Abstract
This article reconstructs Thomas Kuhn’s theory of scientific revolutions through the distinction between cosmic fit, mesocosmic fit, and intersymbolic fit. Kuhn’s central achievements, that normal science depends on stabilized paradigms, that anomalies acquire significance only against a paradigm, and that revolutionary change reconstructs a field rather than merely extending it, are retained and sharpened rather than discarded. His categories of fact, data, observation, theory, and paradigm are shown to be too coarse to capture the heterogeneity of the symbolic systems through which scientific realities become available. A paradigm is redescribed as an L4 stabilization that renders heterogeneous L3 articulations sufficiently coherent for a community to recurse upon them as one intelligible world, and a scientific revolution as the large-scale reconstruction of intersymbolic fit this stabilization requires once it can no longer hold. Dark matter, traced from Zwicky’s 1933 discrepancy through rotation curves, gravitational lensing, and the cosmic microwave background to contemporary particle searches, supplies the central historical case, and displays nearly every stage a revolution requires without settling whether it has completed one.
PART ONE: RECONSTRUCTING KUHN
I. What Exactly Changes in a Scientific Revolution?
When people say that science has undergone a revolution, what exactly has changed? The cosmos has not changed. The gravitational relations that prompted Fritz Zwicky’s calculations were operating long before he calculated anything. Galaxies were rotating before their rotation curves were plotted. Light was being gravitationally lensed before lensing maps existed. The cosmic microwave background existed before it became a map, a power spectrum, or evidence bearing on cosmological parameters.
Nor does a scientific revolution necessarily begin with entirely new empirical phenomena. The same experimental articulation can survive intact across a theoretical transformation that reorganizes everything around it. Measurements can remain. Equations can sometimes remain unaltered in their mathematical form. Instruments remain the same pieces of apparatus. What changes is that a discrepancy long tolerated within an existing framework can suddenly acquire a radically different significance. Two further observations sharpen the puzzle before Kuhn’s own account is reconstructed properly. What changes in a revolution is not simply added to what came before, since calling something revolutionary rather than merely additive implies that large portions of an existing framework are reorganized rather than extended. And the change is not experienced by scientists as a change of opinion about an unchanged world, but as something closer to a change in what the world itself contains, an experience this article will eventually have to explain rather than merely note.
Thomas Kuhn grasped this better than any cumulative account of science that came before him. A scientific revolution, on his account, changes the conditions under which scientific problems, observations, classifications, and explanations hang together, and it does so in a way ordinary talk of steady progress cannot capture.
But hang together needs a theory of its own, and this is exactly where Kuhn’s account, for all its insight, runs out of resources. If science consists of many symbolic systems coordinating with one another rather than one language facing one world, then perhaps what changes in a scientific revolution is not primarily the relation between theory and nature. Perhaps it is the pattern of fit among the symbolic systems science has built.
State the article’s destination without yet fully justifying it. A scientific revolution is an intersymbolic reconstruction. That formulation should sound strange on first encounter. The rest of this article exists to earn it.
II. What Kuhn Saw
Present Kuhn sympathetically before turning to what he lacked, because the critique that follows only has force against an account taken seriously on its own terms.
Kuhn does not present paradigms as fully articulated theories waiting to be applied. He describes them as accepted models or exemplary solutions that earn a community’s allegiance through demonstrated problem-solving success, and that then become the object of further, increasingly detailed articulation. Normal science is the sustained working-out of a paradigm’s promise rather than a search for fundamental novelty. Redescribed in the vocabulary of recursive stabilization, an L4 stabilization becomes secure enough to function as ground for further L3 recursion. The paradigm does not terminate questioning. It makes a highly specific range of further questions askable, and closes off others, which is exactly what allows research to proceed with the depth Kuhn thought characterized mature science.
Kuhn’s famous description of normal science as puzzle-solving deserves to be taken seriously rather than caricatured as mere conservatism. A stable paradigm tells a scientific community what matters, what can be measured, which instruments are relevant, what counts as an admissible problem, which mathematical operations are appropriate, and what a successful result would even look like. Kuhn is emphatic that these restrictions are not merely tolerated costs of normal science. They are what permits investigation of nature at a depth no unconstrained inquiry could reach, because unconstrained inquiry has to keep re-deciding, on every occasion, questions a paradigm has already settled. In the vocabulary developed here, L4 closure is what permits L3 proliferation, and this is one of the article’s central agreements with Kuhn rather than a preliminary to disagreement.
A third Kuhnian insight matters just as much. Every paradigm, on Kuhn’s account, lives continuously with anomalies, and scientists working within a paradigm cannot stop to reconsider its foundations every time some observation resists easy accommodation. Most anomalies are absorbed, set aside, or quietly tolerated. Only certain anomalies become consequential enough to produce what Kuhn calls a crisis. Kuhn recognizes this distinction as central to his account and recognizes, with characteristic honesty, that he lacks a fully general theory of what makes one mismatch disposable and another revolutionary. This is precisely where intersymbolic fit enters, and closing the gap Kuhn leaves open is the task of the sections that follow.
None of this yet requires abandoning Kuhn’s own vocabulary. It requires only noticing that fact, data, observation, theory, and paradigm, the very terms Kuhn relies on to state his insight, are themselves symbolic categories of a particular, and, as the next section argues, insufficiently differentiated, kind.
