Why Do Bees Not Turn Into Ghosts?

Abstract

The anthropology of human-animal relationships has been organised, from Durkheim and Mauss to Lévi-Strauss and the ontological turn, around what animals mean to humans. This article argues that the tradition selected the one form of recursive activity that nonhuman animals do not share, institutionally stabilised symbolic classification, and made it the key to them, which makes the classificatory approach the most anthropocentric option available to the field, however much it presents itself as the study of other ways of thinking. Living Value Theory (LVT) offers a reconstruction that makes the similarity between humans and other animals as large as the evidence allows. All animals, humans included, share three zoetic mediations, coordinate for most of their lives at the recursivity levels L1 and L2, and engage nonrecursive, selfrecursive, and interrecursive counterparts. Interrecursivity is a mode that most animals have, ants included, and it ranges from low-energy forms such as relaxed co-presence and grooming to maximal and antagonistic engagements that are metabolically draining and that humans, like other animals, avoid wherever they can. Rudimentary multimateriality, multisymbolization, and symbolic articulation are present among animals as well. The only substantial difference lies in the recursivity levels L4 and L5, which require a very elaborate multisymbolization together with multimaterial means of inscribing, storing, and transmitting symbols, and which no other animal exhibits. Further claims are developed: that metabolism, above all the difference between beings that feed on light and beings that must eat other lives, shapes relationships with animals before any classification does; that recursive fluidity, the capacity to return to absorbed coordination, matters more than any capacity for higher recursion; that recursive attribution is graded by the mediations actually extended to a counterpart, from a name to food and sleep; that "the human-animal relationship" is as empty a category as "the human-human relationship" would be; and that the radical ontological differences claimed by perspectivism are small symbolic variations against a vast shared ground. The argument moves from cave art and Göbekli Tepe through ants, gulls, hamsters, street dogs, shepherds, Yukaghir hunters, and temple deities to the industrial shed.

I. The Animals on the Cave Walls

Why are animals so prominent in some of the earliest representations that human beings left behind? At Leang Karampuang, in the limestone karst of South Sulawesi, a painted scene dated to at least 51,200 years ago shows three humanlike figures interacting with a wild pig (Oktaviana et al. 2024). At Chauvet, in the Ardèche, painters working some 36,000 years ago filled the walls with lions, woolly rhinoceroses, mammoths, and cave bears (Clottes 2003). At Lascaux, many millennia later, horses, aurochs, and stags crowd the walls and ceilings, while the reindeer whose bones dominate the food refuse found in the cave appears perhaps once. At Göbekli Tepe, in south-eastern Türkiye, people who had domesticated neither crops nor livestock carved foxes, boars, snakes, cranes, vultures, and scorpions into limestone pillars weighing several tonnes, more than eleven thousand years ago, while the gazelle that supplied most of their meat is scarcely shown (Peters and Schmidt 2004).

The record needs two caveats. The oldest securely dated rock art currently known is a hand stencil and no animal: at Liang Metanduno on Muna Island, off south-eastern Sulawesi, a stencil at least 67,800 years old was made by spraying pigment around a hand, and its fingers were afterwards narrowed until they resembled claws (Oktaviana et al. 2026). The T-shaped pillars of Göbekli Tepe, for their part, are widely read as stylised human bodies, so that the carved animals crawl across figures that are themselves humanlike. Animals were never the only subject. They nevertheless became subjects of durable material and symbolic investment remarkably early and remarkably often, and across many sites they outnumber human figures by a wide margin.

Social anthropology rarely treats this record as its own. It confines itself, for the most part, to people who are alive or recently dead, and leaves the deep past to archaeology. For the study of relations between humans and other animals the division is costly. Homo sapiens has existed for some three hundred thousand years, and for nearly all of that time its members lived among large predators, followed migrating prey, and depended on animals they could neither control nor ignore. A theory of human-animal relationships built from the ethnography of the last century, with its pastoralists, pet owners, and laboratory scientists, describes a thin and recent stratum of that history. The painted and carved animals are the oldest surviving testimony of what animals were to people, and they show something that the ethnographic archive tends to hide: the animals that people ate received far less attention than the animals that could eat them.

That observation raises more questions than it settles. Why these animals and not others? Why do dangerous animals appear alongside hunted ones, and sometimes in place of them? Why do different caves, and at Göbekli Tepe different buildings on the same hill, select different repertoires? Two kinds of question can be put to a painted lion. The first asks what it signified, what it stood for, and what system of categories it belonged to. The second asks what part lions played in the lives of the people who painted them: what they had to do about lions, at which hours, in which places, and at what risk. Anthropology has habitually asked the first question, and the prevalence of the imagery is usually taken to confirm that habit. I take it to challenge the habit, because nothing in a painted rhinoceros shows that the most important thing about rhinoceroses, for those who painted them, was their place in a classification. Once the second question is asked, the first loses its priority. Animals did not enter into relationships with human beings so that humans could symbolise something with them. They were food, predators, competitors, companions, carriers of disease, and signs of the season, and only within those relationships did they also become signs of anything else.

This article therefore asks what forms of living coordination and vital relevance made particular animals consequential enough to become subjects of such extraordinary attention. Archaeology cannot read vital stakes directly off an image. The task is to reconstruct the object of inquiry so that differentiated interpretations can be generated and tested, and that reconstruction has to begin with the tradition that taught anthropology to look at animals through human categories.

II. How Structuralism Lost the Animals

The modern anthropology of animals has a founding text that is not about animals. In Primitive Classification, Durkheim and Mauss (1963 [1903]) argued that the classification of things reproduces the classification of people: among the Zuñi and in the Australian section systems, animals, plants, winds, and colours are distributed among clans and moieties, so that the order of nature is modelled on the order of society. Animals enter the argument as items sorted. A decade later, in The Elementary Forms of Religious Life, Durkheim (1995 [1912]) made the totemic animal the emblem of the clan, the form under which a society represents itself to itself. The reasoning is simple. Human beings resemble one another too closely to supply the marks by which groups are told apart, while an eagle, a kangaroo, and a snake differ unmistakably, so the differences between animals are borrowed to mark the differences between groups. Durkheim noted that the engraved image of the totem is more sacred than the animal it depicts. The remark deserves more attention than it has received, because the theory states in it, quite candidly, that the living kangaroo matters less than its sign.

Lévi-Strauss completed the movement in two books of 1962. Totemism dissolved the institution into a way of thinking: natural species are chosen because they are good to think, and not because they are good to eat, and what resembles one another across the human and the animal series is the differences between their terms (Lévi-Strauss 1963). The Savage Mind turned the species into a logical operator through which any domain can be coded into any other (Lévi-Strauss 1966). From sorted items to emblems to operators, the lineage moves in one direction: from living beings to social classifications, and from social classifications to relations among classifications. The three authors did not make identical arguments, and a great deal of careful work on herding, hunting, and husbandry was done during the same century. The hypothesis advanced here is narrower and, I think, defensible. Their combined legacy made the symbolic classification of animals appear more theoretically fundamental than the living processes through which humans and other animals coordinate.

The lineage did not end in 1962. The most influential recent account of how humans relate to animals, Amerindian perspectivism, is structuralist in method and acknowledges the debt (Viveiros de Castro 1998). It takes a body of indigenous statements about souls, bodies, and points of view and arranges them as a system of inversions, in which humans see themselves as humans and jaguars as jaguars, while jaguars see themselves as humans and humans as peccaries. Whatever its ethnographic merits, the analysis again finds the decisive reality of the relationship in a symbolic scheme, now raised to the dignity of an ontology. The structuralist approach to animals has thus been the dominant one for well over a century, and it still shapes how anthropologists frame the question.

That this tradition is anthropocentric is no objection in itself, since any study of human representations is that to some degree; the objection is more specific. Lévi-Strauss takes human symbolic classifications, interprets the relations among their terms, and presents the result as a foundational account of human relations with other beings. Institutionally stabilised, binary classification is, however, a particular and late achievement of human multisymbolization, located at what Living Value Theory (LVT) calls the fourth recursivity level, L4. No nonhuman animal takes part in anything resembling it, for reasons that will be set out below. Structuralism thus selects the one operation that humans do not share with other animals and makes it the organising principle of the anthropology of animals. Two displacements follow. The coordination of the animal with its own world disappears behind the category assigned to it, and the coordination of the human with the animal disappears behind an analysis of relations among categories. The animal is absent twice.

The celebrated formula should accordingly be refused. Animals are living beings engaged in the recursive coordination of their own worlds long before they are good to think with, and most of them never are. A fox is finding its way along a hedge. A ewe is keeping her footing on a wet slope. A hunter is answering the movement of his prey, and a child and a dog are absorbed in a game. Nothing in these processes asks to be organised around the intellectual distinctions that animals allow humans to draw, and many of the most consequential coordinations between humans and animals succeed precisely when nobody is thinking about them. The formula also privileges, without argument, the activity in which human beings spend least of their time. People can think perfectly well without animals, and they can live beside animals for decades without ever thinking with them in the structuralist sense. People do reason with animal images, and the error lies in mistaking one downstream use of animals within human symbolic life for the reason animals matter. Many animals that carry no representation at all, among them pollinators, soil fauna, parasites, and the vectors of disease, are consequential for human living to the highest degree, and an approach oriented to symbolic elaboration cannot reach them.

