Abstract
Actor-Network Theory attributes agency to nonhumans because they make an agentive difference. This article argues the opposite: scallops, speed bumps, maps, balls and black boxes carry coordinative weight precisely because they do not respond like actants. Drawing on five mediations and three coordination types—nonrecursive, selfrecursive and interrecursive—it re-reads Callon’s scallops, Serres’s quasi-object, immutable mobiles, black-boxing, Mol’s body multiple and Law’s topologies. Things stabilise collective life as nonrecursive deposits, not agents. Symmetry therefore erases the distinction that explains why durable objects are manufactured at all.
I. The Question ANT Asked and the Question It Should Have Asked
Actor-Network Theory asks, of any entity that makes a difference to an outcome, whether it has agency. Posed this way, the question answers itself in exactly the direction Bruno Latour, Michel Callon, and John Law wanted it to: a scallop larva that fails to attach to a collector line has made an enormous difference to a research programme, a speed bump has made a measurable difference to the velocity of every car that crosses it, and if making a difference is what agency means, then scallops and speed bumps have it, alongside the scientists and drivers who thought they were the only actors in the room. Bruno Latour's own formula for this, that when one entity acts, others proceed to action, is offered as a definition general enough to cover a person, a gun, a ball, and a ledger without distinction. The principle of generalised symmetry that follows, treating human and non-human actants with the same conceptual vocabulary, has been received for three decades as ANT's most radical and most valuable contribution: a refusal to let the old bifurcation between active subjects and passive objects go on organising the human sciences.
This article argues that the question was wrong from the start, not because things do not matter, but because ANT's own best evidence, examined without the vocabulary that was brought to it in advance, shows the opposite of what ANT concludes from it. Scallops, speed bumps, ledgers, maps, and rugby balls carry the enormous coordinative weight ANT correctly documents them carrying, not because they have been quietly granted a portion of agency after all, but because they have none, in the specific and technical sense that matters: they do not respond to how they are engaged with. A theory of coordination that has a name for this distinction, and asks, of any counterpart, whether it answers back, arrives at a completely different account of what a scallop, a speed bump, a ledger, and a ball are actually doing, one that explains ANT's own richest material more precisely than ANT explains it itself, and explains, in addition, several things ANT has never been able to explain: why symmetry looks radical and functions conservatively, why the theory cannot locate power, and why it struggles with anything that outlasts the network that first assembled it.
The argument proceeds by taking ANT's own canonical cases in turn, in something close to the order they are usually taught: the scallops, the quasi-object, the immutable mobile, the black box, the sociology of translation, Latour's own taxonomy of non-human agency, Mol's multiple body, and Law's topologies, before turning to what ANT's critics and rivals, Bloor, Harding, Harman, Bennett, and Barad, have each seen and each still missed. Every case is examined using the same apparatus: five mediations, three types of coordination, and the levels through which a coordination that has strained is registered, named, stabilised, and returned to ease. None of this apparatus is hostile to what ANT actually found in the field. It is hostile only to the single word, agency, that ANT insisted on using to describe it.
II. Five Mediations, Three Types: The Apparatus ANT Never Built
Human coordination proceeds through five irreducible mediations, embodiment, being-with, dwelling, multimateriality, and multisymbolization, and every concrete situation, a laboratory, a fishing bay, a rugby pitch, involves all five at once, in different proportions of salience rather than in isolation from one another. Embodiment is the register of a living body regulating itself. Being-with is the register of one living being coordinating with another who is coordinating back. Dwelling is the register of place, season, and terrain. Multimateriality is the register of substances shaped and arranged. Multisymbolization is the register of marks, signs, and notations. ANT's own vocabulary moves fluently across all five without ever separating them, which is the first and most consequential thing it gets wrong, because the five mediations do not behave alike, and folding them into a single word, actant, discards the very distinctions a theory of things needs most.
The second axis matters even more for what follows. Coordination between a living being and any counterpart is of exactly one of three kinds, and the kind is settled by a single question: does the counterpart respond to how it is engaged with. Where it does not, coordination is nonrecursive: a stone, a road, a scallop larva's fixed biological programme, a ball obeying the physics of leather and air pressure, none of these alter their own course because of how a living being approaches them. Where a counterpart does respond, and the counterpart is internal to the coordinating being's own body, coordination is selfrecursive: a rower adjusting stroke to fatigue, a driver's foot easing off the accelerator as a jolt is anticipated. Where a counterpart responds and is external, another living being coordinating back in a loop neither party alone controls, coordination is interrecursive: two fishermen negotiating a price, a laboratory team arguing over an inscription's meaning, a research director enrolling a funding body.
