Abstract
"The unreasonable effectiveness of mathematics" names a puzzle that has occupied physicists and philosophers since Wigner posed it: why should an abstract, self-contained symbolic system correspond so precisely to physical reality? This article argues that the puzzle has been posed backwards. Mathematics is a purely multisymbolic system, and every one of its objects, relations and operations is purely nonrecursive: nothing mathematical alters its course in response to being engaged, measured or investigated. This nonrecursivity is categorical, not a matter of degree, and it must not be confused with the entirely different sense in which mathematics and computer science speak of "recursive" functions and definitions. Because mathematics is purely nonrecursive, its applicability beyond itself cannot be presumed from its internal consistency. It depends on cosmic fit: a correspondence between the formal organisation of mathematics and the organisation of whichever aspect of the cosmos it is applied to. Nonrecursivity is identified as the primary ontological condition for direct mathematical fit, with discreteness, continuity, invariance, symmetry and related properties determining which particular mathematics a given nonrecursive domain affords. The uneven success of mathematics across the sciences therefore requires no mysterious mathematical universe. The effectiveness of mathematics is not a universal property of mathematics. It is an empirical property of those aspects of the cosmos with which mathematics achieves fit.
1. The Question Has Been Asked Backwards
1.1 The traditional astonishment
Galileo's image of nature as a book written in the language of mathematics, its characters triangles, circles and other geometric figures, established one of the most durable convictions in the history of science: that the physical world is, at some level, already mathematical, and that mathematics discovers rather than imposes its order upon it. Three centuries later Eugene Wigner gave the conviction its canonical modern formulation, describing the fit between mathematical formalism and physical law as an unreasonable effectiveness, a gift neither understood nor deserved. The puzzle has lost none of its force since. Why should a system of abstract symbolic operations, developed largely for its own internal consistency and elegance, correspond with such extraordinary precision to the behaviour of planets, particles and fields?
Several answers have accumulated around this question over the past century. Mathematics might be discovering structures that were already present in reality, waiting to be found rather than made. Mathematical objects might possess an independent existence of their own, prior to and independent of any physicist's use of them. Reality itself might be, at bottom, a mathematical structure, so that physical law and mathematical law are simply the same thing described from different vantage points. Mathematics might instead be a human technology, selected and refined because it works, its apparent inevitability an artefact of a long history of pragmatic trial and error. Or human cognition itself might have evolved to detect mathematical regularities because organisms that could track periodicities, quantities and spatial relations survived better than those that could not.
These candidate answers disagree sharply with one another, and each has its own defenders and its own difficulties. But they share a structural feature that is more consequential than any of their disagreements. All of them begin from mathematics and ask how it relates to reality. Mathematics is treated as the fixed term of the puzzle, the given whose success then has to be explained by appeal to platonism, pragmatism, cognitive evolution or metaphysical identity. The reality side of the relation is left comparatively unexamined.
1.2 Reverse the direction
An alternative approach is available, and it is simple enough to state. Instead of asking why mathematics describes the cosmos, ask what kind of cosmos would have to exist for mathematical description to be possible at all. This reverses the direction of the original question without denying anything that any of the standard answers assert. It shifts the explanandum from mathematics to the world.
On this approach there may be nothing especially mysterious about mathematics itself. Mathematics is what it is: a formal system built from symbolic distinctions, definitions and rules of inference. Whatever mystery remains attaches not to the system but to the aspects of the cosmos in which that system finds purchase. The question becomes an ontological one about the organisation of reality rather than an epistemological one about the powers of a symbolic technique.
1.3 The empirical unevenness of mathematics
That reversal gains its force from an observation too obvious to have been ignored, yet too rarely built into the architecture of an explanation. Mathematical applicability is not evenly distributed across the phenomena of the world. Mathematics describes planetary trajectories with an accuracy sufficient to land a spacecraft on a comet decades after launch. It describes electromagnetic relations, chemical stoichiometries, rates of change, spatial transformations and probability distributions with comparable precision. Across an enormous range of physical phenomena, mathematical description is not merely useful. It is exact, in a sense that borders on the uncanny.
No comparable mathematics exists for the full trajectory of a friendship, the course of a political negotiation, the unfolding of a psychotherapy session, an ethnographic encounter between fieldworker and interlocutor, or a person's gradually changing understanding of themselves. It would be a mistake to say simply that these phenomena are more complex than planetary motion and leave the matter there. A hurricane involves vastly more interacting variables than a single conversation between two people, and its computational demands dwarf anything a linguist or a therapist deals with, yet a hurricane remains squarely within the domain that mathematics, in the form of fluid dynamics, can model with real if imperfect success. Complexity in the sense of sheer number of interacting components is not what distinguishes the domains where mathematics thrives from the domains where it does not. Whatever the relevant distinction is, it lies elsewhere.
1.4 The first hypothesis
The uneven success of mathematics is not, on this view, an embarrassment to be explained away. It is data. It is the clue that should organise an account of why mathematics works where it works and falls short where it falls short. The question this article pursues can now be stated precisely: what properties must some aspect of reality possess in order for a purely mathematical representation to achieve fit with it? Everything that follows is an attempt to answer that question properly, beginning with an account of what kind of thing mathematics actually is.