III. Where Kuhn’s Vocabulary Becomes Too Coarse
The critique can now become precise, and it should not be mistaken for the more familiar charge that Kuhn believed in naive, paradigm-free observation. He did not. Kuhn repeatedly insists that paradigms shape what counts as a datum in the first place, which instruments are considered appropriate, and how raw observation gets interpreted before it ever reaches the page as a result. Nobody who has read him carefully could accuse him of theory-free empiricism.
The problem lies elsewhere. Kuhn continues to work with epistemological categories that are aggregated well beyond what his own insight should permit: fact, observation, data, experiment, theory, paradigm, nature. Each of these names is treated as though it picked out one kind of thing, and this concealment of internal heterogeneity is precisely what a properly differentiated theory of symbolic fit exists to undo.
Dark matter makes the point concrete immediately. There is no single symbolic substance that could honestly be called dark matter data. There are optical images, spectra, redshift measurements, radio-frequency measurements, inferred velocities, rotation curves, gravitational mathematics, gravitational-lensing reconstructions, X-ray maps, cosmic microwave background temperature and polarization maps, angular power spectra, statistical likelihoods and parameter estimates, numerical simulations, particle-detector event distributions, and exclusion curves. These are not interchangeable instances of one thing called data but differently constituted symbolic products, each preserving certain distinctions with precision while losing others, each transformable into the next only at a cost a flat appeal to data conceals rather than reveals.
This changes the philosophical problem Kuhn thought he was solving. A scientist does not simply compare the data against the theory, as though a single symbolic register on one side answered to a single symbolic register on the other. Scientists continually coordinate heterogeneous symbolic articulations with one another, and it is this coordination, not a dyadic comparison, that does the epistemic work.
The Kuhnian dyad, paradigm on one side and nature on the other, should therefore be replaced with a considerably more heterogeneous architecture. A cosmic process becomes available only through an instrumentally mediated interaction, which is rendered through one symbolic system, which is transformed into a second, and then a third, while those articulations are independently compared with further symbolic systems generated through entirely different instrumental and mathematical routes. The problem of scientific knowledge becomes relational rather than dyadic the moment this architecture is taken seriously, and everything Kuhn wanted to explain about paradigms, anomalies, and revolutions has to be reconstructed within it rather than alongside it.
PART TWO: THE THREE FITS AND SCIENTIFIC KNOWLEDGE
IV. Cosmic Fit, Mesocosmic Fit, and Intersymbolic Fit
Three distinctions do most of the work in what follows, and each deserves a precise statement before the historical argument can proceed.
Cosmic fit concerns how successfully a symbolic articulation tracks realities and relations in the cosmos, independently of whether those realities happen to be organized at scales available to ordinary human coordination. It should not be defined merely as fit with distant astronomy, since a sofa, a protein, a galaxy, and a quantum field are all, in the relevant sense, equally part of the cosmos. The distinction cosmic fit marks is aspectual rather than spatial. It asks whether an articulation answers adequately to reality beyond whatever convenience or organization human mesocosmic coordination happens to impose upon it, a question that can be asked of the very small and the very distant, but just as legitimately of a fact close at hand.
Mesocosmic fit concerns fit within the recursively inhabitable human-scale world sustained through five mediations, multisensorial embodiment, being-with, multimaterial forming, multiversal dwelling, and multisymbolism. Mesocosmic fit can succeed or fail differently across these five mediations rather than as a single undifferentiated verdict, and it possesses, for that reason, an internal coherence of its own. Scientific concepts need mesocosmic fit precisely insofar as they must become intelligible and recurseable for embodied human beings working, teaching, and arguing together, rather than existing only as marks on a page a machine could process without anyone understanding them.
Intersymbolic fit concerns relations among symbolic systems themselves. Can a spectroscopic articulation be translated into a velocity. Can a velocity be transformed into a rotation curve. Can a rotation curve be made compatible with a gravitational model. Can a lensing reconstruction and a cosmic microwave background parameter estimate be made compatible with the same underlying ontology. Intersymbolic fit is the relation that has to hold, and continually be tested, for any of these translations to succeed, and it possesses an internal coherence of its own in the fullest sense, since intersymbolic fit just is a relation among symbolic systems, and the question of how well several such systems cohere with, translate into, or contradict one another is the question intersymbolic fit exists to answer.
The central insight the three-part distinction makes available is this. Science does not merely test symbols against reality in one dyadic operation. It tests symbols against other symbols, through independently constituted routes that remain, throughout, answerable to reality rather than sealed off from it. What convergence among those routes licenses is greater confidence in their shared cosmic fit. The convergence itself is intersymbolic fit operating across several pathways at once, not a separate cosmic species of coherence standing beside the other two. Cosmic fit itself has no internal structure to parallel mesocosmic and intersymbolic fit. It is the answerability of an articulation to the cosmos, not a relation among articulations, and keeping this distinction sharp is what allows the framework that follows to do real theoretical work rather than merely redescribing Kuhn in new vocabulary.
The three fits, finally, can diverge from one another, and this divergence is what makes the framework necessary rather than ornamental. Increasing cosmic fit at L3 can reduce mesocosmic intelligibility, trading resemblance to lived experience for a precision only a specialized notation can supply. Increasing precision within one L3 symbolic system can damage intersymbolic fit with an established L4, generating exactly the local misfit later sections will need to distinguish from genuine crisis. An L4 theory can possess extraordinary mesocosmic and intersymbolic coherence for generations while eventually proving cosmically inadequate, tracking a reality it never actually answered to as well as its institutional stability suggested. Scientific change, on the reconstruction this article proposes, occurs through movement among these partially independent dimensions rather than through simple accumulation along a single axis of improving accuracy.