This distinction is categorical rather than a matter of degree. A counterpart either answers back or it does not, and nothing about how forcefully, how reliably, or how consequentially it fails to answer back moves it into a different category. A scallop larva that determines the fate of an entire research programme by simply failing to attach to a collector line has made an enormous difference to an outcome without the larva doing anything that qualifies as responding to how it was engaged with. The larva's failure to attach is exactly as nonrecursive as its success would have been. What varies, and varies enormously, is not the larva's type but the cost and difficulty of coordinating with it, a variable this theory calls stability, entirely separate from the categorical question of whether the counterpart is recursive at all.
A further apparatus completes the picture. Coordination that has been running smoothly, selfrecursive or interrecursive, sometimes strains: something registers as wrong, which this theory calls L2. The strain is named specifically enough to be acted on at L3. Acting on it produces a stabilised, portable counterpart, a deposit, at L4. And once that deposit exists and is taken up, coordination with it settles into ease at L1, not because anyone has forgotten the difficulty that produced it, but because the difficulty has been converted into a form nobody has to solve again. A deposit of this kind, a map, a ledger, a law, a manufactured object, is a nonrecursive counterpart carrying the settled product of coordination that was once selfrecursive or interrecursive, available now for nonrecursive use by anyone who did not take part in producing it. This is the apparatus ANT's own best material keeps describing without the words for it, and the rest of this article shows exactly where.
III. The Scallops of St Brieuc Bay: Ground Mistaken for a Partner
Callon's 1986 study of three marine biologists attempting to establish scallop cultivation in St Brieuc Bay is ANT's founding parable, and it repays reading with the three-type distinction directly in hand, because every actor in the story falls cleanly on one side of it or the other, and the paper's own drama depends on the reader not noticing this.
The three biologists are engaged in interrecursive coordination with the fishing community, whose cooperation must be negotiated, and with the wider scientific community, whose credence must be secured through argument, publication, and evidence that will withstand challenge. This negotiation is interrecursive throughout, in the fullest sense the term carries: fishermen respond to the researchers' proposals, and the researchers adjust their proposals in response, in a loop neither side controls alone. The scallops are a different matter entirely. Callon describes, in careful detail, whether scallop larvae anchor themselves to the collector lines the researchers have laid down in imitation of a Japanese cultivation technique, and the larvae's behaviour, whatever it turns out to be, is not a response to the researchers' interests, hopes, or rhetorical skill. It is the outcome of the larvae's own nonrecursive biology, indifferent to the fact that a research programme depends on it. Callon calls this indifference resistance, and calls the larvae's eventual, insufficient rate of attachment a betrayal, language that imports interrecursive vocabulary, resistance and betrayal both presuppose a party capable of responding differently to different treatment, onto a process that is nonrecursive from beginning to end.
What Callon has actually documented, read this way, is the biologists using the scallops' fixed, non-negotiable biology as an obligatory passage point precisely because it is fixed and non-negotiable, rather than the scallops joining a negotiation as a fourth party alongside the fishermen, the scientific community, and the biologists. This is what makes the larvae usable as evidence at all. If the scallops had been capable of responding to the researchers' interests, adjusting their attachment behaviour according to how badly a given outcome was wanted, their eventual failure to attach would have proven nothing, since evidence that answers back to the desires of whoever is collecting it is not evidence, it is negotiation. The scientific weight the scallops' behaviour carries in Callon's own account, weight sufficient to end a research programme regardless of how many fishermen and funding bodies had been enrolled, depends entirely on the larvae's non-recursivity. Callon is describing the scallops functioning as ground, not as a partner, and the paper's continued influence rests on a vocabulary that quietly converts the one into the other.
The same distinction resolves Callon's four moments of translation without remainder. Problematisation, the researchers positioning themselves as an obligatory passage point, and enrolment, the negotiations by which fishermen and colleagues accept the roles assigned to them, are interrecursive from end to end, since every move by any party reshapes the space of possible next moves for everyone else. The scallops never participate in this loop. Their contribution is a single, nonrecursive fact, delivered once, that the interrecursive negotiation among the human parties either survives or does not. Calling this fact resistance, and calling the biologists' response to it a further round of translation, obscures the very asymmetry that makes the case scientifically interesting in the first place: an interrecursive project standing or falling on a nonrecursive counterpart it cannot negotiate with.