2. What Kind of Thing Is Mathematics?
2.1 Mathematics is multisymbolic
Human beings possess a distinctive capacity to detach a distinction, once drawn, from the particular occasion of its drawing, so that it remains available across absence, delay, distance and even the death of whoever first drew it. A name persists across every change undergone by the person who bears it. A written law continues to bind long after everyone present at its drafting has died. A tally mark scratched onto bone preserves a count that no living memory needs to retain. Call this capacity multisymbolization: the deposit of a distinction into a symbolic form durable enough to be encountered again by parties who were not present at its making.
Mathematics is among the most extraordinary elaborations of this capacity that any lineage has produced. Numerals, operators, variables, equations, functions, axioms, proofs and diagrams are all symbolic technologies that allow distinctions and relations to be manipulated independently of the circumstances in which they were first produced. The symbol '2' is not two objects; it is a durable mark that stands in for a distinction about quantity that can be redeployed indefinitely, in any context, by anyone trained to recognise it. The operator '+' is not a physical act of combination; it is a rule governing how such marks may be manipulated. The relation '=' is not itself an instance of equivalence occurring anywhere in the world; it is a symbolic convention specifying when two expressions may be substituted for one another. Mathematics exists entirely through relations among symbolic distinctions of this kind, deposited, combined and manipulated according to rules that hold independently of any particular occasion of their use.
2.2 Mathematics is purely nonrecursive
A further distinction is needed before the central claim of this section can be stated. Some things respond to how they are engaged: how they proceed afterwards is altered by the engagement itself. A person who is asked a difficult question is not the same, afterwards, as they would have been had the question not been asked; the asking becomes part of what shapes their subsequent course. Call this property recursivity. A counterpart is recursive when the course it takes is affected by how it is engaged, and nonrecursive when nothing about the engagement alters its course. The distinction is categorical rather than a matter of degree. A counterpart either answers back, in this specific sense, or it does not.
Nothing mathematical does. The number seven does not respond to the number three. A triangle does not modify its angles because someone measures them. An equation does not learn from the solutions previously found for it. A theorem does not reconsider its own truth after being criticised. A proof does not become defensive when a gap is found in it, though the mathematician who wrote it certainly might. Mathematical entities, relations and operations are purely nonrecursive. The claim is categorical, not a generalisation from cases examined so far that some future counterexample might overturn: mathematical entities are, by construction, symbolic constructions whose behaviour, once their governing rules are fixed, is entirely internal to those rules and entirely unaffected by any subsequent engagement with them. There are no partly recursive mathematical entities. The category admits no degrees.
2.3 Mathematical recursion is completely different
This claim will strike anyone trained in mathematics or computer science as strange, because both fields use the word 'recursion' constantly, and use it to describe some of their most powerful tools. A recursive function is one defined in terms of its own previous values. A recursive algorithm calls itself, with a reduced input, until a base case is reached. A recurrence relation specifies each term of a sequence in terms of the terms that precede it. Recursive definitions are common throughout logic and formal grammar. It would be easy to conclude that mathematics is shot through with the very property this article has just denied it possesses.
The conclusion would rest on a false friend. Formal recursion, in the mathematical and computational sense, names a symbolic operation specified through reference to previous applications or states of that same operation. A Fibonacci sequence does not respond to itself in the sense at issue here; its next term is formally determined by a fixed rule that refers to preceding terms, and nothing about calculating one term changes what the rule says about the next. The entire apparatus of formal recursion remains, without exception, nonrecursive in the sense this article is using. What varies is only how many steps of a fixed, nonresponsive rule must be unfolded before an output is produced. Formal recursion is a symbolic operation specified through previous applications or states of that operation. The recursivity this article is concerned with is the responsive alteration of a process through its engagement with itself or with another responsive process. The two concepts share a word by historical accident. They do not share an ontology, and nothing follows from mathematics for the second by way of the first.
2.4 Doing mathematics is recursive; mathematics is not
An obvious objection remains to be addressed before this section can close. Mathematicians are living human beings, and everything true of living human beings is, in the relevant sense, true of them while they work. Mathematicians struggle with a proof, feel confusion give way to understanding, revise a conjecture in light of a counterexample, argue with colleagues, collaborate across years, teach students who misunderstand and eventually understand, and are persuaded, or fail to be persuaded, by an argument's force. The entire activity of doing mathematics is saturated with selfrecursive and interrecursive process: a mathematician's relation to their own developing understanding, and their relation to the community of other mathematicians whose responses shape what they attempt next.
None of this touches the claim made above. It shows that doing mathematics is a recursive human activity while leaving entirely intact the separate claim that mathematics, the object of that activity, is not. A mathematician's relationship to an equation can be fully recursive from the mathematician's side, involving confusion, insight, revision and persuasion, while remaining entirely nonrecursive from the equation's side. The equation itself never answers back. It does not notice being struggled with, does not reward persistence with an altered structure, does not withhold its implications from a mathematician it dislikes. This asymmetry between a living, responsive investigator and a nonresponsive mathematical object is close to the centre of the argument of this article, and the sections that follow show how much depends on keeping the two sides of the asymmetry from collapsing into one another.
