Abstract
Comparative psychology keeps asking whether animals possess the “same” cognitive capacities as humans: representation, planning, theory of mind. This paper argues that the question itself is malformed. There is no universal set of cognitive capacities independent of the how a species is embodied and how it lives in its own specific world. Drawing on embodied cognition's best critiques, it pushes further: cognition is never a portable faculty that different species possess in varying amounts; it is always inseparable from a specific mesocosm: a lived world constituted through a particular body, niche, and coordination. What evolves is not context-independent intelligence but metabolic stakes within mesocosms. I argue that what we see as higher cognitive capacities of humans are forms of being in the world that became indispensable for survival. I further argue that different forms of cognitive capacities have different metabolic costs, and that all living beings try not to exhaust themselves. Advanced thinking is tiring, and the tendency is to return to easy being as soon as possible. Highly developed cognitive skills also tend towards ease: a chess grandmaster calculates less, not more; an accomplished animal's skill becomes invisible to apparatus-based tests precisely because it works too well. Experimental failures often reflect a recursivity misfit, experiments test capacities that never mattered to an animal in the first place. This reframing replaces the scala naturae with a comparative framework grounded in what actually sustains life.
The Wrong Explanatory Variable
Comparative psychology keeps arriving at the same impasse. Do scrub jays represent what happened, where, and when, when they adjust their cache-recovery behaviour to the perishability of what they buried, or do they merely behave as if they did? Do New Caledonian crows plan for tools they will need tomorrow, or does their behaviour only look like planning from the outside, so that every ingenious experiment designed to settle the question generates an equally ingenious re-analysis showing the result is also consistent with simpler mechanisms? Do great apes anticipate that a companion holding a false belief about the location of an object will search for it where they last saw it, or does anticipating another’s search path require nothing beyond associative tracking of where that companion was looking? Do rats that free a trapped cage-mate, or vervet monkeys that produce acoustically distinct alarm calls for different predators, deserve to have their behaviour described in the vocabulary of empathy and semantic reference, or does that vocabulary import more than the behaviour warrants? These questions have been asked, in one form or another, for well over a century, and the honest verdict on the literature that has accumulated around them is not that the evidence remains inconclusive. It is that the questions themselves do not resolve, however much evidence is brought to bear on them, and that this persistent irresolution is itself a finding that deserves an explanation.
Robert Barton and Louise Barrett have recently supplied the clearest diagnosis yet of why (Barton & Barrett 2025). The problem, they argue, is not that comparative psychology lacks sufficiently clever experiments. Ingenious experiments are exactly what the field has produced in abundance: elaborate designs that rule out one confound after another, that isolate variables with real methodological sophistication, that have taught us an enormous amount about what animals actually do. The problem is that the question those experiments are built to answer is malformed at its foundation. Comparative psychology imports folk-psychological categories, representation, episodic memory, theory of mind, executive control, dresses them in the vocabulary of computation, and then goes looking for their presence or absence across the tree of life as though each category named a single, portable kind of thing that a given species either possesses or lacks, in the way a computer program either does or does not implement a given algorithm. Barton and Barrett call this the cognitivist gambit, and their case against it deserves to be taken with full seriousness: cognition, on this framework, is what is left over once the ordering of sensory information is subtracted, a set of internal representations that can be manipulated and transformed independently of the sensorymotor systems that produced them. Once cognition is defined this way, it becomes, in principle, the kind of thing that could be implemented by very different physical systems, a feathered wing and a membranous wing both solving the problem of flight, an octopus’s ganglia and a primate’s cortex both solving the problem of intelligence, and the entire research programme becomes an exercise in trying to detect this substrate-independent capacity through behaviour, an exercise that has generated exactly the kind of intractable, permanently reopened debate the field is now known for.
It is worth noting what Barton and Barrett do not do with this diagnosis, because it clarifies how careful their target is. They do not propose associative learning as the deflationary alternative to cognition, the humble mechanism that will replace the overreaching one. They are explicit that this move, tempting as it is, trades one confusion for another: contemporary associative learning theory, through processes like autoshaping, chaining, and the hierarchical organisation of associations among associations, already produces behaviour that is complex, multi-component, flexibly expressed, and goal-directed, behaviour that integrates multiple sources of information in ways indistinguishable, from the outside, from what the cognitivist framework calls cognition. Associative learning is not an alternative to cognition that a parsimonious researcher can retreat to when representational explanations overreach. It is, in most respects, already thoroughly cognitive in exactly the sense that makes the cognitive/associative distinction unable to do the work comparative psychology has long asked of it. This is the right place to begin, because it shows that Barton and Barrett’s target is not a particular answer within comparative psychology’s debates but the framework that generates the debates in the first place.
Their remedy is to reattach cognition to the body. Sensory-motor specialisation is not a confound to be statistically controlled for on the way to some purer computational essence; it is constitutive of what cognitive evolution actually is. A primate’s visuo-motor system, a
bat’s echolocating auditory apparatus, an octopus’s decentralised arm ganglia are not peripherals feeding information to a central processor that does the real cognitive work. They are where the cognition happens, in a sense strong enough to make the phrase “embodied cognition” strictly redundant, a necessary redundancy, they argue, only because psychology remains stubbornly attached to the possibility that cognition could, in principle, be disembodied. This is a genuine advance on the field it addresses, and it deserves to be taken further than embodiment alone is equipped to take it.
The advance stops one step short of where it needs to go. Barton and Barrett still ask what cognition is doing, and answer: something inseparable from a particular body in a particular ecological niche. Cognition remains the explanandum, the thing whose evolution is being tracked, now furnished with a far better account of its material substrate than the cognitivist framework ever offered. Living Value Theory proposes the deeper substitution their own argument gestures toward without completing. The question that carries the real explanatory weight is not what cognition is, embodied or otherwise, but when a particular recursive coordination becomes metabolically indispensable within a specific mesocosm. Once that substitution is made, intelligence stops being the primitive that explains anything and becomes something itself in need of explanation: one manifestation, among others, of a mesocosm that has already been transformed by what it had to coordinate in order to persist.