The particular instrument of structuralist analysis, the binary opposition, has been examined in detail elsewhere (Une Erreur Sauvage), and the results need only be applied. Classification is a late output of coordination and not its beginning. Differentiation, which is everywhere in living worlds, is not opposition. Binaries are produced when ordinary language is put to work in governance, where a gradient has to be converted into a decision, and structuralism mistook that affordance of one symbolic technology for the structure of mind. The pairs through which animals are habitually approached, wild and domestic, human and animal, predator and prey, edible and inedible, sacred and profane, pure and polluting, are of this kind. A structuralist can chart their relations, and the analysis is often considered complete once it has been settled on which side of each pair a species falls. As descriptions of animals these are the poorest available. LVT asks how each pair emerged, which coordinations it stabilises, where it ceases to fit, and what happens to humans and animals when it governs action. A fox does not become a different living being when an institution classifies it as vermin and not as wildlife. The classification nevertheless decides whether it may be snared, when, and by whom, and so reorganises its possibilities of continuing to live. What a binary means matters less than what it does.

Totemism is the most revealing case. For more than a century it taught anthropologists to ask how a named species represents a human collective, and its dissolution by Lévi-Strauss left that way of asking intact, since the species merely passed from emblem to operator. On the structuralist account the choice of species is in principle arbitrary: a clan could as well be the eagle clan as the bear clan, because what the animal is and does never enters the analysis, only its difference from the other terms. The association between a group and a species name is, however, what stands in need of explanation, and it cannot serve as the foundation of the relationship. Why that animal, in that country, with those practices and that history of encounter? A people who call themselves after the wedge-tailed eagle live among eagles, watch them, are watched by them, and move through places that eagles also use. The name is a deposit of that coordination, to be investigated like any other deposit.

Later anthropology recognised part of the problem. Tapper (1988) showed that relationships with animals vary with hunting, herding, farming, and industry, and that several coexist in one society, but he ordered the variation by analogy with human relations of production. Descola (2013) demonstrated that the modern division of nature from culture is one arrangement among several, and then rebuilt the field as four ontologies generated by two axes of classification, one of which is still called totemism. Each loosens the structuralist frame without leaving it, and both belong to a wider habit in which anthropology treats symbolic deposits as the primary reality of what it studies (Culture Is Not a Text).

The thesis of this article can now be stated. The long anthropological preoccupation with what animals mean to humans has displaced the more fundamental question of how animals live, how humans and other animals become consequential to one another, and what particular vital stakes their coordination sustains or threatens. LVT replaces the primacy of symbolic classification with an investigation of recursive life in the mesocosm, and reconstructs the continuities and differences between human and nonhuman animals without making either human symbolic thought or the human-animal distinction the foundation of inquiry.

III. The Ultimate Anthropocentrism

The anthropology of animals has worried for decades about anthropomorphism. Projecting human intentions onto a dog, a whale, or a crow is treated as the characteristic error of the untrained observer, and recent calls to decentre the human have made the avoidance of anthropocentrism something close to a disciplinary virtue. The structuralist lineage has largely escaped this suspicion, because it appears to do the opposite of projecting human qualities onto animals. It studies how people think, and it leaves animals alone.

That appearance is misleading. An approach that admits animals into the analysis only as items in a human classification, and only to the extent that they help human beings think about human beings, has decided in advance that nothing an animal does on its own account is relevant. It is difficult to imagine a more thoroughly anthropocentric procedure. Anthropomorphism at least credits the animal with a life, even if it misdescribes that life. Structuralism credits the animal with a position in a scheme. The way out of anthropocentrism, accordingly, lies in the study of animals in living coordination: who they are, what they are doing, and what humans are doing with them.

To see why this matters, the recursivity levels that LVT distinguishes need to be set out, since they show both what structuralism took and what it left behind. They can be illustrated from any ordinary room. At L1, coordination proceeds in absorbed, transparent fashion: one gets out of bed, the floor is where it should be, gravity behaves as expected, the table holds up what is placed on it, the air can be breathed, and nothing stands out. Most of any living day is spent here, and it is good that it is, since nobody wishes to attend to a table while trying to do something else. At L2, coordination becomes salient because it has been disturbed, and the disturbance is felt: a patch of pavement is slippery, a chair has become uncomfortable. The disturbance is often remediated before anything has been formulated, as when someone who has sat too long simply shifts position. At L3, salience becomes symbolically articulable: one says that one has been sitting too long, or warns somebody of the ice. At L4, articulations become portable and institutionally stabilised, as categories that apply regardless of the particular case: raw or cooked, clean or polluting, predator or prey, wild or domestic. At L5, stabilisations themselves become open to revision, as when one asks what a whole classificatory system is doing and whether it should be kept.

Seen through the levels, structuralism is a theory of L4 constructed at L5, and it is unsurprising that it found L4 everywhere it looked. Its characteristic move is to take L4 stabilisations, the raw and the cooked, left and right, male and female, and declare them to be the structure of mind, and so the shape of the world as it is lived. Its recent ontological heirs make a parallel move when they treat the way a people classifies beings as that people's world. In both cases a human symbolic operation, late and specialised, is mistaken for the ground of everything else.

The same levels expose the mirror image of the structuralist error, which runs through ordinary thinking about the difference between humans and other animals. Humans think, the familiar argument goes; they have reason, language, and tools, and animals do not. Each item on such lists has rudimentary counterparts among other animals, which signal to one another, make and use tools, and find their way round obstacles. What remains distinctively human is the elaboration of these capacities into L4 and L5, and those are the levels at which human beings spend the smallest share of their lives. For most of every day, and for all of every night, humans do what other animals do: they coordinate in absorbed fashion, register disturbances, remediate most of them without articulating anything, and stay in relaxed, low-energy responsiveness to the beings around them. A comparison that begins from what humans have and other animals lack therefore begins from the narrowest region of the human. It should begin instead from the overlap, which is where living concerns are shared and where most of the relationships between humans and animals are conducted.

The levels also cut through the debate about anthropomorphism. Attributing L1 and L2 to a dog projects nothing, because the dog plainly coordinates at both, and so does its owner. Nor does attributing interrecursive coordination to the dog, which answers the responses of its household all day long. Attributing to the dog an L4 classification of the household into kin and strangers, or an L5 reflection on its own situation, would be projection, and an unfounded one. The question of anthropomorphism is a question about levels, and once it is posed in that way it becomes answerable case by case.

IV. The Zoetic Continuity of Animal Life

LVT starts from life. Its basic proposition is that living and recursive valuing are one continuous process (Ecks 2022). A living being continually sustains, repairs, and transforms the coordination on which its continuation depends, and it discerns differences according to what they do to that coordination. The world so coordinated is its mesocosm: larger than the organism, far smaller than the cosmos, and consisting of whatever has become relevant to that life. The mesocosm is the world as it is lived, and it is no construction. A lecture hall full of students is a shared mesocosm in this sense: everyone in it breathes the same air, sits in the same gravitational field, and has the same metabolic needs, which is why the lecturer can judge, without asking anybody, that after an hour a break will be welcome. No divide between biological needs and cultural meanings is required at the start. That divide is a historically recent symbolic stabilisation, and the comparative habits of anthropology were built on it (The Space of Difference).

Living coordination runs through mediations that are irreducible to one another, each defined by the kind of counterpart it coordinates (What Are the Five Mediations?). Three are zoetic, in that they belong to living as such. Multisensorial embodiment coordinates the recursive continuation of the living being itself: orientation, movement, sensing, feeding, pain, fatigue, repair. Multispecies being-with coordinates other living beings, through encounter, proximity, avoidance, cooperation, reproduction, predation, and learning, and those others are rarely of one species only. Multiversal dwelling coordinates found ground: terrain, water, climate, gravity, light, and seasonal rhythm, which no living being made and with which all must come to terms.

These are dimensions of a single process and cannot be measured as separate components. A swallow returning to a nesting site coordinates the mechanics of its own flight, the presence of rivals and of its mate, and the wind and failing light, all at once. To ask which mediation the swallow is really engaged in is to misunderstand them. Each names something that cannot be derived from the other two: no account of the body of the bird yields the rival, and no account of the rival yields the wind.

Of the three, dwelling is the least appreciated, and for the anthropology of animals it may be the most important. Every animal in Scotland coordinates with the same photoperiod, the same run of seasons, the same tides along the coast, and the same lunar cycle. Humans do so as well, though they seldom notice it. In an Edinburgh December the sun rises shortly before nine and sets before four, and the short days are felt in mood, sleep, and appetite, sometimes to the point of a clinical diagnosis. Weather, altitude, distance, visibility at night, and the rhythm of light and dark organise a human life as surely as they organise the life of a heron. Humans lose sight of this because they insulate themselves from dwelling through multimateriality, with heating, glazing, electric light, and roofs, so that found ground recedes behind made ground. Other animals have no such insulation, and an account of them that omits dwelling omits much of what they are doing.

