Few words in the methodology of social science carry more weight with less agreement about their meaning than microfoundations, and few debates show the cost of that disagreement better than the argument over whether Marxism needs them. A reader who meets the demand in economics or sociology usually meets it as a single requirement, stated with the confidence of a standard: an explanation of a social phenomenon is incomplete until it has been shown how the actions of individuals produce that phenomenon. Pressed against Marxism from the late 1970s onward, the demand helped to found a school, analytical Marxism, provoked some of the sharpest methodological writing the tradition has produced, and left behind a residue of slogans on both sides. Defenders of the classical tradition heard an alien standard imported from economics and refused it wholesale. Its proponents heard the refusal as an admission that Marxist explanation could not survive scrutiny. Both hearings rested on the same mistake, and the mistake is the subject of this article.

Two demand ledgers open on a seminar table beside annotated volumes on method, Insight Crunch

The mistake is that two different demands travel under the one name. One demand insists that explanations lay out a working process, the sequence through which an alleged cause brings about its alleged effect, instead of resting on correlation or on the observation that somebody benefits. The other insists that the process be followed down until it ends with individuals, with what they believe and what they want, so that persons and their reasons form the floor of every finished explanation. The first of these is reasonable, general across the sciences, and was pressed against Marxism with genuine effect; accepting it improved the tradition’s explanations. The second is methodological individualism, a substantive commitment about what exists and what can explain, wearing the clothes of a neutral standard of rigour. Separating the two demands, which this article calls the separable demands, dissolves most of the famous quarrel and relocates the remainder where it belongs: in a substantive dispute about the priority of preferences over the relations that form them. Neither side performed the separation consistently, which is why the argument lasted so long and settled so little.

Does Marxism need microfoundations?

it needs mechanisms, meaning an account of how a cause produces an effect, and accepting that requirement improved the tradition. It does not need reduction to individual preferences, which is a substantive theoretical commitment rather than a methodological standard, and which Marxism may reject on argument.

The canonical statement of the question is therefore double. The microfoundations debate concerns whether explanations of social phenomena must specify the individual actions through which those phenomena are produced, and whether such specification requires reduction to individual preferences or admits structural starting points. A structural starting point takes a position within a set of relations, the wage labourer’s position or the capitalist’s, and explains conduct from the options that position makes available and the penalties it attaches to each option. Whether that is a legitimate resting place for explanation, or merely a postponement of the individual level, is the live question underneath the settled one. The pages that follow take the demands in turn: first the separation itself and the table that orders it, then the case for the demand at full strength, then the case against reduction, then the collective action problem that any honest treatment must face without flinching, and finally the resolution the separation makes available and the influence the whole dispute exerted far beyond the tradition that hosted it.

The microfoundations demand and its two meanings

Few terms in social theory carry quite so much quiet confusion as microfoundations. A reader encountering the word in debates about Marxism might reasonably assume a single, settled requirement hiding behind it, something like a foundation stone that either exists beneath an explanation or does not. Closer inspection dissolves that comfortable picture. Two distinct expectations have travelled together under one label, sharing prestige, sharing advocates, and sharing a polemical history, while asking for quite different things. Pulling them apart changes what counts as meeting the challenge, what counts as evading it, and what a Marxist explanation owes to critics who speak in its name.

The first expectation concerns process. An explanation, on this reading, should show how an alleged cause brings about an alleged effect, step by connected step, rather than resting content with correlation, coexistence, or alleged necessity. A claim that some institution persists because a social system requires it invites a further question about transmission: through whose decisions, under which constraints, amid what incentives, with what unintended consequences, did persistence actually happen? Answering that further question means describing a mechanism, a sequence or configuration of activities linking starting conditions to outcomes in a manner open to scrutiny. Nothing in the request itself dictates the vocabulary of the final link in the chain. Mechanisms might run through organisations, through legal rules, through market pressures, through habits reproduced in families and schools, or through combinations of such levels woven together.

Quite another expectation concerns termination. On this second reading, a satisfactory account must end its explanatory regress with individuals, with their beliefs, preferences, capacities, and situations, treated as the ultimate building blocks from which larger patterns are composed. Social entities may appear along the way, yet they enjoy no independent explanatory standing at the bottom. Classes, states, firms, and cultures must in principle be unpacked into facts about persons, and any residue left unpacked marks incompleteness. Philosophers call the underlying doctrine methodological individualism, a commitment about the proper form of social explanation rather than a mere plea for detail. Someone can honour process while rejecting termination, and someone else can demand termination while offering only sketchy processes. Conflating the pair therefore misdescribes both agreement and disagreement.

Keeping that distinction visible from the outset matters because the history that follows otherwise looks paradoxical. Marxists who welcomed sharper causal storytelling sometimes found themselves accused of resisting rigour altogether, merely because they questioned whether every legitimate story must bottom out in individual psychology. Critics who had performed a genuine service by exposing lazy functional reasoning sometimes slid, almost imperceptibly, into treating a controversial ontology as though it were simply good method. Sympathetic readers inherited the slide. Textbooks compressed decades of argument into a slogan about foundations, and the slogan did argumentative work that careful distinctions would have blocked. Recovering the separation restores room for a more discriminating verdict: one demand largely vindicated, absorbed, and put to productive use; the other demand still contestable, still requiring defence on its own merits, still open to principled refusal.

What does it mean to specify a mechanism?

Specifying a mechanism means describing the connected steps through which an initial condition produces an outcome, naming the actors, constraints, interactions, and feedback effects involved, so that each link can be examined, tested against evidence, and compared with rival accounts of the same outcome rather than merely asserted as necessary.

Elaboration helps because mechanism talk itself can become a slogan, repeated without content. A mechanism is not a decorative middle paragraph inserted between cause and effect, nor a list of intermediate variables measured in a dataset. Genuine specification involves identifying entities engaged in activities, organised in particular ways, such that their operation regularly or plausibly generates the phenomenon under investigation. Sometimes the relevant entities are persons deliberating inside committee rooms. At other moments they are bureaucracies applying rules, creditors enforcing contracts, editors selecting stories, or algorithms ranking information for millions of users. Activities include persuading, coercing, imitating, investing, exiting, voting, striking, hiring, and learning, each conditioned by resources and expectations that themselves call for description. Organisation matters as much as inventory: a crowd, a hierarchy, a network, and a market can contain similar persons yet produce divergent results because relations among participants differ.

Several virtues follow when specification succeeds. Hidden assumptions surface, because a writer forced to narrate transmission must decide who learns what, who bears costs, who can block change, and who benefits from delay. Scope conditions become visible, since a process that operates under competitive product markets may stall under monopoly, and a process relying on credible threats may collapse when enforcement weakens. Points of empirical vulnerability multiply in a healthy way, giving historians and social scientists places to test the story against archives, interviews, statistics, and comparisons across cases. Rival mechanisms can be pitted against one another, not as competing labels for the same correlation but as divergent accounts predicting different intermediate traces. Failure also becomes informative: if an expected intermediary step leaves no trace where traces should abound, confidence in the whole account ought to fall.

Generality in mechanism talk deserves separate attention. Some mechanisms recur across settings, such as selection under scarcity, imitation under uncertainty, or escalation under mutual fear. Others remain stubbornly local, tied to a particular legal code, kinship structure, or technological arrangement. Neither kind enjoys automatic superiority. Recurrent patterns offer portability and cumulative learning, while local patterns offer fidelity to cases where portability would distort. A mature explanatory practice moves between levels of abstraction deliberately, stating plainly whether a proposed process is meant as a universal template, a family resemblance among cases, or a singular reconstruction of one episode. Confusion about intended generality has fuelled many sterile disputes, with critics demanding universal templates from historians who promised only singular reconstruction, and historians dismissing templates that sociologists valued for comparative purposes.

Mechanisms also differ in temporal shape. Certain processes unfold as chains, one event triggering another in sequence. Others operate as simultaneous constraints, holding behaviour within bounds without any single triggering moment. Feedback loops complicate both shapes, as outcomes alter the conditions that produced them, strengthening or undermining the originating process over time. Threshold effects introduce nonlinearity, where gradual accumulation yields sudden transformation once a critical point passes. Path dependence adds memory, so early contingencies lock in later options even after original reasons fade. A rich Marxist historiography has long dealt in such shapes, even when authors lacked the later vocabulary, describing how enclosures reshaped labour markets, how factory discipline reshaped household life, and how fiscal crises reshaped state capacities. Vocabulary arrived late; causal narration itself has older roots within the tradition.

Who pressed the mechanism demand against Marxism?

Jon Elster pressed the demand most forcefully, arguing that Marxist functional explanations lacked credible intervening processes, while other analytical Marxists joined related criticisms from different angles, and defenders such as G. A. Cohen answered by clarifying consequence explanations and the evidence that could support them without supplying detailed mechanisms.

Understanding the force of that intervention requires returning to functional explanation and its attractions. Marxist writing had often explained institutions, beliefs, and policies by pointing to their beneficial consequences for capital accumulation, class domination, or system reproduction. Legal forms endured because they secured property relations. Educational arrangements persisted because they supplied disciplined labour. Ideologies flourished because they legitimated inequality. Such claims captured something real about patterned outcomes in capitalist societies, where beneficial consequences for powerful actors rarely count as coincidence. Yet pattern alone does not explain persistence. Beneficial effects might be accidental, might be noticed by nobody, might be outweighed by costs borne elsewhere, or might depend on fragile conditions that a serious account should identify. Treating benefit as explanation risks a teleological shortcut, as though societies possessed purposes and quietly arranged their parts to serve them.

Biology offered a tempting analogy that ultimately misled when imported carelessly. Traits in organisms can be explained by consequences through natural selection, because variation, inheritance, and differential reproduction supply a known feedback process connecting benefit to prevalence. Hearts exist in a lineage partly because pumping blood aided ancestors, and selection translates that aid into proliferation across generations. Social life contains selection processes too, in markets, in electoral competition, in organisational ecology, but each alleged instance requires demonstration rather than assumption. A firm practice spreading because profitable firms expand differs from a law persisting because judges anticipate promotion, and both differ from a belief spreading because media owners filter dissent. Invoking selection in general terms merely relocates the unanswered question. Which variant spreads, through what channel, at whose expense, against which resistance?

G. A. Cohen mounted the most influential defence of functional reasoning inside Marxism, most fully in his 1978 book Karl Marx’s Theory of History. Cohen argued that consequence explanations could be legitimate even before the underlying process was known, provided evidence supported a robust connection between an item and its effects. His treatment drew careful distinctions among types of functional claim, insisted on precision about what was being explained, and resisted crude caricatures of historical materialism as prophecy or metaphysics. Readers wanting a fuller reconstruction can consult the discussion of ga cohen functional explanation, which traces his argument in detail. Cohen’s position had a plausible analogue in sciences where consequence laws were accepted ahead of detailed process knowledge, Darwin himself lacking genetics while defending selection. On that basis, demanding a finished mechanism before granting any explanatory standing seemed excessively strict.

Elster answered that the analogy concealed more than it revealed. In his writings on Marx and on social explanation more broadly, surveyed in the profile of jon elster marxism, he pressed for intervening processes with unusual clarity and persistence. Where Cohen saw a legitimate explanatory form awaiting elaboration, Elster saw a promissory note too often treated as payment. Without a feedback loop, an intentional agent sustaining the item because of its effects, or some other identifiable transmission route, the claim that an institution exists because of what it does remained, in his judgement, incomplete to the point of failure. He catalogued alternative routes that could redeem particular cases: deliberate design by beneficiaries, reinforcement through rewards and punishments, competition eliminating rivals, socialisation shaping dispositions, and various unintended aggregation effects. Each route transformed a functional slogan into a testable narrative. Each route also exposed cases where no plausible route existed and the functional claim collapsed.

Statements cast in the idiom of systemic need illustrate the problem starkly. Capitalism requires a reserve army of labour, therefore unemployment persists. Capitalism requires expanding consumption, therefore advertising saturates culture. Capitalism requires legitimacy, therefore schools teach deference. Verbs such as requires smuggle agency into an abstraction, while the inference from requirement to existence skips every actor who might fail, resist, miscalculate, or pursue competing goals. Sometimes the alleged requirement is itself dubious, since capitalist economies have operated under widely varying unemployment levels, consumption patterns, and legitimation strategies. Sometimes the requirement holds loosely yet multiple incompatible arrangements could satisfy it, leaving the actual arrangement unexplained. Sometimes beneficiaries lack the coordination, information, or power needed to install what would help them collectively, a familiar collective action difficulty. In every such variant, need talk postpones explanation precisely where explanation should begin.

Accepting the mechanism demand improved Marxist scholarship in tangible ways. Historians became warier of inferring intention from outcome, distinguishing between effects that actors engineered, effects they welcomed after the fact, and effects nobody foresaw. Theorists of the state learned to specify channels through which business influence travels: lobbying expenditures, revolving doors, structural dependence on private investment, party finance, media ownership, expert networks, and anticipated capital flight, each channel carrying different evidence and different limits. Theorists of ideology moved toward accounts of production and reception, asking who funds messages, who crafts them, which audiences encounter them, how prior beliefs filter uptake, and why some messages resonate while others fall flat. Theorists of class formation examined workplaces, neighbourhoods, unions, and parties as sites where shared position might, or might not, become shared organisation. In each domain, questions grew sharper and answers grew more vulnerable to evidence, which counts as progress by almost any standard.

Improvement did not mean unanimity about how much mechanism is enough. Cohen’s defenders noted that explanation comes in degrees, that a well supported consequence pattern can guide research productively, and that premature demands for micro level detail can paralyse inquiry into large structures. Elster’s defenders replied that degrees of explanation still require honesty about gaps, and that labelling a gap as a distinct explanatory form risks dignifying ignorance. Between those poles, a working consensus gradually formed in much analytical writing: functional claims may serve as hypotheses, as summaries of patterns, or as placeholders, but they earn full explanatory credentials only when a sustaining process can be sketched with enough specificity to be checked. Even scholars who rejected Elster’s broader philosophy often absorbed that narrower lesson. The tradition emerged from the exchange with better habits, a point worth granting without reservation before any criticism of his further commitments begins.

A further refinement concerns intentional explanation and its limits. Many sustaining processes involve actors who do intend the beneficial effect: employers designing incentive schemes, governments crafting industrial policy, parties building electoral coalitions. Intentions, however, explain only when beliefs attributed to actors are plausible, information channels exist, and capacities match ambitions. Powerful actors regularly misunderstand their own interests, pursue short term gains that undermine long term stability, and sabotage one another through rivalry. Working class actors likewise act on partial knowledge, divided loyalties, and organisational constraints. Mechanism specification disciplines intentional attribution by forcing the analyst to model information, incentives, and capacities together rather than reading minds from outcomes. Unintended consequences then cease to be an embarrassing residue and become a central category, covering market crashes, arms races, bureaucratic drift, and cultural backlash, all generated by interacting actions that no participant chose as a package.