3. Why Mathematics Is Unlike Other Symbolic Systems
3.1 Ordinary symbolic systems remain recursively open
Mathematics is not the only multisymbolic system human beings have built. Ordinary language is another, and the comparison between the two throws mathematics's peculiar character into relief. Nothing internal to the alphabet requires that the letter B follow the letter A in any word. Nothing about the sequence of letters composing the word 'cat' necessitates that it refer to a particular kind of animal rather than any other. Words acquire and change their significance through long, ongoing histories of use among people who respond to one another's usage, argue about it, extend it by metaphor, invert it through irony, and gradually shift what a term is taken to pick out. A word spoken today carries the sediment of every previous occasion on which it was used, misused, contested and redefined, and it remains open to being used differently tomorrow. Ordinary language, in other words, remains continuously and constitutively dependent on interrecursive coordination among its users. Its meanings do not stand still because the community that sustains them does not stand still.
3.2 Mathematics minimizes recursive dependence
Mathematics does something radically different, and its difference from ordinary language is a difference of ontological kind rather than merely of degree of precision. Once the primitive distinctions, definitions, axioms and operations of a mathematical system have been fixed, everything that follows from them follows without renegotiation. No vote taken by a community of mathematicians, however large or however unanimous, can make the number seventeen cease to be prime within ordinary arithmetic. A disliked solution to an equation does not become false because mathematicians find it inconvenient or unwelcome. Mathematics therefore achieves an extreme form of symbolic closure that ordinary language, by its very nature as a living, interrecursively sustained system, cannot achieve and does not aim at. The closure is not absolute across the entire history of mathematical practice, since mathematicians remain free to invent new definitions, new axioms and entirely new formal systems whenever they choose. But within any specified formal system, once specified, the consequences that follow from it are settled independently of anyone's subsequent engagement with them.
3.3 Made systems, found consequences
This observation resolves a piece of the long-running dispute between mathematical invention and mathematical discovery, though not the whole of it. Human beings make the notation, the definitions, the axiomatic formulations and even the questions that a given branch of mathematics pursues. Nobody discovered the symbol for zero waiting somewhere in nature; it was invented, and its adoption transformed what arithmetic could efficiently express. But once a notation, a definition and a set of axioms have been specified, their consequences are constrained by the system itself, and working those consequences out has the unmistakable phenomenology of discovery rather than invention. A mathematician proving a theorem does not feel that they are deciding what the theorem will say. They feel that they are finding out. The symbolic architecture of a mathematical system is made. Its consequences, once the architecture is fixed, are found. Recognising this distinction dissolves a great deal of the apparent tension between mathematics as human construction and mathematics as objective discovery, because the two descriptions are true of different moments in a single process rather than competing accounts of the same moment.
3.4 But internal necessity proves nothing about external reality
One further distinction closes this section and opens the one that follows. A conclusion can follow with the fullest possible necessity from within a mathematical system without anything in the wider cosmos being obliged to instantiate it. Non-Euclidean geometries are exactly as internally rigorous as Euclidean geometry, and for over two thousand years only the latter was thought to describe physical space. Internal mathematical necessity, the fact that a theorem follows inexorably from a system's axioms, is one kind of relation. External ontological applicability, the fact that some aspect of the physical world happens to be organised in a way that a particular mathematical system fits, is an entirely different kind of relation, and the first does not entail the second. This is the gap across which the remainder of this article has to build a bridge.
4. From Mesocosmic Fit to Cosmic Fit
4.1 The existing concept of fit
Distinguishing the lived world of coordination, the mesocosm, from the cosmos considered independently of any living being's engagement with it, has already proved useful for diagnosing when a symbolic account overreaches the coordination it claims to describe. A concept suffers from mesocosmic misfit when it claims authority beyond what the coordination it is meant to serve can actually support, as when a diagnostic category is applied to an experience it does not capture, or a legal classification is imposed on a relationship whose real structure it distorts. That diagnosis has so far been developed for symbolic operations addressed to living, coordinating beings. Mathematics forces a generalisation of it, because mathematics is regularly applied to aspects of the cosmos that involve no living coordination at all, and no mesocosm in the relevant sense, and yet the same underlying question of fit arises.
4.2 Define cosmic fit
Call the broader relation cosmic fit: the correspondence between the formal organisation of a symbolic system and the organisation of some aspect of the cosmos to which that symbolic system is applied. Cosmic fit does not presuppose life, and it does not presuppose a coordinating perspective of any kind. The orbit of a planet could have afforded an exact mathematical description in a universe that contained no astronomer to write that description down, just as it did, for billions of years, in this one. Fit is a relation between a formal system and an organisational property of some domain. The recognition of fit, by contrast, is something only a living, symbolising being can achieve. The two must be kept apart. A domain can possess mathematical fit long before anyone notices it, and a domain's fit does not depend on anyone ever noticing it at all.
4.3 Mathematics receives no ontological privilege
The generalisation carries a consequence that is easy to state and difficult to resist. Mathematics is a symbolic system, no more and no less, and like any other symbolic system its applicability to some aspect of the cosmos has to be earned empirically rather than assumed from its own internal properties. Extraordinary internal consistency, of the kind mathematics achieves to a degree no other symbolic system rivals, grants mathematics no automatic jurisdiction over anything outside itself. A given mathematical model may fit its target domain extremely well. It may fit only approximately, capturing the coarse behaviour of a system while missing its finer structure. It may fit only selected aspects of a domain while leaving others entirely uncaptured. Or it may fail outright. Nothing about mathematics considered purely as a formal system determines in advance which of these outcomes will obtain for any given domain. That determination belongs entirely to the organisation of the domain itself.