This is not a broader version of Barton and Barrett’s argument, and it is not offered as a rival to it. They are doing decisive work clearing ground within comparative psychology: showing that the cognitivist gambit is quietly anthropocentric even in its convergence-minded forms, that functionalist portability across brains does not survive contact with what nervous systems actually are, that the scala naturae survives underneath frameworks explicitly built to dissolve it. What follows asks a different question, using their critique as its strongest possible ally at one specific hinge, and then following that hinge somewhere their own vocabulary has no way to go.
What Barton and Barrett Get Right, and Where They Stop
Their targets are precisely chosen, and a fair reconstruction has to hold onto that precision rather than flattening it into a general complaint about anthropocentrism. The first target is the idea of portable cognitive faculties: capacities like working memory, inhibitory control, or theory of mind conceived as domain-general and substrate-independent, implementable by any sufficiently complex nervous system the way a computation can, in principle, run on different hardware. The second is the reduction of cognitive evolution to brain-size or neuron-count proxies, on the assumption that the relevant variable is quantity of processing power rather than the specific organisation of a specific nervous system in relation to a specific body. The third, and the deepest, is evolutionary functionalism itself, borrowed more or less directly from philosophy of mind: the assumption that the same cognitive process can in principle be realised by different physical systems, so that theory of mind or episodic memory counts as “the same thing” whether it is carried out by a primate brain or by an octopus’s decentralised nervous system, in the way that flight can be carried out with feathered or membranous wings, the physical implementation being incidental to the function it serves.
Their treatment of the neural-basis literature is where the target becomes most concrete, and it deserves to be reconstructed in some detail because it shows how far the portability assumption has travelled even into work that presents itself as biologically sophisticated. The standard account treats the neocortex in mammals, the pallium in birds, and the vertical lobes in cephalopods as convergent solutions to the same underlying problem: modalityindependent, associative computational devices, “onboard computers,” on one memorable framing, whose cognitive operations can be detached from sensory-motor and affective processing and whose expansion, across independent lineages, tracks the expansion of intelligence. This picture survives despite evidence that keeps not fitting it. Primary visual cortex is proportionately small in larger primate brains relative to “association” regions, which has led most accounts of brain evolution to disregard visual processing almost entirely, yet visual projections ramify throughout the cortex, including regions nothing to do with vision in the textbook sense, and a recent comparison of macaque and human visual cortex found that the further a visual region sits along the processing hierarchy, the more its relative size has expanded in humans, which is exactly what a story centred on “non-visual,” domain-general association cortex should not predict. Ape brain evolution deviates from the pattern the neocortex-centred story requires in an even more direct way: across several hominoid lineages, including our own, the cerebellum, a structure the standard story treats as “lower” and non-cognitive, has expanded rapidly relative to the neocortex, precisely the structure that is supposed to be doing the cognitive work. Natural selection, as Barton and Barrett put it, does not recognise the distinction between “higher” and “lower” brain regions that the cognitivist framework keeps presupposing; cortical and subcortical regions evolve and function as tightly coordinated networks, and cognitive capacity, where the term is even usable, emerges from that coordination rather than from any single structure’s expansion.
Their handling of brain-size comparisons across species carries the same discipline. Crossspecies tests of self-control built on visuo-motor tasks generate results that correlate with absolute brain size, a correlation regularly read as evidence for a unitary faculty of selfcontrol tracking overall computational capacity, but the same tasks depend on visual acuity and the ability to track hand movements, capacities that vary enormously and independently across the species being compared, and hand-tracking training alone has been shown to change performance on canonical tests of self-control in ways that have nothing to do with inhibitory capacity as such. A parallel and more recent line of argument, from researchers who otherwise accept much of the embodied-cognition programme, proposes that fish, reptiles, birds, and mammals might possess broadly similar amounts of cognition while differing mainly in their sensory-motor mechanisms and consequent brain size, a claim Barton and Barrett single out for disagreement precisely because it smuggles the portability assumption back in through a side door: it still treats “cognition” as a separable, comparable quantity sitting behind whatever sensory-motor apparatus happens to carry it in a given lineage, exactly the separation their whole argument is built to dissolve.
Their positive account follows directly from this discipline. Sensory-motor and cognitive processes cannot be separated, because minds are not the kind of thing that floats free of a particular body plan and a particular way of moving through the world; as the philosopher of mind Peter Godfrey-Smith has put it in a related context, minds exist in patterns of activity that are considerably less portable across different kinds of physical and biological organisation than researchers habitually assume. An octopus’s soft, unjointed body, with most of its neurons distributed across ganglia in the arms rather than centralised in a single brain, and with amputated arms still capable of complex, chemically guided exploratory behaviour on their own, is not incidental to its cognition, not a strange chassis for an otherwise comparable mind housed somewhere central. The organisation of the nervous system tracks the organisation of the body it guides, at every level, and the same discipline that self-recognition tests apply to primates ought to apply to the octopus, and to every other lineage in turn, including our own: a series of experiments on manual action expectation in three New World monkey species found significant differences tracking differences in thumb opposability, a result the standard cognitivist framework has no reason to predict but which fits comfortably with an account in which cognition and the hand coevolved as a single system rather than as an organ serving a separate faculty. Human infant object permanence tells the same story from the other direction: performance on the canonical A-not-B test, long treated as a direct probe of an internal representational concept of object permanence, can be manipulated by changing the physical dynamics of the infant’s own body, attaching weights to the arms, or shifting the infant from sitting to standing, manipulations that should make no difference at all if what the test measures is a stable, disembodied representation sitting in the infant’s head.