Selection mechanisms received equally useful discipline. Competition among firms can select for cost cutting practices, but only under stated conditions regarding entry, exit, finance, and information. Competition among political parties can select for policies attractive to pivotal voters, but only where electoral rules, party discipline, and media environments take particular forms. Competition among ideas can select for memorable slogans rather than accurate beliefs, depending on attention economies and institutional gatekeeping. Specifying conditions converts selection from metaphor into model. Marxist political economy, with its long attention to competition as a coercive force acting behind the backs of individual capitalists, was well placed to appreciate that conversion once the vocabulary was clarified. Indeed, some of the strongest mechanism writing in the Marxist tradition concerns precisely such impersonal compulsion: firms compelled to innovate or perish, workers compelled to sell labour power under propertylessness, states compelled to court investment under fiscal dependence. Compulsion language itself still needs unpacking, yet the underlying insight, that structured pressure can generate regular outcomes without conspiracy, survived the analytical critique and emerged sharper.

Why were the two demands so easily confused?

Confusion arose because both demands arrived in the same books, from overlapping critics, during the same controversies, sharing a vocabulary of rigour and science, so resistance to individualist reduction sounded like resistance to careful explanation, and acceptance of mechanisms appeared to commit an author to individualist foundations as well.

Historical packaging explains a large share of the muddle. Analytical Marxism emerged in the late nineteen seventies and nineteen eighties as a self conscious movement importing tools from analytical philosophy, rational choice theory, game theory, and neoclassical economics into Marxist questions. Its leading figures shared seminars, journals, and an annual gathering, and their books addressed functional explanation, exploitation, class, history, and justice in a recognisably common idiom. A reader meeting the mechanism critique inside a volume that also championed individualist modelling naturally experienced the pair as one programme. Bibliography reinforced psychology: citations clustered, opponents clustered in reply, and conference panels treated the package as a package. Intellectual movements often cohere socially before they cohere logically, and later readers inherit the social grouping as though it were a logical entailment.

Vocabulary deepened the entanglement. Words such as rigorous, scientific, clear, and precise attached themselves to the whole programme, while words such as vague, metaphysical, holistic, and Hegelian attached themselves to resistance in any form. Once praise and blame distribute along a single axis, distinctions along other axes become hard to hear. A scholar objecting that classes possess causal relevance not reducible to member properties could be heard as defending obscurity, even when the objection came packaged with meticulous archival evidence. A scholar urging detailed process tracing could be heard as endorsing rational choice modelling, even when the processes traced were cultural, organisational, or ecological. Tone amplified the effect. Polemical writing rewards sharp binaries, and the binary between modern rigour and traditional murk left little room for a position that welcomed one innovation while declining another.

Defenders of older Marxist styles contributed to the blur from the opposite side. Some treated any request for mechanisms as a hostile import, a bourgeois demand alien to dialectical thinking, and answered detailed criticism with invocations of totality that never descended to particulars. Such responses confirmed critics in their suspicion that holism functioned as a protective device against falsification. When the most visible resistance to reduction also displayed resistance to specification, observers could be forgiven for treating the two resistances as one. Forgiveness has limits, however, because careful defenders existed as well, scholars who supplied richly detailed histories while maintaining that structural concepts earned their keep. Their existence proves separability in practice, whatever polemics suggested. Practice nevertheless speaks quietly beside manifestos, and the manifestos set the terms in which many newcomers learned the debate.

Institutional incentives pushed in the same direction. Economics enjoyed high prestige among policy makers and among philosophers seeking formal models of social life, and its dominant explanatory style centred individual choice under constraints. Borrowing prestige by borrowing method offered tangible rewards: publications in generalist journals, invitations across disciplines, and a reputation for hardness in a field often accused of softness. Graduate training followed prestige, so cohorts learned individualist modelling as the default meaning of mechanism itself. Within such training, a mechanism almost by definition meant a choice theoretic story, and alternatives registered as incomplete sketches awaiting translation. Translation sometimes illuminated, yet sometimes it merely recoded organisational or cultural processes into preference language without adding evidence, while the recoding itself counted as theoretical progress. Prestige can masquerade as argument when a community stops noticing the substitution.

Philosophical ambiguity in the word foundation did further damage. Foundations can mean evidential support, causal underpinning, conceptual primitiveness, or ontological dependence, and arguments sliding among those senses gain illicit momentum. Showing that individual actions causally underpin a strike, in the trivial sense that strikes involve acting persons, does not show that strike explanations must terminate in individual properties. Showing that concepts such as class require definition partly through relations among persons does not show that class level patterns lack explanatory autonomy. Showing that evidence about behaviour often arrives at person level does not show that higher level variables mislead. Each slide converts a modest truth into a strong doctrine. Much confusion in the literature consists of exactly such slides, repeated until they feel like common sense. Naming the senses separately, whenever foundations are invoked, remains a cheap and effective prophylactic.

Generational dynamics also played a part. Younger scholars defining themselves against teachers often adopt the full toolkit of whichever school promises emancipation from inherited vagueness. Older scholars defending life work often reject the full toolkit to protect genuine insights buried inside unfashionable prose. Both reactions favour packages over distinctions. Publishing formats reinforce the tendency, since review essays and textbook chapters compress positions into camps with memorable labels. Analytical Marxism became one camp, its critics another, and internal variation faded from summaries even though primary texts contained plenty of it. Elster himself criticised aspects of rational choice theory, Cohen defended consequence explanation against Elster, John Roemer modelled exploitation with general equilibrium tools while others in the same circle doubted such tools, and still the camp label smoothed everything flat. A history attentive to internal friction resists the smoothing, and resistance matters for the argument developed here because separability was visible inside the camp from the beginning.

Language about levels added one more layer. Micro and macro suggest a spatial picture, small particles below and large structures above, with explanation ideally travelling downward. Spatial metaphors smuggle in assumptions: smaller seems more real, lower seems more basic, bottom seems like bedrock. Alternative pictures yield different intuitions. One might picture explanation as a map at varying scales, each scale revealing patterns invisible at others, none simply unreal. One might picture levels as interlocking gears, where motion at any gear depends on neighbours in several directions. One might picture social reality as fields of relations, with persons themselves constituted by positions they occupy. None of these pictures settles substantive disputes, yet each loosens the grip of the particle image that makes reduction feel inevitable. Scholars who sensed that looseness sometimes expressed it through metaphor alone, which critics easily dismissed, while scholars who formalised reduction gained an aura of literalness. Metaphor auditing, unfashionable as it sounds, belongs in methodological training for precisely this reason.

What does the reduction demand add beyond mechanisms?

Beyond mechanisms, reduction adds a stopping rule about admissible explanatory bedrock, requiring that social processes terminate in individual beliefs, preferences, and properties, thereby excluding irreducible structural, relational, or institutional causes even when a detailed process has already been described in fully intelligible and empirically checkable terms.

The addition deserves precise statement because caricature helps nobody. Methodological individualism, in its stronger forms, holds that explanations of social phenomena are fully satisfactory only when they refer exclusively, at the fundamental level, to individuals, their properties, their relations understood as aggregations, and their actions. Weaker forms treat individual level accounts as an ideal toward which inquiry should strive, allowing higher level shorthand under licence. Stronger forms again distinguish ontology from method: some proponents concede that social entities might exist in some sense while insisting that method should proceed as if only individuals counted. Others assert the ontology directly. Family resemblance unites these variants: a suspicion that structural language hides either confusion or unfinished work, plus a conviction that finishing the work means descending to persons. Steelmanning the position means acknowledging its motivations. Individual level discipline can expose ecological fallacies, prevent reification, clarify aggregation problems, and reveal how macro correlations dissolve or reverse under disaggregation. Those gains are genuine, and any Marxism dismissing them wholesale would impoverish itself.

Aggregation problems in particular gave reduction its strongest evidence. A correlation between regional unemployment and regional voting patterns does not by itself show that unemployed persons vote in the suggested direction; employed persons in depressed regions might drive the pattern, or migration might sort populations in ways that manufacture the correlation. Similar warnings apply to claims about class voting, ethnic mobilisation, and religious affiliation. Simpson type reversals, where a trend visible in aggregate data flips within every subgroup, dramatise the danger. Game theory added another class of examples where individually rational conduct aggregates into collectively perverse outcomes: bank runs, tragedy scenarios in shared resource use, arms build ups, and segregation emerging from mild preferences. Such models taught a generation that macro patterns cannot be read off micro intentions in any simple direction, and that lesson cuts both ways. If intentions do not transparently explain patterns, patterns do not transparently reveal intentions. Both directions require explicit bridging work.

Reification worries likewise carry weight. Social theory abounds in sentences where abstractions acquire verbs of wanting, deciding, and punishing. Capital seeks, the state decides, the market punishes, tradition demands. Sometimes such sentences compress well understood processes into shorthand that specialists can expand on request. Sometimes expansion fails because nobody has done the underlying work, and shorthand conceals a void. Individualist critics performed a service by demanding expansion and refusing to accept compression as depth. Yet expansion need not end at individuals to count as expansion. Expanding state decides into a narrative about cabinet procedures, bureaucratic incentives, judicial review, and party discipline genuinely deepens understanding, even though every element in the narrative remains institutional. Demanding one further descent, into the psychology of each official, may add detail without adding insight, and may subtract insight by scattering attention away from procedures that explain why different officials, with different psychologies, produce similar decisions. Position explains what personality cannot in such cases, a point organisational sociology established long before analytical Marxism arrived.

Relational properties pose a deeper difficulty for termination talk. Being unemployed, being a creditor, being married, being disenfranchised, and being a tenant are properties of persons, grammatically speaking, yet each property essentially involves institutions and other persons arranged in particular ways. An explanation terminating in such properties has not escaped the social; it has smuggled the social in through adjectives. Preferences raise a parallel difficulty, since preferences form inside families, schools, workplaces, media environments, and legal regimes. Treating preferences as bedrock means treating as given precisely what a critical social theory often wants to explain. Elster himself, to his credit, wrote extensively about adaptive preferences, sour grapes reasoning, and preference formation under deprivation, thereby acknowledging that bedrock shifts. Once bedrock shifts under social pressure, the line between foundation and superstructure blurs, and reduction becomes a research programme with porous borders rather than a clean termination rule. Porous borders may be acceptable, but they surrender the strong claim that made reduction distinctive.

Emergence and multiple realisability supply further considerations, familiar from philosophy of science beyond Marxism. A pattern can be stable across wildly different micro realisations, and stating the pattern at its own level can capture generalisations that micro description fragments. Traffic jams, market crashes, and linguistic changes exhibit structure recognisable regardless of which particular drivers, traders, or speakers participate on a given occasion. Temperature in physics offers the classic analogy: mean kinetic energy underlies it, yet thermodynamic generalisations retain autonomy for most explanatory purposes. Analogies do not settle social questions, since social patterns depend on meanings in ways gases do not, but they undermine any blanket assertion that higher level explanation is merely provisional. Provisional status must be argued case by case, discipline by discipline, pattern by pattern. Blanket assertions in either direction, holist or individualist, substitute temperament for inquiry.

The separable demands thesis, which organises everything argued in later sections of this article, can be stated plainly at this point. Two requirements travelled together historically; they remain distinct logically and methodologically. Requirement one, specification of process, states a reasonable standard that Marxist explanation often failed, was challenged to meet, and in large measure learned to meet, to its lasting benefit. Requirement two, termination in individuals, states a substantive theoretical commitment about the architecture of social reality and the proper shape of social science. Substantive commitments require substantive defence, and they remain open to substantive refusal. Marxism, or any research tradition, may reject the second requirement on argument without thereby rejecting rigour, evidence, clarity, or process tracing. Conversely, a tradition may accept individualist termination for some domains while denying it for others, provided the sorting follows reasons rather than fashion. Separability thus protects both critics and defenders from a false bargain in which clarity must be purchased with reduction, or structural insight must be purchased with vagueness. False bargains impoverish every party to them.

Framing the thesis as namable matters for scholarly hygiene. Debates drift when central claims lack handles, because participants attack shifting versions and defend whichever version survives the sharpest objection. A named thesis can be quoted, tested against texts, applied to cases, and revised in public. Students encountering the microfoundations literature for the first time gain a sorting device: for each author, for each argument, ask separately whether process is being demanded and whether termination is being demanded, then evaluate each answer independently. Sorting exercises of that kind often reveal unexpected alliances. A structural Marxist supplying meticulous process narratives may stand closer to Elster on requirement one than a rational choice author whose models posit mechanisms without empirical tracing. A liberal individualist sceptical of rational choice formalism may share more with holistic historians on requirement two than camp labels suggest. Alliances rearranged by sorting are evidence that sorting captures something real.

Opponents of separability deserve a hearing before the thesis settles. One objection holds that mechanisms, properly understood, just are individual level by definition, so separating the demands commits a conceptual error. Replying requires inspecting definitions rather than trading intuitions. Major accounts of mechanisms in philosophy of science define them through entities and activities organised to produce a phenomenon, with no restriction that entities be persons. Cells, organs, firms, ecosystems, and machines all figure as entities in respected mechanism literature. Restricting the term to persons therefore stipulates a narrower usage, and stipulation cannot do argumentative work against broader usage without independent justification. Another objection holds that structural mechanisms inevitably smuggle in unexamined black boxes, whereas individual level stories force completeness. Completeness, however, is relative to a question. A story about neurons is incomplete relative to chemistry, and a story about choices is incomplete relative to neuroscience, yet incompleteness relative to a deeper level does not invalidate explanation at the level where the question lives. Every explanation stops somewhere; the dispute concerns where stopping is principled, and principled stopping can invoke relevance, manipulability, invariance, and evidence, not merely size.

A third objection appeals to parsimony. Individualist accounts, it is said, explain social patterns with fewer kinds of entities, and parsimony favours them. Parsimony arguments require care, because counting kinds depends on description choices, and because a lean ontology paired with baroque bridging assumptions may cost more overall than a richer ontology with simple bridges. Explaining a bureaucratic decision through preferences alone can require elaborate auxiliary hypotheses about belief formation, information costs, and strategic misrepresentation, while a spare institutional account invokes rules and roles directly. Which total package is simpler remains an empirical and theoretical judgement, not a counting exercise settled in advance. Marxist theory, with its emphasis on historically specific institutions, will often find the institutional package leaner in practice. Simplicity in use, rather than simplicity in slogan, should guide the comparison.

The comparison basis rule adds a necessary guardrail to the whole discussion. Explanatory standards differ across disciplines, legitimately, because disciplines ask different questions about different kinds of systems with different evidentiary resources. Importing a requirement from one discipline into another is itself a move requiring justification, not a neutral application of universal reason. Physics envy has distorted social science repeatedly, encouraging imitation of surface features, mathematical dress, reduction to smallest components, while neglecting whether the features responsible for success in physics have analogues in social life. Biology offers a contrasting lesson, since biological explanation thrives at multiple levels, from molecules to ecosystems, with level choice guided by the phenomenon rather than by a standing order to descend. Economics itself contains multitudes: industrial organisation studies firms as actors, economic history studies institutions across centuries, behavioural research studies heuristics that violate canonical choice axioms. Citing economics as though it spoke with one individualist voice misrepresents the cited discipline as well as the receiving one.

Justification for importation can take several forms, each assessable on its merits. An importer might show that the target discipline suffers from a specific failure that the imported standard demonstrably repairs elsewhere. An importer might show that systems studied by both disciplines share relevant causal structure, making transfer plausible. An importer might argue from general philosophy of science that a standard binds all empirical inquiry regardless of subject matter. Each argumentative route carries burdens. Repair elsewhere does not guarantee repair here, shared structure must be specified rather than gestured at, and universal claims must survive counterexamples from successful sciences that violate them. Geology explains mountain formation without terminating in particle physics, epidemiology explains transmission patterns without terminating in biochemistry for every purpose, and linguistics explains sound change without terminating in acoustics alone. Successful practice across the sciences displays plural standards coordinated by problem type. Social theory asking for similar pluralism is not asking for exemption from rigour; it is asking for rigour of the kind that fits its problems.