A recent dissertation in social anthropology at Edinburgh, on dog walking in the city, shows what follows when dwelling is taken seriously. Nothing could seem less promising as an object of study. Yet a person and a dog crossing the Meadows on a wet morning share the same terrain, the same rain, the same light, and the same distances. What differs is salience. Through its own multisensorial embodiment the dog finds relevant what the walker scarcely registers, the scent at the base of a lamppost, the trace of another dog along the edge of the grass, while the walker attends to the time, the traffic, and the path. The two are in one dwelling and in two mesocosms, and the walk is the daily coordination of the two. Shared ground and divergent relevance is a pattern that recurs throughout relationships between humans and animals.

Three consequences follow for the anthropology of animals. The first is that a zoetic world is not an impoverished one. A wolf lacks architecture and law, and its world is not thereby incomplete (What Are the Five Mediations?). In particular zoetic domains many animals exceed human beings by a wide margin, as a swift does in flight, a dog in scent, and a migrating godwit in its coordination with wind and season, and nothing in the framework treats zoetic coordination as simple (Zoetic and Poietic Mediations).

The second is that the zoetic mediations are not a foundation that humans have left behind. Human beings remain constituted, every hour, through embodiment, being-with, and dwelling. An infant coordinates zoetically for many months before any symbol is available to it, and an adult asleep does so every night. The human-animal distinction therefore does not divide two kinds of living existence, and humans are animals in the precise sense that they share the whole of this ground.

The third is methodological. Anthropology has attended to the mediations very unevenly. Being-with among humans and symbolisation dominate its archive, and dwelling has been nearly erased from it (An Emporium of Benevolent Knowledge). An anthropology of animals built from that archive inherits the imbalance, which is one reason why it found animals chiefly where humans talk about them. Starting from the zoetic mediations reverses the order of inquiry: bodies, other living beings, and found ground come first, for the animal and for the human alike, and what is said and depicted comes afterwards.

V. Metabolism Before Meaning: Why Are There No Tree Ghosts?

Among the zoetic facts that shape relationships between humans and other animals, one is so basic that it is easily overlooked. Some living beings are autotrophic: plants and many microorganisms produce what they need to live from light, water, and air. Others are heterotrophic and must consume other living beings to continue. Every animal is a heterotroph. A vegan human being eats other lives as surely as a lion does, only plant lives, and there is no way of living as an animal without doing so. No society has constructed its way out of this condition, and no classification can alter it. The names autotroph and heterotroph are nineteenth-century scientific coinages, while the condition they name has been coordinated by every animal that ever lived.

A leftfield question: why are there no tree ghosts? Why are ghosts mostly human, with a few dog, cat, or horse ghosts, but not plants? Trees are inhabited by spirits in many places, from the yakṣīs of early Indian sculpture to the kodama of Japanese folklore, yet the ghost of a felled tree, returning to trouble those who cut it down, is a figure I could not find documented in any culture. Nobody expects a ghost from a mown lawn, and a person who mows the grass does not feel like a mass killer. Why should that be?

The question is sharper than it looks, because it tests the structuralist assumption of arbitrariness. If the distribution of ghosts were a matter of free symbolic classification, there would be no reason for haunting to follow any particular line through the living world. It does follow one. The beings that return after death are, overwhelmingly, beings that during life ate, moved, fled, bled, and responded to those who responded to them, and above all beings with whom people had been engaged interrecursively. A ghost is a counterpart whose responsiveness outlasts its death. That the ghostly dog is common and the ghostly oak rare is what one would expect if the attribution of afterlives tracks heterotrophic kinship and histories of interrecursive coordination, and it is hard to see how one would expect it on any other account. The hypothesis needs comparative testing, but its form is instructive. A classification that looks symbolic turns out to be constrained by metabolism and by the kind of counterpart a being was in life.

The same constraint appears wherever one looks at relationships between humans and other animals. It is no accident that the large mammals domesticated for meat are herbivores, which convert plants that humans cannot digest into flesh that humans can, or omnivores such as the pig, which thrive on what humans leave over, while the carnivores that live most closely with people, dogs and cats, entered human mesocosms as hunters, guards, and scavengers of human refuse. Grazing animals compete with one another and with human cultivators for pasture. Predators compete with hunters and prey on herds. Rats, mice, and house sparrows live from human stores, and lice, fleas, and mosquitoes live from human bodies. Each of these relationships is defined, before anything else, by what the animal eats, where, when, and at whose expense.

A methodological maxim follows. To understand any animal and its relations with humans, begin by asking what it needs to consume and what consumes it. The answer will not settle everything, but it settles a great deal of what matters, and it leaves very little room for social construction. Eating is accordingly one of the privileged entry points of any mesocosmography, for animals as for people. The cave walls point the same way from the opposite direction. What fed people, the reindeer at Lascaux and the gazelle at Göbekli Tepe, was coordinated so routinely that it scarcely needed depicting. What might eat people was depicted again and again.

VI. Recursivity in Animal Mesocosms

Recursivity is the capacity of a living process for the consequences of its previous mediation to alter the conditions of its subsequent mediation (What Is Recursivity?). It is no synonym for reflection, and it does not require language. A tree that grows in relation to earlier damage is recursively organised. The concept of consciousness is of little use here, because it covers everything from absorbed movement to institutional critique in one word and conceals the differences that matter. LVT distinguishes instead three types of recursive counterpart and the five recursivity levels already introduced.

A nonrecursive counterpart, such as a slope, a current, a branch, or a chair, does not alter its course because it has been engaged. One can sit on a chair and the chair will not mind, flee, or be afraid. That does not make a nonrecursive counterpart stable or predictable, as anyone who has watched a deer on river ice will know. Selfrecursive coordination is the response of a living process to its own changing state, in metabolism, pain, repair, learning, memory, and anticipation: whether one is hungry, thirsty, tired, or in need of a break. Interrecursive coordination occurs when two or more recursive beings respond to each other's responses, so that the next move of each depends on the other. To sit on a stranger's lap would be a different act from sitting on a chair for exactly this reason. These three do not mean unthinking, thinking about oneself, and thinking about others. Selfrecursivity has nothing to do with possessing a concept of self. It concerns the way prior coordination modifies the conditions of later coordination in the same living being, and every animal has it necessarily: a rat that avoids a food which once made it ill, a salmon returning to its natal stream, and a hamster working night after night at the weak corner of its cage are all selfrecursive. Nonrecursive and selfrecursive counterpart relations are universal in animal life, interrecursive ones very nearly so, and animals themselves draw the distinctions: a cat that sleeps on a cushion and keeps an eye on the dog has sorted its counterparts with some precision.

Interrecursive coordination is a mode that most animals have, and no achievement reserved for a few clever species. Ants on a trail follow the pheromone laid by nestmates who followed theirs, pass food to one another mouth to mouth, and touch antennae in passing. In tandem running, an ant that knows the way to a food source slows and waits whenever its follower loses contact, and the follower signals with taps of its antennae that it is ready to go on, so that each adjusts its pace to the other's (Franks and Richardson 2006). Fish in a school and starlings in a murmuration turn as their neighbours turn, and their neighbours turn as they do. Wolves hunt in differentiated roles, a scrub jay re-caches food after being watched by a rival, and primates keep grooming partnerships over years. Nothing in the definition confines interrecursivity to humans, to mammals, or even to animals, so that responsive coordination between plants would have to be assessed by the same standard. Interrecursivity is neither a human prerogative nor a rung on an evolutionary ladder.

What varies is the intensity of interrecursive engagement and what it costs. At one end lie low-energy forms: resting together, grooming, relaxed co-presence, the mutual spacing of a flock. Gulls spaced at even intervals along a roof ridge in northern Norway, watched over many hours, kept their distances from one another with great regularity. Each registered and answered the small movements of its neighbours, shuffling along when one landed too close, without anything that would count as close engagement (Being With Others Doesn't Mean Having to Think about Others). Being with others does not require thinking about others, and it requires very little energy. Students in a lecture hall do the same. They distribute themselves evenly across the rows, settle, and thereafter keep a background responsiveness to one another that costs almost nothing, which is why a neighbour who kept changing places would be so irritating: the arrangement exists to keep interrecursivity at its lowest setting, and the restless neighbour raises it. Grooming among primates is the classic case of a low-energy interrecursivity that sustains relationships over long periods, and Dunbar (1996) has argued that much human talk does the same work, as a vocal grooming that maintains bonds at little cost. The three criteria proposed for interrecursive coordination, differential engagement with particular individuals, adjustment to what a particular other is about to do, and sensitivity to how that other will respond to one's own move, are best read as markers of rising intensity within a mode that is already present.

At the other end lies maximal interrecursivity. In the most casual conversation, each participant answers what the other has just said, revises in response to the answer, and is revised in turn, so that neither can predict what will be said a minute later, let alone further on. The unpredictability marks the interrecursive relation, since the future of the exchange is generated jointly and belongs to neither party in advance. Maximal interrecursivity is absorbing and metabolically draining, and it cannot be sustained for long. Its most draining form is antagonistic: engagement with a counterpart whose aims run against one's own, whose every move has to be anticipated, and who is anticipating one's own. The gulls on the ridge entered this mode from time to time, when one contested another's place or a scrap of food, with posturing, calling, and occasionally a fight, and afterwards returned to their spacing. Animals avoid antagonistic interrecursivity as far as they can. Tinbergen (1953) showed how much of the herring gull's threat behaviour settles disputes without fighting, and Maynard Smith and Price (1973) argued that ritualised contests are favoured because escalated fights are costly to winners and losers alike. Predator and prey are each other's antagonistic interrecursive counterparts, which is one reason why every hunted animal spends so much of its life keeping out of the way.