This is where the account reaches its limit, and the limit is precise enough to state in a single clause. Barton and Barrett move from brain to organism-in-environment. They do not move from organism to mesocosm. Their octopus is embodied, ecologically situated, phylogenetically specific, but it remains, in the final analysis, an organism whose cognition is to be explained by facts about that organism’s body and the niche it occupies. What never enters the account is the field of coordination the organism is caught up in: the other beings it must respond to, the material scaffolds it depends on or builds, the symbolic operations, where present, that stabilise coordination beyond what any single organism’s body could hold on its own. They overcome brain-centrism decisively. They do not overcome organism-centrism. That clause is the precise hinge on which everything that follows here turns.
Cognition Is Never Portable Because Mediation Is Not Portable
Embodied cognition, even in Barton and Barrett’s careful and well-evidenced version, still privileges cognition as the thing to be explained and asks how the body shapes it. Living Value Theory reverses the priority. Living, multimediated coordination is primary; what gets called cognition is one aspect of that coordination, not its container and not its ultimate referent. The formulation that carries the weight is this: there is no neutral environment sitting beneath species differences, waiting to be perceived differently by different nervous systems that happen to be built differently. An octopus does not receive different sensory inputs from the same world that a human receives and then process them through a different piece of hardware. It inhabits an octopus mesocosm, a lived field constituted through octopus embodiment, octopus dwelling, octopus metabolism, and octopus sensory range, in which reality is available in precisely the way that octopus coordination makes it available. The mesocosm is not a layer of interpretation laid over a shared substrate that different species merely tint differently according to their sensory equipment. Differences between mesocosms are differences in coordination itself, in how life actually proceeds, not differences in representation added on top of one common, pre-given world that all species are equally in contact with.
This is why “general intelligence” is a category mistake and not merely an empirical overreach that better data could eventually correct. It is not that intelligence fails to exist as a real property of living systems; it is that intelligence is never intelligence-in-general, only ever intelligence-through-a-mesocosm, inseparable from the specific coordination it belongs to, unable to be abstracted out and placed on a single comparative scale without losing the very thing that made it intelligence in that particular form of life. And this is exactly where Barton and Barrett become the strongest possible ally, read one further step than they read themselves. Their own comparative cases, the octopus’s distributed arm cognition, the echolocating bat’s auditory world, the primate’s visuo-motor specialisation shaping the very structure of its cortex, are not simply well-chosen examples of embodied variation within a single underlying phenomenon called cognition. They are an implicit pluralisation of mesocosms that their own vocabulary lacks the resources to name as such. Nagel’s famous argument that we cannot know what it is like to be a bat, because projecting human consciousness into a bat’s body only ever concocts a human idea of bat experience rather than the thing itself, is gesturing at exactly this point without the apparatus needed to develop it. What a bat knows and how a bat knows it are not separable questions, because the bat’s entire manner of being coordinated with its world, sonar-guided movement through a multidimensional field of prey, obstacles, and roosts, foraging conducted through an auditory rather than a visual architecture of space, constitutes what there is for the bat to know in the first place. The claim that echolocation guides bat cognition understates the relationship; echolocation, in the relevant sense, is not a channel feeding data to a bat mind that could in principle have arrived at the same understanding of its world by other means. It is the shape that the bat’s dwelling stake takes: the specific way spatial and temporal orientation is achieved by a creature whose mesocosm has never had any use for a primarily visual architecture of space.
The octopus case makes the same point with unusual clarity precisely because its nervous system so visibly refuses the vertebrate template that comparative psychology, even in its embodied form, keeps reaching for. Most of an octopus’s neurons sit in the arms rather than the central brain; an amputated arm goes on probing, grasping, and exploring on its own, guided by a chemo-tactile system built into the skin that lets the arm taste what it touches independently of any central command; the motor regions of the central brain themselves lack the kind of somatotopic mapping, a stable, body-part-specific representation of the limbs, that organises motor control in mammalian brains. If cognition is treated as a portable computational achievement that this animal happens to run on unusually distributed hardware, the octopus becomes a standing puzzle: how can something so decentralised achieve anything recognisable as intelligent, flexible, problem-solving behaviour at all, and where, in this distributed architecture, is the achievement actually located? This is the question that has led some researchers to ask, quite seriously, not only what it is like to be an octopus but where it is like to be an octopus, whether octopus consciousness, if the term even applies, might itself be disunified across the arms rather than centred, the way vertebrate consciousness is typically assumed to be centred, in a single locus. If cognition is instead read as one aspect of a mesocosm constituted by that very decentralisation, a form of life in which the boundary of the coordinating self does not track the boundary of a central nervous system the way it does for a primate, the puzzle dissolves rather than deepens. There is no question left over about where octopus cognition “really” happens once it is no longer assumed that cognition must happen somewhere central for it to happen at all, and no question left over about a disunified consciousness that needs an equally disunified answer, because the coordinating unit was never single-centred to begin with. The arms are not instruments an octopus mind uses from a command centre elsewhere. The distributed arms, coordinating with the central brain and with each other, are the shape the octopus mesocosm takes.