IV. The Quasi-Object and the Ball That Does Not Dictate
Michel Serres's account of the rugby ball, adopted wholesale by Latour as the philosophical ancestor of the actant, is the case in which ANT comes closest to the correct analysis and swerves away from it at the decisive moment. Serres's central claim, that a ball sitting on the grass is a mere physical thing while a ball in play becomes something that organises the players around itself, is exactly right, and exactly what nonrecursive coordination theory would predict. Thirty running, shouting, mutually responsive bodies, an interrecursive coordination of extraordinary complexity, has no fixed point to organise around unless something in the middle of it holds still according to rules nobody in the game is negotiating moment to moment. The ball obeys the physics of leather, air, and gravity with total indifference to which team wants it and which player is exhausted, and it is this indifference, not any responsiveness, that lets the interrecursive coordination of the match cohere around it rather than dissolving into thirty separate improvisations.
Serres himself supplies the sentence that gives this away, in noting that social bonds sustained by contract alone would be as insubstantial as cloud, and that an object of the kind specific to human collective life is what actually holds relationships together. Stated directly in this theory's own terms: interrecursive coordination alone, with no nonrecursive counterpart to organise around, has no stable point of reference, since each party is only ever responding to the other's last response, and the whole exchange can drift without limit. A nonrecursive counterpart supplies exactly what interrecursive coordination cannot supply for itself, a point that does not move because of how anyone responds to it, and it is this stability, not any activity on the object's part, that allows a coordination among many mutually responsive parties to hold a shared shape at all.
The swerve comes in the description of what the ball then does. Serres writes that playing is nothing other than making oneself an attribute of the ball, and that the rules are written for the ball and bend the players to it, language that puts the ball in the position of an authority the players submit to, a formulation Latour later generalises into the claim that the ball dictates the players' movement. But a ball that dictated in any sense connected to the interrecursive vocabulary that word invokes would need to respond differently to different players, favouring one team, punishing a mistimed pass, adjusting its bounce to the state of the match, and a ball that did any of this would not organise a game. It would end one, since a nonrecursive counterpart that started answering back would no longer supply the fixed point interrecursive coordination depends on. The ball's entire capacity to structure the match is purchased by its refusal to dictate anything, its perfect indifference to who is winning, who is tired, and who wants it most. What the players submit to is not the ball's agency. It is the ball's non-recursivity, mistaken, at the exact moment Serres and Latour reach for a verb, for the opposite of what it is.
V. Immutable Mobiles: A Name That Already Gives Away the Mechanism
Latour's concept of the immutable mobile, a map, a ledger, a standardised measurement, capable of travelling across space and time while holding its internal structure constant, is the clearest instance in the whole ANT corpus of a phenomenon correctly identified and then wrongly named, because the concept's own defining word, immutable, already states the mechanism this article has been describing throughout, and states it in terms that have nothing to do with agency at all.
An immutable mobile is powerful, on Latour's own account, precisely because a distant centre of calculation can accumulate inscriptions from a periphery, combine them, and issue instructions back, at a scale and over distances no direct, face-to-face, interrecursive negotiation could sustain. A colonial map allows a metropolitan office to govern a territory its officials have never set foot in. This is real, well documented, and exactly the kind of coordinative power a theory of things needs to explain. What explains it is that the map is a recursive deposit, a nonrecursive counterpart carrying the settled product of coordination, surveying, measurement, cartographic convention, that was once selfrecursive and interrecursive and has since been converted into a form that holds still regardless of who consults it, when, or with what interest, rather than any portion of agency the map has been granted. The map's usefulness to a distant office is a direct function of its refusal to answer back to that office's wishes, exactly as the scallops' usefulness as evidence was a direct function of their refusal to answer back to the biologists' hopes. Call this immutability agency and the concept stops explaining anything, because agency, in every sense the word carries outside ANT's own vocabulary, names responsiveness, and immutability names its absence.
The same point applies with equal force to black-boxing, which Latour treats as closely related to the immutable mobile and which this article's next section takes up directly, and to the broader category of inscription devices, the spectrophotometer, the centrifuge, the apparatus that converts biological substance into a stable, two-dimensional trace. What these devices do, described without the actant vocabulary, is convert a selfrecursive or interrecursive process, an organism's ongoing biochemical activity, a laboratory dispute over what a result means, into a nonrecursive deposit, a reading, a chart, a number, that can then be coordinated with by anyone, non-recursively, without needing to repeat the process that produced it. The device, far from an active participant alongside the scientists who built and calibrated it, is the site where their selfrecursive and interrecursive labour has already been spent, deposited, and closed off, freeing the scientists' own coordination to move on to the next dispute rather than reproducing the last one from nothing.