Disciplinary difference also reflects difference in intervention. Fields oriented toward engineering control prize manipulable micro variables, because intervention operates there. Fields oriented toward historical understanding prize narrative fidelity across long durations, because their questions concern how a particular configuration arose and why alternatives vanished. Fields oriented toward critique prize exposure of interests and ideologies embedded in categories themselves, because their questions concern whose perspective a description encodes. Marxism spans all three orientations, sometimes uneasily, which helps explain why methodological debates inside Marxism grow so heated: participants are often optimising for different purposes while using identical vocabulary. Clarifying purpose before adjudicating method reduces heat without reducing substance. A mechanism adequate for policy intervention may be inadequate for historical understanding, and a narrative adequate for understanding may be useless for intervention, with neither failure impugning the other.

Consider three concrete domains where separability does visible work. Theories of capitalist crisis invite functional shortcuts, profits restored because crisis clears excess capacity, yet mechanism specification forces attention to bankruptcy procedures, credit networks, state rescue decisions, and labour resistance shaping which capacities actually clear. None of that specification dictates whether crisis theory must terminate in firm managers’ expectations or may legitimately operate with value categories at system level. Theories of revolution invite heroic shortcuts, classes rising because history assigns them a mission, yet specification forces attention to organisation, communication, repression, defection within armed forces, and food supply logistics. Detailed tracing of that kind leaves open whether revolutionary agency should be modelled as aggregated choices by separate persons or treated as emergent collective capacity built through organisation itself. Theories of everyday consent invite manipulation shortcuts, beliefs installed because domination requires them, yet specification forces attention to schooling, workplace experience, family conversation, entertainment industries, and lived contradictions that sometimes reinforce and sometimes corrode dominant messages. In each domain, the first demand transforms practice while the second remains an open theoretical wager. Observing the transformation and the openness together is exactly what separability predicts.

Historical materialism’s core claims provide a larger testing ground. The proposition that productive forces and relations shape legal, political, and ideological forms has been read functionally, causally, and as a research heuristic, with very different fortunes under each reading. Read functionally without process, the proposition invites the objections Elster catalogued. Read as a directive to investigate how property regimes channel innovation, how fiscal needs shape state building, how labour processes shape skills and solidarities, the proposition generates mechanism rich research across economic history. Whether resulting accounts must end with personal attributes at their base remains a further question, separable in every investigation. A historian might trace how enclosure altered village power through court records, rent rolls, and riot prosecutions, producing an account dense with process and sparse with individual psychology, and the account’s quality would turn on evidence and coherence rather than on its level of termination. Insisting otherwise would discard some of the finest historical writing in the Marxist tradition on a priori grounds, a cost that reduction’s defenders rarely acknowledge openly.

Class analysis sharpens the same point. Classes in Marxist theory are defined relationally, by position within production relations, not by income brackets, cultural tastes, or self identification alone, although all of those may correlate with position. Relational definition already resists simple aggregation, because a relation cannot be decomposed into properties of relata without loss, just as siblinghood resists analysis as a trait belonging to one child alone. Mechanisms linking position to action run through workplaces where cooperation and conflict are daily experiences, through labour markets where insecurity disciplines choices, through organisations that interpret interests and coordinate strategy, and through states that define the legal terrain on which classes meet. Specifying those mechanisms satisfies the first demand in full. Insisting that class disappear at the explanatory bottom, replaced by a distribution of personal attributes, would not add specification; it would change the theory’s subject. Critics are entitled to argue that the subject should change, but the argument must be made as theory against theory, with evidence about which framework illuminates more, not as method against sloppiness.

Rational choice Marxism illustrates both the gains and the costs of accepting the second demand voluntarily. Roemer and others rebuilt exploitation theory using models of property distribution and strategic interaction, yielding precise results about how asset inequality generates class like outcomes even without labour market imperfections. Precision came at a price: historical process receded, power inside production received thinner treatment, and normative conclusions shifted as modelling assumptions did the heavy lifting. Admirers celebrate the clarity; detractors mourn the thinning; both reactions concern a substantive theoretical wager with substantive payoffs and losses. Treating the wager as mere methodological hygiene misdescribes what admirers gained and what detractors lost. Other analytical Marxists chose different wagers, Erik Olin Wright building class typologies for empirical sociology, Cohen defending historical materialism through conceptual analysis, others pursuing game theoretic models of revolution or bargaining. Diversity inside the movement, again, demonstrates that shared commitment to clarity never entailed a single foundation. The movement at its best practised separability without naming it; naming it protects the practice against later simplification.

Attention to language completes the groundwork laid in this section. Demands, foundations, bases, levels, reduction, and mechanism each carry multiple senses, and authors slide among senses under argumentative pressure, sometimes innocently, sometimes strategically. A reader equipped to pause at each term, asking which sense is operative and whether evidence offered matches that sense, will find many celebrated refutations losing their target and many celebrated defences losing their shield. What survives the pausing is more valuable than what dissolves: a genuine challenge about process that transformed a tradition for the better, and a genuine dispute about theoretical architecture that remains open for argument. Later sections take up the dispute in detail, examining candidate structural mechanisms, assessing individualist models case by case, and asking what a Marxism that learned the first lesson while refusing the second looks like in practice. Separation is the precondition for that examination, because without separation every case becomes a referendum on rigour itself, and referendums of that kind reward slogans over evidence. With separation in hand, evaluation can proceed piecemeal, domain by domain, mechanism by mechanism, which is how explanatory traditions actually improve.

One further caution belongs at the foundation of the discussion, concerning the politics of method. Methodological positions acquire political colouring through association: individualism coded as liberal, holism coded as authoritarian, mechanism coded as technocratic, structural talk coded as radical. Colouring by association has a shabby history on every side. Liberal societies have produced structural sociology of great power, Marxist scholars have produced individual level models of great subtlety, and authoritarian regimes have sponsored whatever method flattered their administrators. Political evaluation of a theory should rest on its claims, its evidence, and its practical entanglements, not on the company its method keeps at conferences. Separability helps here too, since it breaks the chain of associations at its weakest link, the assumption that a package must be swallowed or rejected whole. Readers who break that assumption gain freedom to assemble positions argument by argument, which is a modest freedom with large consequences for intellectual honesty.

The stakes extend beyond Marxism, although Marxism supplies the central case. Feminist theory, critical race theory, postcolonial studies, and ecological political economy all face versions of the same paired demand: specify processes, certainly, and then a further insistence that legitimate processes be cast in some favoured vocabulary, whether individual choice, discourse, or network interaction. Each field benefits from the first insistence and deserves space to contest the second on evidence. A general lesson about explanatory pluralism emerges from the particular history traced here, namely that demands arriving together need not stand or fall together, and that intellectual progress often consists precisely in prising apart what polemic joined. Prising apart is slow work, requiring textual patience, conceptual care, and willingness to credit opponents with partial victories. This article undertakes that work for one influential pairing, in the belief that Marxism emerges from the operation neither acquitted on every count nor condemned on every count, but sorted, clarified, and better able to defend what remains defensible in its explanatory ambitions.

Pedagogy offers a quiet test of whether separation has been genuinely absorbed. A seminar leader presenting the functionalism controversy can organise readings in two columns, one column gathering texts about process, evidence, and feedback, the other gathering texts about ontology, levels, and reduction. Students asked to summarise a single author twice, once per column, quickly discover that verdicts diverge: a writer praised in the first exercise may falter in the second, and a writer criticised for ontology may shine on process. Assessment rubrics built around the double summary reward discrimination rather than loyalty. Graduates carrying that habit into archival work, policy analysis, or journalism possess a transferable skill, namely hearing two questions inside one slogan and answering them at different speeds. Speed difference deserves emphasis. Process questions often yield to patient empirical labour within a single project, whereas questions touching explanatory bedrock implicate whole frameworks and may stay contested across careers. Expecting both to resolve together manufactures frustration, and frustration tempts scholars into declaring victory by definition. Definitions declared in frustration rarely survive contact with neighbouring disciplines.

Textbook writers bear special responsibility here, because compression is their craft and distortion rides along easily. A paragraph summarising analytical Marxism for undergraduates must choose which distinctions survive abbreviation. Leaving separability behind during abbreviation produces cohorts who believe mechanism talk and individualism stand or fall as a unit, and those cohorts later reproduce the belief in lecture halls and review essays. Retaining one sentence about separability costs little space and prevents a durable misunderstanding. Editors of companions and handbooks face a parallel choice when commissioning entries: pairing a historian of process rich Marxist scholarship with a philosopher of reduction in a single entry, under instructions to disagree openly, yields livelier and more accurate coverage than assigning the topic to one camp. Accuracy in reference works matters disproportionately, since reference works teach readers what the questions are before readers learn any answers.

Translation across languages introduces yet another site where the pairing can fuse or separate. Terms for mechanism travel reasonably well among European languages, while vocabulary for foundations carries distinct philosophical resonances in German, French, and English traditions. A German discussion echoing debates about Grundlage, a French discussion shaped by arguments over fondement, and an English discussion organised around foundations may sound identical in summary while diverging in connotation. Comparative readers alert to separability will ask, for each translated text, whether complaints target missing process or missing bedrock, instead of assuming the target matches the familiar English controversy. Such alertness also guards against importing polemical heat across borders, where a term carrying mild methodological meaning in one literature arrives carrying heavy ontological freight in another.

Archival scholarship supplies reassuring evidence that process rich, nonreductionist explanation predates the controversy and thrived outside its spotlight. Labour historians reconstructed hiring halls, piece rate bargaining, and neighbourhood credit networks without waiting for philosophical permission. Historians of slavery traced plantation accounting, kinship resilience, and market integration through records that reward patience over modelling. Historians of decolonisation followed party branches, printing presses, and radio broadcasts as they carried ideas across continents. Feminist historians recovered household labour, informal economies, and organisational work inside movements that male centred narratives had rendered invisible. Each body of scholarship satisfies the mechanism expectation in substance, whatever vocabulary its authors preferred, and each resists translation into preference talk without remainder. Citing such scholarship during methodological disputes changes the burden of proof: reduction must explain away accomplished work, not merely promise superior work in the future. Promises have their place in research programmes, yet accomplished explanation sets a standard promises must meet.

Future disputes will surely generate fresh pairings that resemble the pattern examined here. Calls for computational foundations, neural foundations, or data foundations already circulate in neighbouring fields, each call bundling a plausible request for explicit process with a contestable claim about privileged bedrock. Scholars trained on the Marxist case possess a template for response: welcome explicitness, interrogate privilege, demand justification for any importation across disciplines, and insist that bedrock claims compete as theory rather than masquerade as hygiene. Applying the template does not prejudge outcomes. Some bedrock claims may earn their privilege through demonstrated fertility, while others may fade once their bundling advantage disappears. Either result counts as learning. What the template prevents is the oldest failure mode in the story told here, the failure mode where a genuine improvement in explanatory practice becomes hostage to an unrelated doctrine, and where defending the doctrine becomes indistinguishable from defending improvement itself. Marxism paid a price for that confusion in lost decades of cross talk. Other traditions, warned by the example, might pay less.

A closing observation concerns intellectual generosity and its methodological payoff. Reading Cohen generously meant granting consequence explanation a hearing before demanding elaboration, and reading Elster generously meant granting that elaboration could transform a claim rather than merely decorate it. Generosity of that reciprocal kind made the original exchange fruitful despite sharp disagreement, because each participant answered the strongest version of the opposing case available. Later polemics often abandoned reciprocity, answering caricatures and celebrating easy victories, with predictable sterility as the harvest. Reviving reciprocity around separability offers a practical programme: state the mechanism demand at full strength, concede where functional writing failed it, then present the case for termination in its strongest form, acknowledging where individual level analysis exposed real fallacies, and only afterwards argue that concession one does not entail concession two. Argument sequenced in that order earns trust from sceptical readers, because trust follows visible fairness more reliably than it follows declarations of fairness. Trust, once earned, permits genuine persuasion about bedrock, evidence, and the proper architecture of Marxist explanation, which is where the interesting disagreements have lived all along.

The two demands table sets the separation out in the form the rest of the article uses. Each row states what a demand requires, the reason its proponents give for insisting on it, whether a Marxist explanation can meet it without abandoning the tradition’s characteristic claims, and the commitment a reader takes on by accepting it. Reading the rows against each other is the quickest way to feel the asymmetry the debate obscured: the first row asks for work, the second asks for allegiance.

Demand What it requires Why it is made Whether Marxism can meet it What accepting it commits one to
The mechanism demand An account of the process by which a cause produces its effect Explanation without transmission is correlation or teleology Yes, and meeting it strengthened the tradition’s arguments Diligence in specifying processes, nothing about ontology
The reduction demand Mechanisms that terminate in individual beliefs and preferences Only individuals act, so only their reasons finally explain Only by surrendering the claim that relations form preferences Methodological individualism as a doctrine about what exists

The table also exposes why the quarrel generated more heat than its substance warranted. A critic who demanded mechanisms and a defender who refused reduction were never answering the same question, yet each heard the other as refusing rigour or imposing dogma respectively. Once the rows are separated, the mechanism row can be conceded in full, with gratitude for the improvements it forced, while the reduction row is argued on its merits as one theoretical commitment against another. That argument, about the formation of preferences and the autonomy of relational explanation, is where the genuine disagreement lives, and it is an argument Marxism can conduct without methodological embarrassment.

The case for the demand, stated at full strength

The demand for microfoundations begins with a modest observation that carries disruptive force once its implications are followed patiently. Large outcomes in social life, whether a collapse in investment, a wave of strikes, a shift in voting, or a long decline in profit rates, do not descend from abstractions acting upon passive material. Each outcome arrives through countless decisions made under pressure, through refusals and accommodations, through imitation and miscalculation, through habits repeated until they harden into routine. An account that moves directly from one aggregate to another, from accumulation to crisis or from class position to political allegiance, leaves the connecting tissue invisible. Readers are asked to trust that a pathway exists because the endpoints appear together with regularity. Defenders of the demand reply that trust of that kind belongs to description, not to explanation. Explanation starts when the pathway itself becomes an object of inquiry, when somebody specifies who perceived which signal, who bore which cost, who copied a neighbour, who feared dismissal, who calculated a margin, and how those separate moves combined into a pattern nobody individually intended.

Stated generously, the position does not deny structure, history, or constraint. Property relations shape the menu confronting each person. Law narrows permissible moves and raises penalties for others. Inherited technology fixes certain costs before any manager opens a ledger. Culture supplies meanings through which options become thinkable or shameful. None of that is disputed by serious proponents of microfoundations. Their insistence concerns the direction of explanatory travel. However powerful a structure may be, its influence reaches an outcome only by altering what particular persons expect, want, fear, know, or can afford to attempt. A factory closure does not migrate into household behaviour by osmosis. It travels through a notice pinned beside a time clock, through a foreman repeating rumours, through a spouse recalculating rent, through a creditor shortening terms, through a union meeting where anger competes with exhaustion. To skip those mediations is to treat structure as a spell whose efficacy requires no carrier. The demand asks for carriers, named and situated, whose conduct can in principle be examined, disputed, and revised.