4.4 The master principle
The argument of this section can now be compressed into a single governing proposition, one that the remainder of the article exists to unpack and defend: mathematics has direct cosmic fit only with nonrecursive aspects, relations, or symbolic deposits. Wherever mathematics achieves striking success, on this proposition, it is because it has made contact with something in the relevant domain that does not answer back. Wherever mathematics falls short, the first question to ask is not what better mathematics might be found, but whether the domain in question possesses the kind of nonresponsive organisation that any mathematics, however sophisticated, would need in order to gain purchase on it.
5. What Kind of Cosmos Allows Mathematics?
5.1 Level one: nonrecursivity as the ontological entry condition
The master principle stated at the end of the previous section is the foundation on which everything else in this article rests, and it is better established as a hierarchy than as a flat list of requirements, because not every property relevant to mathematical fit carries the same weight. The primary, entry-level condition is nonrecursivity itself. For a mathematical relation to map directly onto some counterpart in the world, the relevant aspect of that counterpart cannot alter its own course in response to the mathematical representation being constructed of it. The moon does not adjust its orbit after Newton's laws are published. A physical length does not reconsider itself because a ruler has been laid against it. A radioactive isotope does not alter its probability of decay because a physicist has calculated that probability. In every case where mathematics achieves its most exact and durable successes, the represented property does not respond to being represented.
This entry condition should not be confused with several properties that are often run together with it, and separating them out matters enough to receive its own full treatment in the section that follows this one. Nonrecursivity does not require stability, simplicity, predictability, determinism or controllability. A domain can be wildly unstable, extraordinarily complex, only partially predictable and entirely beyond anyone's control while remaining completely nonrecursive, and mathematics can still, in principle, achieve fit with it. Those further distinctions belong to a later stage of the argument. At this first stage, only one question is being asked: does the relevant aspect of the domain answer back, or does it hold still while being engaged?
5.2 Level two: which structure does the nonrecursive domain possess?
Nonrecursivity clears the ground for mathematical fit without yet determining what particular mathematics that fit will take. A second level of structure has to be present, and several distinguishable properties belong to it.
Differentiability is the most fundamental of these, more basic even than discreteness. Before anything can be mathematically represented there must be distinguishable states, positions, magnitudes or relations within the domain; without difference of some kind there is nothing for a symbol to track.
Discreteness supports arithmetic specifically. Counting requires individuable units: one sheep, another sheep, two sheep. But the abstraction this involves should not be passed over. The two sheep counted are never identical to one another in every respect; counting works only because most of their differences, in colour, in temperament, in precise weight, have been treated as irrelevant to the operation being performed. Every act of counting is already an act of selective abstraction.
Equivalence deserves particular emphasis, because it is doing more work in mathematical practice than is usually acknowledged. Mathematical operations require criteria under which entities that differ in other respects can nonetheless be treated as equivalent for the purpose at hand. Two red apples added to two green apples make four apples only because colour has been excluded from the equivalence relation governing the operation of addition as applied to apples. Mathematisation therefore always proceeds through a prior act of selective irrelevance, deciding what counts as the same for the purposes of the calculation and setting the rest aside. This point matters a great deal once the discussion turns to the social sciences.
Continuity supports an entirely different family of mathematical tools from discreteness, and its presence shows that discreteness cannot be treated as a universal prerequisite for mathematical fit. Calculus depends on continuous variation rather than on individuable units, and it achieves its own extraordinary fit precisely in domains where discreteness would be the wrong description. The deeper requirement running beneath both discreteness and continuity is simply that a domain afford formally representable differentiation and relation, in whichever form that differentiation happens to take.
Invariance matters especially for mathematical physics. Some property or relation within a domain must remain invariant across the relevant class of transformations if change within that domain is to become mathematically tractable; it is invariance within change, rather than the absence of change, that mathematical physics actually requires.
Symmetry, where a domain exhibits transformations that leave its relevant structure unchanged, is a special and unusually productive case of invariance, and mathematics maps symmetrical domains with exceptional power.
Periodicity, the recurrence of a pattern across time, allows mathematical coordination across temporal distance in a way few other properties permit; the astronomical cycles that gave humanity its earliest calendars are the paradigm case.
Relative separability requires that some variable or relation within a domain be isolable for the purposes of analysis, without this implying that the isolated element is metaphysically independent of everything else in the domain. Separability means only that a particular relation can be treated on its own, for analytic purposes, without the phenomenon under study collapsing into incoherence.
Compositionality, finally, requires that relations within a domain remain sufficiently coherent when combined into larger formal structures, so that simple mathematical relations can be built up into more complex ones without the combination itself destroying what made the simple relations tractable in the first place.