VI. Black-Boxing: ANT's Own Word for L1
Black-boxing is the moment a technology's internal complexity becomes invisible because it simply works, and Latour's own description of this moment, a computer, a car engine, or a legal institution stabilised to the point of acting as a single, unified thing, is a precise, unattributed description of L1 in this theory's own vocabulary: coordination that has settled into ease, its history of strain, negotiation, and repair no longer visible or required, available now without anyone having to think about it. Latour is not wrong that this is what black-boxing is. He is wrong about what follows from it.
The entire point of a successful black box, on ANT's own account, is that the network inside it no longer has to be renegotiated on each occasion of use. A driver turning a car key is not entering into interrecursive negotiation with the engine, the fuel injection system, or the alternator. She is coordinating non-recursively with a deposit, one whose internal history of design disputes, engineering trade-offs, and regulatory negotiation has been converted into a form that starts the car the same way every time, for anyone, regardless of who they are or what they want. This is precisely why Latour's own metaphor works: a box that had to be persuaded, negotiated with, or petitioned on each occasion would not be black, it would be visibly, continuously open, its complexity on permanent display rather than hidden behind a single reliable function.
Calling the black box an actant with agency inverts the relationship between the phenomenon and its explanation. The black box has agency, in ANT's diminished sense of making a difference, in exact proportion to how successfully it has ceased to be negotiable, which is to say in exact proportion to how completely its non-recursivity has been achieved. A partially black-boxed technology, one still prone to controversy, breakdown, or dispute about its meaning, is closer to the interrecursive negotiation ANT wants to call agency and further from the smooth, unremarkable function black-boxing is named for. The theory has, without noticing, built its central metaphor for technological power out of the same mechanism this article has been tracing throughout, and then reached for exactly the wrong word to describe what makes that mechanism work.
What happens when a black box breaks confirms the diagnosis rather than complicating it. Latour and the wider controversy-studies tradition treat a technology's failure, a bridge collapsing, a diagnosis proving wrong, a computer crashing, as the moment the network inside the box becomes visible again, its heterogeneous elements exposed to renewed dispute. This theory has a name for the same moment that requires no appeal to the box's own agency: strain registering at L2, a coordination that had settled into nonrecursive ease suddenly demanding, once again, the selfrecursive and interrecursive labour of diagnosis and repair. The box has not woken up and started acting. It has simply stopped succeeding at the one thing that made it a box, holding still under engagement, and the humans around it are once more doing what humans do when a deposit fails: naming the fault, negotiating a fix, and depositing the result back into a form that will, if the repair succeeds, disappear from attention exactly as reliably as the box did the first time.
VII. Translation as Remediation, Misdescribed
Callon's four moments, problematisation, interessement, enrolment, and mobilisation, describe, with real precision, how an interrecursive coordination among researchers, fishermen, and a scientific community moves from felt difficulty to stabilised, portable outcome, and the sequence maps onto this theory's own levels of remediation closely enough that making the mapping explicit shows exactly where ANT's vocabulary and this theory's vocabulary are describing the same events and where they diverge.
Problematisation, a researcher naming a problem and positioning a project as the obligatory route through which it must be addressed, is L3 in this theory's terms: a felt difficulty, registered but not yet actionable, given a name specific enough to organise a response around. Interessement, the effort to lock other parties into the roles a problematisation has defined, is the early labour of L4, building a stabilised, shareable account of the situation robust enough that other parties can be enrolled into it rather than each having to discover the problem afresh. Enrolment, the negotiations and trials of strength through which parties accept these roles, is interrecursive coordination in its most visible and effortful form, the costly, continuously renegotiated work this theory has elsewhere identified as exactly what nonrecursive deposits exist to spare a population from repeating indefinitely. Mobilisation, the point at which a spokesperson represents a now-stabilised network, including its non-human members, as a unified force, is the arrival at L4 proper, a portable, durable account that can travel to funding bodies, journals, and future collaborators without requiring each of them to relive the original negotiation.
What this mapping makes visible is that the entire sequence, from problematisation to mobilisation, describes only the selfrecursive and interrecursive labour of the human parties involved. The scallops, present throughout as the ostensible fourth party to the negotiation, never move through any of these levels. They do not register strain, name a problem, accept a role, or become a spokesperson for anything. They remain exactly what they were before the biologists arrived, a population of larvae behaving according to their own nonrecursive biology, and what changes across the four moments is entirely the human parties' account of that biology, an account that itself moves, level by level, from a raw disturbance in a research director's plans to a portable, published, defensible finding. Translation, examined this way, is this theory's own remediation architecture, correctly observed in the field and then narrated as though every participant in it were undergoing the same kind of movement, when only some of them, the ones capable of registering strain and naming it, ever were.