A further attraction lies in intellectual honesty about failure. Aggregate theories often survive every disappointment because their vocabulary bends to absorb contrary evidence. If workers support a reformist party, false consciousness explains the result. If workers abandon that party, awakening explains the reversal. If a state aids capital during expansion and also during contraction, functional need explains both generosity and austerity. Such flexibility feels like depth while quietly removing risk from the argument. Microfoundational writing accepts a different discipline. Once an author commits to a sequence involving identifiable actors with limited information and conflicting aims, the sequence can break. Perhaps employers lacked the knowledge attributed to them. Perhaps officials pursued prestige rather than accumulation. Perhaps households valued security above income in ways the model omitted. Exposure to refutation is not a defect imported from laboratory science. It is the price of saying something definite enough to teach a reader anything beyond loyalty to a tradition.

The strongest version also resists a lazy division between levels of analysis. Macro language remains indispensable because persons encounter one another in patterned ways and because totals possess properties absent from isolated acts. Nobody denies that unemployment can rise even while each employer sheds labour for reasons that look prudent in isolation. Emergence of that sort is precisely why transmission matters. A fallacy of composition sits at the heart of many social processes, and only a careful reconstruction of interaction reveals how locally sensible conduct yields collectively ruinous results. The demand therefore does not reduce society to psychology. It asks how interaction, aggregation rules, market clearing, organizational hierarchy, and network position transform scattered conduct into durable regularities. A bank run illustrates the point with cruel clarity. Depositors need not be irrational, bankers need not be villains, and regulators need not be conspirators for a solvent institution to collapse under synchronized withdrawal. The mechanism is the explanation, and the mechanism lives between persons rather than inside any solitary mind.

Why is an unexplained aggregate claim incomplete?

An aggregate pattern gains explanatory force only when the steps linking separate choices to the shared result become visible. Without those steps, correlation between large forces remains description dressed as cause. Showing how beliefs, constraints, incentives, and interactions generate the pattern turns a floating regularity into an account that can fail, improve, and guide intervention.

Consider two histories that produce identical statistics through opposite routes. In one district, a strike wave spreads because shop stewards coordinate across plants after a wage cut breaches a customary fairness norm. In another district, similar stoppages follow unrelated grievances that happen to coincide after a transport failure delays shifts and pay packets. The headline rate of days lost matches, yet the causal stories diverge completely, and remedies suited to one setting would misfire in the other. Aggregate equivalence conceals explanatory difference. Only reconstruction at the level where stoppages were proposed, debated, endorsed, or abandoned separates coincidence from coordination. The same lesson applies to consumption booms built on confidence, to property bubbles fed by imitation among lenders, to electoral landslides driven by local brokers rather than national mood. Wherever similar totals can arise from dissimilar processes, a theory satisfied with totals alone has stopped one step before understanding begins.

Philosophers of science sometimes call the missing element a mechanism, meaning an organized arrangement of entities and activities that regularly brings about a phenomenon. The term can mislead if it suggests clockwork or physical gears. Social mechanisms include deliberation, bargaining, queuing, signalling, sanctioning, learning, and emulation. A queue outside an employment office is a mechanism because position within it alters chances and expectations. A rumour moving through a trading floor is a mechanism because timing and credibility reshape orders within minutes. A promotion ladder inside a firm is a mechanism because anticipated advancement disciplines effort long before any bonus arrives. Requiring mechanisms does not impose a single template borrowed from physics. It imposes completeness: enough intermediate structure that a sceptical reader can follow the production of the result without leaping across a void labelled with a theoretical noun.

Floating regularities also create a problem of scope. A correlation observed during one institutional settlement may dissolve after rules change, yet an aggregate law stated without transmission gives no guidance about which alteration matters. If rising education accompanied rising productivity across several decades, did schooling transmit skill, sort workers by perseverance, delay labour market entry, or merely certify family advantage to employers who rewarded pedigree? Each possibility implies a different response when expansion of schooling ceases to raise output. Policymakers who inherit only the regularity inherit a riddle. Historians face a parallel difficulty when comparing societies whose surface indicators converge while underlying arrangements differ. Transmission detail travels better than correlation because it identifies dependencies: dependence on a particular credit system, on a specific form of household organization, on a legal definition of contract, on a technology for monitoring effort. Explanation improves when its fragility becomes locatable.

There is also a moral dimension that proponents rarely state crudely but often feel strongly. Attributing outcomes to vast forces without mediation can erase responsibility and agency at once. If capital requires a reserve of unemployed labour, who keeps wages low, who designs benefit rules, who decides against hiring, who benefits from fear among applicants? Questions of accountability vanish when necessity speaks in the passive voice. Microfoundational inquiry restores grammatical subjects to sentences about power. Somebody drafted the regulation. Somebody chose the cheaper supplier despite known hazards. Somebody spread a story about migrants to divide a bargaining unit. Somebody else resisted, organized, leaked documents, or quietly sabotaged a speedup. Restoring subjects does not settle blame automatically, since constrained actors deserve different judgment from architects of constraint. Still, an explanation that cannot locate decision points cannot support serious judgment about praise, blame, reform, or resistance either.

Critics sometimes reply that every explanation stops somewhere, that infinite descent into smaller components would paralyze inquiry. Defenders accept the premise and deny the conclusion. A stopping point earns legitimacy when further descent would add detail without altering the causal claim, or when evidence at a chosen level is stable across contexts. What defenders reject is stopping at exactly the place where the theory faces its hardest test. If a Marxist account asserts that competition compels firms to cheapen labour, the behaviour of firms under competitive pressure is not decorative detail beneath the main claim. It is the main claim in operational form. Showing managers comparing unit costs, fearing rivals, borrowing for machinery, and dismissing workers after demand falters does not distract from the theory of competition. It vindicates or damages that theory where vindication counts. Depth is therefore selective, guided by the location of doubt rather than by a metaphysical preference for small things.

A related argument concerns unintended consequences, which occupy a central place in Marxian thought itself. Exploitation, crisis, alienation, and ideological distortion are rarely presented as products of a committee designing misery for its own sake. They emerge, on the classic telling, from relations that reproduce themselves behind the backs of participants. That insight becomes sharper, not weaker, when unintended aggregation receives explicit treatment. Thomas Schelling demonstrated how mild preferences about neighbours can generate severe residential segregation without any resident desiring the final map. Robert Merton catalogued self defeating prophecies and latent functions long before analytical Marxism adopted similar tools. A bank panic, a credential arms race, an overfished commons, a bidding war for scarce housing: each shows order arising from interaction rather than intention. Proponents argue that Marxism, committed to explaining how domination persists without constant conspiracy, needs such models more urgently than doctrines that locate outcomes in deliberate design. Refusing to specify aggregation leaves the most distinctive Marxian claim, that systems can oppress through ordinary conduct, as an evocative slogan instead of a demonstrated process.

Precision about information deserves separate emphasis. Aggregate accounts frequently smuggle omniscience into the heads of actors or into the structure itself. Capital is said to move toward higher returns as if investors survey the entire economy with perfect maps. States are said to secure conditions for accumulation as if officials receive reliable signals about what accumulation requires and possess instruments equal to the task. Workers are said to recognize interests as if experience translated itself into belief without mediation by press, party, priest, family, or fatigue. Microfoundational discipline forbids those shortcuts. Each actor receives a biography of knowledge: partial, delayed, distorted by position, filtered through trusted intermediaries, revised after costly mistakes. Error becomes explicable rather than embarrassing. So does learning, which aggregate language often invokes without describing any learner. When a minister misreads a balance of payments crisis, when a personnel manager overestimates the deterrent effect of dismissal, when a tenant mistakes a temporary rent freeze for permanent security, the resulting behaviour follows intelligibly from perceived circumstances. Situational reconstruction honours the rationality of persons while exposing the poverty of their situations.

Another strength involves counterfactual discipline. To claim that one factor caused an outcome is implicitly to claim that matters would have unfolded differently in its absence, yet aggregate narratives often leave the alternative world unspecified. Microfoundations force the comparison into the open. Remove the closed shop, and would recruitment for a strike collapse, or would informal networks among kin sustain solidarity? Introduce deposit insurance, and does the incentive to join a queue vanish, or does distrust of government preserve panic? Replace piece rates with hourly wages, and do workers slow down, or do peer norms maintain pace while reducing injury? Each counterfactual becomes testable in principle because the causal joints are exposed. Vague holism can gesture toward alternative totalities, capitalism versus socialism, tradition versus modernity, while evading the intermediate comparisons that historical evidence can actually adjudicate. The demand narrows some questions to a size evidence can grip, without pretending that large comparisons are illegitimate. It simply insists that large comparisons decompose, at least partially, into smaller contrasts whose plausibility can be weighed.

Heterogeneity supplies yet another reason. Aggregates average away differences that often drive the phenomenon under study. A working class contains skilled and unskilled labour, recent migrants and settled residents, workers in expanding sectors and workers trapped in declining trades, households with savings and households one illness away from destitution. Treating the class as a uniform respondent to structural pressure predicts a uniformity of conduct that history rarely delivers. Splits, delays, sectional bargains, and betrayals then appear as anomalies requiring ad hoc repair. Begin instead with differentiated positions inside the aggregate, and variation becomes data rather than embarrassment. Why did printers organize earlier than dockers in one port, while the sequence reversed elsewhere? Answers may involve literacy, casual hiring, ethnic segmentation, employer concentration, or the geography of pubs where recruitment occurred. Those answers enrich class analysis by showing how a shared relation to capital is refracted through uneven lives. The demand, at its best, protects Marxism from a schematic version of itself in which categories march across history like chess pieces, identical within each colour and moved by an unseen hand.

Who acts when a class is said to act?

Only persons deliberate, sign, march, strike, hire, invest, obey, or resist. Class language remains useful shorthand for shared position and aligned interest, yet explanation must identify organizers, members, officials, employers, and intermediaries who translate position into conduct through meetings, orders, habits, sanctions, and expectations inside concrete settings.

The slogan that collective actors do not act compresses a grammatical warning into an ontological challenge. Sentences attributing desire, belief, strategy, or guilt to a class borrow predicates that belong literally to minded beings. A class has no ears for a speech, no hand for a ballot, no memory except through archives maintained by persons and institutions staffed by persons. None of this makes class talk useless. Shorthand economizes when an audience can unpack it on request. Trouble begins when shorthand graduates into explanation and the unpacking never occurs. To say the bourgeoisie demanded tariff protection leaves open whether manufacturers petitioned, bankers acquiesced, newspaper proprietors campaigned, civil servants drafted schedules, or legislators traded votes for unrelated favours. Each unpacking tells a different story about power. One reveals a cohesive lobby with disciplined messaging. Another reveals fragmented interests stitched together by a minister seeking a parliamentary majority. A third reveals policy drifting forward because no official wished to own repeal. Political consequences differ accordingly, and so do opportunities for opposition.

Firms provide a parallel case familiar from everyday reporting. A company announces, decides, apologizes, or invests, yet every such verb conceals a chain of authorization. Directors weigh legal advice, executives defend budgets, middle managers translate targets into rosters, frontline staff improvise around shortages, and shareholders exert pressure through exit threats or board elections. Sometimes the chain is tight enough that corporate agency functions as a reliable approximation. Legal personhood, accounting unity, and brand identity all support that approximation for many purposes. Approximation, however, should remain answerable to internal process when internal conflict explains the outcome. A merger attributed to a firm seeking growth may in fact reflect a chief executive protecting tenure against a rival division, with growth language serving as respectable cover. A safety failure attributed to corporate greed may involve a bonus formula designed three levels above the supervisor who skipped inspections, plus a procurement rule that rewarded the cheapest contractor. Locating those links does not exonerate anyone. It distributes causal responsibility with enough accuracy to prevent both scapegoating and absolution by abstraction.

States attract the most inflated agency language of all. A state intervenes, modernizes, represses, or retreats, as if a single will inhabited ministries, courts, barracks, and central banks simultaneously. Defenders of microfoundations ask which coalition held office, which agency controlled implementation, which judges interpreted the statute, which officers refused an order, which local officials diluted a directive for fear of unrest. Bureaucratic politics is not a distraction from state theory. It is where state theory meets evidence. Even highly centralized regimes depend on informants who exaggerate threats, clerks who misfile petitions, commanders who hoard supplies, and censors who misunderstand satire. Coercion itself requires persons willing to aim, arrest, guard, and record, plus many more willing to look away. A theory of repression that cannot discuss recruitment, pay, indoctrination, career incentive, and fear within the repressive apparatus has described an effect while skipping its machinery. The point generalizes: collective nouns summarize coordination, and coordination is an achievement requiring explanation whenever it succeeds, let alone whenever it fractures.

None of this requires denying that organizations possess emergent capacities. A university can remember procedures after every employee who designed them has retired, because files, rituals, software, and training transmit practice across generations of staff. An army can execute manoeuvres no soldier grasps in full. A market can process information dispersed among millions of traders into a price that guides strangers. Emergence, properly understood, strengthens the demand rather than weakening it. If an organization remembers, the storage medium and retrieval routine deserve identification. If an army coordinates, signals, doctrine, logistics, and command authority deserve reconstruction. If a market aggregates knowledge, bidding rules, credit constraints, and information asymmetries deserve attention. Emergent language becomes evasive only when it names the mystery instead of guiding inquiry toward the arrangement producing it. Proponents happily grant that wholes exceed sums. Their question is how excess arises, through which connections, under which conditions, and with what characteristic failures when connections snap.

Class formation offers a testing ground where the demand earns its keep most visibly. Objective position in production may be shared by thousands who never meet, speak different languages, worship in separate buildings, and compete for the same jobs. Turning shared position into shared action requires translation work: leaflets that name a common enemy, strike funds that make solidarity affordable, halls where grievances can be compared, martyrs whose funerals convert private anger into public commitment, songs that make an abstract identity singable. Historians who study labour movements already perform this translation work in rich detail. The demand simply asks theory to catch up with historiography, to stop treating formation as an automatic sequel to position. E. P. Thompson portrayed class as a relationship made through experience over time, a portrayal congenial to microfoundational reconstruction even though Thompson wrote in a different idiom. Where formation fails, the same lens explains failure without invoking deficient consciousness as a residual bin. Competition for housing, employer paternalism, craft exclusiveness, racial segmentation, religious loyalty, and state patronage each raise the cost or lower the appeal of collective identification in traceable ways.

A sympathetic reading also notices what the demand does for solidarity as a practical project. Organizers cannot mobilize an abstraction. They persuade individuals one conversation at a time, addressing doubts about lost wages, police attention, childcare during meetings, scepticism toward leaders, memories of earlier defeats. A theory that begins where organizers begin, with persons weighing risks inside relationships, speaks more usefully to strategy than a theory announcing that history has assigned a class its mission. Mission language may inspire, yet inspiration that ignores free riding invites disappointment. Mancur Olson formalized a dilemma that union activists had long recognized in practice: benefits won by a strike accrue to nonstrikers as well, so each participant has a temptation to let others bear the cost. Solutions are equally concrete: social honour for pickets, sanctions against strikebreakers, selective benefits for members, small group monitoring, moral commitments made publicly before peers. Listing such solutions does not cheapen solidarity into calculation. It respects solidarity as something built under adverse incentives rather than conjured by rhetoric, and it honours builders by taking their obstacles seriously.