V. Levels of Recursivity: Why L3 Can Outrun L4
Bring the levels of recursivity explicitly into the argument, since the asymmetry between L3 and L4 is what ultimately explains why scientific revolutions take the shape they do.
L3 scientific symbolization permits extraordinarily differentiated articulation. Equations can add variables without limit. Spectroscopy can distinguish wavelengths no human eye could ever discriminate. Statistics can discriminate relations entirely invisible to unaided observation, extracting a signal from what looks, to any embodied observer, like pure noise. Numerical models can coordinate quantities across spatial and temporal scales no single human life, or human institution, could otherwise hold together. Here is the crucial point. Specialized L3 symbolisms can increase cosmic fit precisely by decreasing their dependence on mesocosmic intelligibility, trading the requirement of resembling lived experience for a precision the cosmos, at scales mesocosmic coordination cannot itself register, actually demands.
L4 performs a different operation entirely, and this is where the asymmetry begins to bite. L4 stabilizes classifications, ontologies, generalized entities, and, crucially, accounts of what kind of world the proliferating L3 articulations actually disclose. It does not merely aggregate L3 results. It answers a question L3 specialization was never built to answer on its own, namely what all this differentiated symbolization is actually about.
One useful hypothesis, offered here as deliberately falsifiable rather than as an established principle, is that L4 recursive stabilization may be disproportionately, perhaps uniquely, afforded by ordinary language rather than by specialized notation, because ordinary language alone emerged through, and remains saturated by, mesocosmic human existence in a way purpose-built formalisms do not. This article does not need to settle that question to make use of it. It needs only the asymmetry the hypothesis, if correct, would explain.
The asymmetry is this. L3 can become extraordinarily non-mesocosmic, travelling as far from embodied intelligibility as a research programme requires. L4 must nonetheless render the result recurseable, available to a community of embodied scientists who need to argue about it, teach it, and build further L3 work on top of it. Scientific ontology therefore repeatedly returns to mesocosmically derived terms even after its underlying mathematics has left mesocosmic intelligibility far behind. Matter, particle, field, wave, halo, structure, force, information, and code are none of them native to the formalisms that use them. Each is borrowed from categories embodiment, being-with, dwelling, multimateriality, and multisymbolism jointly make available to ordinary language long before any of them is pressed into technical service.
Dark matter supplies the cleanest example available anywhere in contemporary science, because it shows the entire sequence, from initial discrepancy through L4 naming to full ontological thickening, in unusually legible steps. The remainder of this article follows that sequence in detail.
PART THREE: DARK MATTER AS A HISTORY OF INTERSYMBOLIC STABILIZATION
VI. 1933: Zwicky Discovers a Discrepancy, Not a Substance
Begin the historical reconstruction with Fritz Zwicky’s 1933 study of the Coma cluster, a paper conventionally credited as the first evidence for what Zwicky himself called dunkle Materie.
Resist narrating this, as textbooks so often do, as Zwicky discovered dark matter. What actually happened is better reconstructed through the symbolic relations involved rather than through the vocabulary of discovery. One symbolic chain concerns the observed velocities of galaxies within the cluster, obtained through spectroscopic measurement. A second concerns estimates of visible luminosity and the mass conventionally associated with it. A third is purely mathematical, relating the dynamics of a gravitationally bound system to its total mass through the virial theorem. Zwicky combined these three chains and found that they did not cohere. The mass required to bind the cluster gravitationally, as the virial calculation demanded given the observed velocities, vastly exceeded the mass the luminosity estimate could account for.
That incoherence, not any perceived object, is the actual event of 1933. The first discovery, properly described, is an intersymbolic discrepancy, three independently generated symbolic articulations, velocities, luminosities, and a mathematical relation binding them, refusing to fit together under the ontology then available.
Ordinary language then supplied a remarkably economical stabilization. Something is there. It is matter, because matter is the mesocosmic category ordinary language reaches for whenever a quantity of stuff needs positing. It does not shine, because nothing in the discrepancy required it to. Dark matter, in other words, names an L4 solution to a problem of fit among symbolic articulations before anything about its own ontology was independently established, and this is the critical philosophical point the rest of this article depends on. The entity was introduced to stabilize a mismatch, not encountered as an object and then named.
Kuhn’s category of anomaly, for all its explanatory power, cannot adequately specify what happened here. An anomaly, on Kuhn’s account, is a mismatch between paradigm and nature, but Zwicky’s discrepancy was not a mismatch between one paradigm and one nature. It was a failure of fit among three independently generated symbolic articulations, resolved by the L4 introduction of a fourth term built to absorb exactly the misfit the other three had generated between them. This is a more precise description of what actually occurred than Kuhn’s own vocabulary can supply, and the difference matters more with every subsequent stage of the story.
VII. Rotation Curves: When the Discrepancy Becomes Systematic
Move now to Vera Rubin and Kent Ford, whose 1970 spectroscopic study of the Andromeda galaxy, M31, marks an important step in the observational history of galactic rotation. Subsequent rotation-curve work across a growing number of spiral galaxies strengthened the finding considerably. Orbital velocities at large radii did not decline the way they should have if the visible matter alone dominated a galaxy’s gravitational mass.