Humans behave in the same way, spending relatively little time in maximal interrecursivity and still less in its antagonistic form, which is among the most exhausting experiences most people know: the quarrel, the hostile negotiation, the cross-examination. Most human interrecursivity is relaxed: sitting together, sharing a meal, walking side by side, the half-attentive companionship of a household. Dunbar (1992) related the size of the neocortex in primates to the size of their social groups and predicted that human beings can sustain stable, individually differentiated relationships with about one hundred and fifty others, a figure that survived the arrival of online networks in which people count their friends in thousands (Dunbar 2016). The limit applies to intense, individually tracked relationships, and relaxed interrecursivity extends much further at much lower cost. Relationships between humans and other animals follow the same pattern. Anyone who lives with a dog will notice that the dog is around most of the time in a relaxed, mutually adjusted co-presence, each keeping half an eye on the other, and that there are moments, a look across the room, a nudge towards the treat cupboard, a game of chase, when each reads the other closely. Much writing on humans and animals has concentrated on direct encounter, as though the intense face-to-face meeting were the paradigm of being with an animal. Most of being with animals, as of being with humans, is interrecursivity at its lowest setting, punctuated by brief episodes of the most intense kind.

The levels need the same care as the counterpart types. L1 and L2 are fundamental to animal living. A fox that stops short before a newly fallen tree, sniffs, circles, and tries another line has formulated nothing in words, and the obstruction has nonetheless become relevant to it as a disturbance. Pain is the plainest L2 phenomenon, and every animal capable of pain has L2. The levels are no fixed properties of activities or species. The same ground is crossed at L1 on a hundred evenings and becomes an L2 matter on the evening when the stream has risen, and a living being is in good order when the passage between the levels is fluid (Recursive Mediation as Continuous Discernment). L2 is metabolically costly, which is why an animal held there without relief, in chronic pain or in confinement that blocks every familiar route, deteriorates.

At L3, salience becomes symbolically articulable. Since articulation is a symbolic act, L3 can occur only where some form of multisymbolization is available, and rudimentary forms therefore occur in those animals that have rudimentary signs: the predator-specific alarm calls of vervet monkeys, which distinguish leopard, eagle, and snake (Seyfarth, Cheney, and Marler 1980), are an example, and birdsong and whale song are others. Such a capacity must not be conflated with ordinary human language, and its absence from an experimental setting does not establish its absence from the living repertoire of the animal. In animals as in humans, L3 occupies a small share of coordination, which runs overwhelmingly through posture, movement, scent, and touch. L4 and L5 are the exception. They are found in no nonhuman animal, for reasons taken up below, and they constitute the one substantial difference between humans and other animals. No animal other than the human sorts the world into the raw and the cooked. Table 1 summarises the position and marks what humans share with other animals and what is distinctively human.

Table 1. Recursive availability in animal life: what humans share with other animals and what is distinctively human.

The table is deliberately not a grid. It would be easy to cross three mediations with levels and counterpart types and to announce a fixed number of cells that every animal fills. That would be a mistake of exactly the kind this article criticises, because the dimensions are not independent variables. Embodiment is paradigmatically selfrecursive, dwelling coordinates nonrecursive ground, and being-with is where interrecursive counterparts are found, at intensities that change from moment to moment. A classification scheme laid over animal life in advance would repeat, in the vocabulary of LVT, the structuralist substitution of a table for a living process. What the theory proposes for comparative purposes is a minimal zoetic profile, followed by an inquiry into the intensities of interrecursive engagement a particular animal enters in a particular situation, and the rudiments of poiesis and articulation it has. Capacity, current availability, and the capacity that others attribute to the animal are three different things.

How, then, is an animal mesocosm to be investigated? Mesocosmography is the systematic description of consequential coordination (What Is Mesocosmography?). Applied to a relationship between humans and animals, it reconstructs the ordinary organisation of coordination, the participating beings and their respective mesocosms, the mediations involved, the counterpart relations that are in fact established, and the interruptions that make transparent coordination visible. It treats what humans say about animals as one access among several and gives it no priority. A flock of sheep cannot be studied through interviews with shepherds about what sheep mean to them. It requires observing sheep, people, dogs, land, hurdles, weather, illness, feeding, and the timing of all of these, and it requires labelling each inference as direct, necessary, probable, or unsupported.

VII. Recursive Fluidity: The Mouse and the Nut

The levels are easily read as a ladder, with the human at the top and other animals arranged on the lower rungs. A children's book found in a great many Scottish households shows why that reading is wrong. In Julia Donaldson and Axel Scheffler's The Gruffalo (1999), a mouse walks through a deep, dark wood and is approached in turn by a fox, an owl, and a snake, each of whom invites him home with the evident intention of eating him. The mouse declines each invitation by claiming an appointment with a gruffalo, a monstrous creature he describes in terrifying detail and whose favourite food, he adds, is the animal in front of him. Each predator flees. The mouse then meets a real gruffalo, who proposes to eat him in turn. The mouse announces that he is the most feared creature in the wood and invites the gruffalo to walk behind him and see. The fox, the owl, and the snake flee at the sight of the gruffalo, the gruffalo concludes that they are fleeing from the mouse, and when the mouse mentions that he is now hungry the gruffalo runs away too. The book ends with the mouse alone in the quiet wood, eating a nut.

The story is about predation, about who eats whom, and in that respect one of the most widely read picture books in Britain shares its preoccupation with the walls of Chauvet. A remarkable share of the animal stories told to small children concern the animals that eat, far more than the animals that are eaten. The story is also a precise study of recursivity. The fox, the owl, and the snake each deploy an interrecursive deception, an invitation that conceals an intention, and the mouse answers each with a symbolic articulation that refers to an absent being. Reference to what is absent occurs in rudimentary form among animals, as when a honeybee's dance directs nestmates to forage out of sight, and the book gives an elaborate version of it, with evident delight, to the smallest animal in the wood. With the gruffalo the mouse goes further. He anticipates how the other animals will respond to the gruffalo, and how the gruffalo will read their response, and he arranges the walk so that each reading produces the next. The gruffalo, strongest of all, fails as an interpreter: he attributes the flight of the others to the wrong counterpart.

Every encounter in the book is antagonistic interrecursivity, the most draining kind, conducted at its highest intensity, with being eaten as the price of a misreading. What makes the book more than a fable of cleverness is its ending. The mouse does not remain at the height of his achievement. Once the danger is past he returns to the most ordinary coordination available to a mouse, eating, in a wood that is quiet again. The book teaches young children, among other things, that high-level recursion is for emergencies and that the point of mastering an emergency is to be able to stop attending to it.

Psychology has reached a related conclusion by another route. Kahneman (2011) distinguished fast, automatic, and largely effortless processing from slow, deliberate, and effortful reasoning, and showed how quickly the second is exhausted and how readily people fall back on the first. LVT's levels are not two systems, and the movement among them matters more than either pole. The comparative point is nonetheless the same. Higher recursion is expensive, it cannot be sustained for long, and every living being, human or not, must return from it to coordination that needs no attention. LVT calls the capacity for that movement recursive fluidity: the ability to rise to the level a situation requires and to come back down when the situation allows. Vitality consists in recursive fluidity, and never in remaining at any one level. The most successful animal is the one that knows, without needing to articulate it, when to be alert and when to rest.

This changes the question put to animal minds. Comparative cognition has largely proceeded by listing human accomplishments, such as planning, mirror self-recognition, and false-belief tracking, and asking which species climb how far. Cognition cannot be detached from the mesocosm and the metabolic stakes of the animal that exercises it, and higher recursion is not better recursion (Easy Does It). Living beings avoid exhausting themselves. Mastery drives recursivity down, as when a migrating bird or an experienced sheepdog coordinates superbly at L1, and an apparatus that forces an animal into effortful L2 processing measures the misfit between the laboratory and its world as much as any capacity. The aim of a living being is viable coordination, and never maximal complexity.

Human beings are distinctive in how far up the levels they can go, and in how much of their made world is devoted to coming back down. Houses, beds, sofas, and the evening meal are arrangements for the return to L1, places where nothing needs watching. The human capacity for L4 and L5 is a remarkable achievement, and it remains, like the mouse's stratagem, something that has to be set down again. What humans share with every other animal is the need to set it down.

VIII. Rudimentary Poiesis and the Human Evolutionary Threshold

Two further mediations are poietic, in that they concern making. Multimateriality coordinates made material ground, and multisymbolization coordinates made symbolic ground. Both are nonrecursive counterparts, since a made thing does not answer back, and that is the source of their usefulness.

Animals are not confined to the zoetic mediations. Chimpanzees strip and shape stems to fish for termites, with techniques that differ between populations. Beavers dam streams and weaver birds knot grass. New Caledonian crows manufacture hooked tools (Hunt 1996), and octopuses carry coconut shells about as portable shelter. Rudimentary multisymbolization is equally real: the alarm calls of vervets, the dance by which a honeybee conveys the direction and distance of forage (von Frisch 1967), regional dialects in birdsong, and the structured vocalisations of whales. Chimpanzee tool techniques are passed on by observation and birdsong dialects are learned from neighbours, so that animal poiesis already has a history. Neither mediation, taken separately or together, distinguishes human beings in kind.