Similar reasoning applies to capitalists, who appear in schematic accounts as a unified subject executing a coherent plan. Actual owners quarrel over trade policy, compete for labour and customers, lobby against one another, cheat on cartels, and misjudge innovations that later destroy their own firms. Exporters may welcome a weak currency that importers dread. Creditors may demand austerity that ruins debtors whose purchases sustain retailers allied with the same governing party. A theory that starts from a single capitalist interest must then explain away perpetual discord as superficial. A theory that starts from situated firms can explain both discord and occasional unity: unity as a fragile product of peak associations, shared threats, interlocking directorates, revolving doors into ministries, and crises that temporarily align survival strategies. Class power looks less magical and more formidable once its organizational scaffolding is visible, because scaffolding can be mapped, contested, infiltrated, or dismantled. Abstraction flatters power by making it seem effortless. Reconstruction shows the labour power requires, and every labour process contains points of vulnerability.

What is wrong with saying capitalism requires something?

Requirement language confuses effect with origin. An institution may stabilize accumulation, discipline labour, or lower uncertainty while arising from accident, conflict, imitation, or unrelated reform. Unless benefit feeds back through selection, reward, coercion, learning, or deliberate design to reproduce the institution, necessity has been asserted rather than demonstrated through evidence.

Functional explanation is the target where the demand cuts deepest, because functional idiom pervades Marxist writing and ordinary political speech alike. A prison system gets attributed to an alleged requirement for managing surplus populations. School curricula are said to exist because employers want punctual, literate, obedient workers. Welfare programmes supposedly arise from legitimation pressures or from purchasing power that must be preserved. Family forms get credited to reproduction of labour power at private cost. Each claim may contain insight about effects. Effects, however, are not causes unless a loop returns them to the scene of production. In biology, natural selection supplies the loop that licenses functional talk about hearts and camouflage: traits contributing to reproductive success become more common across generations because organisms bearing them leave more descendants. Society possesses no single universal selector with comparable reliability. Markets eliminate some firms, elections remove some governments, wars destroy some states, fashions exhaust some customs, yet each selector operates unevenly, with lags, protections, and perverse side effects. Invoking function without naming a selector borrows biological authority while declining biological rigour.

Jon Elster pressed this objection with unusual clarity, insisting that a beneficial consequence explains persistence only when the consequence causally contributes to maintaining its own cause through a feedback channel. The channel might run through conscious design, as when legislators preserve a subsidy because recipient firms fund campaigns and threaten relocation. It might run through blind selection, as when firms adopting a cost cutting technique survive downturns that bankrupt rivals who retain older methods. It might run through reinforcement, as when a practice that rewards its practitioners with status encourages repetition and teaching. It might also fail to exist, leaving benefit as coincidence. A rainy season benefits umbrella sellers without explaining rain. A war benefits undertakers without explaining war, unless evidence shows undertakers influencing mobilization. Benefit alone proves compatibility, perhaps even convenience for the powerful. Compatibility is a weak relation, compatible with origin in accident, in compromise among enemies, in imitation of foreign prestige, in a reform won by subordinated groups and later turned to new uses. Functional writing habitually upgrades compatibility into necessity with a single verb, requires, and the upgrade is exactly what the demand refuses to grant for free.

The refusal protects Marxist analysis from a flattering mirror image of apologetic economics. Apologists sometimes argue that whatever survives in a market must serve efficiency, treating survival as a certificate of social value. A careless functional Marxism argues symmetrically that whatever survives under capitalism must serve capital, as if endurance itself stamped an institution with proof of domination. Both moves foreclose investigation into waste, drift, contradiction, and unintended protection for subordinate interests. Some institutions persist because dismantling them would cost more than tolerating them. Some persist because beneficiaries are diffuse and victims are organized, or the reverse. Some persist because law entrenches them behind veto points guarded by courts and upper chambers. Some persist because nobody with power has imagined an alternative, a failure of imagination rather than a triumph of function. Distinguishing persistence through inertia, entrenchment, selection, design, and genuine contribution to system stability yields a richer politics. Reformers need to know whether they confront a load bearing wall or a decorative facade protected by habit. Revolutionaries need the same knowledge even more urgently.

Teleology is the philosophical name for the error under attack. A teleological account explains an event by a future state it helps bring about, as if the future reached backward to arrange its own preconditions. Human action is legitimately teleological because persons aim at goals and adjust conduct in light of imagined outcomes. Systems lack minds, so system level purpose requires translation into either the purposes of designers and maintainers or a selection process indifferent to purpose. Without translation, phrases about capital needing a docile workforce conceal a missing subject. Did employers recognize docility as a need, coordinate schools to produce it, punish educators who resisted, and fund alternatives when public provision failed? Perhaps in some places, under some regimes, with documentary traces awaiting the historian. Elsewhere, schooling expanded through religious rivalry, nation building, parental demand for credentials, or teachers pursuing professional status, with workplace discipline arriving as a side effect later appreciated by employers. Appreciation after the fact is not authorship before the fact. The demand separates those temporal relations that functional prose habitually fuses.

A mature version of the argument concedes that functional hypotheses can serve as fruitful starting points. Noticing that an institution benefits a dominant group generates a research question: through what channels, if any, does that benefit secure the institution against challenge? The hypothesis directs attention to lobbying records, appointment patterns, budgetary dependence, revolving personnel, legal immunities, and cultural prestige. It may also direct attention to absence: a practice beneficial to capital that capitalists never defend, fund, or even notice carries a weaker claim to functional explanation than one surrounded by vigilant guardians. Used heuristically, function talk opens inquiry. Used conclusively, it closes inquiry at the moment inquiry should begin. Proponents of microfoundations present themselves as rescuing functional insight from functional dogma, converting a rhetorical full stop into a comma followed by evidence. That self presentation deserves a fair hearing, especially from readers who value the insights while regretting the dogma.

Examples sharpen the distinction between benefit and explanation. Unemployment may discipline wage demands by weakening bargaining confidence, yet the level of unemployment in any given year reflects central bank choices, fiscal stance, exchange rate commitments, commodity shocks, and electoral calculations, many of them aimed at inflation or currency prestige rather than labour discipline. A functional reading that leaps from disciplinary effect to disciplinary purpose misdescribes policymakers and misleads opponents about where pressure might alter policy. Similarly, a legal doctrine may shield employers from liability while originating in judicial concern for contractual certainty, professional ideology among lawyers, or deference to precedent laid down under different economic conditions. Employers who later exploit the doctrine did not necessarily create it, though their defence of it once challenged becomes part of its persistence story. Splitting origin from maintenance is another gift of the demand. Different mechanisms may govern birth, diffusion, entrenchment, and decay of the same institution, and a single functional label flattening all four phases sacrifices exactly the variation that historical materialism, at its best, exists to capture.

Consider also the fate of institutions harmful to powerful groups that nevertheless endure. Usury restrictions constrained lenders for centuries under religious sanction. Guild rules limited masters who might have profited from unrestricted entry. Hereditary succession placed incompetent rulers above competent administrators while elites grumbled and adapted. Environmental hazards poison the neighbourhoods where executives themselves live, albeit less severely than districts inhabited by the poor. A theory committed to reading persistence as service to capital must either stretch service beyond recognition or classify every awkward survival as a lag awaiting correction, a promissory note that history may never honour. Microfoundational analysis expects friction, veto, misperception, and sunk cost as normal features of social life. Persistence becomes a puzzle with multiple candidate solutions rather than a verdict delivered by system need. That expectation aligns surprisingly well with Marxian themes of contradiction and uneven development, once those themes are cashed out in processes instead of invoked as atmosphere.

The demand also disciplines claims about ideology. To assert that a belief system legitimates domination invites a sequence of questions that separate plausible influence from magical diffusion. Who produces the message, through which medium, financed by whom, selected by which gatekeepers, received by which audiences, interpreted through which prior commitments, repeated in which rituals, and enforced by which sanctions against dissent? A sermon, a textbook, a news bulletin, and a workplace training video each travel differently. Reception can fail, backfire, or mutate into parody. Believers may comply outwardly while privately mocking the doctrine, a split that matters when a crisis suddenly reveals how thin allegiance had become. Regimes have collapsed with startling speed after decades during which official ideology appeared hegemonic, precisely because preference falsification concealed private dissent behind public conformity. Timur Kuran developed that insight into a general account of revolutionary surprise. Microfoundational attention to expression, concealment, and cascade effects explains both the appearance of stability and the velocity of collapse within a single framework, a unification aggregate ideology talk struggles to achieve.

Named proponents give the demand a lineage that predates analytical Marxism and extends beyond any single school. J. W. N. Watkins, writing during methodological debates of the 1950s, defended a thesis he called methodological individualism: explanations of large social phenomena should, in their finished form, refer to the dispositions, beliefs, resources, and interrelations of individuals. Watkins distinguished his position from crude psychologism. Institutions, roles, and social relations could enter explanations as conditions confronting actors, provided the final account did not attribute conduct to collectivities as if collectivities possessed minds of their own. He also allowed unfinished explanations that stopped at organizational regularities, treating them as promissory sketches rather than completed science. That allowance matters because it shows the demand functioning as a standard of completeness, not as a ban on macro vocabulary. A sketch can guide research honourably, provided its status as a sketch is remembered when conclusions are drawn and policy is urged.

Karl Popper supplied an influential ancestor for this line of thought through situational logic. In works including The Poverty of Historicism and The Open Society and Its Enemies, Popper argued that social explanation should reconstruct the situation facing an actor, including institutional constraints and the aims attributed on evidence, then show how the action made sense within that situation. The rationality principle at work was not a psychological law claiming universal wisdom. It functioned as a modelling device: assume actors respond appropriately to their situation as they perceive it, then let discrepancies direct attention toward mistaken beliefs, hidden constraints, or flawed reconstruction. Popper aimed his criticism at historicist prophecies that treated nations and classes as protagonists in a scripted drama. Situational logic returned drama to human scale without denying that institutions shape situations profoundly. Indeed, Popper insisted institutions could be analysed as embodied rules and expectations, created by persons, maintained by persons, and reformable by persons who understood them as human products rather than natural fate.

Watkins and Popper wrote within liberal traditions often hostile to Marxism, yet hostility of origin does not settle validity of argument. A tool forged by an opponent can still cut a genuine knot. Marxists who dismiss situational reconstruction because Popper opposed historical materialism commit a genetic fallacy, confusing biography with logic. The serious question is whether reconstructing situations illuminates processes that Marxist theory itself posits: competition compelling cost reduction, insecurity compelling job seeking, monopoly enabling price setting, debt compelling asset sales. In each case, compulsion operates through a situation: alternatives narrowed, penalties attached, information arranged, time horizons shortened. Situational language gives compulsion a mechanism instead of a metaphor drawn from physical force. Ironically, the ancestor supplied by an anti Marxist philosopher can help Marxists state their own structural claims with greater precision, provided the borrowing is acknowledged and the differences about which situations matter remain open for argument.

Jon Elster brought these older debates directly into Marxist territory. A Norwegian philosopher and social scientist trained across several disciplines, Elster had engaged Marx seriously rather than treating Marxism as a museum piece. His Making Sense of Marx, published in 1985, subjected central Marxian concepts to patient reconstruction, accepting some, revising others, and rejecting those that survived only through ambiguity. Elster endorsed functional explanation in biology while challenging its unlicensed transfer to society. He advocated methodological individualism as a regulative ideal for Marxist theory, argued that game theory could clarify coalition, bargaining, and exploitation, and explored how preferences themselves form under adaptive pressure, sour grapes being his memorable label for desires trimmed to fit deprivation. Admirers praised the book for taking Marx seriously enough to disagree in detail. Detractors accused Elster of dissolving dialectics and totality into fragments. Even detractors, however, often answered on terrain Elster had marked out, supplying mechanisms where earlier writing had relied on assertion, which is itself evidence that the demand altered practice.

Elster at full strength deserves presentation as a reformer addressing fellow readers of Marx, not as a customs officer imposing foreign rules. His complaint was that Marxist theory possessed powerful insights about exploitation and historical change trapped inside explanatory habits that weakened them. Dialectical language, in his reading, sometimes licensed contradiction talk where a precise account of conflicting mechanisms was needed. Holistic language sometimes protected claims from evidence by relocating them to a totality no observation could touch. Functional language sometimes converted consequences into causes without the feedback that alone could justify conversion. Against those habits, Elster proposed virtues any tradition can recognize: define terms so that disagreement becomes possible, separate distinct claims bundled inside slogans, follow causal arrows in both directions before declaring necessity, and prefer a modest mechanism demonstrated in one setting to a grand law asserted for all settings. The proposal threatened establishments because establishments benefit from ambiguity that lets incompatible audiences hear agreement. Clarity creates losers among interpreters whose authority rests on mediating obscurity. That sociological fact about reception helps explain heat surrounding the book beyond its technical arguments.

Game theory formed part of Elster’s toolkit, and its appeal illustrates the positive side of the demand. A game specifies players, available moves, information, timing, and payoffs, then examines which combinations of choices can persist when each participant anticipates the responses of others. Applied to class conflict, such specification prevents a common equivocation between what would benefit a group collectively and what each member has reason to do individually. Applied to bargaining between a union and an employer, it highlights credibility of threats, cost of delay, outside options, and the shadow cast by future negotiations. Applied to competition among firms, it shows how an innovation can spread even when every adopter would prefer a world where nobody adopted it, because unilateral restraint invites ruin. Nobody needs to believe that life is literally a parlour game for these models to earn their place. Models simplify in order to isolate a tension. The demand asks critics of a model to identify which simplification distorts the tension, rather than rejecting modelling as such because social reality overflows every representation.

Rational choice, as used by the strongest proponents, is thinner and more flexible than caricature suggests. It need not assume selfishness, perfect foresight, stable tastes, or calculation performed with pencil and spreadsheet. A person may act from duty, spite, love, habit disciplined by anticipated regret, or loyalty to a group whose approval matters more than money. The core commitment is intelligibility: conduct should be reconstructible as a response to beliefs and evaluations within a situation, so that changing beliefs or evaluations would intelligibly change conduct. Even apparently irrational phenomena become researchable under that commitment. Panic, addiction, self deception, and weakness of will fascinated Elster precisely because they strain simple rationality while still involving persons who deliberate, precommit, rationalize, and relapse. Ulysses binding himself to a mast supplied Elster with an enduring image of a person who anticipates his own future failure and designs a constraint against it. Institutions can be read similarly, as precommitments built by groups aware of their own temptations. That reading humanizes institutions instead of treating them as alien forces.

Other voices strengthened the demand from inside Marxist scholarship. John Roemer reconstructed exploitation using general equilibrium models and property relations, showing how unequal asset distribution could generate exploitative transfers without relying on a labour theory of value in its classical form. Whatever verdict readers reach about Roemer’s reconstruction, its clarity forced disagreement onto identifiable assumptions about assets, markets, and alternative allocations. G. A. Cohen, though defending functional explanation in a sophisticated form, sharpened debate by distinguishing elaboration from mere assertion and by asking what kinds of consequence laws could be legitimate. Adam Przeworski examined how electoral participation reshaped workers movements through strategic choices under institutional rules, linking class compromise to party organization and anticipated policy. Erik Olin Wright mapped class locations with attention to contradictory positions occupied by managers and credentialed employees, grounding typology in relations that could be surveyed and compared. These projects differ sharply from one another. Their shared legacy is a style in which claims wear their assumptions visibly enough for strangers to test them.