5.3 Level three: different structures afford different mathematics
These properties do not simply add up to a single undifferentiated condition called mathematical fit. Different combinations of them afford different branches of mathematics, and the correspondence is close enough to function as a kind of map. Discreteness affords arithmetic. Spatial invariance affords geometry. Continuous change affords calculus. Symmetry affords group-theoretic treatment. Repeated stochastic events afford probability theory. Network relations afford graph theory. The cosmos, on this account, does not generically fit mathematics as a single undifferentiated whole. Particular aspects of the cosmos afford particular mathematical forms, according to which of the level-two properties those aspects happen to possess.
5.4 Mathematical fit is aspect-specific
The final move in this section is the one on which the rest of the article depends most heavily, and it follows directly from everything established so far. Mathematical fit should never be assessed for an entity taken as a whole. It should be assessed aspect by aspect. A human being has a height, a mass, an age, a body temperature, a number of children and a set of geographical coordinates, all of which are mathematically tractable properties, each affording precise numerical representation. The same human being also interprets a diagnosis they have been given, changes their behaviour after learning a prediction made about them, responds to how another person has responded to them, and reconsiders a relationship in light of something newly understood about it. Those processes are recursive in the full sense this article has been developing: their course is altered by the very engagement that constitutes them. Mathematics can therefore achieve perfect fit with a nonrecursive aspect of a recursive being without thereby achieving any fit at all with the recursive organisation of that same being. This is the foundation on which the critique of mathematisation developed later in the article rests, and it needs to be held firmly in view before that critique is made, because without it the critique would look like an attack on mathematics itself rather than what it actually is, an insistence on tracking fit aspect by aspect rather than assuming it wholesale.
6. Nonrecursivity Is Not Simplicity, Stability, Determinism or Predictability
The distinctions drawn in the previous section presupposed a separation that this section now makes fully explicit and defends against the most natural objections to it. Nonrecursivity is not the same property as simplicity, stability, determinism or predictability, and confusing any of these with nonrecursivity produces exactly the kind of error that has made the argument of this article harder to see than it needs to be.
Complexity is the first and most tempting confusion. A system can be fantastically complex, involving vast numbers of interacting components, while remaining completely nonrecursive throughout. Complexity concerns the number of components in a system and the density of relations among them. Recursivity concerns something else entirely: whether the system's course is altered by its own engagement with itself or by another responsive system's engagement with it. The two properties vary independently, and nothing about a system's complexity settles the question of its recursive status.
Stability produces the second confusion. A mountain is stable and nonrecursive. An avalanche is unstable and equally nonrecursive. Instability, in other words, does not imply recursivity, however tempting it might be to associate responsiveness with change and nonresponsiveness with permanence. The mountain's stability and the avalanche's instability are both properties of nonrecursive matter behaving according to fixed physical law; neither the mountain nor the avalanche answers back to being engaged, regardless of how much or how little either one changes over time.
Predictability produces the third and perhaps the most consequential confusion, because it bears directly on how physical science has often been popularly understood. Chaotic systems can be entirely nonrecursive while remaining, in practice, unpredictable, because their governing equations are exquisitely sensitive to initial conditions that can never be measured with sufficient precision. Weather does not become recursive merely because tomorrow's exact state resists calculation three weeks in advance. The governing physics of the atmosphere never answers back to the meteorologist attempting to forecast it; what defeats the forecast is sensitivity, not responsiveness. This distinction deserves complete explicitness because so much confusion in public discussion of chaotic and complex systems rests on exactly this conflation.
Determinism produces a fourth and closely related confusion. Randomness and indeterminacy do not, on their own, create recursivity. A stochastic process, one whose outcomes are governed by probability rather than by fixed necessity, can remain entirely nonrecursive throughout its operation; the dice do not respond to being rolled, even though which face lands upward cannot be predicted with certainty in advance. Quantum indeterminacy, whatever the correct interpretation of it eventually turns out to be, should not by itself be taken to threaten mathematical fit, because indeterminacy of this kind is a different property from responsiveness, and only the latter is at issue in the account this article is defending.
Interaction produces a fifth confusion that is worth making fully explicit, because otherwise critics of this article's argument will be tempted to claim that every physical feedback loop qualifies as recursive in the relevant sense, which would collapse the distinction the entire article depends upon. Physical entities interacting with one another does not constitute recursivity as this article has defined it. Two billiard balls collide and exchange momentum according to fixed laws. Particles interact through fields. Gravitational bodies affect one another's trajectories. Chemical compounds react, sometimes violently, releasing or absorbing energy in the process. None of these interactions makes any of the entities involved recursive in the relevant sense, because none of them involves a course of action being altered through engagement with a response as such, in the way a person's subsequent behaviour is altered by having been asked a question, or a relationship's future course is altered by how a difficult conversation within it unfolds. Causal interaction, however intricate, is not recursivity.
The precise boundary that all five of these clarifications converge upon can now be stated plainly. That boundary runs not between simple and complex systems, ordered and chaotic ones, deterministic and stochastic processes, or isolated and interacting entities, but between processes whose course is altered through recursive responsiveness and processes that, however complex, unstable, unpredictable, indeterminate or thoroughly interacting they may be, do not answer back in this specific sense. That is the ontological boundary that mathematical fit has to respect, and every subsequent claim in this article depends on keeping that boundary exactly where this section has drawn it.