None of this is confined to the exotic end of the comparative range; it holds for the primate hand and the human infant’s body just as fully as for the bat’s ear and the octopus’s arm, and this is the point at which embodied cognition’s own best evidence, read carefully, starts to outrun the framework built to explain it. If manual action expectation tracks thumb opposability across species that differ in exactly that anatomical respect, and if a human infant’s grasp of object permanence can be switched on and off by changing what its arms weigh, then what is being tracked in each case is not a general cognitive capacity differentially expressed through different bodies. It is a coordination that is partly constituted by the specific body doing the coordinating, a coordination that would not be the same achievement, would not even be intelligible as the same kind of achievement, if extracted from that body and relocated somewhere else. What recursive attribution and recursive relevance describe for symbolic mesocosms applies with equal force here, at the level of bare sensorimotor engagement: before any question can be asked about what a scrub jay’s cache-recovery behaviour or a crow’s tool selection demonstrates about internal representation, a prior question has already been settled at L1, in the bird’s actual bodily practice, about which entities in its world are worth coordinating around at all, which caches, which competitors, which materials are recursively relevant to this particular bird’s ongoing life, given the body and the niche that bird has. That settlement is not a preliminary to the bird’s cognition. It is inseparable from what the bird’s cognition, whatever the word is finally taken to mean, actually consists in.
Recursivity Levels and the Collapse of the Reflective-Ascent Picture
Barton and Barrett’s own critique of comparative psychology’s inherited vocabulary already senses that something is wrong with treating cognitive evolution as a climb toward ever more explicit reasoning, even though their framework has no architecture for saying precisely what. They trace the ascent picture back to Romanes, whose nineteenth-century account held that the lower one descends in the animal kingdom, the more reflex action and non-mental adjustment predominate over volitional, deliberate behaviour, and they show that this picture’s vocabulary survives largely intact in contemporary research on inhibitory and executive control, however thoroughly that research now disavows its Victorian ancestry. What embodied cognition cannot do, on its own, is say why a trend toward “higher” cognition, understood as more explicit representation and more deliberate control, is not merely anthropocentric in its labelling but backwards in its direction of travel.
Living Value Theory’s levels of recursivity supply exactly this missing architecture, and it is worth being precise about what each level names before applying the architecture to the evolutionary question. L1 is vital coordination: the level at which coordination across a living system’s relevant domains proceeds without demanding explicit attention, the smoothly functioning background against which everything else takes place. L2 is the registration of threatened vitality, the point at which something in that smooth coordination becomes unsettled and begins to require notice, whatever form “notice” takes in a given nervous system. L3 is the level at which a disturbance becomes articulated as this specific thing, this specific problem, still embedded in the coordination that generated it rather than detached into a portable category. L4 is abstraction and stabilisation: forms that become portable across coordinations, generalisable enough to guide response beyond the single episode that first called for them. L5 is meta-recursive reflection, in which L4 systems themselves become objects of further assessment. What matters for cognitive evolution is the claim buried, largely unexamined, in comparative psychology’s inherited vocabulary: that “cognition proper” is quietly identified with L3 through L5, representation, deliberation, executive oversight, while L1 coordination, however accomplished, is treated as mere reflex, the thing cognition is supposed to have risen above on its way to becoming worth studying.
This identification gets the relationship backwards, and it gets it backwards for every species examined this way, our own included. Humans, no less than octopuses, bats, or crows, spend the overwhelming majority of their lives absorbed in L1, punctuated occasionally by L2 disturbance, stabilising only sometimes into L3 articulation, and rarely needing L4 or L5 at all in the ordinary course of a day. Even the species that comparative psychology has long treated as the ceiling of cognitive achievement is not, on the evidence of how it actually lives, a primarily reflective creature. It is an absorbed coordinator that resorts to higher recursivity only when L1 coordination has broken down and needs repair, when the habitual path through a familiar landscape is blocked, when a well-practised social exchange goes wrong, when a tool fails in a way its user has not encountered before. This is not a minor correction to the standard picture; it inverts its direction of travel entirely. The standard picture imagines evolution trending toward increasingly explicit, increasingly deliberate processing, with humans arrayed at the far end of that trend and other species positioned at various distances behind. Living Value Theory’s architecture predicts the opposite trajectory for accomplished coordination within any single lineage or any single lifetime: mastery drives recursivity down, not up. A novice chess player calculates effortfully, working through positions with laborious, explicit deliberation that sits closer to L3 and L4 activity than to L1. A grandmaster does not do more of this as expertise accumulates; she does dramatically less of it, because years of practised coordination have returned the skill to L1, where good moves are simply seen rather than computed move by move. The same descent characterises a forager who has learned a landscape over years, a surgeon whose hands have absorbed a procedure to the point of no longer narrating it, a fluent speaker moving between two languages without translating. Accomplishment, wherever it is found, is not the intensification of explicit processing. It is the disappearance of the need for it.
Nothing about this collapses the distinction between accomplished L1 coordination and mere behavioural rigidity, and the distinction matters enough to state carefully, because without it the whole inversion risks proving too much, risks treating any unreflective, stereotyped behaviour as evidence of mastery simply because it runs without deliberation. Smoothness at L1 counts as vitality only while disturbance remains registerable and remediation remains reachable when it is needed; a coordination that has become naturalised past the point of being able to register that anything has gone wrong is not displaying accomplished mastery, it is displaying a failure mode that happens to look like mastery from the outside. The distinction is empirically tractable, and it gives the inversion argued for here a falsifiable edge rather than leaving it as an unfalsifiable reinterpretation of whatever a given species happens to do. A genuinely accomplished L1 coordination, the octopus’s practised approach to a familiar prey item, the crow’s habitual route to a cache, remains capable of redirection the moment the situation departs from what the coordination was built to handle: a prey item that behaves unexpectedly, a cache site that has been disturbed, a social partner who has broken an established pattern. A merely rigid, canalised behaviour pattern, the kind an evolutionary trap exploits, where an organism’s fixed response to a formerly reliable cue persists even after the cue has become disconnected from the outcome it once tracked, drawing moths to flames or sea turtles to artificial light sources on the shoreline, does not redirect under the same pressure, because there was never a live L2 registration capacity behind the L1 smoothness in the first place, only a fixed
pattern that had not yet been tested by a mismatched environment. Comparative psychology’s laboratory paradigm, ironically, is often well suited to detecting exactly this difference, once it is looking for the right thing: an animal whose response to a novel perturbation of a familiar task reorganises fluidly is demonstrating something the framework here would call recursive fluidity, the capacity to move between levels in the direction repair requires; an animal whose response to the same perturbation persists unchanged, degrading rather than adapting, is demonstrating the opposite. The error in the standard paradigm is not that it measures the wrong thing when it measures flexibility under perturbation. The error is in treating explicit, effortful, L3-and-L4-level processing during the perturbation itself as the sign of the underlying capacity, when the sign that actually matters is whether L1 is eventually restored, and restored in a form that has absorbed the lesson the disturbance offered.