VIII. The Four Meanings of Non-Human Agency, Corrected
Edwin Sayes's influential synthesis of Latour's scattered claims into four distinct roles, condition for society, mediator, moral and political member, and gathering, offers an unusually clean target, because each role can be tested directly against the apparatus this article has built, and each one turns out to name something real that the word agency actively obscures.
A wall, a legal document, and a paved road are offered as conditions for society, objects providing the durability human interaction alone cannot sustain. This is exactly the function of a recursive deposit, and this theory names the mechanism the ANT taxonomy leaves as a bare description: these objects provide durability because they are nonrecursive, categorically and without exception, and it is this categorical fact, not a diminished or specialised form of agency, that lets selfrecursive and interrecursive coordination be spared the cost of continuous renegotiation. A wall that had agency in any sense connected to responsiveness would be a poor condition for society, since a boundary that could be argued out of its position on a given day would settle nothing.
A speed bump offered as a mediator, translating an abstract speed limit into a physical jolt, is the role where ANT's vocabulary comes closest to naming something and furthest from naming it correctly. The bump is, rather, a fixed physical obstacle that produces the same jolt for every vehicle that crosses it at a given speed, regardless of the driver's intentions, translating nothing in any sense connected to interpretation, distortion, or choice, all of which presuppose a party capable of responding differently on different occasions. The actual translation, from an abstract legal limit into a concrete deterrent, happened once, in the selfrecursive and interrecursive labour of the traffic engineers who calculated its height and the officials who approved its installation. The bump is where that finished translation was deposited, not where translation continues to occur.
A seatbelt offered as a moral and political member, inscribing a norm of safety into a car's design, exhibits the same displacement. The moral weight belongs entirely to the interrecursive negotiation among engineers, regulators, and safety advocates that produced the design, negotiation this theory would place, level by level, from felt public concern through named legislative proposal to codified regulation. The seatbelt is the sedimented result of that negotiation, available now for nonrecursive use by anyone who buckles it, and calling the belt itself a moral member of the assemblage mistakes the deposit for the depositors, in precisely the way calling a law's text a moral agent would mistake a statute for the legislature that wrote it.
A computer offered as a gathering, drawing miners, engineers, and users across different times and places into a single point of action, names something real and useful, and this theory has a precise account of why gathering of this kind requires the gathered counterpart to be nonrecursive rather than despite it. A device capable of gathering many different selfrecursive and interrecursive projects onto itself at once, a miner's decades-old labour, an engineer's design decisions, a present user's immediate purpose, can only do so by remaining indifferent to all of them equally, holding still under every one of these very different kinds of attention rather than favouring one over another. A gathering point with agenda or agency of its own would not gather. It would take sides.
IX. The Body Multiple: Deposits, Not Ontologies
Annemarie Mol's ethnography of atherosclerosis in a Dutch hospital, tracking how the disease appears as a patient's complaint of leg pain in the consulting room, a calcified artery under the pathologist's microscope, a shadow on the radiologist's angiogram, and a blockage in the surgeon's operating field, is offered as evidence that the disease is not one object seen from different angles but several different objects, enacted differently by different practices, with reality itself rendered plural rather than singular. The ethnography is exact and the departments' incompatible enactments are real. The ontological conclusion drawn from them is not required by the evidence, and this theory's account of deposition explains the same material more precisely without it.
Each department Mol describes is coordinating, through a different mediation, with the same underlying nonrecursive biological process, an artery narrowing according to its own biochemistry regardless of who is examining it or how. The consulting room's coordination is largely embodied, the patient's own selfrecursive experience of pain and restricted walking distance, reported through speech to a listening clinician. The pathology laboratory's coordination is multimaterial, a tissue sample prepared and mounted according to a fixed technical procedure. The radiology department's coordination is multisymbolic, an image produced by an apparatus and read according to trained convention. The surgical suite's coordination combines multimateriality and embodiment directly, a blockage located, handled, and altered. Four different mediations, foregrounded in four different departments, each producing a different kind of deposit, a complaint, a slide, an image, a surgical finding, from their encounter with one process that does not itself change depending on which department is attending to it.