7. Why Mathematization Varies Across the Sciences
With the boundary drawn precisely, it becomes possible to explain why mathematics succeeds so differently across the sciences without appealing to any ladder of increasing recursive density running from physics at one end to anthropology at the other. The variation is real, but its explanation is more specific than any such ladder could capture.
Physics studies relations that are overwhelmingly nonrecursive. Particles, fields, forces and the great majority of the quantities physics is built to track do not respond to being measured, calculated or theorised about, and this affords physics an exceptionally extensive mathematical treatment. The mathematical success of physics reflects not merely a superior method, though physicists have certainly refined their methods across centuries, but a remarkable degree of fit between the symbolic technologies mathematics provides and the actual organisation of the objects physics studies.
Biology occupies a more mixed position. Living beings possess an enormous number of nonrecursive aspects: mass, chemical concentrations, gene frequencies, core temperature, anatomical dimensions, population counts, all of which submit readily to mathematical treatment, and biology has built extraordinarily productive mathematics around exactly these aspects. But life also involves recursivity throughout, in development, in behaviour, in the immune system's responsiveness to what it encounters. Mathematics can therefore describe an enormous amount about living beings without thereby mathematically exhausting the recursive coordination that constitutes their being alive in the first place.
Economics presents an especially interesting case, and it repays more careful treatment than the simple observation that economic behaviour resists full mathematisation. Economic systems do not merely happen to contain some nonrecursive aspects among many recursive ones. They actively manufacture nonrecursive aspects as part of their basic operation: prices, currencies, quantities, balances, interest rates, inventories and accounting categories are all deliberately constructed nonrecursive deposits, built precisely so that they can be manipulated with the extraordinary mathematical power those deposits afford. This is why so much of economics achieves genuine mathematical tractability. But economic actors also respond to prices, to forecasts, to one another's behaviour, and to the institutional classifications and economic models built to describe them, and this responsiveness means that the mathematical representation itself can become recursively relevant, feeding back into the behaviour it was built to describe. The correct diagnosis is therefore not that economics is too recursive for mathematics to handle, but that economics contains both mathematically tractable nonrecursive aspects, deliberately manufactured as such, and recursively organised processes built around and through those aspects, and that serious error arises whenever the former are mistaken for an exhaustive representation of the latter.
Psychology exhibits a closely related structure. Scores, response times, counts of correct answers and other recorded outputs of an experiment are straightforwardly mathematically tractable, and psychology has produced a great deal of rigorous, useful mathematics on this basis. But a score is not identical to the recursive process through which a person encountered a question, interpreted what it was asking, deliberated about how to answer it, and finally produced the answer that got recorded. Mathematics may describe the nonrecursive deposit that process leaves behind with perfect accuracy while saying nothing at all about the recursion that produced it.
Anthropology, finally, is distinctive among the sciences discussed here because it so often investigates precisely those processes in which interrecursivity is constitutive of the phenomenon under study: kinship, ritual, exchange, negotiation, the ongoing coordination between fieldworker and interlocutor that produces ethnographic knowledge in the first place. This does not make quantification illegitimate within anthropology, and anthropologists have produced valuable quantitative work on demography, migration and much else. It means that quantitative representation cannot be assumed to substitute for analysis of the recursive organisation that is often what makes an anthropological phenomenon what it is.
A single methodological principle can now replace the mistaken idea that mathematisation simply declines as some notion of recursive density rises across the sciences. The scope of adequate mathematisation in any domain depends on how much of the phenomenon under investigation consists in nonrecursive properties, and how much depends constitutively on recursive coordination that no nonrecursive deposit can substitute for. The characteristic error of mathematisation, wherever it occurs, is therefore not a mathematical error in any technical sense, but an ontological extrapolation: the mistake of treating an accurate mathematical representation of a domain's nonrecursive aspects as though it were an exhaustive representation of a phenomenon whose organisation also depends, in whole or in part, on recursion that no amount of further mathematical refinement can capture.
8. The Apparent Counterexample: Population Statistics
An obvious challenge to the argument developed so far deserves direct treatment before the article proceeds further, because if it succeeds it would undermine everything claimed above. If mathematics achieves direct fit only with nonrecursive counterparts, population statistics looks like an immediate counterexample. Demographers study mortality, fertility, illness, crime, voting behaviour, income and belief, and every one of these concerns recursive living beings behaving in recursive ways. Statistics appears to mathematise recursive life directly, and with considerable success. Has mathematics, in this domain at least, crossed the boundary this article has drawn around it?
It has not, and seeing why requires tracing the actual sequence through which a statistical figure comes into being. Consider a census. The population being counted remains fully recursive throughout the counting: people marry, migrate, change their minds about how to answer a question, and respond to the very fact of being counted in ways that can alter their subsequent behaviour. But the census itself does not produce a mathematical representation of that recursive population directly. It produces entries: discrete records, each one the nonrecursive deposit of a recursive encounter between an enumerator and a respondent. Consider a survey. The interaction between interviewer and respondent may be richly interrecursive, each shaping how the other proceeds through the exchange. But once an answer has been recorded, coded as, say, 'strongly agree equals five', the resulting datum is no longer recursive in any sense. It sits on the page, or in the database, entirely indifferent to whatever the respondent goes on to think, feel or do next. Consider mortality statistics. A person's life, right up to its final moment, was thoroughly recursive. The death certificate that records its ending is not.
