The story of the sea squirt larva that settles on a rock, transforms into its sessile adult form, and absorbs its own brain, is famous and oft-repeated, as is the inference that the function of brains is to guide movement (Llinas, Reference Llinas2002). More broadly, the independent origins of complex brains are consistently associated with complex motility. The message and full implications of the story, however, are yet to substantially influence mainstream cognitive science, neuroscience and – the field we focus on here – comparative psychology. These fields tend to treat cognition as occupying a separate realm to the sensory-motor mechanisms that guide animals’ behaviour in their real-world interactions with the environment; a dualist position whereby cognition floats free of an animal’s bodily engagement with the world. By linking cognitive evolution to details of sensory-motor adaptations, the target article provides a welcome antidote and a stimulus for a more biologically realistic, phylogenetically informed and embodied grounding for comparative psychology.
Elsewhere we argue that, in taking its concepts and ontology from mainstream cognitive science, comparative psychology has turned into a blind alley (Barton & Barrett, Reference Barton and Barrett2025). Cognition is framed as the computation of abstract representations, and cognitive evolution is characterised as a process of increasing the power of generic computational devices independent of mechanisms of sensory-motor control. Indeed, some authors are very explicit about the need to separate cognition from sensory-motor control (e.g., Chittka et al., Reference Chittka, Rossiter, Skorupski and Fernando2012). This has led to a search for “species-neutral” capacities: those that could theoretically exist in any species with a large enough brain or enough neurons, regardless of its phylogenetic history, body plan, organisation of its nervous system and the manner in which it senses, explores, and learns about the world. Comparative psychology is now a quagmire of apparently insoluble debate about cognition and intelligence (e.g., whether other species can reason about unobservable phenomena) and, more generally, whether and when the cognition of other species counts as “human-like” (Barton and Barrett, Reference Barton and Barrett2025).
The target article plots a route out of the quagmire by emphasising that cognition is grounded in the details of how animals perceive, explore, and learn about their worlds. Although we are wary of the potentially distorting effects of terminology such as “advanced” or “apical” versus “basal” traits (target article, Figure 5) and the notion of progression through evolutionary stages (target article, Figure 6), we are fundamentally supportive of the major goal: to link the evolution of bodies, sensory-motor mechanisms, brains, and cognition.
The target article makes clear how the evolution of complex behaviours is associated with sophisticated mechanisms that mediate the animal’s direct interaction with its environment. One area that needs more clarity, however, is what, exactly, is meant by “cognition”? This is something on which there has been a surprising lack of consensus (e.g., Bayne et al., Reference Bayne, Brainard, Byrne, Chittka, Clayton, Heyes, Mather, Ölveczky, Shadlen, Suddendorf and Webb2019). Coombs & Trestman rather hedge their bets, referring throughout, perhaps with deliberate ambiguity, to “cognition/behaviour.” They offer two sets of defining traits, one behavioural (central place foraging and constructive behaviour), and the other a set of learning processes under the umbrella of Unlimited Associative Learning (UAL). Cognition is not usually conceived in either purely behavioural or learning terms. Indeed, behaviour is usually taken as the outward manifestation of cognition. This being the case, it becomes circular to define complex cognition as complex behaviour. Circularity aside, maintaining the distinction between behaviour and mechanism is important because behavioural complexity can arise in different ways, including through the use of relatively simple mechanisms (Barrett, Reference Barrett2011). While the links made by C&T between behaviours and underlying processes are plausible, care is needed not to simply equate them.
The emphasis on UAL is nevertheless welcome, as it dispels the common notion that cognition and associative learning (AL) are alternative explanations for behaviour. C&T derive their concept of UAL from Ginsburg and Jablonka (Reference Ginsburg and Jablonka2021), who show how associative learning, when implemented by complex nervous systems, can produce flexible behaviour through the integration of compound stimuli and chaining, in ways that look thoroughly “cognitive” (see also Heyes, Reference Heyes2012). We should, therefore, be asking not whether learning is involved, but how learning builds cognitive processes – that is, how much integration, and of what kind, across which learning contexts is needed to explain behaviour? AL is often characterised as a “domain general mechanism” but it is not, because AL is not something that can occur independently of specific mechanisms – indeed, it is the specific mechanisms that do the learning. AL has common features across these mechanisms but there’s no AL part of the brain – it involves the whole brain, and brains are structured in distinctive ways.