What distinguishes human beings is what the elaboration of the two poietic mediations made possible. In the human lineage, material change reshaped symbolic possibilities and symbolic change reshaped material ones, and the process kept expanding instead of converging on a fixed repertoire (Zoetic and Poietic Mediations). Reference to what is absent, present in rudimentary form in the honeybee's dance, became detachable from the occasion, so that a symbol could stand for something not currently there and serve for planning, teaching, and remembrance. The evolutionary account ties this elaboration to practices, such as seasonal and long-distance hunting, that stretched coordination across absent times and distant places (The Genesis of the Five Mediations).

The elaboration matters because it made L4 and L5 possible, and these depend on it for a precise reason. An L4 stabilisation is an articulation that has become portable: it holds regardless of the particular case, speaker, and occasion, so that a category such as clean or polluting, or a rule about who may kill which animal and when, can be applied by people who never met those who first formulated it. Portability of this kind requires a very elaborate multisymbolization, and it requires multimaterial means of inscribing, storing, and transmitting symbols beyond the moment of their utterance. Writing is the most powerful of these means, and it was preceded and is still accompanied by many others: marks on bone and stone, painted and carved images, ritual objects, tallies, ornaments, and the layout of built places, each of which holds a symbolic distinction steady and carries it to others. L5 makes still greater demands, since stabilisations can be revised only when they are held still long enough to be set side by side, compared, and criticised, which in practice means written, archived, and institutionally transmitted. No animal other than the human has symbolisation of this elaboration, and none has the material means to store and transmit it. That is why L4 and L5 belong to humans alone, and it is the only substantial difference between humans and other animals.

The recursive coupling of the two poietic mediations has a recognisable shape. A symbolic distinction guides a construction, the constructed thing preserves the distinction, and the arrangement becomes the object of renewed symbolic and material revision. A carved animal is the plainest case. A trap is another, since it holds in wood and cord a body of accumulated knowledge about how a particular animal moves. A register of livestock, a consecrated image, and a deed of ownership over a herd carry the coupling into arrangements that outlast those who made them.

It is tempting to summarise the difference temporally, as though zoetic life were bounded by the lifespan and poietic life escaped it. That is too compressed. Dwelling coordinates conditions that run far beyond any individual life: a salmon run or a migration route is older than every animal that uses it, and mountains and solar cycles persist independently of all organisms. Animal living also leaves consequential material transformations of its own. A beaver meadow, a termite mound, or a game trail may organise the lives of generations that had no part in making it. The mediations have different temporal profiles, and the differences are relational, not simply matters of duration (A Process Ontology of Differential Temporalities). What human poietic coupling adds is the sustained recursive elaboration of durable material-symbolic ground, deposit laid upon deposit, each available for revision. Human beings did not simply adapt to a given dwelling; they began manufacturing new ground, and the dead continue to organise the living through what they left (The Ground We Made, the Ground We Found).

Two qualifications keep the argument from becoming a ranking. The difference is developmental and evolutionary and carries no hierarchy of value. L4 and L5 make human mesocosms differently built, able to be constituted across absence, distance, and generations, and vulnerable in ways no other animal is vulnerable. The capacity is also directionless: the same elaboration that paints a lion and builds a fold writes the regulation for the battery shed. Humans share with other animals the zoetic mediations, all three types of recursive counterpart, the levels L1 to L3, and the rudiments of poiesis, and they differ by the two levels that the elaboration of the poietic mediations made possible. The rest of the argument depends on holding these statements together.

IX. What Is at Stake? Mesocosmic Relevance and Living Value

What exactly is at stake in a given encounter between a human being and another animal? The phrase "human-animal relationship" suggests that the answer has a general form. Consider what it would have to cover: a child playing with a dog, a herder moving sheep to summer pasture, a person slapping a mosquito, a hunter following an elk through snow, a veterinarian palpating the leg of an injured horse, a visitor standing before a painted bison. The category names some of the participants. It establishes nothing about what matters in any of the encounters. The prior question is what is at stake, and what something means comes a long way after it (Living Value Theory as a Discipline of Attention).

LVT has a precise account of stakes. What is ultimately at stake for a living being is the continuation of successful, transparent coordination across its mediations at L1, and value is whatever sustains, repairs, intensifies, or improves that coordination (What's Vital?). When embodiment is in good order, when one is neither exhausted nor in pain, when the people and animals around one are as they should be, and when dwelling, made things, and signs hold steady, life goes well, and nobody notices. L2 is the registration that such coordination is threatened, and the higher levels, where they exist, are increasingly elaborate forms of remediation whose aim is a return to coordination that no longer needs attention. This changes the description of familiar scenes. When a child who was happily playing with a dog becomes anxious because the dog has stopped responding, the event is poorly described as a change from one emotion to another: a coordination that worked is under threat, and what the child wants is its restoration. The same holds when a shepherd notices that a ewe is moving oddly.

Three questions can therefore be put to any relationship between a human and another animal, and to each participant separately. What coordination must ordinarily work for this mesocosm to remain viable? What becomes salient when that coordination is threatened or interrupted? What forms of remediation are available, and at what cost? No predefined list of biological needs, economic functions, emotional attachments, or symbolic meanings answers them. Such labels may describe certain stakes once the work has been done and cannot fix them beforehand.

Three properties of a counterpart also have to be kept apart, because they vary independently (Recursive Openings). Recursive attribution is the settled sense of what kind of counterpart a being is engaged as: nonrecursive, selfrecursive, or interrecursive. Recursive relevance is how much of ongoing coordination depends on it. Recursive opening is whether the relation is operative at this moment or sedimented. An animal can be vitally consequential and scarcely salient. Darwin (1881) devoted his last book to earthworms and showed that they turn over the soil of the whole cultivated earth, yet few people spend any part of a day thinking about them. Pollinating insects are in the same position: bee populations have declined sharply under the combined pressure of parasites, pesticides, and the loss of flowering plants (Goulson et al. 2015), with consequences for the crops on which human beings depend, and it takes a recurring scare for the decline to register at all. Nobody keeps a pet bee, while many people treat their dogs as members of the family, and relevance follows the dog. The cattle inside an industrial supply chain are vitally consequential and scarcely salient in the same way. An animal can also be intensely salient while very little depends on it, as a spider on the wall is to a person who fears spiders. An animal can be engaged through an attribution that held for years and suddenly becomes questionable, as when a farmer sees grief in a cow. An anthropology that follows only what people notice about animals will miss the animals on which they most depend, and one that follows only attribution, as the anthropology of ontologies does, will miss both relevance and opening.

Stakes may be shared, divergent, or directly opposed. For the hunter, the encounter with the elk may bear on food, on practised skill, on standing among other hunters. For the elk, it bears on evading a threat. Coordination between them can become extraordinarily fine without either sharing the stakes of the other. Predation, parasitism, competition, captivity, domestication, and care each distribute stakes differently, and mutual responsiveness guarantees nothing about mutual benefit.

Most animals, moreover, are not relevant to a given mesocosm most of the time, and this holds in both directions. A person may be flattened by grief at the death of a companion animal while hardly registering the deaths of animals elsewhere. No inconsistency in the assignment of importance is involved, because importance was never assigned. The companion took part in thousands of daily coordinations: the weight on the bed, the sound at the door, the walk that structured the evening. Its absence disrupts an established mesocosm at L1, where the loss is felt before it is thought. From the other side, the nonhuman animal is never merely a participant in a human-animal relationship. It is engaged in the coordination of its own mesocosm, into which humans may enter as dangers, companions, sources of food, predictable presences, or beings with whom interrecursive coordination becomes possible. Humans may also never become relevant to it at all, and it remains a living being regardless.

Stakes, finally, are generated and transformed in the course of coordination. A fleeting encounter becomes play, play establishes familiarity, familiarity becomes dependence, and dependence brings responsibility, vulnerability, and eventually grief. The sequence can also run down, until what once demanded close attention is habitual, then uninteresting, then out of reach. Value, on this account, is the living significance of ongoing coordination, differentiated through particular mesocosmic stakes. Usefulness, exchange, affection, and encounter are among its occasions and are never its source.

X. Recursive Attribution in Practice: Names, Food, and Sleep

Recursive attribution varies between societies and within them, and it varies historically. The differences are real, and they are consequential above all where violence is concerned. A chair can be kicked, broken, and thrown out of a window without the act returning to trouble anyone's sleep, because a chair is engaged as nonrecursive. A being engaged as interrecursive is a different matter: what is done to it may be answered.

Modern Europe narrowed the range of beings engaged in that way to a degree that has few parallels. Weber (1946 [1919]) called the outcome disenchantment. The sociologist Jonathan Hearn has described the same process in sharper terms, as a culling of nonhumans: in the course of the eighteenth century, ghosts, gods, ancestors, fairies, and many animals lost their standing as counterparts that had to be minded, until interrecursive standing was reserved, in principle, for human beings alone. Whatever one makes of the larger history, its effects on animals are plain. Industrial livestock farming engages chickens, pigs, and cattle as nonrecursive counterparts, things that happen to breathe and have a heartbeat, while the animals themselves remain selfrecursive, and interrecursive with one another and often with their keepers. Attribution and capacity have come apart on an enormous scale.