A sequence can be laid out explicitly to make the pattern visible. A recursive process gives rise to a recursive act of measurement, classification or recording, undertaken by a living investigator. That act of measurement produces a nonrecursive multisymbolic deposit: a datum, an entry, a coded response. Nonrecursive mathematical operations are then performed on that deposit, or on collections of such deposits, yielding a nonrecursive statistical output: a mean, a rate, a correlation, a regression coefficient. That output may then be encountered by living beings, whose recursive response to encountering it, changing behaviour in light of a published statistic, for instance, restarts the cycle from a new recursive starting point.
Statistics, properly understood through this sequence, demonstrates not that mathematics has successfully captured recursion after all, but the opposite: the prior conversion of selected aspects of recursive processes into nonrecursive symbolic counterparts, upon which nonrecursive mathematics is then free to operate with all its usual power and all its usual limits.
One further temptation should be explicitly refused here, because it would otherwise reintroduce exactly the confusion this article has been at pains to dissolve. It might seem tempting to say that a sufficiently large population somehow becomes nonrecursive in aggregate, even though each individual within it remains recursive. This formulation should be rejected outright. Nonrecursivity does not emerge from recursion by any process of aggregation, however large the aggregate. The population being studied remains recursive throughout, at every scale. What accumulates in a statistical dataset are nonrecursive symbolic traces of past recursive processes, not a new, emergent nonrecursive entity somehow constituted out of recursive parts. Keeping this distinction sharp is what keeps the ontology developed in this article categorical rather than a matter of scale or aggregation, and the question of exactly how the transformation from recursive event to nonrecursive datum occurs, and what is preserved and what is lost in that transformation, is an important question in its own right, one that repays careful treatment beyond the scope of the present discussion.
9. Cosmic Boundaries: Does Mathematics Fit Everywhere?
The argument so far has been building toward an empirical claim, and this section states it directly. There is no theoretical reason to assume, in advance of investigation, that every aspect of the cosmos must possess the kind of organisation that affords mathematical fit. Perhaps every aspect does. Perhaps some do not. The matter cannot be settled by mathematics itself, precisely because mathematics's own internal consistency, however complete, says nothing about what exists outside it or how that exterior is organised.
Black holes offer a useful test case, though they need to be handled with some care. It would be a mistake to claim that mathematics simply breaks down inside a black hole, as though the formalism itself failed. What actually happens is that general relativity produces singularities under certain conditions, points at which its equations yield infinite or undefined values, and physicists generally interpret this as evidence that the theory has reached the limit of its own applicability rather than as proof that mathematics as such has ceased to function. The distinction matters. A theory reaching its limit is a claim about that particular formal system's fit with a particular domain. Mathematics failing as such would be a much stronger and much less well-supported claim.
This case allows the more general question to be posed with proper precision. When a mathematical model fails to fit its target domain, at least two possibilities are always on the table. The mathematics being used might simply be the wrong mathematics for the domain, in which case a different formal system might succeed where the first one failed. Alternatively, the domain itself might lack the organisation required for that particular kind of mathematical representation to achieve fit with it, regardless of which formal system is tried. A third possibility, more radical than either of the first two, cannot be ruled out in advance: some aspect of a domain might resist mathematical representation as such, not because the right mathematics has not yet been found, but because the aspect in question is not the kind of thing any nonrecursive formal system could fit. Whether this third possibility is ever actually realised is an empirical question, not one that can be settled from the armchair, and this article does not claim to know the answer for any specific domain.
A final methodological point closes this section, and it matters more than it might initially appear to. Cosmic fit, as this article has defined it, has to remain falsifiable if it is to do any genuine explanatory work. If every apparent failure of mathematical fit is attributed, without exception, to the mathematics not yet being sophisticated enough, then the claim that the cosmos is fundamentally mathematical becomes immune to any conceivable evidence against it, and an unfalsifiable claim of this kind explains nothing, however impressive it sounds. The account developed in this article deliberately keeps a different possibility open: that some domains, or some aspects of some domains, may possess an organisation to which no mathematical representation, however refined, could ever achieve more than partial or approximate fit. Leaving that possibility open is not a weakness in the argument. It is what keeps the argument answerable to evidence rather than immune to it.
10. Beyond Platonism Versus Invention
The classic dispute in the philosophy of mathematics, over whether mathematical objects are discovered or invented, can now be addressed directly, and the framework developed above suggests that the dispute has persisted for so long partly because it compresses three distinct questions into a single binary choice.
The first question asks where mathematical symbolic systems come from in the first place. The answer this article's framework supplies is that they are multisymbolically constructed through living, recursive coordination among mathematicians: someone proposes a definition, someone else extends it, a community argues over which axioms are worth adopting, and a formal system gradually takes shape through this thoroughly recursive process of proposal, criticism and revision.
The second question asks why the consequences that follow from a given mathematical system feel discovered rather than decided, even though the system itself was made. The answer is that once a set of formal commitments has been fixed, whatever follows from those commitments is nonnegotiably constrained by them, in a way no subsequent act of preference or persuasion can alter. This is why proving a theorem feels like finding something out rather than making something up, even though the axioms from which the theorem follows were themselves a human construction.