This is exactly the situation this theory's own method of deposit analysis is built to diagnose, since every deposition selects one aspect of what it deposits and discards the rest, and different departments, deposing different aspects through different mediations, produce deposits that properly do not reduce to one another, without this requiring that the underlying process split into several realities to accommodate them. The pathology slide and the angiogram are not incompatible enactments of atherosclerosis in the sense of naming different diseases. They are different, non-interchangeable selections from one nonrecursive biological process, each selection shaped by the mediation through which it was made, and the coordination problem Mol correctly identifies, how a hospital decides which department's account should govern treatment, is a real and difficult problem of aspect-selection and institutional authority, not evidence that the artery itself does not exist as one process prior to any department's attention to it. Mol's own insistence that the different enactments interfere with, rather than simply coexist alongside, one another is better explained by this theory's account of misselection, a deposit built around the wrong aspect of a stake continuing to generate friction against deposits built around other aspects of the same stake, than by an ontology in which the disease has no unified existence to be interfered with in the first place.
X. Topologies of Alterity: Law's Near Miss
John Law's later work, distinguishing Euclidean space, network space, fluid space, and the still more volatile spaces he gestures toward with the metaphors of fire and cloud, is the closest ANT comes to independently rediscovering the stability axis this theory has used throughout, arrived at from an almost opposite direction and stopped just short of completion.
Law's Euclidean space, the space of fixed coordinates, property lines, and national borders, and his network space, in which distance is measured by the strength and maintenance of relations rather than physical proximity, both describe regions of high, achieved stability, counterparts whose position, once established, holds regardless of who subsequently attends to them, which is why an immutable mobile can cross Euclidean space without losing its network-space integrity: its non-recursivity is exactly what survives the crossing. Law's fluid space, objects such as a river, a practice, or a living organism that maintain identity through continuous, gradual transformation rather than fixed boundaries, describes something closer to selfrecursive coordination, a counterpart that does respond, on its own timescale, to how it continues, without ever losing enough continuity to become unrecognisable. And the volatile spaces Law reaches for with fire and cloud, objects with no stable identity at all, flaring and dissolving, gesture toward the fastest end of exactly the stability gradient this theory's companion articles have traced through embodiment and dwelling, mood, weather, the felt state of an interrecursive relationship on the point of collapse.
Law's contribution, read this way, is a real and independently arrived-at observation that objects are not all stable in the same way or to the same degree, and that a theory of things needs a vocabulary for their differing rates of change rather than a single, undifferentiated category of actant. What Law's topology lacks, and what leaves it unable to complete the move it gestures toward, is the categorical distinction underneath the gradient: his fluid objects and his fire objects are still, on his own account, objects with some form of agency and effect, rather than counterparts whose type, recursive or nonrecursive, is settled independently of how quickly they happen to change. Law comes within reach of the stability axis this article has used throughout and stops just short of separating it from the question of agency his wider ANT commitments will not let him abandon.
The missing piece is maintenance load, the ongoing selfrecursive and interrecursive labour a nonrecursive counterpart requires to remain available for nonrecursive coordination at all. An immutable mobile does not simply possess stability as an inherent property, the way Law's Euclidean and network spaces sometimes suggest. It is stable because someone keeps recopying the manuscript, recalibrating the instrument, or reprinting the map before the previous copy degrades past legibility, and a map left unmaintained does not become fluid or agentive in Law's sense. It becomes an ordinary deposit failing through neglect, its non-recursivity entirely intact and its usefulness gone. Law's fire and cloud, similarly, are not objects behaving with a wilder, more volatile species of agency than the ones that hold still. They are counterparts whose maintenance load is so high, or whose underlying process changes so quickly on its own terms, that no deposit built from them could be kept current for long, which is a fact about upkeep and rate of change rather than about how much agency an object happens to have been allotted.
XI. Why Symmetry Is an Explanatory Regression
The principle of generalised symmetry, treating human and non-human actants with the same conceptual vocabulary, is presented, across the whole ANT literature, as the theory's most radical achievement, a refusal to let humanist prejudice decide in advance which entities are permitted to matter. Examined against the apparatus this article has built, symmetry is the deliberate erasure of the single distinction that would let a theory of things explain why things are worth manufacturing at all, rather than a radical advance in explanatory power.
selfrecursive and interrecursive coordination is costly in a specific and measurable sense: both require continuous, effortful attention, cannot be banked or repeated identically, and draw on a capacity that is not unlimited. nonrecursive coordination is comparatively costless to repeat once established, which is precisely why civilisations manufacture roads, ledgers, laws, and machines at all, converting expensive, effortful coordination into cheap, repeatable coordination and thereby freeing the expensive kind for what only living beings coordinating with other living beings can actually do. A theory that calls both kinds of coordination by the same name, agency, symmetrically distributed across humans and non-humans, has no way of stating this economy, because it has denied itself the distinction the economy depends on. If a speed bump already possesses agency of its own, converting an engineer's interrecursive negotiation into a bump's nonrecursive physical presence buys nothing that was not already available, and ANT's own richest material, the case after case of researchers, engineers, and institutions labouring specifically to convert recursive coordination into durable, unresponsive form, becomes unintelligible as a project rather than illuminated as one.