The theoretically interesting development is not the familiar textbook narrative in which better observations simply revealed more dark matter, as though the entity had been sitting there waiting to be seen more clearly. What actually happened is that an entirely new set of symbolic and material mediations reproduced a structurally similar discrepancy to Zwicky’s, through routes independent of his. Optical spectroscopy produces wavelength shifts. These are symbolically transformed into velocities. Velocities become functions of radial distance from a galaxy’s centre. These functions become rotation curves. Rotation curves are compared with mass distributions inferred from the distribution of luminous matter. Once again, and now across dozens of galaxies rather than one cluster, the symbolic systems fail to fit one another without positing an additional gravitational component.
A new L4 term becomes important at this stage, halo. Dark matter, previously a comparatively unlocalized stabilization, becomes spatially specific, a dark matter halo surrounding a visible galaxy in three dimensions. This deserves to be analysed as a mesocosmic return rather than treated as an incidental detail of scientific vocabulary. What began as a non-mesocosmic relation, a discrepancy among velocities, luminosities, and a mathematical theorem, becomes a thing, surrounding another thing, occupying space, possessing an inferred distribution and a mass.
That mesocosmic stabilization is not an embarrassment for the science, whatever a purist theory of representation might want to say about it. It is precisely what makes further recursive scientific inquiry possible. Once dark matter has a halo, scientists can ask how large halos typically are, how dense they are at different radii, how they form over cosmic time, how they interact with one another during galaxy mergers, and how their presence shapes the visible galaxies embedded within them. None of these questions were askable of a bare discrepancy among three numbers. All of them become askable the moment L4 stabilization renders the discrepancy as a mesocosmically intelligible thing occupying a place.
VIII. From Anomaly to Convergence: Dark Matter Becomes Robust
This section is the pivot of the historical reconstruction, because it is where dark matter stops being merely a repeatedly confirmed discrepancy and becomes something considerably stronger.
Dark matter did not become increasingly compelling merely because more rotation curves accumulated, though they did accumulate, and their accumulation mattered in its own right. What mattered more was the entry of new symbolic pathways, generated through instruments, mathematics, and assumptions substantially independent of galactic rotation.
Gravitational lensing supplied the first of these. Lensing allows a gravitating mass distribution to be reconstructed from the way it distorts the light of more distant background objects, a route that depends on general relativity and on background source populations rather than on stellar or gas velocities at all. The Bullet Cluster became particularly influential within this pathway because its weak-lensing mass reconstruction was spatially displaced from the cluster’s dominant X-ray-emitting baryonic gas, showing that whatever was doing the gravitating had separated, during the cluster collision, from the ordinary matter that had collided and slowed. This is evidence for an unseen gravitating component generated through a chain almost entirely unlike the rotation-curve chain that had generated the earlier evidence.
The cosmic microwave background supplied a further, independent pathway again. Temperature and polarization measurements are transformed into an angular power spectrum, and the angular power spectrum is transformed, through a substantial apparatus of cosmological modelling, into estimates of the universe’s fundamental parameters. The Planck collaboration’s final analyses found strong consistency with the standard six-parameter flat cosmological model and tightly constrained the density of cold dark matter within that model, a result obtained through instruments, physics, and assumptions that share almost nothing with either rotation curves or gravitational lensing.
Large-scale structure surveys and cosmological simulation supplied a fourth, again largely independent, pathway. Simulations specify initial conditions and physical rules and generate predicted distributions of matter and structure that can then be compared statistically against what astronomical surveys actually observe, a route that tests dark matter’s consequences for the universe’s largest visible architecture rather than for any single galaxy or cluster.
State the central epistemological proposition this convergence licenses. Dark matter becomes robust through intersymbolic convergence, and it matters considerably to be precise about what kind of robustness this is. Repetition within a single symbolic pathway, more rotation curves of the same general kind, genuinely increases precision and reliability, and nothing here denies that. But convergence across substantially independent symbolic pathways provides a further and distinct kind of warrant, because what survives that convergence has survived routes that could, in principle, have diverged and did not. Rotation curves, lensing, the cosmic microwave background, and large-scale structure were generated through different instruments, different underlying physics, different mathematical assumptions, and different spatial and temporal scales, and yet they converge on a compatible cosmic picture. This is why scientific robustness frequently depends on relations among symbolic articulations rather than upon the sheer quantity of isolated results within any one of them. Call this intersymbolic triangulation, and treat it, from this point in the argument onward, as the primary reason dark matter earned the extraordinary confidence with which most of contemporary astrophysics now treats it.
PART FOUR: THE KUHNIAN PROBLEM
IX. Is Dark Matter Actually a Paradigm Shift?
This section should be the most surprising in the article, and it earns that status by refusing an answer the argument so far might seem to have already supplied.
Ask the question directly. If dark matter emerged from a sequence of accumulating anomalies stretching from Zwicky through rotation curves to lensing and the cosmic microwave background, does its acceptance therefore count as a textbook Kuhnian scientific revolution? The answer is not straightforwardly yes, and getting this wrong would cost the article something important, because if every well-evidenced scientific advance counted as a revolution, this reconstruction would merely redescribe Kuhn in new vocabulary rather than sharpening his distinctions.
Consider, first, how much dark matter initially preserved rather than destroyed. Zwicky’s discrepancy did not force anyone to abandon Newtonian or Einsteinian gravitational theory. Quite the opposite. If existing gravitational dynamics are retained exactly as they stood, an additional quantity of gravitating mass reconciles the symbolic systems that had failed to fit, and nothing else in the surrounding architecture needs to move. Dark matter was, at this stage, an unusually conservative L4 stabilization. The anomaly was absorbed by expanding the ontology rather than by revising the gravitational relation the anomaly seemed to threaten.
