for the building of complex hierarchical structure that is crucial to cultural evolution, and is not present in the other cultural learning components that Heyes analyses.
If language is so different, how can it be built up from existing cognitive mechanisms? The answer might be found in the construction grammar framework, which characterizes language as a structured inventory of form-to-meaning mappings. Part of the beauty of this characterization is that the mechanism that learns and generalizes over these mappings between form (words, sentences) and meanings (e.g., event representations) can indeed be from a more general-purpose toolbox. The cue competition hypothesis (Bates & MacWhinney Reference Bates, MacWhinney, MacWhinney and Bates1987; Bates et al. Reference Bates, Wulfeck and MacWhinney1991, Li & Macwhinney Reference Li and Macwhinney2013) holds that across languages, cues including word order, grammatical marking, and intonation are used to code thematic roles and other dimensions of meaning. Using this hypothesis, we re-used a well-characterized model of the primate cortico-striatal sensorimotor sequence learning system (Dominey Reference Dominey1995) in order to learn grammatical constructions (Dominey Reference Dominey2013; Dominey et al. Reference Dominey, Hoen, Blanc and Lelekov-Boissard2003; Dominey et al. Reference Dominey, Inui and Hoen2009; Hinaut & Dominey Reference Hinaut and Dominey2013). Cues including word order and grammatical marking are used by the model to appropriately map sentence form to predicate-argument meaning. Thus, grammatical function words indicate the relations between open class words in the sentence. We then extended this notion to the narrative level, where narrative function words (such as because, before, after) indicate relations between elements of events in a narrative (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). This introduces the concept of narrative construction, which extends the notion of grammatical construction as a form-to-meaning mapping at the sentence level to the narrative level, consistent with the construction framework, where there is a continuum between lexical elements and constructions.
Part of the novelty of this system is that it easily allows a form of narrative enrichment, whereby new narrative relations, such as causal relations between a mental state and action can be created. For example, I want to grasp a glass but it is out of reach, so I ask you to give it to me. In your subsequent narration, you say “I gave you the glass because you wanted it,” and establish a causal link with an unseen mental state (Mealier et al. Reference Mealier, Pointeau, Mirliaz, Ogawa, Finlayson and Dominey2017). Once the construction was learned, it could then be instantiated with different arguments in a form of analogical mapping, thus allowing expression of a narrative where I am trying to learn math but I don't get it, so I ask you to teach it to me, and you teach me, because I wanted you to. The use of such patterns can provide a basis for powerful mechanisms for enrichment including metaphor (Lakoff & Johnson Reference Lakoff and Johnson2008).
It can also provide a simple but powerful tool in the development of theory of mind. In this context, Gallagher and Hutto (Reference Gallagher, Hutto, Zlatev, Racine, Sinha and Itkonen2008) and Hutto (Reference Hutto2007) have developed a narrative practice theory of folk psychology. They noted that humans appear to have powerful mechanisms for intersubjectivity, like the detection of intentions in expressive movement and eye direction, and from an early age we are actively involved in dyadic relations with others. These interactions form the meaning component of a form to meaning mapping in our narrative construction model. Gallagher and Hutto (Reference Gallagher, Hutto, Zlatev, Racine, Sinha and Itkonen2008) then held that in addition, narrative patterns accompany this intersubjective behavior. In our narrative construction model, these narrative patterns contribute to the form component of the form-to-meaning mapping. Then, children understand why people act as they do by being exposed to psychological narratives of intersubjective relations. Through narrative practice they come to understand the norms of behavior in their society, as the narrative forms a structure in which the behavior can be interpreted, from experience (practice). Thus, as similarly noted by Heyes, in the development of theory of mind, language comes first.
In summary, although I applaud Heyes’ analysis which argues for the use of generalized cognitive mechanisms in the development of higher cognitive functions, here in the service of cultural evolution, I fear that the status of language as a unique mechanism for making meaning has been ignored, which gives a (perhaps unintentional) misconstrual. Characterizing language as one of multiple mechanisms resulting from cultural evolution is misleading, as it obscures its role as a principal vehicle of cultural learning and evolution.