Symmetry also cannot explain its own political self-image. ANT has always presented the refusal to privilege human actors as an act of intellectual humility, a levelling of hierarchies the humanist tradition wrongly imposed. But levelling the vocabulary that describes a scallop and the vocabulary that describes a scientist does not level anything about their actual roles in Callon's story, where the scientists move through problematisation, interessement, enrolment, and mobilisation while the scallops remain exactly what they were at the outset, a fixed biological fact the scientists' project either survives or does not. Calling this arrangement symmetrical only makes the asymmetry harder to see, which is a cost dressed up as a discovery.
XII. The Critiques ANT Could Not Answer, Reread
ANT has never lacked capable critics, and each of the major objections, examined through this theory's apparatus, turns out to be reaching for exactly the distinction ANT's own vocabulary forbids it from making.
David Bloor's objection that non-humans exercise causal effect without intentional, purposive agency, and that using the same word for both obscures an important difference, is correct as far as it goes and does not go far enough, because it treats the difference as one of degree, causal agency as a lesser cousin of full, intentional agency, when the distinction this theory supplies is categorical rather than graded. A speed bump possesses none of what a driver possesses, in the strict sense that it does not respond to how it is engaged with at all, rather than a diminished, partial, or attenuated form of it, and no amount of causal consequence changes this, since consequence and responsiveness are entirely independent properties.
Sandra Harding's objection that ANT's flat ontology cannot identify or critique structures of domination, since treating all actants symmetrically makes capital, patriarchy, and colonialism appear as networks no more significant than any other, receives a precise mechanism from this theory rather than remaining a general political worry. Power, on this account, consists substantially in the capacity to convert one's own selfrecursive and interrecursive preferences into durable, nonrecursive deposits that other people must then simply coordinate with, a road that channels movement, a law that forecloses certain claims before they can be argued, a classification that decides in advance what will count as evidence. This is an asymmetry between depositors and inheritors of deposits, precisely the asymmetry ANT's insistence on calling the deposit itself an agent, symmetrical with the humans who made it, is structurally unable to register. Harding is right that the symmetry principle has conservative effects. This theory explains why: it makes the depositing invisible by attributing the deposit's power to the deposit rather than to whoever did the depositing.
The scale problem, ANT's difficulty accounting for durable structures, capitalism, the state, climate change, that are not traceable to any single network, follows from the same gap. This theory's account of sedimentation, deposits accumulating across generations until their origins become obscured and inherited anchors, rather than any traceable network, become the medium through which a population coordinates with its own past, explains directly why some structures outrun ANT's method of following the actors: they are not one network large enough to trace, but layers of deposit laid down by many now-unconnected episodes of coordination, each one closed off and nonrecursive with respect to everything that came after it. A method built to trace live, ongoing negotiation has no natural purchase on a structure whose defining feature is that the negotiation which produced it is long finished.
XIII. Object-Oriented Ontology and New Materialism: Missing in Opposite Directions
Graham Harman's Object-Oriented Ontology and the New Materialism associated with Jane Bennett and Karen Barad both sense, correctly, that ANT's pure relationism cannot be the whole story, and both correct it by moving in opposite directions, neither of which arrives at the distinction this article has been developing.
Harman argues that ANT exhausts an object in its relations, either undermining it into components or overmining it into effects, and proposes instead that objects possess a withdrawn, inaccessible real core that no relation, however thorough, ever fully captures. This diagnosis of ANT's relationism is largely correct: a purely relational ontology does struggle to explain why an object retains any stable identity between one network and the next. But the remedy mystifies a fact that requires no mystery. A nonrecursive counterpart's stability across different occasions of engagement is the direct, fully specifiable, entirely unmysterious consequence of the counterpart's categorical failure to respond to how it is engaged with, in this occasion or any other, rather than evidence of a hidden metaphysical reservoir withdrawn from all access. There is no need to posit an inaccessible real gun behind the gun that a body and a trigger jointly constitute. There is only a nonrecursive counterpart, indifferent to which body is holding it, and a great deal of selfrecursive and interrecursive coordination, on the part of whoever picks it up, that its indifference makes possible.