Elsewhere attribution is broader. Among the Rock Cree of northern Manitoba, as Brightman (1993) describes them, animals give themselves to hunters who treat them with respect, and the disrespectful treatment of a kill can lead animals to withhold themselves in future, so that the relationship between hunter and prey continues through and beyond the death of the animal. Similar practices of address, apology, and care in killing have been reported across the circumpolar and North American hunting world. In Hinduism, the doctrine of karma and the value of ahiṃsā, non-violence, extend the range of beings whose injury matters, which is one reason why the distinction between vegetarian and non-vegetarian food carries so much weight. Jain renouncers carry the principle furthest, sweeping the ground before them, straining water, and avoiding food and movement that might harm the smallest living beings (Laidlaw 1995). In each case the scope of non-violence follows the scope of interrecursive attribution. Violence becomes a moral and cosmological problem exactly where a counterpart is engaged as one that can respond.

Attribution is not exhausted by the question whether a being counts as recursive. It has a profile across the mediations, and the profile is visible in what people do. In Hindu temple worship the deity, present in its image, is woken in the morning, bathed, dressed, fed, and put to sleep at night (Fuller 2004). The deity is engaged, in other words, as a metabolically embodied being with the daily needs of a living one. Ask whether the goddess actually eats the food that is offered, and the question is usually waved aside as missing the point; the food returns to worshippers as prasād, transformed by her acceptance of it. What is at stake is an extension of embodiment, of feeding and sleep, to a counterpart that no one claims eats as a human eats. The attribution is enacted in the coordination of a temple day, and it would be poorly described as a belief held about the goddess.

Compare a student who has given her scooter a name, and who speaks of it with some affection. On a classification of whole societies by their attributions of interiority, such as Descola's (2013), this might be counted as a trace of animism, and its owner as a modest animist. Yet nobody puts a scooter to bed, bathes it, or offers it food. The name extends a fragment of multisymbolization, and perhaps a fragment of being-with, to a nonrecursive counterpart, and it stops there. The temple deity is engaged through embodiment, being-with, and symbolisation together, with all the coordination of a household built around its daily needs. Treating the two as instances of a single ontological type erases the differences that matter. Recursive mediational attribution, the attribution of particular mediations to a counterpart, is graded, and the grades are found in practice: whether a being is named, addressed, fed, put to sleep, mourned, apologised to before it is killed, or protected from violence.

The history of Christianity offers a further test. The Reformation controversies over the Eucharist turned on whether communicants eat and drink the body and blood of Christ or a sign of them, and a great deal of European history followed from the positions taken. Douglas (1970) read the dispute as one between ritualist and anti-ritualist cosmologies. In the terms used here, it was a dispute over which mediations the rite engages: whether the sacrament is a metabolic event, in which substance is taken into a body, or an act of multisymbolization alone. That the question could divide a continent shows how much can depend on the mediational profile of an attribution.

Two conclusions follow for the anthropology of animals. The first is that labels such as animism, totemism, and naturalism are too coarse to capture what is going on, since they classify societies by attributions of interiority and leave out the mediations through which attribution is in fact carried out. The second is that comparison remains possible, and has hardly begun. It has to be built from what people do with and for other beings, in feeding, sleep, naming, address, mourning, killing, and the avoidance of killing, and it requires very close ethnography of each case. The broad patterns, if there are any, will emerge from such work and cannot be presupposed by it.

XI. Mesocosmographies of Particular Animal Relationships

To speak of "the human-animal relationship" is about as informative as it would be to announce a single account of the human-human relationship. Nobody would expect one theory to cover a mother and her infant, two strangers on a train, a judge and a defendant, and two old friends; the relationships between humans and other animals are at least as various. They differ by species, by situation, by history, and above all by individual. Three cases show what finding out about them involves. In each, the question is which coordinations are in fact established, for whom, and with what at stake.

When my children were about five and seven, we acquired two hamsters, chosen over a dog or a cat because the family travelled often (Of Hamsters, Dogs, and Peccaries). The children expected companions: animals that would respond to them, play with them, and enjoy being held. The hamsters did not get on. One bullied the other and, within a few weeks, escaped and was never seen again. The survivor lived two more years, by every sign content in his cage. He disliked being held and took no interest in the children, whose interest in him faded in turn. He died, was buried in the garden, and was forgotten so thoroughly that I can no longer recall his name.

The case is instructive because the expected relationship failed. The children attributed interrecursive capacity to the hamsters, and the attribution lacked mesocosmic fidelity: it did not hold up in coordination with the animals, and repeated L2 disappointment wore it down. It does not follow that hamsters lack interrecursive capacity. The bullying suggests that the two animals were responding to each other only too well. Nor did the survivor lack a mesocosm of his own, with his wheel, his hoard, his nocturnal rounds, and his steady selfrecursive learning of the cage. What was missing was the particular interrecursive relationship that the children sought. Capacity, availability, and attribution came apart, and a low recursive relevance followed: the animal lived in the house for two years and very little came to depend on him.

A household dog differs at each of these points. Dogs respond to human responses, single out individuals, and adjust to what a particular person is about to do, so that the attribution of interrecursive capacity is sustained by coordination and deepens over the years. Layer builds upon layer: bodily familiarity in infancy, games of pursuit and retrieval in which each reads the readiness of the other, L2 disturbances when an expected response fails, and later the L3 questions of what the dog thinks and why it must die first. The earlier layers do not disappear when the later ones arrive, and they are what is torn at the end. The grief that follows the death of a dog and the shrug that follows the death of a hamster reflect neither sentimentality nor callousness; they are accurate registrations of two very different histories of coordination. The mesocosm of the dog, meanwhile, still has to be observed in its own right, since the dog is not passing through anyone's childhood.

Even "the dog" is too general. The Kolkata painter Shanu Lahiri kept a changing household of street dogs she had taken in, usually four or five at a time. One of them, Goli, would recognise me after absences of several years and make it plain, in the way dogs do, that she remembered me and approved. No other dog has ever done that for me, and the satisfaction of being accepted by Goli was considerable. A friend who has owned two shih tzus in succession describes the first as attentive and companionable and the second as opinionated, obstinate, and largely indifferent to her. Breed is an L4 classification. The relationship is with an individual, and the individual has to be discovered. What kind of being is this one? What does she find interesting, alarming, or funny? What does she want, and when? Recursive discernment of this kind is the ordinary work of every relationship between beings that can respond to one another, among humans as much as between humans and dogs. Some people one meets are immediately easy company and others never become so, whatever they share in background and interests, and no available vocabulary captures very well why. The same discernment, carried out with a dog, is how a relationship with that dog comes to be.

A shepherd and a ewe coordinate daily over years. The shepherd knows the animal by her gait, her place in the flock, her lambing history; the ewe knows the whistle, the dog, the gate, the hill. Much of this runs at low intensity. The ewe grazing is engaged with grass, slope, and her own appetite, and with the rest of the flock through the relaxed, mutual spacing that the gulls on the ridge also keep. Interrecursivity rises to its most intense between shepherd and dog when they work, between dog and sheep at the moment of the gather, which for the sheep is an antagonistic engagement with a stand-in for a predator, and between ewe and lamb when they are separated. Then, one morning, the ewe hangs back. For her, something in her own movement has gone wrong. For the shepherd, a pattern has broken, and questions of diagnosis, cost, and contagion open. For the veterinarian, she is a case among cases. One event runs through several mesocosms with different stakes in each, and the aim in all of them is that things should go on as before. "Pastoralism as a system of production" is an L4 summary of thousands of such coordinations and is silent on how any of them is formed.

Among the Yukaghir of the Upper Kolyma, as Willerslev (2007) describes them, a hunter approaching an elk takes on the bearing of the animal. He sways in the rolling gait of an elk, on skis faced with the skin of elk legs so that they sound like the animal moving through snow, and the elk, drawn by what it takes for one of its own, comes towards him. That the scene is interrecursive should be shown and not announced, and it can be shown against the criteria. The hunter engages one particular animal and adjusts moment by moment to what it is about to do. The elk adjusts to him, though to him as an elk: what it models is a conspecific, so that its side of the coordination rests on an attribution the hunter has manufactured. The interrecursivity is real and asymmetrical, and deception is its instrument, much as it is in the dark wood of the picture book. It also has limits, since a shift of wind returns the hunter to the mercy of nonrecursive ground. Attunement of this intimacy serves a kill, and the hunter's own continuation depends on the continuation of the elk population he hunts.

The children, the friend with her shih tzus, the shepherd, and the hunter take part in radically different mesocosms. All of them can be investigated in the same way, and none yields its stakes to a prior category.

XII. From Prehistoric Images to Modern Classifications

The prehistoric material can now be read without the expectation that it conceals a code. A painted rhinoceros shows that rhinoceroses became subjects of durable representation. It does not show that the painter sorted rhinoceroses into a binary taxonomy with other species, and a fox carved into a pillar is no evidence of an abstract relation between foxes and human clans. Readings of that kind require additional evidence which the images do not supply. In the terms of mesocosmography, the images give depositional access. They are deposits, and archives of deposits overrepresent what was articulated and made durable, so that absence of depiction can never be read as absence of coordination.