The demand also travelled through historical sociology and political economy beyond self described analytical Marxism. Douglass North explained institutional change through relative prices, enforcement costs, and organizations learning within rules they also sought to alter. James Coleman proposed a boat shaped diagram linking macro conditions to individual situations, individual conduct to interaction, and interaction back to macro outcomes, giving teachers a durable picture of what completeness requires. Peter Hedström later codified mechanisms of desire, belief, and opportunity, insisting that beliefs about opportunities can move conduct even when objective opportunities diverge. Each framework translates the same intuition into distinct vocabulary: no macro to macro shortcut without a passage through situated action and aggregation. Convergence across rival traditions counts for something. When scholars who disagree about capitalism, rationality, and revolution nevertheless converge on transmission as a criterion, the criterion likely tracks a genuine explanatory need rather than a factional fashion.

A steelmanned case should also acknowledge the rhetorical discipline the demand imposed on its own advocates. Early polemics sometimes treated every macro concept as suspect and every formal model as progress, a philistinism that invited justified backlash. Mature proponents learned to distinguish levels of completeness appropriate to different questions. A demographer projecting pension liabilities may legitimately use aggregate transition rates without reconstructing each retirement decision, because the rates have proven stable and the projection makes no deep causal claim. A historian explaining why a pension reform provoked riots faces a heavier burden, since the causal claim concerns perception, organization, and timing. Context sets the required depth. What the demand forbids is borrowing the authority of causal explanation while supplying only the evidence of projection or correlation. That distinction, modest as it sounds, would clean a remarkable amount of confusion from textbooks, manifestos, journalism, and policy memoranda if applied consistently across ideological camps.

How did the demand change Marxist writing in practice?

Writers began naming decision makers, specifying information available to them, tracing unintended consequences, and testing claims against alternative choices. Functional shorthand survived less easily. Debates over exploitation, state policy, ideology, and class formation gained clearer mechanisms, explicit counterfactuals, and closer attention to organizations that mediate between structural pressure and lived conduct.

Concrete achievements can be traced debate by debate. Discussion of the capitalist state changed once writers were pressed to identify the channel through which business influence operated. Structural dependence became a precise hypothesis: governments restrain policies hostile to investment because officials anticipate capital flight, falling tax revenue, rising unemployment, and electoral punishment, regardless of personal sympathy toward business. Instrumentalist research documented personnel links, lobbying, and campaign finance. Poulantzas inspired accounts of relative autonomy gained specificity when successors asked which fractions of capital benefited from a policy, which state agencies mediated the benefit, and which feedback rewarded officials who complied. Even readers who reject every one of those hypotheses benefit from their separability. Evidence can support structural dependence in monetary policy while supporting instrumental capture in regulatory appointments and supporting bureaucratic self interest in procurement. A single label for state power dissolves into a map of channels, each with its own strength, scope, and vulnerability to democratic pressure.

Exploitation debates gained similar clarity. Once Roemer and others reformulated exploitation around asset inequality and counterfactual withdrawal, arguments could focus on whether labour markets, credit constraints, or skill ownership carried the explanatory weight in a given setting. Defenders of the classical labour theory answered with renewed attention to production, surplus extraction at the point of work, and the difference between generating value and redistributing claims. The exchange improved both camps because each had to state where exploitation happens, who transfers what to whom, under which rules, and what alternative arrangement would count as its absence. Moral condemnation acquired an address. Instead of denouncing a system in terms so broad that every misfortune becomes evidence, writers could distinguish exploitation from domination, from unfair bargaining, from monopoly rent, from simple misfortune, while still arguing that several of those wrongs interlock in practice. Distinctions of that kind strengthen rather than dilute critique, because a remedy matched to a specific wrong has a chance of working.

Class formation research also absorbed the lesson productively. Studies of union density began comparing workplace size, employer concentration, legal recognition procedures, ethnic composition, and community institutions with a precision that earlier class analysis sometimes disdained as empiricism. The payoff was explanatory surprise: solidarity flourished in some unlikely places because a chapel, a football club, a migrant association, or a company town street plan lowered the cost of meeting and raised the cost of betrayal. Defeat became intelligible without condescension toward the defeated. Where organization failed, researchers could point to shift patterns that prevented assembly, to turnover that dissolved trust, to piecework that set worker against worker, to supervisors recruited from the same neighbourhood who blurred lines of antagonism. Such findings return agency to history in a double sense. They show ordinary persons making movements under constraints, and they show constraints operating through arrangements that other persons designed, inherited, contested, and occasionally transformed.

Ideology critique underwent a comparable tightening. Assertions that schooling reproduces hierarchy became investigations into tracking decisions, parental strategies, teacher expectations, credential inflation, and employer screening, each open to comparison across regions and periods. Assertions that media manufacture consent became studies of ownership concentration, advertising dependence, source routines among journalists, audience scepticism, and the economics of attention. Some investigations confirmed strong versions of the older claims. Others found leakage, resistance, professional norms, and commercial incentives pulling against simple domination. Either result counts as progress because belief replaced incantation. A critique that can discover its own limits is more trustworthy when it reports a genuine case of manipulation. Readers learn to distinguish propaganda systems that punish deviation from noisy media environments where power tilts visibility without dictating belief. Political strategy benefits likewise, since countering a tilt requires different tools from countering a command: subsidies for alternative outlets, transparency about ownership, training in source evaluation, protection for whistleblowers inside newsrooms.

Perhaps the deepest achievement was pedagogical. Students encountering Marx through mechanism conscious writing learn a transferable habit of questioning that survives disagreement with any particular conclusion. Faced with a claim about globalization disciplining governments, a trained reader asks which investors move first, which signals ministers watch, which domestic groups amplify the threat, and which institutions could insure against exit. Faced with a claim about technology deskilling labour, the reader asks which managers chose the machinery, which skills became redundant, which new skills appeared at the margins, and how training systems responded. The habit democratizes theory by lowering the deference owed to fluent abstraction. A beginner can challenge a famous name by requesting a missing link, and the request is legitimate even before the beginner masters the canon. Authority shifts, partially and healthily, from mastery of vocabulary toward ability to reconstruct a process in public. Seminar rooms become workshops where arguments are disassembled and rebuilt, rather than temples where interpretations are recited.

Honesty requires a boundary marker at the close of the case, placed there by its fairest advocates themselves. Establishing that mechanisms are owed does not establish where mechanisms must terminate. A chain of explanation might pass through individual conduct and then through the formation of the preferences, beliefs, and capacities that conduct expresses, and that formation might involve language, upbringing, institutions, and relations irreducible to a solitary chooser confronting a fixed menu. Nothing in the critique of floating aggregates proves that desire arrives preformed, that rationality exhausts motivation, or that the individual is a natural atom prior to society. Those stronger theses require separate argument and separate evidence. A mechanism can run through norms internalized so deeply that compliance feels like identity rather than choice. A mechanism can run through emotions, addictions, loyalties, and fantasies that rational choice language handles awkwardly. A mechanism can even run through organizations whose rules, once specified, do explanatory work without descending further into the psychology of every clerk. The demand, strictly understood, governs the presence of transmission, not the metaphysics of the final stopping place.

That boundary matters for fairness in both directions. Opponents of microfoundations sometimes refute the stronger thesis, that explanation must end with given preferences inside isolated heads, and then announce victory over the weaker thesis, that transmission is owed. The announcement is premature. A critic can reject atomism, reject selfishness, reject perfect information, reject equilibrium modelling, and still owe an account of how a purported macro force reaches an outcome through situated conduct and aggregation. Conversely, advocates sometimes slide from the weaker thesis, which commands wide assent once stated plainly, to the stronger thesis, as if accepting mechanisms logically compelled accepting a particular theory of the person. The slide is a fallacy, and naming it strengthens the case by trimming it to defensible size. What survives trimming remains formidable: no unexplained leaps from structure to outcome, no collective subjects without unpacking on request, no functions without feedback, no necessity without a selector or a designer or a maintainer, no correlation promoted to cause without a pathway that evidence could in principle disturb. A Marxism that accepted those rules would lose some rhetorical shortcuts. It would retain exploitation, crisis, class, and ideology as research programmes, each rendered more dangerous to complacent power because each could show, link by link, how domination is produced by identifiable arrangements and therefore how different arrangements might interrupt the production.

The case against reduction to preferences

The case against reduction begins with a concession, offered without reluctance. Individuals act, choose, deliberate, refuse, comply, organise, and desert, and no social explanation earns credibility by pretending otherwise. Armies do not march unless soldiers move. Markets do not clear unless traders place orders, accept prices, extend credit, or withhold goods. A strike exists because particular workers stay away from particular workplaces on particular mornings, often after arguments in kitchens, union halls, and yards. Any structural account that floats free of such conduct invites a fair charge of mystification. The dispute examined in this section concerns a different claim, namely whether acknowledgement of acting individuals obliges explanation to terminate in their preferences, beliefs, and opportunities, redescribed until every social pattern dissolves into an aggregate of separate decisions. Opponents of reduction answer that acknowledgement and reduction are separate commitments. One can grant that nothing social happens except through people while denying that people, taken one by one with their wants treated as given, supply the deepest or most illuminating level of explanation available.

Respect for the individualist programme sharpens rather than softens the disagreement. That programme achieved much by insisting that large nouns do no causal work by themselves. Class, capital, the state, and ideology explain nothing when invoked as hovering agents with intentions of their own. Demanding a pathway from macro claim to situated conduct exposed loose functional talk, forced hidden assumptions into daylight, and replaced many slogans with mechanisms that could be checked, modelled, and criticised. Critics of reduction who forget those gains relapse into a vocabulary their opponents rightly distrust. Yet a discipline that began as a demand for clarity can harden into a restriction on what counts as real, and the structural reply starts precisely at that hardening. It asks whether mechanisms must always run through conscious preference, whether regularities at one level require derivation from a lower level before they qualify as explanations, and whether wants themselves can be treated as an unexamined starting point when the formation of wants belongs among the phenomena under investigation.

Consider first the simplest structural observation, stated plainly. Some relations bind persons who never chose them, dislike them, misunderstand them, or deny their existence outright. A tenant who regards rent as theft still faces eviction after nonpayment. A debtor convinced that interest is usury still watches arrears compound under a signed contract enforced by courts. A small producer who despises price cutting still lowers prices when rivals do, because unsold stock pays no wages and services no loan. In each instance compulsion travels through a position occupied within an arrangement of property, law, and exchange, not through approval in the mind of the occupant. Belief colours experience of the constraint, alters tactics for evading or softening it, and sometimes fuels collective attempts to abolish it, but belief alone does not suspend operation of the constraint while the arrangement persists. That distinction between experiencing a relation and constituting it carries much of the anti-reductionist argument.

Marxist theory gives the point a specific form through the wage relation. Separation of most producers from land, tools, and raw materials sufficient for independent production leaves them with labour power as the principal asset they can bring to market. Ownership concentrated elsewhere controls access to workplaces where that capacity becomes productive. Neither side invented the configuration single-handedly, and neither side can dissolve it by a private change of heart. A worker persuaded by radical criticism of wage labour still requires food, housing, heat, and care for dependants, and those requirements arrive on a schedule indifferent to conviction. Refusal directed at one employer leads, in ordinary circumstances, to search for another. Refusal directed at employment as such leads toward destitution, dependence on kin or charity, subsistence activity at the margins, or confrontation with poor relief regimes designed to make dependence unattractive. Exceptions exist, and margins matter historically, but exceptions priced in hunger confirm the pressure they escape only partially.

Symmetry of compulsion on the other side deserves equal emphasis, because popular versions of structural argument sometimes picture capitalists as free authors of a system that merely imprisons workers. Competition disciplines owners and managers with comparable impersonality, albeit from a position of far greater security and reward. A firm that pays well above prevailing rates, shortens hours unilaterally, forgoes profitable but harmful products, or declines techniques adopted by rivals may flourish through reputation, loyalty, or niche demand, yet it operates against a gradient. Costs feed into prices, margins, creditworthiness, and capacity to reinvest. Rivals gaining advantage through harsher methods can undercut, expand, attract capital, and eventually purchase or marginalise the reluctant competitor. Bankruptcy law, shareholder pressure, managerial labour markets, and takeover threats translate that gradient into personal consequences for decision makers. A chief executive animated by sincere distaste for relentless accumulation discovers that boards, lenders, and competitors supply motives regardless of private distaste. The system selects for behaviour compatible with its continuation, rewarding compliance and penalising deviation, even when selection operates through regretful agents who describe their own conduct as forced upon them.

Selection language matters here because it separates structural explanation from conspiracy. No committee needs to design outcomes that emerge from many agents adjusting to shared constraints under conditions of mutual dependence. Grain prices in a famine, rents in a housing shortage, and wages in a slack labour market can move against the wishes of nearly everyone involved while still displaying intelligible pattern. Each participant pursues intelligible aims, yet the outcome belongs to the interaction structure rather than to any participant considered alone. Individualist modelling captures such emergence elegantly in many domains, through game theory, price theory, and simulation, and structural critics lose credibility when they ignore those successes. Their narrower claim survives the concession: emergence acknowledged in practice is often denied in methodological pronouncement, where explanation is said to be complete only after social predicates have been eliminated in favour of individual predicates. If interaction terms, institutional rules, and positional constraints do indispensable work inside the model, then reduction has already been qualified beyond recognition, whatever label the modeller prefers.

Property illustrates why elimination proves so difficult. To say that a person owns a factory is not to report a physical relation between a body and a building, nor a belief inside a single head. Ownership bundles rights to exclude, to direct use, to appropriate income, to transfer, and to call upon courts and police for enforcement, all distributed across officials, neighbours, employees, buyers, and rivals who recognise, contest, or evade those rights in patterned ways. Remove every attitude held by the named owner and the bundle can persist through trustees and heirs. Remove recognition by everyone else and the title becomes paper. The fact is therefore irreducibly relational and institutional, although it supervenes upon countless acts and dispositions spread across a population. Translating ownership into a long conjunction of individual statements may be possible in principle for a frozen moment, but the translation explodes in length, loses the generalisations that make ownership explanatory, and must be rebuilt whenever personnel change while the institution endures. Persistence through complete turnover of occupants is a hallmark of structure: firms survive their founders, ministries survive their ministers, and classes survive every individual member, because positions are defined by relations to other positions rather than by the psychology of whoever fills them for a season.

Can a relation hold whatever individuals believe about it?

Yes, within limits set by collective recognition and enforcement. A wage contract, debt, or property title binds a sceptical participant because courts, employers, creditors, and other workers act upon shared rules regardless of private dissent. Belief influences compliance, resistance, and reform, yet isolated disbelief leaves positional compulsion intact.