This carries a direct methodological consequence for the discipline Barton and Barrett are addressing, and it sharpens rather than softens their own critique of anthropocentric measurement. An executive-control task presented to an animal in an unfamiliar laboratory apparatus forces the animal out of whatever L1 coordination governs its ordinary life and into effortful, apparatus-specific processing almost by design, and the resulting performance is then read as a direct measure of the animal’s cognitive ceiling, on the tacit assumption that more visible deliberation signals more underlying capacity. This is what LVT would name a recursivity misfit: an instrument built to detect L4-level processing applied to a domain whose real achievement, in the animal’s own mesocosm, is L1 mastery, mistaking the absence of visible deliberation in the animal’s ordinary environment for the absence of anything worth calling cognition, when the absence of deliberation is precisely what accomplished coordination in that species’ own terms looks like. This should not be confused with the ordinary distinction between automatic and controlled processing that dual-process accounts already draw and that no one, least of all Barton and Barrett, needs LVT’s help to notice. The point is not that some behaviours are automatic and others controlled; every dual-process framework grants that much and stops there. The point is that the direction of travel between the two, across a lifetime of skill acquisition or across a lineage’s accumulated coordination with its niche, runs from effortful toward absorbed, and no dual-process account, built to sort a snapshot of behaviour into one bin or the other, predicts or even has the conceptual room to notice this descent as the thing worth explaining.
Recursive Relevance and the Problem of the Apparatus
There is a further consequence of taking mesocosm-specificity seriously that comparative psychology’s own results have already been pointing toward, without quite finding the vocabulary to state it as a general methodological principle. A phenomenon becomes recursively relevant within a mesocosm when the living beings that inhabit it come to treat it as consequential enough for their ongoing coordination to organise attention, habit, and response around it. Relevance, in this sense, is not itself the ground of anything; it is downstream of whatever vitality stakes a given entity does or does not bear on for a given form of life. But it is precisely because relevance is downstream of stakes, rather than a free-floating property that any sufficiently interesting object automatically acquires, that a novel object placed into an animal’s environment by an experimenter does not arrive preloaded with the relevance it would need in order for the animal’s response to it to be diagnostic of anything at all. Before a mirror, a lever, a puzzle box, or an unfamiliar conspecific’s vocalisation can register as behaviourally significant, it first has to clear a threshold that has nothing to do with the cognitive capacity the experiment is designed to probe: does this object matter enough, within this animal’s actual mesocosm, for engaging with it in the way the experiment requires to be worth the coordination cost of doing so?
This reframes a body of otherwise puzzling results that the standard paradigm has struggled to make sense of on its own terms. Mirror self-recognition, treated within the cognitivist framework as a graded marker of self-awareness that should track something like general intelligence across species, has produced a pattern of positive and negative results that maps poorly onto any plausible intelligence ranking: several primate species show inconsistent or absent evidence of mirror-guided self-directed behaviour, while cleaner fish, animals that occupy almost no place at all in the traditional hierarchy of cognitive sophistication, have been shown to use their reflection to guide inspection of their own bodies in a manner functionally similar to what mirror tests are usually taken to demonstrate in apes. Read within a general-intelligence framework, this is an anomaly that keeps needing to be explained away. Read as a question about recursive relevance, it stops being anomalous. Cleaner fish make their living by inspecting and removing ectoparasites from the bodies of other fish, a foraging ecology already organised around close visual attention to bodily surfaces, including, plausibly, their own; a reflective surface has a route into that fish’s mesocosm that it simply does not have into the mesocosm of a primate whose visual attention is organised around faces, hands, and social gesture in an entirely different configuration. The fish is not passing a test of self-awareness that most mammals fail. It is bringing a foraging-relevant visual competence to bear on an object that happens, for ecological reasons specific to that species, to have acquired relevance within its ordinary coordination, in a way the same object may never acquire for a species whose mesocosm has no comparable use for it.
The methodological upshot is not that comparative cognition experiments are uninformative, and it would be a mistake to read this as scepticism about the value of the discipline’s careful empirical work. It is that a negative result, an animal’s failure to engage with a novel apparatus in the way a test predicts, is systematically ambiguous between two very different explanations that the standard paradigm has no principled way to distinguish: the animal may lack whatever underlying capacity the test is designed to probe, or the apparatus may simply never have crossed the threshold of recursive relevance that would make engaging with it, in the way the test requires, a coordination worth the animal’s investment. A positive result is not immune to the same ambiguity in the other direction: an animal that does engage with a novel apparatus may be demonstrating exactly the general capacity the experimenter hoped to find, or it may be demonstrating a highly specific, ecologically narrow competence that happens to transfer to this particular object for reasons that have nothing to do with the general capacity being claimed. Comparative psychology has developed considerable sophistication in controlling for perceptual and motor confounds between species being compared. It has developed almost no comparable apparatus for asking, before the experiment is run, whether the object being presented is likely to clear the relevance threshold for the mesocosm of the species being tested, and this is not a gap that better statistics or larger sample sizes can close, because it is a gap in what the experiment is asking in the first place rather than in how well it answers the question it does ask.