The question accordingly changes. One no longer asks which symbolic oppositions explain the images, but which stakes, encounters, and poietic practices made these particular images possible and worth sustaining. The selections point away from the larder: reindeer at Lascaux and gazelle at Göbekli Tepe fed people and were scarcely depicted, while lions, bears, leopards, snakes, and scorpions were. What sustains a population tends to be coordinated at L1 and receives little symbolic investment, and what demands attention, including what can kill, receives a great deal (From the Cave to the Algorithm). A cave lion was a counterpart whose attention to oneself mattered absolutely, an antagonistic interrecursive counterpart of the most draining kind. For people who slept in the open or in rock shelters, the nights belonged to such animals, and no one could simply walk out into the dark. That such encounters fed the images is a probable inference and no more, yet it predicts what the dietary and the taxonomic readings do not, namely that the repertoire should track histories of attention more closely than histories of consumption.

What the images show directly is the elaboration of the poietic mediations. A painted animal required a prepared surface, ground pigment, light carried underground, trained hands, and a figure that others could recognise. Once made, it persisted independently of the animal and of its maker, and it became a component of later human mesocosms, including our own. Images of this kind are among the earliest surviving means of storing and transmitting a symbolic distinction beyond the occasion of its making, which is the material precondition of L4. At Göbekli Tepe the elaboration is built into architecture, with snakes predominating in one enclosure, foxes in another, and boars in a third (Peters and Schmidt 2004), and with interpretations ranging from feasting and ritual to doubts whether these were temples at all (Banning 2011; Dietrich et al. 2012). Whether the differences answered to distinct groups, practices, or locally established stakes is a question the framework can pose with some precision. The decisive novelty is that animals could now be consequential for people who never met them.

One further possibility deserves to be stated, because it follows from the argument about recursive fluidity. The builders of Göbekli Tepe were hunters and gatherers, without towns, fields, or herds, and why they assembled in numbers to raise pillars of several tonnes remains unexplained. The site has yielded evidence of the large-scale processing of cereals, which its excavators connect with feasting and possibly with fermented drinks (Dietrich et al. 2012), though the interpretation is debated. To feast and drink with pleasure, people need a place where vigilance can lapse, and in a landscape where predators had to be reckoned with every night such places had to be made. The enclosures, walled and roofed or not, may have served among other things as places where people could gather, eat, drink, and let their attention go: a collective return to L1, secured by poietic means against the antagonistic interrecursivity that predators imposed every night, and opening onto the relaxed interrecursivity of eating and drinking together. The carved animals would then show the beings whose attention had to be reckoned with outside, on the walls of a place made for the relief of not reckoning with it. The hypothesis is speculative, and many archaeologists, including some who have worked on the site, prefer ritual readings. Its merit is that it connects the selection of dangerous animals with the architecture of enclosure and with a need that every animal shares, which the ritual readings leave unconnected.

That novelty, the consequence of animals for people who never meet them, has a modern form. Most of what governs animals today is stabilised at L4 and operates without reflection by those who carry it out: welfare codes, lists of protected species, veterinary protocols, slaughter standards, and the law of property. A stocking density written into a regulation fixes how many bodies share a floor, and with it what the embodiment, being-with, and dwelling of each bird can be, for billions of animals whose keepers follow the figure as routine. The distinction between animals as living beings and animals as objects of L4 governance is the ethical centre of the matter. The stakes of a hen do not begin when someone recognises them, and her coordination is disturbed by a cage whether or not she is granted the capacity to suffer.

Two concepts sharpen the analysis. Symbolic overreach occurs when a classification is taken to exhaust what it classifies. One animal may be property in law, a model organism in science, livestock in the accounts, and sacred in ritual; each description licenses a treatment, and none describes what is occurring in the mesocosm of that animal. L2 buffering is the interposition of material, spatial, and symbolic distance between an attribution regime and the feedback that would correct it (Of Hamsters, Dogs, and Peccaries). Industrial farming is its great instance: the shed, the supply chain, and the label ensure that almost no one who lives by these animals encounters the disturbances that their classification produces. Animal ethics, so understood, investigates how symbolic-material arrangements reorganise the vital stakes of living animals and when those arrangements require reopening. The reopening has its own temporality. A regulation stands for decades and a broiler lives about six weeks, so that correction arrives, when it arrives, for animals other than those whose coordination prompted it.

XIII. Animal Anthropology Beyond Structuralism

If the classificatory lineage is the problem, the later anthropology of animals can be sorted by how far each author leaves it and what each recovers.

Tapper (1988) recovers multiplicity. That relationships with animals vary historically, that one community sustains several, and that animal stories are instruments of moral teaching are observations LVT retains. What it sets aside is the typology modelled on human relations of production, which lays an L4 economic vocabulary over coordinations that have not yet been described.

Descola (2013) recovers the contingency of naturalism. His evidence is richer than his vocabulary: the Achuar do not merely credit peccaries with an interior, they credit them with leaders, quarrels, techniques, and rules of marriage, which is to say with the full range of mediations. LVT redescribes his four ontologies as regimes of recursive mediational attribution, tested for mesocosmic fidelity over the long ecology of the hunt (Of Hamsters, Dogs, and Peccaries). The scheme pluralises the boxes and leaves box-making intact. It cannot distinguish the named scooter from the fed and sleeping goddess, and it cannot follow the contested, historical revision of attribution within a single mesocosm, of which the recent career of the urban dog is an example.

Haraway (2003, 2008) recovers becoming-with. Her accounts of training show responsiveness built hour by hour in two bodies, and LVT would call this the establishment of interrecursive coordination. Her encounter value, proposed alongside use value and exchange value, is less successful, since living value is generated well beyond recognised encounters, in dependence on animals one never meets and in routines nobody would call an encounter. The preference for encounter also mistakes the punctuation for the text: most of being with an animal is relaxed, low-energy interrecursivity without anything one would call an encounter.

Ingold is the principal theoretical ally. He rejected the separation of organism and environment, treated skill as coordination and not as a stored capacity, and insisted that the world is encountered before it is represented (Ingold 2000). His contrast between the trust of hunters and the domination of pastoralists (Ingold 1994) displaced the view of domestication as a technical conversion of resources, and he has objected to later writing on humans and animals for its distance from practical knowledge (Ingold 2013). LVT builds on all of this and differs at three points (Dwelling in the Mesocosm). Dwelling in his sense covers what LVT separates into three zoetic mediations and one poietic one. Recursivity is underweighted, so that trust and domination name broad characters of relationship without distinguishing levels, counterparts, or the stakes of each participant. Symbolisation appears as a layer superimposed on practical engagement, when it is in fact a force that reorganises it, as the stocking density shows.

Willerslev (2007) is the principal ethnographic ally, for the closeness with which he follows imitation, danger, and deception. He organises his findings through personhood and perspectivism (Viveiros de Castro 1998), so that the question becomes how persons of different kinds exchange points of view. The same observations support a plainer description: an asymmetrical interrecursive coordination, with opposed stakes, sustained in posture and rhythm. The danger that a hunter may not return from the elk then describes how far a human mesocosm can be reorganised by prolonged coordination with another animal, and it needs no doctrine of souls to be taken seriously.

The ontological turn deserves a separate word, because LVT stands at the opposite end of the spectrum from it. Perspectivism and its relatives insist that the worlds of different peoples are different worlds, that the differences between them are ontological and no mere matter of culture or belief. LVT proposes a process ontology common to all living beings, and on that basis it finds the claimed differences much smaller than they are said to be. They are real, and they are located almost entirely in a narrow band of multisymbolization and recursive attribution, which the ontological literature magnifies until it fills the picture. Everything else is shared. The Amerindians among whom perspectivism was formulated have the same bodies as their ethnographers, give or take minor variation, with the same hungers, fatigues, and vulnerabilities. Their being-with varies within a range that is small compared with what all human groups have in common, and it differs from the being-with of whales or lions far more than from that of the ethnographer's own society. Their dwelling is the found ground of a particular region, and an account of its weather, distances, rivers, and visibility at night would say much of what matters about life there. Their made things and their signs are different in detail and continuous in kind with everyone else's, which is why the ethnographer could learn their language at all. Even symbolisation, the mediation on which the ontological argument rests, is overwhelmingly shared.

The claim can be stated as a challenge. A human world in which any of the five mediations was absent, a world without embodiment, being-with, dwelling, made things, or signs, would refute it, and none has been described. The five mediations are therefore offered as a description of what any human world must contain, and of how much of it is shared with other animals, rather than as one more perspective among the theories on offer. To say that something is real remains, in much of anthropology, a provocation. The anthropology of animals is where the provocation is most needed, because an over-investment in symbolic difference has kept the discipline from seeing how much humans share with each other and with every other animal.

The sorting has a pattern. The authors who remain closest to classification, whatever their intentions, tell us most about humans. Those who follow practice tell us most about animals, and they lack only the differentiated vocabulary that would let their findings be compared.