Modified gravity reveals the alternative structure this conservatism concealed, and it does so with unusual clarity. Mordehai Milgrom’s 1983 proposal made the underlying choice explicit by refusing dark matter’s route entirely. Rather than add an unseen mass, Milgrom proposed modifying the dynamics themselves at the very low accelerations characteristic of galactic outskirts. The same broad family of discrepancies could now participate in two radically different L4 stabilizations. One preserves the established dynamics and adds an ontology, an unseen population of matter, to reconcile them with observation. The other preserves the visible ontology and revises the dynamics instead. This is precisely an intersymbolic choice rather than a straightforward empirical verdict, because the discrepant measurements themselves carry no label instructing a physicist which of the two repairs to prefer.
This forces a distinction the article needs from here onward, one Kuhn’s own vocabulary elides. Three scales of scientific change should be kept separate rather than treated as points on a single continuum. Local accommodation adjusts parameters or auxiliary assumptions while leaving an existing L4 organization entirely intact, the kind of routine repair every functioning paradigm absorbs continuously. L4 restabilization introduces or substantially revises particular entities or classifications, dark matter itself is an instance, while preserving much of the larger intersymbolic architecture surrounding it, gravitational theory, the general practice of inferring mass from dynamics, the broader observational apparatus of astronomy. Scientific revolution, reserved for a considerably rarer and more demanding event, reconstructs the relations among several load-bearing symbolic systems at once, such that a different L4 ontology reorganizes what problems, entities, measurements, and explanations mean across an entire field rather than within one corner of it.
Dark matter’s history, read this way, includes major episodes of L4 restabilization without its cumulative acceptance obviously amounting to one completed Kuhnian revolution. Whether it eventually will remains an open question. That openness is not a weakness in the reconstruction offered here. It is exactly the theoretical payoff the distinction between accommodation, restabilization, and revolution was built to deliver, and later sections return to it directly.
X. From “Dark” Matter to “What Is the Matter?”
Examine now the peculiar current structure of the dark matter problem, without announcing a revolution that has not, in fact, occurred.
The astronomical and cosmological convergence around an unseen gravitating component remains extraordinarily strong, arguably stronger than at any earlier point in the story told above. But one further, and rather different, translation the ontology seemed to promise remains stubbornly incomplete, the translation from dark matter, as an astronomical inference, to a population of particles physics could detect directly in a laboratory.
If matter is taken sufficiently literally, as the ordinary-language term chosen at Zwicky’s stage always implied it should be, direct laboratory detection ought in principle to supply another independently generated symbolic pathway, a further route of the same general kind as lensing or the cosmic microwave background, converging on the same underlying ontology from an entirely different direction. Decades of increasingly sensitive experiments have pursued exactly this translation. The most sensitive of them to date, the LUX-ZEPLIN experiment operating nearly a mile underground in South Dakota, reported in 2025 a search covering hundreds of live days and several tonne-years of exposure that set world-leading limits on weakly interacting massive particles without finding any candidate signal, and an extended analysis released later that year pushed the excluded mass range still lower without altering the basic result. Each successive search narrows the available parameter space for the leading particle candidate. None has yet delivered the signal the astronomical convergence would, on the most literal reading of matter, seem to promise.
This does not refute dark matter, and nothing in the argument here should be read as suggesting that it does. What it weakens is one particular expected intersymbolic translation, the link between an astronomically inferred gravitating component and a directly detectable weakly interacting particle, rather than the astronomical inference itself. This distinction is vital to the theory of scientific change this article is building. One symbolic connection within a larger architecture can fail, repeatedly and increasingly stringently, while the larger L4 stabilization around it remains entirely intact, because the astronomical convergence traced in the previous section does not depend on the particle-detection pathway succeeding.
This gives a considerably more precise account of scientific tension than the bare observation that a paradigm has anomalies. It licenses two questions this article’s remaining sections are built to answer properly. At what point would repeated failures of this kind become load-bearing enough to reopen the ordinary-language noun matter itself, rather than merely narrowing which particle candidates remain viable. And what further symbolic systems, generated through what further independent routes, would have to fail to converge before the existing L4 stabilization around dark matter became unsustainable rather than merely incomplete in one of its expected translations. Those two questions prepare the general theory of scientific crisis the next part of the article develops.
PART FIVE: REDEFINING KUHN’S CORE TERMS
XI. What Is an Anomaly?
Kuhn observes that an anomaly becomes visible only against the background of a paradigm precise enough to make deviations detectable in the first place, a paradox he treats correctly, the more exact a paradigm’s predictions, the more conspicuous its failures become. That insight should be retained in full. What needs redefinition is the phenomenon itself.
An anomaly is not simply a mismatch between nature and theory, as though two homogeneous terms had failed to correspond. It is, more precisely, a failure of expected fit among symbolic articulations under an existing L4 stabilization. Dark matter supplies three examples of exactly this structure, each already examined above in its historical setting. Visible mass fails to fit the dynamical mass inferred through the virial theorem, which is Zwicky’s original anomaly. Rotation curves fail to fit the mass distribution expected from visible matter under established gravitational dynamics, which is Rubin and Ford’s anomaly and its many successors. Particle-detection expectations fail, so far, to produce the translation anticipated between an astronomically inferred gravitating component and a directly detectable particle, which is the anomaly examined in the previous section.