The Evolution of Cognition Rewritten as the Evolution of Metabolic Stakes
Every species has recursive coordination of some kind; that much embodied cognition already establishes convincingly, and nothing here disputes it. What differs, and what actually does the explanatory work that “intelligence” is wrongly asked to do, is when a given coordination becomes indispensable to a mesocosm’s continuation rather than merely available to it. The word at stake here carries two registers that need to be held together rather than allowed to drift apart, because each does distinct work. In the ordinary sense, what is at stake is what is crucially important, what matters most to the outcome. In the etymological sense, what is at stake is what pertains to life, to its bare continuation. A metabolic stake, in the sense this argument needs, brings these two registers together with enough precision to do explanatory work rather than merely gesture at importance:
A metabolic stake exists when failure to stabilise a particular recursive coordination reliably reduces survival, reproduction, or the long-term viability of a mesocosm.
This definition does the anti-scalar work that a convergence-only framework, however embodied, cannot do on its own, and it does so by making the comparison between species a comparison of stakes rather than a comparison of quantities of a shared underlying capacity. The octopus did not fail to build libraries because it fell short on some general intelligence scale that humans happened to clear through a larger brain or a more elaborate cortex. Nothing in the octopus mesocosm, a form of life that is largely solitary, comparatively short-lived, and in which, notably, brain size does not correlate with sociality or life history the way it does in long-lived, socially embedded mammals, ever made a cumulative, transgenerational symbolic archive a metabolic stake in the sense defined above. And the asymmetry runs deeper than a simple absence of opportunity or motivation: even granting an octopus literacy by some thought experiment, it could not use a library, because the mesocosm that would make a library’s use possible, sustained multigenerational teaching, stable dwelling organised around fixed sites returned to across years, a being-with structure organised for the transmission of stored knowledge to others who were not present when the knowledge was gained, is simply not there to be drawn on. Libraries, writing systems, mathematics, universities are stabilised responses to stakes that became indispensable within one particular mesocosm’s history. They are not monuments to a superior cognitive faculty that other lineages simply possess less of.
This gives human distinctiveness an explanation that is genuinely non-scalar rather than merely dressed in convergence-friendly language while still smuggling a hierarchy back in. The relevant sequence begins from the ground every form of life shares: embodiment and dwelling emerging together, since no organism exists first in itself and only afterward enters an environment, a bounded metabolic process is already, from its very first instant, a body-in-place, and this holds as fully for the earliest autotrophic life, drawing energy directly from sunlight or inorganic compounds without needing to move or coordinate with other organisms to survive, as it does for anything that comes after. Being-with then emerges not primarily from sexual reproduction, which deepens and complicates it considerably but does not originate it, but from the broader fact that collective coordination becomes indispensable wherever the fate of one body becomes bound up with the movements and responses of others, schools of fish, flocks of birds, herds and swarms, in which coordinated plurality is not an intimate supplement to survival but survival’s actual condition, a moving protection device and sensory multiplier before it is ever a reproductive arrangement. Sexual reproduction then enters as a major intensifier, adding durable asymmetry, temporal extension, and intergenerational complexity to a form of coordination that was already indispensable on its own terms.
What group hunting cannot, on its own, explain is why any one lineage crossed from this shared ground into the specifically human architecture, because too many species hunt, and too many hunt cooperatively, for the pressure of coordinated predation alone to be sufficiently discriminating. Wolves, lions, dolphins, and chimpanzees, the clearest case of socially coordinated nonhuman primate hunting among living apes, targeting red colobus monkeys and other prey through pursuit with differentiated roles among the hunters, all show that coordinated predation is a real and significant pressure without, on its own, generating the human trajectory. What group hunting does is approach, from one historically important direction, a deeper threshold that is not about predation as such: the threshold at which coordination must be stabilised across absent times, distant spaces, delayed returns, and remembered patterns, a threshold reached most clearly not by hunting in the heroic, immediate sense but by seasonal hunting, which requires anticipation of cyclic abundance and scarcity, route memory, timing, transport, and preservation integrated into a broader annual rhythm. At that threshold, embodied dwelling and being-with, however elaborated, no longer suffice on their own to carry the coordination a temporally extended, heterotrophic form of life now depends on for its survival.
This is the specific point at which multimateriality and multisymbolisation co-emerge, and co-emerge rather than arrive in sequence, because neither alone can solve the coordination problem that a lineage dependent on temporally extended predation generates. A sharpened stone is never a single event; it presupposes a model of the finished tool that exists before the tool itself, a purpose that precedes the shaping, and, in every case that matters for a social species, a community in which such tools are taught, passed on, improved, and deployed collectively, which means the tool cannot be consistently made, transmitted, or used without some form of symbolic coordination capable of naming what a spear is, when it is to be used, by whom, and against what, and capable of warning, teaching, and signalling the moment of coordinated action. Equally, symbolic coordination without material supports would have nothing durable to organise: a gesture that vanishes the instant it is made, a call that leaves no trace, cannot do the work that reference to absent intentions, future plans, and remembered techniques requires. Symbols gain their traction precisely because they can organise around persistent things, the hearth, the cache, the tool, the marked path, and this is why the archaeological record shows tool traditions and symbolic behaviour, ochre use, engraved objects, ornaments, becoming elaborated together rather than in sequence: the same coordination pressure that drives elaboration in one mediation drives elaboration in the other, because a hunting group approaching a seasonal migration must coordinate equipment preparation, route selection, role assignment, and timing in ways that exceed what embodied attunement and copresence alone can carry, while simultaneously teaching younger members how to make the tools, recognise the signs, and handle the outcome, a demand that requires a symbolic capacity able to stabilise reference across absence at the same moment it requires material forms able to carry the techniques and stored resources that symbolic coordination organises. Language, on this account, is not an autonomous achievement that arrived to narrate an intelligence already otherwise in place; it is the coordination system that this specific metabolic stake required, arising because the joint production and deployment of persistent material tools within collective, temporally extended heterotrophic life could not proceed without it.