XIV. Conclusion: How Living Beings Become Consequential

At Sulawesi, Chauvet, Lascaux, and Göbekli Tepe, particular animals became the subjects of extraordinary material and symbolic investment. Which stakes made them significant is not fully known. What can be said is that the images are deposits of human poietic coordination, connected in ways still to be worked out to long histories of zoetic engagement with lions, bears, pigs, foxes, and snakes, and that nothing obliges us to read them as a taxonomy. Even the oldest of them, a human hand reworked until it resembled a claw, was made by an animal among animals.

The anthropology of animals took a wrong turning when it made human classification its object. From Durkheim and Mauss to Lévi-Strauss, and through perspectivism into the present, the one form of recursive activity that nonhuman animals do not share was treated as the key to them, and the result was the most anthropocentric account of animals the discipline could have produced. The living relationships are far more remarkable than the tables drawn up to represent them: a hunter drawing an elk towards him, a bird at its nest, a herder catching the first sign of illness, a street dog recognising a visitor after years, a child finding out what play with another animal can be.

In place of the classificatory object, this article has set out an account that makes the similarity between humans and other animals as large as the evidence allows. All animals share the three zoetic mediations, coordinate at L1 and L2, and engage nonrecursive, selfrecursive, and interrecursive counterparts, and humans spend most of their lives in exactly this shared condition. Interrecursivity is a mode that most animals have, from ants to gulls to dogs, and in every species, the human included, it runs mostly at low intensity, in relaxed co-presence and grooming, while its maximal and above all its antagonistic forms are metabolically draining and avoided wherever possible. Rudimentary multimateriality, multisymbolization, and articulation at L3 occur among animals as they do among humans. Metabolism, and the difference between beings that live from light and beings that must eat other lives, shapes relationships with animals before any classification does, down to the question of which beings return as ghosts. Recursive fluidity, the return from higher recursion and from intense interrecursivity to absorbed coordination, matters more than any capacity for higher recursion, for a mouse in a picture book as for the builders of Göbekli Tepe. Recursive attribution is graded by the mediations actually extended to a counterpart, and the grades, from a name to food and sleep, are found in practice. The one substantial difference is that humans have elaborated multisymbolization, together with multimaterial means of inscribing, storing, and transmitting it, to the point where L4 and L5 became possible, and no other animal exhibits either. That difference allows animals to become enduring elements of human ground as images, monuments, stories, institutions, and legal categories, and it creates the specifically human risk of mistaking what has been made prominent for what is vitally consequential. A society may possess an elaborate mythology of the jaguar while its most consequential coordination runs through chickens, lice, and earthworms.

The contribution claimed here is the extension of mesocosmography and of the analysis of vital stakes to the differentiated worlds of humans and other animals. The empirical task is to find out how particular animals become relevant to one another, what coordinations must work for each, what is registered when they fail, and how arrangements made by humans reorganise the lives of animals who never chose them. The programme reaches from predation and play to conservation and the industrial shed, and from the cave wall to the temple and the dog walk, and it has scarcely begun.

The anthropology of human-animal relationships has too often asked what animals are to humans: resources, companions, persons, symbols, kin, or others. Living Value Theory begins with animals as living beings that coordinate their own mesocosms. It asks how these lives become consequential to one another, what is sustained or endangered through their coordination, and how particular relationships are established, transformed, or ended. Humans share the zoetic ground of animal life and have built upon it worlds of made things and made signs. The task is to understand the endlessly differentiated ways in which living beings come to matter to one another, and the meaning of animals in general was never the right thing to look for.

References

Banning, E. B. 2011. So Fair a House: Göbekli Tepe and the Identification of Temples in the Pre-Pottery Neolithic of the Near East. Current Anthropology 52 (5): 619–60.

Brightman, Robert. 1993. Grateful Prey: Rock Cree Human-Animal Relationships. Berkeley: University of California Press.

Clottes, Jean, ed. 2003. Chauvet Cave: The Art of Earliest Times. Salt Lake City: University of Utah Press.

Darwin, Charles. 1881. The Formation of Vegetable Mould, through the Action of Worms, with Observations on Their Habits. London: John Murray.

Descola, Philippe. 2013. Beyond Nature and Culture. Translated by Janet Lloyd. Chicago: University of Chicago Press.

Dietrich, Oliver, Manfred Heun, Jens Notroff, Klaus Schmidt, and Martin Zarnkow. 2012. The Role of Cult and Feasting in the Emergence of Neolithic Communities: New Evidence from Göbekli Tepe, South-Eastern Turkey. Antiquity 86 (333): 674–95.

Donaldson, Julia, and Axel Scheffler. 1999. The Gruffalo. London: Macmillan Children's Books.

Douglas, Mary. 1970. Natural Symbols: Explorations in Cosmology. London: Barrie and Rockliff.

Dunbar, R. I. M. 1992. Neocortex Size as a Constraint on Group Size in Primates. Journal of Human Evolution 22 (6): 469–93.

Dunbar, Robin. 1996. Grooming, Gossip, and the Evolution of Language. London: Faber and Faber.

Dunbar, R. I. M. 2016. Do Online Social Media Cut through the Constraints That Limit the Size of Offline Social Networks? Royal Society Open Science 3 (1): 150292.

Durkheim, Émile. 1995 [1912]. The Elementary Forms of Religious Life. Translated by Karen E. Fields. New York: Free Press.

Durkheim, Émile, and Marcel Mauss. 1963 [1903]. Primitive Classification. Translated by Rodney Needham. Chicago: University of Chicago Press.

Ecks, Stefan. 2022. Living Worth: Value and Values in Global Pharmaceutical Markets. Durham, NC: Duke University Press.

Franks, Nigel R., and Tom Richardson. 2006. Teaching in Tandem-Running Ants. Nature 439: 153.

Frisch, Karl von. 1967. The Dance Language and Orientation of Bees. Translated by Leigh E. Chadwick. Cambridge, MA: Harvard University Press.

Fuller, C. J. 2004. The Camphor Flame: Popular Hinduism and Society in India. Revised and expanded edition. Princeton: Princeton University Press.

Goulson, Dave, Elizabeth Nicholls, Cristina Botías, and Ellen L. Rotheray. 2015. Bee Declines Driven by Combined Stress from Parasites, Pesticides, and Lack of Flowers. Science 347 (6229): 1255957.

Haraway, Donna J. 2003. The Companion Species Manifesto: Dogs, People, and Significant Otherness. Chicago: Prickly Paradigm Press.

Haraway, Donna J. 2008. When Species Meet. Minneapolis: University of Minnesota Press.

Hunt, Gavin R. 1996. Manufacture and Use of Hook-Tools by New Caledonian Crows. Nature 379: 249–51.

Ingold, Tim. 1994. From Trust to Domination: An Alternative History of Human-Animal Relations. In Animals and Human Society: Changing Perspectives, edited by Aubrey Manning and James Serpell, 1–22. London: Routledge.

Ingold, Tim. 2000. The Perception of the Environment: Essays on Livelihood, Dwelling and Skill. London: Routledge.

Ingold, Tim. 2013. Anthropology beyond Humanity. Suomen Antropologi 38 (3): 5–23.

Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York: Farrar, Straus and Giroux.

Laidlaw, James. 1995. Riches and Renunciation: Religion, Economy, and Society among the Jains. Oxford: Clarendon Press.

Lévi-Strauss, Claude. 1963. Totemism. Translated by Rodney Needham. Boston: Beacon Press.

Lévi-Strauss, Claude. 1966. The Savage Mind. Chicago: University of Chicago Press.

Maynard Smith, John, and G. R. Price. 1973. The Logic of Animal Conflict. Nature 246: 15–18.

Oktaviana, Adhi Agus, et al. 2024. Narrative Cave Art in Indonesia by 51,200 Years Ago. Nature 631: 814–18.

Oktaviana, Adhi Agus, et al. 2026. Rock Art from at Least 67,800 Years Ago in Sulawesi. Nature, published online 21 January 2026.

Peters, Joris, and Klaus Schmidt. 2004. Animals in the Symbolic World of Pre-Pottery Neolithic Göbekli Tepe, South-Eastern Turkey: A Preliminary Assessment. Anthropozoologica 39 (1): 179–218.

Seyfarth, Robert M., Dorothy L. Cheney, and Peter Marler. 1980. Monkey Responses to Three Different Alarm Calls: Evidence of Predator Classification and Semantic Communication. Science 210 (4471): 801–3.

Tapper, Richard. 1988. Animality, Humanity, Morality, Society. In What Is an Animal?, edited by Tim Ingold, 47–62. London: Unwin Hyman.

Tinbergen, Niko. 1953. The Herring Gull's World: A Study of the Social Behaviour of Birds. London: Collins.

Viveiros de Castro, Eduardo. 1998. Cosmological Deixis and Amerindian Perspectivism. Journal of the Royal Anthropological Institute 4 (3): 469–88.

Weber, Max. 1946 [1919]. Science as a Vocation. In From Max Weber: Essays in Sociology, edited and translated by H. H. Gerth and C. Wright Mills, 129–56. New York: Oxford University Press.

Willerslev, Rane. 2007. Soul Hunters: Hunting, Animism, and Personhood among the Siberian Yukaghirs. Berkeley: University of California Press.