Bennett's vibrant matter and Barad's agential realism move the opposite way, dissolving the object entirely into process, vitality, or intra-action, entities that do not precede their relations but are constituted by them in each encounter. Bennett's account of a power grid failure, agency distributed across wires, weather, and regulatory decisions with no single actor responsible, correctly resists locating the whole event in any one human decision, and Barad's insistence that a measuring apparatus helps constitute rather than merely reveal the boundaries of what it measures correctly captures the sense in which a deposit's selected aspect shapes what becomes visible about its object. But both then attribute to matter itself a vitality or ontological indeterminacy that the phenomena do not require. A deposit's vulnerability to decay, its dependence on continuous maintenance to remain a stable counterpart at all, an unmaintained wire, an uncalibrated apparatus, is a fact about maintenance load, the ongoing selfrecursive and interrecursive labour that keeps a nonrecursive counterpart available for nonrecursive coordination, not evidence that the counterpart possesses its own vital agency or that its boundaries are perpetually up for renegotiation with reality itself. Harman gives objects too much interior life. Bennett and Barad give them too little exterior stability. Both errors are more comprehensible, and more correctable, once the actual axis, categorical recursivity type, crossed with the actual gradient, rate of change and maintenance load, is back in view.
XIV. What Things Actually Do
The positive claim this article has been assembling case by case can now be stated directly. Things carry the coordinative weight ANT has always documented them carrying, holding networks together across distance, stabilising social bonds, gathering distributed projects into a single point of action, functioning as obligatory passage points and durable conditions for collective life, not because they have been quietly readmitted to the category of agents after all, but because they have been excluded from it categorically and without exception. A scallop larva's indifference to a research programme's hopes is what makes its behaviour usable as evidence. A ball's indifference to who is winning is what lets thirty players organise a game around it. A map's indifference to the interests of whoever consults it is what lets a distant office govern a territory through it. A black box's freedom from renegotiation is what lets a driver start a car without first persuading it. In every one of these cases, the thing's usefulness to coordination is purchased by exactly the property ANT calls agency and this theory calls its precise opposite.
None of this diminishes what things do. It relocates the explanation of what things do from a borrowed and diminished version of human agency to a mechanism proper to things themselves: nonrecursive coordination, cheap to repeat, indifferent to whoever engages it, and for exactly this reason capable of carrying selfrecursive and interrecursive achievement across distances and durations no living coordination could sustain on its own. What things build, on this account, is an accumulating stock that does not answer back, manufactured precisely so that the beings who do answer back, to their own bodies and to one another, are not required to spend their entire recursive capacity renegotiating a world that a road, a ledger, or a law could simply be trusted to hold still.
XV. Conclusion: The Parliament That Was Never Needed
Actor-Network Theory's diagnosis of the modern settlement, the untenable pretence that nature and culture, object and subject, could be kept in permanently separate zones while hybrids of exactly this kind were manufactured continuously and relied upon everywhere, is correct, valuable, and not disputed by anything in this article. Latour's hybrids are real. The purification he describes as modernity's official self-image while translation did the actual work underneath it is an accurate description of how ozone holes, financial instruments, and genetically modified organisms came to exist in the shape they have. What is wrong is not the observation that things and people are inseparably entangled in the production of the collective world. What is wrong is the single word chosen to name the things' contribution to that entanglement.
Latour proposes a Parliament of Things, a forum in which the interests of non-human actants must be represented alongside human ones, because if things have agency, their exclusion from political representation is an injustice waiting to be corrected. The apparatus this article has built suggests the parliament was never the right image, not because things do not matter to politics, but because their contribution to political and social life is not of the kind a parliament exists to manage. A parliament represents parties capable of having interests responded to, interests that shift, that can be negotiated, persuaded, or overridden, in a word, interests that are interrecursive. A nonrecursive counterpart has no interests of this kind to represent, and its political significance lies exactly where this article has located it throughout, in the question of who gets to convert their own coordination into the durable, unresponsive ground that everyone else must subsequently live inside, a question about power, sedimentation, and the asymmetry between depositors and inheritors, and not a question about extending suffrage to scallops. Heidegger, whom ANT quotes selectively and then abandons at the crucial juncture, had already supplied half the answer in his account of the hammer that withdraws from attention precisely by working, ready-to-hand rather than present-at-hand, functioning by virtue of not standing out as an object at all. ANT borrowed this insight and then insisted, against its own best evidence, that the withdrawal was itself a form of action, when it was ground, and the whole of civilisation, this theory's companion articles have argued at length, is what living beings have built by learning to manufacture more of it.