Once this architecture stabilised, the metabolic stake it was built to address did not disappear, and subsequent pressures, climatic shifts, new food sources, more complex social arrangements, were absorbed through redistribution and intensification within the existing five-part architecture rather than through the emergence of some further, more cognitively advanced mediation. What looks, across the archaeological and historical record, like an ascending trajectory toward higher intelligence is better described as the continued elaboration of stakes that were already locked in at the single threshold where multimateriality and multisymbolisation first became jointly indispensable to a lineage’s survival. This is also where the distinction between accomplished L1 coordination and mere naturalisation, introduced above for the case of skill, applies with equal force to evolutionary history itself. A coordination that has become metabolically indispensable and remains vital continues to permit disturbance to register and remediation to reach it, a hunting technique that fails against unfamiliar prey behaviour gets modified rather than persisted with regardless of outcome, a teaching practice that stops transmitting successfully gets revised. A coordination that has become metabolically indispensable and then hardens past the point where disturbance can register at all is no longer functioning as vital coordination in the sense this argument requires, whatever its evolutionary origin; it has become the evolutionary equivalent of harmful naturalisation, a fixed response persisting because it once tracked a stake accurately, now surviving on inertia rather than on continued fit. Distinguishing these two outcomes, genuinely vital metabolic stakes still capable of registering and repairing disturbance, and merely inherited fixities that have stopped being able to do so, is itself an empirical question about any given lineage’s history, not a distinction the framework can settle in advance for every case, but it is the distinction that keeps the metabolic-stakes account from collapsing into a story in which whatever survived must, by definition, have been vital.
“Intelligence” as Explanandum, and What the Discipline Has Been Doing to Animals
Bring the moves together and the shift in explanatory priority becomes exact. Mediation shows that cognition cannot be abstracted from the mesocosm it belongs to, because there is no neutral world beneath species-specific coordination for a portable faculty to operate on. The recursivity architecture shows that even within a single mesocosm, accomplished coordination runs opposite to the ascent picture that treats explicit, deliberate processing as the marker of cognitive sophistication, mastery drives coordination toward L1, not away from it, and the distinction between genuine mastery and mere rigidity is empirically available rather than assumed. Recursive relevance shows that an apparatus, however carefully controlled for sensory and motor confounds, cannot generate a diagnostic result until it has cleared a threshold of mattering that has nothing to do with the capacity it is meant to test. And metabolic stakes show what actually explains a lineage’s distinctive achievements: not a quantity of cognition possessed in greater or lesser measure, but the specific coordinations that became indispensable to survival and reproduction within its specific mesocosm, at the specific thresholds where the mediations already available could no longer carry the coordination load alone.
Once these four moves are in place, intelligence can no longer function as the term that explains why one lineage builds cities and another does not. It becomes, instead, a name applied retrospectively to whatever coordination happens to be visible and symbolically legible once a mesocosm has already been transformed by the stakes that organised it, which means that the concept inherits, in the domain of comparative cognition, exactly the distortion that overvalues the visible and named while undervaluing the coordination that works too well to demand attention. The coordinations that most fully meet a species’ stakes across its relevant mediational domains are, by the nature of L1, the least likely to register as achievements requiring explanation: the octopus’s practised, flexible approach to prey it has hunted a thousand times, the migratory bird’s route memory spanning thousands of miles, the weaning female mammal’s readiness for the next litter. These are treated, within the standard framework, as background biology rather than as cognitive accomplishment precisely because they work smoothly enough not to announce themselves, while the comparatively rare, effortful, L3-and-L4-level problem-solving that a novel laboratory apparatus elicits gets treated as the discipline’s paradigm case of cognition, for no better reason than that it is the visible exception rather than the invisible rule. This is not a minor methodological quirk. It is the same systematic overvaluing of what is symbolically legible over what actually sustains coordination that distorts institutional value assessment in every other domain LVT has examined, transplanted here into a discipline that studies animals rather than human institutions, with the same practical consequence: research attention and resources get redirected toward whatever generates a publishable, apparatus-based result, and away from the far larger, far less visible body of accomplished L1 coordination that constitutes the actual cognitive life of the species being studied.
There is a further consequence worth stating plainly, because it bears on how comparative psychology’s own authority over its subjects should be understood. When an institution, in this case, a research paradigm, holds the recognised authority to determine what an animal’s behaviour can count as, and that authority is exercised through categories (representation, executive control, theory of mind) that did not arise from the animal’s own coordination but were imported wholesale from human folk psychology, something structurally similar to what happens when human institutions relocate the authority to define experience away from those undergoing it is occurring here too, adapted to a domain where the being whose coordination is being assessed cannot contest the categories applied to it in any symbolic register the discipline is prepared to recognise. An animal that fails an executive-control task designed around human-derived assumptions about what counts as inhibition is not thereby shown to lack inhibitory capacity; it is shown to have a form of coordination that the imported category does not fit, and the failure gets recorded as an absence in the animal rather than as a misfit in the instrument. This is not an accusation of bad faith against a discipline that has, as Barton and Barrett’s own careful reconstruction shows, been arguing internally against exactly this tendency for decades. It is a description of what remains structurally at stake even in the field’s best current self-critique, and it is the reason the substitution proposed here, from cognition to metabolic stakes, is not merely a terminological refinement but a change in what the discipline is entitled to conclude from a negative result.