Where does all this leave the received wisdom that cognition is neither sensory-motor control nor learning (MacPhail, Reference MacPhail1982; Chittka et al., Reference Chittka, Rossiter, Skorupski and Fernando2012; Triki et al., Reference Triki, Aellen, van Schaik and Bshary2021)?. We suggest that an even more radical step is required: viz, that cognition encompasses all of the processes that guide and control behaviour as the animal responds to the challenges of surviving and reproducing. That is, cognition is not a distinct phenomenon separable from sensory-motor capacities. Hence, to study cognitive evolution, we need to explicitly integrate sensory-motor specialisations into our conception of cognitive processes – they do not simply set the stage for cognitive evolution to occur. Such a stance would further enrich C&T’s framework by ridding it of the last vestiges of dualism, which will continue to lurk as long as we distinguish between “cognitive” and “non-cognitive” traits.
One implication of this embodied perspective is that it highlights evolutionary divergence as well as convergence. Convergence certainly happens at the behavioural level, but this is where the real questions arise – we need to think more deeply about the mechanisms used by different lineages to produce similar behaviour or solve similar problems; these will surely differ given different bodies, sensory-motor capacities, and ecological niches. We need to consider the whole animal, and ask how it solves problems in the wild, and what it might have been selected to do, instead of asking how much of some human-constructed notion of intelligence it seems likely to possess.
The story of the sea squirt larva that settles on a rock, transforms into its sessile adult form, and absorbs its own brain, is famous and oft-repeated, as is the inference that the function of brains is to guide movement (Llinas, Reference Llinas2002). More broadly, the independent origins of complex brains are consistently associated with complex motility. The message and full implications of the story, however, are yet to substantially influence mainstream cognitive science, neuroscience and – the field we focus on here – comparative psychology. These fields tend to treat cognition as occupying a separate realm to the sensory-motor mechanisms that guide animals’ behaviour in their real-world interactions with the environment; a dualist position whereby cognition floats free of an animal’s bodily engagement with the world. By linking cognitive evolution to details of sensory-motor adaptations, the target article provides a welcome antidote and a stimulus for a more biologically realistic, phylogenetically informed and embodied grounding for comparative psychology.
Elsewhere we argue that, in taking its concepts and ontology from mainstream cognitive science, comparative psychology has turned into a blind alley (Barton & Barrett, Reference Barton and Barrett2025). Cognition is framed as the computation of abstract representations, and cognitive evolution is characterised as a process of increasing the power of generic computational devices independent of mechanisms of sensory-motor control. Indeed, some authors are very explicit about the need to separate cognition from sensory-motor control (e.g., Chittka et al., Reference Chittka, Rossiter, Skorupski and Fernando2012). This has led to a search for “species-neutral” capacities: those that could theoretically exist in any species with a large enough brain or enough neurons, regardless of its phylogenetic history, body plan, organisation of its nervous system and the manner in which it senses, explores, and learns about the world. Comparative psychology is now a quagmire of apparently insoluble debate about cognition and intelligence (e.g., whether other species can reason about unobservable phenomena) and, more generally, whether and when the cognition of other species counts as “human-like” (Barton and Barrett, Reference Barton and Barrett2025).
The target article plots a route out of the quagmire by emphasising that cognition is grounded in the details of how animals perceive, explore, and learn about their worlds. Although we are wary of the potentially distorting effects of terminology such as “advanced” or “apical” versus “basal” traits (target article, Figure 5) and the notion of progression through evolutionary stages (target article, Figure 6), we are fundamentally supportive of the major goal: to link the evolution of bodies, sensory-motor mechanisms, brains, and cognition.