Further precision helps, because the affirmative answer can be caricatured as a claim that ideas never matter. Ideas matter enormously, and structural theorists who treat consciousness as decorative foam on material process betray their own tradition, which assigned organising, interpreting, and imagining a central place in historical change. The point concerns thresholds and distribution. One tenant disbelieving in rent changes little. Tenants organised around a shared redefinition of obligation, willing to act together and withstand sanctions, can change what landlords and courts can enforce. Money offers the classic illustration: a single shopper declaring a banknote worthless accomplishes nothing beyond personal embarrassment, whereas generalised collapse of confidence destroys the value everyone previously took for granted. Social facts often depend on attitudes in aggregate while remaining stubbornly independent of attitudes in any particular breast. Methodological pronouncements that move from dependence on somebody believing something to reducibility to what some specified individual believes commit a fallacy of division. The wage relation likewise rests historically on law, custom, violence, and consent distributed across generations, but dependence on that vast background does not make a given worker’s employment conditional on that worker’s endorsement.

Historical formation reinforces the distinction. Enclosure, colonial expropriation, factory legislation, and the slow destruction of commons and customary rights created populations for whom wage labour became the normal route to subsistence. Those processes involved identifiable agents pursuing identifiable purposes, often with brutal clarity, and a good structural history names them rather than hiding behind abstractions. Once established, however, the resulting configuration confronts later generations as a premise rather than a proposal. Children born into districts shaped by vanished industries inherit housing stock, skill profiles, transport links, debts, and expectations formed around opportunities that may no longer exist. Their preferences adapt to a landscape they did not select. Explaining their conduct by citing those preferences without explaining the landscape mistakes a transmission belt for a motor. Genealogy therefore belongs inside structural explanation, not as ornament but as an account of how positions acquired their powers and why alternatives carry unequal costs.

Nor does compulsion require literal absence of choice, a confusion that weakens many popular presentations. A chess player facing mate in two still chooses among legal moves, and describing the position explains why every available move loses without denying that a move will be selected by a reasoning person. Structural constraint operates similarly across much social life. It narrows feasible sets, attaches radically unequal penalties to options, and makes some choices heroic, eccentric, or self destructive while rendering others routine. Explanation can then proceed at the level of the feasible set itself: why subsistence requires sale of labour power, why firms must valorise capital, why states dependent on private investment court business confidence, why tenants bid against each other for scarce dwellings near employment. Answers invoke distributions of assets, legal entitlements, demographic pressure, and technical organisation. Individual narratives fill in which option a person took and how it felt from inside, and such narratives remain indispensable for biography, moral judgement, and political understanding, but they answer a different question from the one posed by the shape of the menu.

Menus, moreover, are not neutral between occupants of different positions. The same formal freedom to contract conceals substantive asymmetry when one party contracts to avoid hunger and the other contracts to expand an already substantial portfolio. Liberal legal theory prizes the abstraction that treats both signatures identically, and that abstraction delivered real emancipation from status, guild, and feudal obligation. Marxist criticism targets its limit rather than its historical achievement. Formal equality before the law coexists with material inequality in bargaining power, information, reserves, and capacity to wait. A negotiator able to abstain for a year faces another whose reserves run dry within a week, and the resulting agreement records that disparity even when both parties prefer agreement to breakdown and both sign without fraud or threat uttered aloud. Preference based explanation registers satisfaction of wants under constraints and stops. Structural explanation interrogates distribution of the constraints, asking who set them, who benefits from their reproduction, and why the weaker party’s wants adjust so reliably to what the stronger party can offer. Both inquiries can coexist, but the second cannot be translated into the first without remainder.

Discipline also requires specifying mechanisms of reproduction, a demand structural writers should welcome rather than resent. Positions endure because schools train for them, laws enforce them, families socialise into them, infrastructures presuppose them, and daily success within them confirms their apparent naturalness. A labour market reproduces itself each morning when transport runs toward employment centres, when landlords price rooms against expected wages, when shops stock goods priced for weekly pay cycles, and when jobseekers present credentials minted by institutions oriented toward hiring. None of that requires anyone to intend reproduction of the whole. Habits, investments, and built environments lock in patterns whose origins recede from view, so that later conduct guided by local reasons perpetuates arrangements nobody chose as a package. Individualist accounts of path dependence describe similar lock in convincingly. The remaining disagreement concerns whether, after mechanisms are specified, the structural pattern retains explanatory standing or survives only as shorthand. Anti-reductionists argue that shorthand which captures robust generalisations across indefinitely many substitutions of personnel is doing genuine explanatory work, comparable to talk of pressure in a gas, where molecular detail underlies the phenomenon without replacing thermodynamics for most purposes.

Inside Marxism itself, an older parallel argument rehearsed many of these tensions under different banners, and readers interested in that internal history will find the contrast drawn sharply in discussions of humanist vs structuralist marxism where emphasis on lived experience confronted emphasis on impersonal process. Humanist readings stressed alienation, praxis, and the young writings on estranged labour, warning that talk of structures could freeze living struggle into a machine that runs behind human backs. Structuralist readings stressed decentred causality, ideology as material practice, and suspicion toward any subject imagined as origin of meaning. Each side accused the other of losing something essential: agency in one indictment, objectivity in the other. The micro-foundations dispute replays that quarrel with analytic tools, game theoretic models, and a vocabulary borrowed from philosophy of science. Recognising the replay helps explain the heat involved. At stake is not merely modelling technique but the identity of a tradition deciding whether emancipation requires a sovereign subject who makes history, or an account of how subjects themselves get made by history they subsequently learn to contest.

Terrain metaphors risk a fresh error, namely treating social arrangements as fixed geology rather than as reproduced activity. Armies of clerks, teachers, inspectors, programmers, carers, and drivers remake institutions daily through competent performance, improvisation, and occasional sabotage. A rule unapplied decays. A norm universally mocked loses force. Structures therefore possess a dual character: they confront each person as external fact and exist only through ongoing enactment by persons, most of whom aim at local purposes rather than at preservation of the whole. Anti-reductionism that forgets enactment drifts toward reification, picturing capital as a subject with appetites and history as a train on rails. Individualism that forgets externality drifts toward voluntarism, picturing each morning as a fresh social contract among unencumbered choosers. Between those caricatures lies the serious debate, conducted by writers on both sides who accept duality and argue over explanatory priority, permissible idealisation, and the point at which shorthand becomes evasion.

Substitution evidence deserves lingering attention because it offers the structural case an empirical spine rather than a purely conceptual one. When an outcome recurs despite turnover of personnel, variation in stated motives, and differences in local culture, suspicion grows that positional logic drives the recurrence. Hospital managers of divergent politics implement similar triage under similar funding formulas. University administrators with opposite rhetorical commitments raise similar fees under similar revenue constraints. Landlords who inherited property and landlords who speculate aggressively converge on similar rents where demand outstrips supply. Convergence does not prove inevitability, and careful researchers hunt for deviant cases where organisation, law, or solidarity broke the pattern, because deviants reveal which supports were load bearing. Nevertheless convergence shifts the burden of proof. An explanation pitched entirely at motives must multiply coincidence to cover recurrence, whereas an explanation pitched at incentives, constraints, and selection can predict recurrence while leaving motives variable. Prediction of that modest, pattern level kind is a scientific virtue even when precise forecasting of single events remains impossible.

Selection mechanisms supply the bridge between intention and pattern that both camps can inspect. Markets eliminate firms whose costs persistently exceed revenues, whatever their founders dreamed. Electoral systems eliminate parties that persistently fail thresholds, whatever their manifestos promised. Bureaucracies promote officials whose habits fit audit criteria, whatever their private philosophies. Over time the population of surviving organisations comes to display traits adapted to the selecting environment, and observers can explain distribution of traits by reference to selection pressures without claiming that any survivor intended the distribution. Darwinian reasoning transformed biology by legitimating exactly that explanatory shape: design like order without a designer, explained retrospectively by differential survival. Marxist structural argument borrows the shape while substituting historically specific mechanisms: competition, accumulation, class struggle, and state enforcement. Borrowed shape does not guarantee borrowed success, and each claimed selection process must be demonstrated rather than asserted, but the shape itself cannot be dismissed as unintelligible merely because outcomes outrun intentions.

What does supervenience concede and what does it refuse?

Supervenience concedes that social variation requires some individual variation somewhere, in acts, dispositions, or relations among persons. It refuses identity between levels, refusing to treat a class, firm, or state as nothing beyond a list of members. Dependence without reducibility is the position staked out.

That formulation attracted support precisely because it concedes so much to intuition on the individualist side. Nobody defending supervenience imagines a social change floating free of any change in people, their conduct, their expectations, or the material traces those leave behind. If two societies differ in whether private ownership of productive assets prevails, then somewhere there must be differences in how persons behave, what officials enforce, which documents get honoured, and how children are taught to regard taking, lending, and excluding. A ghostly alteration in the social realm with every individual fact held identical strikes most participants as unintelligible, a difference without a bearer. Accepting that verdict costs structural theory nothing it should wish to keep, because structural claims were never meant to posit a separate substance hovering above human activity. They concern patterns within activity, patterns whose description requires vocabulary pitched at the pattern rather than at each contributing act.

Refusal enters at the inference from dependence to identity. A painting depends entirely on pigment, canvas, and the distribution of both, yet a chemical inventory of pigments does not state what the painting depicts, why it moves viewers, or how it relates to a genre. A chess position depends entirely on placement of pieces, yet describing wood grain and coordinates misses check, threat, and sacrifice, concepts defined by rules and possibilities rather than by timber. A sentence depends on ink marks or pixels, while its meaning depends on conventions shared across a language community and cannot be read off the physics of a single inscription. Analogies do not settle social ontology, since societies differ from paintings and games in causal complexity, but analogies break a bad habit of thought: the habit of sliding from every difference needs a base difference to only base vocabulary can express real differences. Multiple realisation blocks that slide. The same institutional fact, a valid contract, a recognised marriage, a legal tender note, can be realised through wildly divergent materials, ceremonies, and mental states across cultures and centuries, and the unity captured by the institutional term disappears when description fragments into realising details.

Multiple realisation also explains why reduction can fail in practice even where a determined reductionist insists it succeeds in principle. Suppose, for argument, that each token social event could be paired with an enormous individual level description true of that occasion alone. Explanation seeks more than token pairing. It seeks generalisations that survive variation in realisation, support counterfactual reasoning, and identify which features mattered. A law connecting profitability pressure to relocation behaviour holds across firms differing in ownership form, managerial biography, and workforce composition, and its power lies exactly in abstraction from those differences. Replacing the law with a disjunction of all its possible realisations produces a statement too unwieldy to test, too heterogeneous to project onto fresh cases, and silent about why the disjuncts belong together. Philosophers of science encountered the same problem with mental states and brain states long before sociologists borrowed the lesson: even convinced physicalists learned to prize psychological generalisations because neural description, however fundamental, scattered the patterns psychology unified. Autonomy of the special sciences rests on that pragmatic and epistemic ground, not on denial of underlying dependence.

Several figures inside the micro-foundations debate gravitated toward supervenience for exactly these reasons, treating it as common ground sturdy enough to build upon. From one direction, individualists could accept that macro predicates pick out genuine patterns provided those patterns never vary independently of micro arrangements. From the opposite direction, defenders of macro explanation could accept micro dependence provided dependence stopped short of eliminability. The truce clarified remaining disagreement wonderfully. Arguments shifted away from existence, where rhetoric easily inflates, toward questions about causal relevance, explanatory depth, and the legitimacy of intervening at one level rather than another. If altering interest rates reliably alters investment across heterogeneous firms, then a macro variable earns causal credentials through interventionist tests regardless of micro heterogeneity beneath it. If altering a single manager’s mood alters nothing beyond anecdote, then that micro variable earns modest credentials despite its ontological concreteness. Concreteness, it turns out, is a poor proxy for importance, since the most tangible item in a system may be among the least consequential for its trajectory.

Causal talk at the social level still provokes a familiar objection concerning exclusion. If individual conduct fully causes an outcome, what causal work remains for a structure to perform without double counting or spooky overdetermination? Responses developed across philosophy of mind transfer with adjustment. One response treats levels as descriptions of a single causal process at different grains, comparable to describing a flood through meteorology, hydrology, or the testimony of residents, where rivalry between descriptions is a category mistake. Another response locates structural causation in the shaping of probabilities and feasible sets rather than in direct pushing, closer to how a riverbank causes a river’s course without supplying its water or its momentum. A third response emphasises constitution: a demonstration is constituted by marching bodies, chanting voices, and occupying space, and asking whether the demonstration or the bodies blocked the street poses a false choice. Each response carries difficulties, and honest structural theorists admit that causal vocabulary borrowed from billiard balls fits social processes imperfectly. Admission, however, cuts both ways, since individualist causation through reasons, expectations, and anticipated reactions fits billiard imagery no better.

Intervention offers firmer ground than imagery. Policies operate habitually at aggregate levels because levers exist there: tax schedules, benefit rules, planning law, interest rates, curriculum requirements, competition policy. Governments that accepted a strict ban on macro causation would paralyse themselves, forced to reconceive every measure as billions of separate nudges with no legitimate summary of expected effect. In practice even administrations rhetorically devoted to individual responsibility forecast outcomes through elasticities, multipliers, and demographic projections, all irreducibly aggregate instruments. Success of such instruments varies, sometimes embarrassingly, yet their indispensability in planning reveals a working commitment to level specific regularities that methodological manifestos disown. A finance ministry does not interview every consumer before estimating response to a fuel duty change. It relies on patterns that hold across populations whose individual stories diverge wildly, and when forecasts fail, analysts revise the aggregate model rather than abandoning aggregation as such. Practice, in this domain, has already voted.

Class provides the test case where supervenience talk must prove itself or fail. A class, on the structural reading, is not a club with membership cards nor a sum of incomes averaged for convenience. It is a set of positions defined by relation to productive assets, authority at work, and exposure to market dependence, occupied by persons whose other attributes vary enormously. Two workers sharing a position may differ in religion, nationality, skill, temperament, politics, and prospects, while a worker and a small proprietor may share income, neighbourhood, and tastes yet occupy divergent positions with divergent vulnerabilities. Classification by position predicts distinctive pressures: susceptibility to unemployment, dependence on overtime, capacity to survive a strike, access to credit, and likely attitude toward collective provision, without predicting any individual’s vote or virtue. When critics object that classes do not act, only persons do, the supervenience reply agrees and redirects attention. Class analysis claims that persons situated similarly face similar problems, acquire similar information, and find similar strategies attractive, so aggregate conduct displays pattern even amid personal variety. Where organisation, leadership, and shared interpretation convert situation into collective action, further individual and organisational facts enter, and class formation becomes a historical achievement rather than an automatic reflex.

That redirection also answers the charge of determinism frequently levelled against structural vocabulary. A position inclines, constrains, and selects; it seldom scripts. Probabilities rather than certainties are the natural output of structural claims, as they are for epidemiology, where exposure raises risk without deciding which exposed person falls ill. Smoking explains lung cancer at population level while leaving many smokers untouched and some non smokers afflicted, and nobody concludes that epidemiology lacks explanatory value because biography retains surprises. Social structural claims deserve assessment by the same standard: do they identify exposures, gradients, and mechanisms that shift distributions in predicted directions? Judged thus, many Marxist claims about commodification, precarious employment, and concentration of ownership have aged respectably as descriptions of distributional pressure, whatever verdicts history delivers on stronger prophecies attached to them. Weaker, probabilistic structuralism loses rhetorical grandeur and gains empirical respectability, a trade its defenders should accept openly.