None of these are mismatches between one paradigm and one nature. Each is a specific, locatable failure of fit between two or more symbolic articulations that an existing L4 stabilization had led a community to expect would cohere. This changes what the important diagnostic question actually is. It is not how many anomalies a field currently tolerates, a question that treats anomalies as interchangeable units to be counted and compared against some threshold. It is where, within the intersymbolic architecture holding a field together, a given misfit actually occurs, since a peripheral misfit and a load-bearing one can look, counted as bare anomalies, identical, while functioning, as the next section argues, in entirely different ways.
XII. What Is a Scientific Crisis?
Kuhn observes that anomalies become crises only under certain conditions, and that a crisis, once under way, blurs an existing paradigm and loosens the rules that had governed normal research. He is right about the phenomenon and short on the mechanism. The theory of symbolic fit can specify the missing condition with considerably more precision than Kuhn’s own vocabulary allows.
A crisis begins when an intersymbolic failure becomes load-bearing rather than local. Several features distinguish load-bearing misfit from the kind every functioning paradigm absorbs as a matter of course. The misfit can no longer be localized to one symbolic connection while everything around it remains stable. Repairing it within one symbolic system generates new problems within others, so that local accommodation, the mildest of the three scales of change distinguished above, is no longer available as a strategy. Multiple, previously independent symbolic pathways become difficult to stabilize simultaneously, so that a repair adequate to one pathway actively worsens the field’s fit with another. And the ordinary-language L4 categories that had rendered the whole field recurseable, the very terms a community had stopped needing to question, themselves become recursively questionable again, forcing exactly the kind of foundational reopening normal science exists to make unnecessary.
This licenses a distinction considerably more powerful than a simple count of anomalies could ever supply, the distinction between anomaly count and anomaly topology. Ten thousand small discrepancies scattered across peripheral locations within a field’s symbolic architecture may not matter at all, however impressive their number looks when counted. One discrepancy that happens to connect several central, load-bearing symbolic systems can destabilize an entire field on its own, however modest it looks as a single data point. Crisis, on this reconstruction, is not a property of anomalies considered individually or even in aggregate. It is a network property of intersymbolic fit, a fact about where a misfit sits within a field’s architecture of dependencies rather than a fact about how large or how numerous the misfit happens to be.
XIII. What Is a Paradigm?
Offer, at this point, the article’s formal reconstruction of Kuhn’s central term. A scientific paradigm is an L4 stabilization that establishes sufficient intersymbolic fit among heterogeneous L3 scientific articulations for a community to recurse upon them as descriptions of one sufficiently coherent world.
A paradigm, on this reconstruction, does not merely contain a set of propositions a community happens to believe. It organizes translations among observations, mathematics, instrumental procedures, classifications, images, statistical conventions, predictive models, and the ordinary-language ontologies through which all of the above eventually become teachable, arguable, and recurseable for embodied human beings. Dark matter’s L4 stabilization, reconstructed across the sections above, organizes exactly this range, spectroscopic velocities, virial mathematics, rotation curves, lensing reconstructions, cosmic microwave background parameter estimates, numerical simulations, and the ordinary-language noun matter that renders the entire apparatus intelligible to anyone outside the specialist community that produced it.
This reconstruction explains one of Kuhn’s own claims that has often struck readers as elusive, his insistence that a paradigm is simultaneously conceptual, methodological, exemplary, and practical, as though a single word were being asked to do implausibly many jobs at once. On the account offered here, those are not miscellaneous components loosely bundled under one label. They are different symbolic and mediational systems, conceptual articulation, methodological procedure, exemplary demonstration, and practical instrumentation among them, held together by a sufficiently stable L4 organization. A paradigm feels unified, to those working within it, precisely because its L4 stabilization has succeeded in making heterogeneous symbolic systems answer to one another well enough that the heterogeneity itself becomes invisible, the same way any successful intersymbolic achievement tends to disappear from view once it no longer requires conscious coordination to sustain.
XIV. What Is Normal Science?
Normal science can now be reconstructed positively rather than merely defended against the charge of conservatism. It is the productive recursive cycle in which an L4 stabilization enables constrained L3 differentiation, constrained L3 differentiation increases cosmic articulation within the bounds the stabilization has set, increased cosmic articulation is continually tested against the existing intersymbolic architecture for fit, and local L4 adjustment repairs whatever small misfits that testing turns up, without ever requiring the larger stabilization itself to be reopened.
This modifies Kuhn without rejecting a single one of his central claims about normal science. The entire point of L4 stabilization is to let scientists stop recursively reopening foundational questions on every occasion a new result comes in. A cosmologist interpreting a fresh lensing observation does not need to ask, from first principles, whether gravity exists, whether general relativity is approximately correct, or whether mass can be inferred from light’s deflection at all. A stable L4 has already provided those grounds, and providing them is not an incidental convenience. It is the condition under which the kind of depth Kuhn attributed to mature, paradigm-governed science becomes achievable at all.
Normal science, understood this way, is not the opposite of creativity, whatever the word’s more casual usage might suggest. It is what recursively stabilized grounds actually make possible, the depth of inquiry available only once a community has stopped needing to settle its foundations anew before every experiment, and has been freed, by that settlement, to differentiate at L3 with a precision no perpetually re-litigated foundation could ever support.
XV. What Is a Scientific Revolution?
Now give the argument’s full definition, built from every distinction the preceding sections have assembled.