Objections and Limits
Five objections deserve to be met directly rather than left for a hostile reader to raise unanswered.
Is “metabolic stake” doing suspiciously much work? The definition has a falsifiable edge precisely because it distinguishes stakes from conveniences by their behaviour under removal, and the same edge distinguishes genuine L1 mastery from mere naturalisation, as argued above. A coordination that is a genuine metabolic stake should show characteristic signs when it is degraded or withdrawn: cascading disturbance across other domains of the same mesocosm, mobilised effort to restore it, persistence of the coordination even once its original triggering conditions have changed. A coordination that is merely an elaboration or a convenience should be absorbable without this cascade, removed, substituted, or left to lapse without generating the same systemic strain. This is testable in both directions, in the archaeological, comparative, and experimental record alike, and it is a substantially stronger claim than simply relabelling whatever happened to persist as having been “at stake” all along.
Doesn’t this just relabel adaptation? No. Adaptation is an outcome-level claim: a trait increased fitness relative to available alternatives. Recursive coordination under metabolic stakes specifies the mechanism and the unit at which selection is doing its work, which coordination, of which recursivity type, became indispensable within which specific mesocosm, and why the mediations already available could no longer carry the coordination load alone at a given threshold. Two traits can be equally adaptive in the outcome sense while differing entirely in this structural sense: one an elaboration within an alreadystabilised architecture, generating marginal fitness gains without closing any genuine coordination gap, the other the closing of a gap whose failure would have reduced viability directly. Adaptation, as ordinarily used, cannot distinguish these two cases from each other. The stakes framework can, because it asks what would have happened to the mesocosm’s viability had the coordination in question not stabilised, rather than only asking whether the trait that did stabilise conferred an advantage once it existed.
Is this a genealogy or a typology? This argument proposes a reframing of what counts as the explanatory variable in cognitive evolution, a typological, ontological claim about what kind of thing is being tracked when cognitive evolution is studied at all. It is not, on its own, a settled account of the historical sequence by which any particular lineage’s stakes came to be what they are; that is a separate, harder, more evidentially demanding claim, pursued through the archaeological and comparative record on its own terms elsewhere. The typological claim does not stand or fall with any particular reconstruction of that sequence, though it is answerable to it and would be weakened by a reconstruction that showed the proposed thresholds did not correspond to anything in the actual historical record.
Does this over-read its own convergences? Barton and Barrett’s own caution applies here with full force, and it would be a failure of the argument’s own standards to exempt itself from it: similar behaviour across lineages does not establish an identical underlying process, and the discipline this argument asks of comparative psychology it must ask of itself with equal rigour. The octopus’s distributed arm cognition and the human hand’s integration with symbolic coordination are not instances of “the same” embodied intelligence differing only in degree of complexity. They are different mesocosms, organised around entirely different stakes, arrived at through entirely different evolutionary histories, and the temptation to read them as points on a single continuum, even a continuum organised around metabolic stakes rather than around intelligence, is exactly the scala naturae returning by the back door, this time dressed in vitality-framework vocabulary instead of computational vocabulary. The stakes framework has to be applied to each mesocosm on its own coordinative terms, case by case, not used to construct a new hierarchy in the old one’s place.
Doesn’t the appeal to protective invisibility versus harmful naturalisation risk becoming unfalsifiable in practice, always available to explain away any inconvenient case as either mastery or mere fixity, whichever the argument needs at the time? This is the sharpest version of the worry and deserves a direct answer rather than a reassurance. The distinction is falsifiable in principle because it makes a specific prediction about behaviour under perturbation that does not depend on which label the theorist wants to apply in advance: genuine L1 mastery redirects when the situation departs from what the coordination was built to handle, while mere fixity persists unchanged and degrades. This prediction can fail. A coordination confidently classified as accomplished mastery that in fact fails to redirect under a well-designed perturbation would count as evidence against that classification, not as a case the framework can simply relabel after the fact. The honest caveat is that classifying any particular case correctly, in practice, requires exactly the kind of careful, perturbation-based experimental design that comparative psychology at its best already knows how to build, which means the framework does not remove the burden of good experimental work from the discipline it is addressing. It relocates what a welldesigned experiment should be looking for.
Conclusion
The reframing offered here is not only a contribution to a debate within comparative psychology, though it is offered as one. It is a demonstration of why the five mediations that organise every mesocosm are not an arbitrary set of categories imposed on living systems from outside a discipline that would otherwise have no use for them, but the specific dimensions along which evolution can do its work once metabolically indispensable coordination, rather than a portable faculty called cognition, is taken as the object requiring explanation. The stakes that made multimateriality and multisymbolisation co-emerge in one particular lineage are not evidence of that lineage’s superior cognitive endowment, measured against some scale every other lineage also stands on somewhere lower down. They are evidence of the specific coordination problem that lineage alone faced, generated by the specific mesocosm it alone inhabited, and solved in the only way available to a form of life whose survival had come to depend on coordination stretched across absences no embodied dwelling and no being-with, however elaborated, could carry on their own.
What this account leaves undeveloped, deliberately, is everything downstream of the threshold it has identified. The interrecursive structure that being-with reaches once symbolic coordination becomes available to it; the institutions that stabilise interrecursive coordination once informal transmission can no longer bear its own weight across the scale a settled, technologically elaborated form of life requires; the cultivation through which culture takes stakes that were once matters of bare survival and turns them into matters of flourishing, obligation, and meaning, each of these is a further mesocosm-specific elaboration of the architecture set out here, not a further increment of intelligence accumulating on some continuous scale. They are where an account of this kind goes next, once the substitution proposed here, from asking what cognition is, to asking when a coordination became indispensable to a mesocosm’s continuation, has done the work of clearing away the question that made those further questions impossible to ask in the first place.