The target article makes clear how the evolution of complex behaviours is associated with sophisticated mechanisms that mediate the animal’s direct interaction with its environment. One area that needs more clarity, however, is what, exactly, is meant by “cognition”? This is something on which there has been a surprising lack of consensus (e.g., Bayne et al., Reference Bayne, Brainard, Byrne, Chittka, Clayton, Heyes, Mather, Ölveczky, Shadlen, Suddendorf and Webb2019). Coombs & Trestman rather hedge their bets, referring throughout, perhaps with deliberate ambiguity, to “cognition/behaviour.” They offer two sets of defining traits, one behavioural (central place foraging and constructive behaviour), and the other a set of learning processes under the umbrella of Unlimited Associative Learning (UAL). Cognition is not usually conceived in either purely behavioural or learning terms. Indeed, behaviour is usually taken as the outward manifestation of cognition. This being the case, it becomes circular to define complex cognition as complex behaviour. Circularity aside, maintaining the distinction between behaviour and mechanism is important because behavioural complexity can arise in different ways, including through the use of relatively simple mechanisms (Barrett, Reference Barrett2011). While the links made by C&T between behaviours and underlying processes are plausible, care is needed not to simply equate them.
The emphasis on UAL is nevertheless welcome, as it dispels the common notion that cognition and associative learning (AL) are alternative explanations for behaviour. C&T derive their concept of UAL from Ginsburg and Jablonka (Reference Ginsburg and Jablonka2021), who show how associative learning, when implemented by complex nervous systems, can produce flexible behaviour through the integration of compound stimuli and chaining, in ways that look thoroughly “cognitive” (see also Heyes, Reference Heyes2012). We should, therefore, be asking not whether learning is involved, but how learning builds cognitive processes – that is, how much integration, and of what kind, across which learning contexts is needed to explain behaviour? AL is often characterised as a “domain general mechanism” but it is not, because AL is not something that can occur independently of specific mechanisms – indeed, it is the specific mechanisms that do the learning. AL has common features across these mechanisms but there’s no AL part of the brain – it involves the whole brain, and brains are structured in distinctive ways.
Where does all this leave the received wisdom that cognition is neither sensory-motor control nor learning (MacPhail, Reference MacPhail1982; Chittka et al., Reference Chittka, Rossiter, Skorupski and Fernando2012; Triki et al., Reference Triki, Aellen, van Schaik and Bshary2021)?. We suggest that an even more radical step is required: viz, that cognition encompasses all of the processes that guide and control behaviour as the animal responds to the challenges of surviving and reproducing. That is, cognition is not a distinct phenomenon separable from sensory-motor capacities. Hence, to study cognitive evolution, we need to explicitly integrate sensory-motor specialisations into our conception of cognitive processes – they do not simply set the stage for cognitive evolution to occur. Such a stance would further enrich C&T’s framework by ridding it of the last vestiges of dualism, which will continue to lurk as long as we distinguish between “cognitive” and “non-cognitive” traits.
One implication of this embodied perspective is that it highlights evolutionary divergence as well as convergence. Convergence certainly happens at the behavioural level, but this is where the real questions arise – we need to think more deeply about the mechanisms used by different lineages to produce similar behaviour or solve similar problems; these will surely differ given different bodies, sensory-motor capacities, and ecological niches. We need to consider the whole animal, and ask how it solves problems in the wild, and what it might have been selected to do, instead of asking how much of some human-constructed notion of intelligence it seems likely to possess.
Acknowledgements
The authors thank the Institute of Advanced Study, Durham University for funding and support and the provision of IAS fellowships (October–December 2024) facilitating their collaboration. LB is supported by NSERC Discovery Grant and the Canada Research Chairs Programs.
Financial support
Funding for LB’s travel, subsistence was supplied by The Institute of Advanced Study, Durham University.
Competing interests
The authors declare no competing interests.