A scientific revolution occurs through a recognizable sequence rather than a single dramatic event. An existing L4 stabilization has, over some period, enabled extensive L3 differentiation, exactly as normal science is supposed to. The specialized L3 symbolic systems this differentiation produces improve their cosmic fit and proliferate increasingly fine distinctions, precisely because L4 closure has freed them to do so. At some point, some of these increasingly fine articulations cease fitting comfortably with other established symbolic systems within the same field, generating the kind of localized anomaly examined above. Local repairs initially preserve the existing L4 intact, absorbing the misfit through accommodation rather than through any deeper reorganization. As differentiation continues, however, the misfit becomes increasingly distributed and load-bearing rather than remaining confined to one symbolic connection, crossing the threshold into crisis as that term was reconstructed above. The existing L4 ontology, previously taken for granted as the ground on which normal science proceeded, becomes recursively reopenable, its own adequacy now an object of the kind of L5 reflection normal science had made unnecessary. Alternative L4 stabilizations, of the kind MOND represented for dark matter, compete to reorganize the symbolic ecology the crisis has destabilized. And eventually, if the field does not simply fracture, one alternative achieves sufficient intersymbolic fit across enough of the field’s load-bearing symbolic systems to provide new stable ground for a fresh round of normal science.
The revolution, on this reconstruction, occurs primarily at L4, since it is the L4 stabilization that gets replaced. But it is driven throughout by relations among L3 systems, since it is the accumulating, increasingly load-bearing misfit among specialized articulations that forces the L4 replacement in the first place. Neither level, taken alone, supplies the whole event.
This licenses the central reversal of Kuhn this article has been building toward from its opening question. A scientific revolution is not primarily the moment at which science achieves better cosmic fit. A great deal of that improvement may already have occurred incrementally, at L3, well before any revolution became visible, exactly as the rotation-curve and lensing evidence for dark matter accumulated steadily within an L4 stabilization that never, at any point in that accumulation, needed replacing. Revolution is instead the reconstruction required when an existing L4 can no longer stabilize what increasingly cosmically successful L3 systems have made available, when the ontology that once rendered a field’s heterogeneous symbolizations mutually intelligible can no longer do the job its own earlier success has made necessary.
Hence the formulation this article exists to establish. A scientific revolution is an event of intersymbolic restabilization, precipitated by cosmic success at L3 that has outrun what the field’s existing L4 architecture can hold together, and resolved only when a new L4 stabilization achieves, across enough of that architecture, the coherence the old one has lost.
PART SIX: WHAT THIS DOES TO KUHN’S STRANGEST CLAIMS
XVI. Incommensurability without Separate Worlds
Kuhn’s most notorious claim holds that competing paradigms can be incommensurable, and that scientists working after a revolution can appear, in some important sense, to inhabit a different world from their predecessors, even while the world beyond their research has not literally changed at all. This claim has attracted more hostile commentary than almost anything else in his work, and much of the hostility is deserved if the claim is read as asserting that paradigms seal scientists into genuinely separate realities.
The theory of symbolic fit gives the same observation a considerably less mysterious reconstruction. Different paradigms establish different patterns of intersymbolic fit among the same underlying pool of instruments, mathematics, and observational technique. They can classify an identical instrumental output differently, connect the same equation to different underlying ontologies, assign different significance to the same visual trace on a detector’s readout, treat as signal what a rival paradigm treats as mere noise, and translate the same technical symbol into different ordinary-language entities entirely. None of this requires two sealed conceptual universes floating past one another without contact.
Incommensurability, reconstructed this way, means that translation between two L4 stabilizations is incomplete, not impossible, because each stabilization organizes the relations among its field’s L3 symbolizations differently, and translation has to cross that organizational difference rather than merely swap one vocabulary for a synonymous other. This retains everything genuinely important in Kuhn’s insight, that paradigm change is not simply the accumulation of new propositions within an unchanged conceptual scheme, without requiring the strongest and least defensible version of his different-worlds rhetoric, the suggestion that two scientific communities working on the same reality might, in some literal sense, no longer share one.
XVII. Why Scientific Revolutions Feel Like World Changes
Return, briefly, to mesocosmic fit to explain a further feature of revolutionary experience Kuhn described vividly without fully accounting for.
A paradigm does more than coordinate formal symbolisms with one another. At L4, it makes an entire scientific world inhabitable, populated with things, processes, causes, places, boundaries, materials, forces, organisms, particles, and fields that a working scientist can think, argue, and act with rather than merely calculate about. That ontology, however technical its formal apparatus, is overwhelmingly expressed through ordinary language, for exactly the reasons the earlier discussion of L3 and L4 established.
When an L4 stabilization changes, scientists therefore do not merely recalculate quantities within an unchanged conceptual furniture. Things become different kinds of things. A quantity once treated as noise becomes signal. An entity once treated as fundamental becomes derivative, or vanishes from the ontology altogether, or is joined by a new entity nobody had needed to posit before. The revolutionary experience Kuhn insisted upon, however overstated its different-worlds formulation, is partly the collapse and reconstruction of mesocosmic intelligibility itself, the felt sense that the kinds of things one’s field is even about have shifted beneath one’s working vocabulary.
This explains, more precisely than Kuhn’s own account manages, why a new paradigm can feel like a change in the world while the cosmos, as the article’s opening insisted, remains entirely unchanged throughout. What changes is not the cosmos. It is the mesocosmically inhabitable ontology through which a scientific community had learned to recurse upon the cosmos at all, and that change, for anyone who has to live and work inside it, is not a small one.