The third question asks why mathematics, once developed, so often turns out to describe aspects of external reality with such precision. The answer this article has defended at length is that this happens only where empirical cosmic fit actually obtains between the mathematical system in question and the organisation of the domain it is applied to, and that no amount of internal mathematical rigour can guarantee, in advance, that such fit will be found.
Compressed together, these three answers yield a formula considerably more precise than the traditional invention-versus-discovery binary allows: mathematical systems are poietically made, in that they are deliberately constructed and deposited into durable symbolic form through recursive coordination among their makers. Their internal entailments are formally discovered, in that whatever follows from a fixed set of commitments is found rather than decided once those commitments are in place. Their external applicability is empirically discovered, in that whether and how well they fit any given aspect of the cosmos cannot be settled from within mathematics itself but has to be established by testing the formal system against the domain in question.
This tripartite formula does more useful work than the binary it replaces, because it separates three genuinely different questions that the older debate had allowed to collapse into one another, and it explains why partisans on both sides of the traditional dispute have always been able to marshal real evidence for their position. Both sides were right about different parts of a single, more complicated process, and wrong only in mistaking their part for the whole.
11. Wigner Reversed
It is now possible to return to the puzzle with which this article began and show how the argument developed since then dissolves it. The effectiveness of mathematics ceases, on this account, to be unreasonable in Wigner's sense. What would actually be unreasonable is the opposite expectation: that mathematics should be equally effective everywhere, regardless of what any particular domain of the cosmos happens to be like.
The original puzzle asked why mathematics describes physical reality so well. The revised version this article proposes asks a different and more tractable question: why do some aspects of the cosmos possess an organisation with which purely nonrecursive multisymbolic systems achieve extraordinary fit? Posed this way, the question is no longer primarily about mathematics at all. It is a question about the cosmos, about which of its aspects happen to be structured in the nonrecursive, differentiable, often discrete or continuous, often invariant or symmetric ways that mathematical representation requires.
And the question, posed this way, need not receive a single answer across every domain. Physics discovers one particular set of mathematical affordances, rooted in the overwhelmingly nonrecursive character of the relations it studies. Chemistry discovers a related but distinct set, built around the nonrecursive behaviour of atomic and molecular interaction. Biology discovers a more mixed set, mathematically tractable in its nonrecursive aspects while remaining constitutively recursive in others. Statistical science obtains its own distinctive access to recursive domains, not by mathematising recursion directly but through the symbolic deposition described earlier in this article, converting recursive events into nonrecursive data before any mathematics is applied to them at all.
The cosmos, on this account, need not be written in the language of mathematics in the sense Galileo originally proposed. Some of its organisation simply permits mathematical mapping, and that permission is unevenly distributed, tracking not some single scalar property like complexity or size but the specific, aspect-by-aspect distribution of nonrecursivity across the different kinds of things the cosmos contains. Where that permission is granted, mathematics performs at a level that can look, from the inside of the discipline, like a kind of miracle. Where it is withheld, no refinement of mathematical technique, however patient or however ingenious, will ever fully close the gap, because the gap in question was never a gap in the mathematics to begin with.
12. Conclusion: Mathematics Does Not Describe Reality; It Describes What Fits
The argument of this article can now be drawn together in compressed form. Mathematics is purely multisymbolic, an elaboration of the human capacity to deposit distinctions into durable form that persists across absence, delay, distance and death. It is purely nonrecursive, in that nothing mathematical alters its course in response to being engaged, measured or investigated, a claim that holds categorically and admits no exceptions once formal recursion has been properly distinguished from the recursivity this article has been concerned with throughout. It is internally constrained, in that whatever follows from a fixed set of definitions and axioms is found rather than decided once those commitments are in place. And it is externally contingent in its applicability, in that whether and how well it fits any given aspect of the cosmos depends on empirical cosmic fit that cannot be guaranteed by mathematics's own internal consistency, however complete that consistency may be.
Its empirical success, wherever it is found, therefore depends on a correspondence between mathematics's formal structures and nonrecursive aspects, relations or symbolic deposits somewhere in the cosmos. The extraordinary effectiveness mathematics achieves in physics, and in many other domains besides, establishes not that reality is, at bottom, mathematical, but that large and consequential aspects of reality afford extraordinarily strong mathematical fit. Other aspects may afford weaker fit, partial fit, or no fit at all, and recursive living process introduces a categorical boundary that no increase in mathematical sophistication can cross: mathematics can represent the nonrecursive properties and deposits of a recursive being with all its usual precision, but it does not thereby capture that being's recursivity itself.
The success of mathematisation has repeatedly encouraged an inference from mathematical tractability to ontological completeness, the assumption that whatever can be mathematised exhausts whatever is really there. The argument developed here reverses that inference. Wherever mathematics works, its success should prompt a question not about the universal authority of mathematics as such, but about the particular organisation of the domain that made such success possible. Wherever mathematics fails, the corresponding question is not automatically what better mathematics might be required. It is also whether the phenomenon in question possesses the kind of organisation to which any mathematical representation could ever achieve fit. The deepest question is therefore not why mathematics describes the cosmos. It is what kind of cosmos allows mathematics.