PART SEVEN: DARK MATTER REVISITED
XVIII. The History of Dark Matter as a Recursive Sequence
Summarize the dark matter case now, not chronologically as the historical sections above proceeded, but theoretically, as a single recursive sequence the preceding argument has assembled piece by piece.
The sequence begins with discrepancy. Different L3 articulations fail to fit one another, as they did for Zwicky when dynamical mass, inferred through the virial theorem from observed velocities, came out far larger than luminous mass could account for. A provisional L4 stabilization follows directly. The word dark matter turns a bare discrepancy into an entity, giving the mismatch a name before its ontology is independently established. New L3 differentiation then proliferates around that provisional stabilization, exactly as normal science should produce, rotation curves, radio observation of neutral hydrogen, gravitational lensing reconstructions, cosmic microwave background analysis, and cosmological simulation each contributing further, increasingly differentiated articulation. Intersymbolic convergence follows as these independently generated pathways become mutually stabilizable through the same underlying entity, exactly the process examined at length above under the name intersymbolic triangulation. Ontological thickening comes next, as dark matter stops being merely unseen matter in general and becomes, specifically, cold and predominantly non-baryonic matter embedded within the standard cosmological model, a stage for which the Planck collaboration’s final cosmological parameter analysis provides a particularly strong instance, constraining the density of this thickened ontology to remarkable precision. Institutional sedimentation follows the thickening, dark matter becomes textbook ontology, simulation input, direct experimental target, a funding category in its own right, and ordinary vocabulary among scientists who will never personally inspect any of the evidence behind it. Residual misfit then appears, as the previous section detailed at length, since direct particle detection has not yet delivered the translation the ontology’s own literal name seemed to promise, and modified gravity remains conceptually available as an alternative L4 stabilization nobody has been forced to abandon. And a final, so far only possible, stage of recursive reopening becomes conceivable at the sequence’s current edge, where the operative question shifts from which dark matter particle explains the astronomical convergence toward the more foundational question of what exactly licenses the category matter here at all.
This final stage requires one qualification stated as plainly as the argument can manage. It does not mean that a scientific revolution is currently underway, still less that one is inevitable. It defines the conditions under which one might eventually become possible, nothing more and nothing less, and the difference between those two claims is precisely what the distinction between accommodation, restabilization, and revolution, established earlier in this article, exists to preserve.
XIX. Scientific Revolutions as the Reconstruction of Fit
Return, in closing, to the question this article opened with. What exactly changes in a scientific revolution?
Not the cosmos, as the opening pages already insisted. Not necessarily the instruments, which can survive a revolution largely intact. Not necessarily all the observations, many of which remain exactly as recorded before and after. Not even necessarily all the equations, some of which can migrate from one L4 ontology into another with their mathematical form barely altered. What changes is the architecture through which heterogeneous symbolic articulations are held together as one recursively inhabitable scientific world.
Kuhn was right that revolutions are non-cumulative in an important sense that ordinary talk of scientific progress obscures. He was right that anomalies exist only against a background of stabilized expectations precise enough to make deviation detectable. He was right that normal science depends on closure, and that the depth of inquiry closure makes possible is not a cost paradigms impose but a benefit they confer. He was right that crises loosen an existing framework’s rules, and that new paradigms reconstruct scientific practice rather than simply appending new propositions to an unchanged one.
But paradigm-nature fit remains too coarse a formulation, because science does not possess one symbolic relation running from one paradigm to one nature. Modern science works through an extraordinary and irreducible heterogeneity of symbolic systems instead. A radio spectrum is not a lensing map. A lensing map is not an equation. An equation is not a numerical simulation. A simulation is not a detector’s event distribution. A detector’s event distribution is not the ordinary-language claim that dark matter exists. The scientific achievement examined throughout this article has consisted, in substantial part, precisely in making these radically heterogeneous articulations fit one another well enough to be treated as evidence for a single, shared entity, a fact this article has now traced through one field’s actual history in detail.
The argument’s final proposition can be stated compactly, gathering every distinction the article has built along the way. Normal science is what stable intersymbolic fit makes possible. An anomaly is a localized failure of expected intersymbolic fit. A crisis occurs when that failure becomes load-bearing across a field’s symbolic architecture rather than remaining confined to one connection within it. A paradigm is an L4 stabilization that renders heterogeneous L3 articulations recursively coherent for an entire scientific community. And a scientific revolution occurs when that stabilization must itself be replaced, because further differentiation at L3 has outrun what the existing L4 architecture can any longer be made to hold together.
Dark matter is the ideal case for this reconstruction because it displays nearly every stage the argument requires, and stops, tantalizingly, just short of the last. It began as a discrepancy. The discrepancy became a name. The name became an entity. The entity coordinated an extraordinary proliferation of independent symbolic systems, each converging on it through routes the others did not share. Those systems, together, have greatly increased the whole apparatus’s cosmic fit. And yet the apparently simplest mesocosmic question the entire edifice was built to answer remains, at the time of writing, unanswered. What is the matter.
That unresolved question is the right place to close, because it captures the entire argument in miniature. The history of dark matter shows that science can know, with extraordinary and well-earned confidence, that some existing symbolic articulations require something, while remaining uncertain what kind of thing that something is. A scientific revolution begins at the point where the answer that once stabilized the question can no longer do so.