Accepting supervenience also disciplines talk about emergence. Emergent, in careful usage, does not mean miraculous or exempt from underlying process. It means that a pattern becomes visible and tractable at a given grain while remaining invisible or intractable at finer grains, like a traffic jam that moves backward along a motorway although every car moves forward. Drivers experience the jam as an external object that swallows them at one point and releases them at another, and traffic engineers model its propagation without interviewing motorists about childhood or ambition. Yet the jam is nothing beyond cars, drivers, road geometry, and reaction times. Social structures resemble the jam more than they resemble a soul: external to each participant, dependent on all participants, patterned in ways no participant intended, and real in the only sense that matters for explanation, namely that ignoring them guarantees predictive and practical failure. A motorist who denied the jam because only cars exist would still arrive late.

Why does preference formation trouble the individualist starting point?

Because explanation that begins with given wants cannot explain how wants arose, yet wants bear the fingerprints of class, schooling, advertising, and workplace discipline everywhere. Treating preferences as bedrock protects them from inquiry, converting a central explanandum into an unexplained premise by methodological decree alone.

The trouble can be stated as a dilemma about scope. Either the individualist programme restricts itself to explaining conduct conditional on wants already formed, in which case its claims are modest, useful, and incomplete in a way Marxist theory exists to remedy, or it claims completeness for preference based explanation as such, in which case it must explain formation of the preferences it starts from, thereby admitting social processes prior to choice that shape the chooser. Sophisticated defenders often embrace the first horn in practice, building models that take wants as parameters while acknowledging that other disciplines study their origins. Difficulty arises when methodological rhetoric upgrades a modelling convenience into an ontological doctrine, declaring that only preference based explanation counts as genuine social science. At that moment a division of labour becomes a border wall, and questions about formation get exiled to psychology, history, or speculation, defined as external to social explanation proper. Marxist objection targets the border wall, not the models built inside it.

Formation operates through channels so familiar that their power hides in plain sight. Infants arrive without opinions about careers, cuisines, deities, or the moral status of wealth, and adults depart childhood with durable dispositions on all those matters, acquired through family speech, neighbourhood composition, religious observance or its absence, peer hierarchies, and the stories a culture repeats about success and shame. Schooling sorts and certifies, teaching punctuality, deference to evaluation, competition for marks, and identification of worth with measurable performance, alongside literacy and numeracy that genuinely liberate. Workplaces continue the education: shift systems train bodily rhythms, piece rates train attention toward speed, salaried autonomy trains self monitoring, and precarious scheduling trains a permanent readiness that colonises private life. Consumption environments train desire with professional skill, associating products with belonging, attractiveness, and care. None of this training requires a puppet master. It proceeds through institutions pursuing their own functions, families pursuing love and security, and individuals pursuing distinction or conformity for reasons that feel intimate. Intimacy of feeling, however, is no evidence of independence from formation, since the deepest social influences present themselves precisely as personal taste.

Adaptive preference poses the sharpest version of the problem. Persons deprived of options frequently cease to want them, and the cessation can reflect wisdom, exhaustion, dignity preserved through revaluation, or injury internalised as contentment. A worker who stops desiring higher education after repeated humiliation at school may report sincere satisfaction with a narrow horizon. A community denied clean water for generations may rank other goods above infrastructure it cannot imagine obtaining. Women socialised to regard ambition as unwomanly may genuinely prefer self effacement, and their sincerity complicates any outsider’s judgement while not dissolving the question of how the preference formed. If explanation stops at revealed or reported wants, deprivation that succeeds in lowering wants registers as welfare, and domination that secures consent registers as freedom from interference. Marxist and feminist critics converge here: a theory unable to distinguish wants formed under abundance of options from wants trimmed to fit deprivation lacks resources for criticising the trimming, however elegantly it models choice among the options remaining.

Elster, whose work on mechanisms and on sour grapes gave individualist social science some of its finest tools, recognised adaptive formation as a genuine phenomenon requiring causal history rather than mere registration of desire. That recognition matters for fairness in this dispute. The strongest individualist literature does not deny formation; it studies dissonance reduction, habit, addiction, and framing with empirical care. Disagreement narrows to whether formation can be handled as one more individual process, a sequence of stimuli and adjustments inside a biography, or whether formation refers explanation outward to relations that organise the stimuli: who owns the media that frame choices, who designs the curriculum, who controls the workplace where habits form, whose interests the built environment serves. A biography level account tells how a particular person came to want particular things. A structural account asks why millions of biographies bend in correlated directions, why messages cluster, why some voices amplify and others whisper, and why resistance takes the forms it does. Correlation across biographies is itself a social fact demanding social explanation, not a coincidence to be noted and set aside.

Advertising furnishes a concrete arena where correlated bending can be measured without metaphysical extravagance. Firms spend vast sums because exposure shifts demand, and they test messages, segment audiences, and exploit knowledge about insecurity, status competition, and parental anxiety produced by academic research. Consumers remain active interpreters who mock, resist, and repurpose campaigns, and campaigns fail regularly despite lavish budgets. Activity of the audience does not restore the image of sovereign preference, because the menu of images, the frequency of repetition, and the settings saturated by commercial speech are supplied asymmetrically. One side commands billboards, platforms, sponsorships, and data harvested at scale; the other side commands attention already fragmented by work and care. To model the outcome as an encounter between preformed wants and neutral information misdescribes the encounter. Wants are partly manufactured inside it, by actors who understand formation better than many theorists and invest accordingly. Political persuasion follows a parallel logic, with parties and movements shaping the identities in whose name demands are later voiced, so that expressed public opinion records both popular feeling and the machinery that organised its expression.

Identity deepens the difficulty beyond taste for goods. Preferences in standard models attach to outcomes, but persons also care about who they are, which group claims them, and what conduct befits that membership. A miner may accept danger and defend an industry harming his health because mining embodies masculine pride, communal memory, and loyalty to ancestors, commitments not reducible to a wage premium without grotesque flattening. A professional may refuse lucrative work that violates a self conception built over decades of training and peer regard. Such cases can be absorbed into utility language by adding identity as an argument of the function, and formal absorption always remains available, yet absorption purchased at that price empties the framework of discipline. When every anomaly becomes a new preference, explanation degenerates into redescription: whatever was chosen was preferred, because choosing reveals preference, because preference explains choice. Structural accounts of identity formation restore content by specifying institutions, rituals, and conflicts through which selves acquire their commitments, and by predicting which identities a given social order makes available, honoured, stigmatised, or literally unthinkable.

Unthinkability deserves separate emphasis, since formation governs not only ranking among options but the boundary of imagination itself. Serfs who regarded hierarchy as divinely ordained did not weigh egalitarian arrangements and reject them after deliberation; the arrangements lay outside the horizon within which deliberation occurred. Employees who regard managerial prerogative as natural as weather may bargain fiercely over pay while never formulating a claim to govern the enterprise, not because the claim was considered and priced, but because available concepts and organisational experience do not sustain it. Conversely, periods of rapid change are marked by ideas migrating from absurdity to common sense within a generation, as horizons expand under pressure from crisis, example, and organisation. A method that samples preferences inside a fixed horizon will register stability where a historical method perceives the horizon itself as the variable requiring explanation. Marx’s interest in ideology, Gramsci’s account of hegemony, and later work on symbolic power all address that variable, analysing how dominance secures itself by shaping what appears realistic, respectable, and sayable before any vote or purchase takes place.

Hegemony language, handled carelessly, invites an individualist rejoinder that should be taken seriously. If subordinate groups consent because their wants have been formed by dominant institutions, how does the critic claim knowledge of their real interests without arrogance, and how could emancipation ever begin among persons whose desires are supposedly manufactured? The rejoinder exposes genuine risks: vanguardist contempt for actual workers, unfalsifiable dismissal of expressed opinion as false consciousness, and a politics that substitutes the theorist’s preferences for those of the people in whose name it speaks. Structural theory at its most careful answers by grounding criticism in tensions inside lived experience rather than in a theorist’s privileged access to truth. Wants conflict within persons: security against dignity, loyalty against resentment, aspiration against fear. Formation is never total, because institutions contradict each other, because material life frustrates official stories, and because subordinate cultures preserve counter memories in jokes, songs, and family warnings. Critique amplifies those internal tensions, offering concepts for grievances already felt inarticulately. Where no tension exists and contentment is complete, criticism gains no purchase and should admit defeat rather than inventing hidden misery. That admission keeps formation theory empirical: it predicts friction, ambivalence, and sudden reinterpretation under changed circumstances, all observable, rather than positing a sealed ideological dome.

Changed circumstances indeed provide the best evidence that preferences are endogenous to relations. Migration transforms wants within a single lifetime as persons encounter different reference groups, price structures, and status orders, often reporting that desires once unthinkable became ordinary and desires once central faded. Wartime mobilisation reshaped attitudes toward women’s employment, state planning, and collective provision across many countries, and peace reshaped them again, unevenly. Deindustrialisation altered what young men in affected towns could plausibly want from work, masculinity, and place, producing political volatility that preference taking models registered as inexplicable mood swings. Digital platforms reorganised attention, sociability, and self presentation for cohorts who grew up inside their incentive structures, generating wants for visibility and immediacy that older cohorts find baffling. In each case the wants are real, sincerely held, and causally potent, and in each case their distribution tracks transformations in material and institutional environment with a fidelity that no theory of autonomous taste can explain away as coincidence.

None of this licenses dismissal of choice, responsibility, or the moral weight of individual decision. Formation explains dispositions statistically and historically; it does not issue excuses wholesale or convert persons into conduits. Courts, friendships, and political movements rightly hold agents answerable for cruelty, courage, honesty, and betrayal, because answerability itself is one of the practices through which selves form and communities endure. The structural claim concerns explanatory order. When a pattern repeats across thousands of answerable agents, citing answerability in each case does not explain the pattern, and when wants themselves display pattern, the pattern asks for causes located in shared conditions. Individualists who accept endogenous preferences inside their models, as many leading practitioners do, have in substance conceded the Marxist point while retaining tools Marxists would be foolish to discard. Remaining dispute then concerns emphasis and completeness: how far back explanation should travel, which causes deserve priority in a given inquiry, and whether a science that brackets formation can describe itself as a general theory of society rather than a powerful partial one.

Are explanatory levels allowed to be autonomous?

Yes, in the sense that matters for inquiry. A level earns standing by capturing robust regularities, supporting interventions, and answering questions framed at its grain. Chemistry, biology, and economics all proceed without waiting for derivation from fundamental physics, and dependence on lower processes does not demote their explanations to mere shorthand.

History of science supplies the precedent, and precedent here carries argumentative weight because reductionist rhetoric often presents itself as the voice of science against metaphysical indulgence. Chemistry before quantum mechanics explained reaction, valence, acidity, and the periodic pattern of elements with immense success, guiding industry and laboratory practice while its foundations in physics remained incomplete and partly mysterious. When quantum accounts eventually illuminated chemical bonding, they enriched chemistry without replacing its classifications, because reactions, reagents, and equilibria remain the working ontology through which chemists think, teach, and intervene. No practising chemist derives a synthesis from wave equations for a complex molecule when deciding conditions in a flask. The higher vocabulary persists not from laziness but because it isolates variables that matter at bench scale and screens off detail that would overwhelm prediction and manipulation. Autonomy, in this lived sense, means justified reliance on level specific concepts whose credentials come from performance, not permission granted after successful translation into a more fundamental idiom.

Biology sharpens the lesson because its central theory explains through history and population rather than through particle trajectories. Natural selection accounts for adaptation by variation, inheritance, and differential reproductive success, concepts defined at the level of organisms and lineages. A complete molecular description of every mutation and every death across a species over millennia is not a deeper version of the evolutionary explanation; it is a different object, unusable for the questions evolution answers about trait distribution, convergence, and extinction risk. Molecular genetics transformed biology by revealing mechanisms of inheritance that Darwin lacked, correcting errors and opening interventions undreamed of in the nineteenth century, yet the encounter between levels produced integration rather than elimination. Population thinking survived its encounter with the gene because selection operates on phenotypes in environments, and environment includes other organisms, climates, and constructed niches described in irreducibly ecological terms. Physicists who once predicted that biology would dissolve into chemistry watched instead a pluralistic science mature, with levels informing each other through bridge principles, models of mechanism, and occasional corrections in both directions.

Economics, whatever its contested status among the sciences, displays the same practical pluralism internally. Macroeconomic reasoning about liquidity, aggregate demand, and financial instability addresses phenomena, bank runs, panics, depressions, whose description requires expectations about expectations, a recursive social structure poorly represented as a sum of isolated choices. Microeconomic reasoning about substitution, marginal cost, and local knowledge corrects macroeconomic fantasy and grounds policy in incentive effects that aggregate models miss. The healthiest episodes in the discipline involve traffic between levels, with each exposing blind spots in the other. The least healthy episodes involve imperial claims that one level alone is real and the other a folk superstition awaiting elimination. Marxist structural argument belongs, at its most defensible, to the traffic rather than to an empire of the macro, insisting that relations, distributions, and institutional configurations capture regularities that preference aggregation misses, while granting that any structural claim unable to survive contact with situated conduct deserves revision or abandonment.

Interlevel relations, on the pluralist picture, run in both directions and take several forms. Lower level findings constrain higher level theory by ruling out mechanisms that could not possibly operate, as neuroscience rules out certain psychologies and as accounting identities rule out certain economic fantasies. Higher level findings constrain lower level inquiry by identifying which among countless micro processes deserve attention, as epidemiology directs laboratory search toward exposures that population data incriminate. Constitution links levels when a higher fact consists in organised lower activity, as a market consists in exchanges conducted under rules. Context links them when higher arrangements alter the effects of identical lower dispositions, as the same ambition produces entrepreneurship in one setting, bureaucratic intrigue in another, and criminal enterprise in a third. A science mature about its levels maps these links case by case instead of announcing a universal direction of explanation in advance of inquiry. Social science, dealing with reflexive and historically variable material, has even stronger reason for such modesty than the natural sciences that reductionist rhetoric invokes as model.

Modesty also clarifies what Marxist explanation claims about laws. Regularities at the structural level are typically tendencies operating through counteracting influences, valid for configurations that persist for a time and dissolve under specifiable pressures. A tendency toward concentration of capital coexists with antitrust enforcement, technological disruption favouring entrants, and political resistance, and the outcome in any sector reflects the resultant of forces rather than the unfolding of an iron necessity. Tendency language has been abused as an escape hatch, invoked whenever prediction fails, and critics are right to demand that counteracting influences be named in advance with their own evidence. Properly used, however, tendency is the honest form of most social generalisation, matching the open character of social systems where learning, invention, and conflict continually alter parameters that natural systems hold stable. Demanding particle style law as the price of admission to science would expel evolutionary biology, geology, meteorology, and medicine alongside sociology, a price reductionist manifestos rarely acknowledge they are asking everyone to pay.

The structural reply, assembled across these arguments, can be summarised without slogans. Relations bind occupants independently of individual endorsement, although they depend on enactment distributed across populations and generations. Positions channel conduct through feasible sets, incentives, and selection pressures that persist through turnover of personnel and variation of motive. Social facts supervene on individual facts while resisting translation into them without loss of the generalisations that make social facts worth naming. Preferences, beliefs, and identities form inside relations, so any method enthroning them as unexamined origins mistakes an intermediate link for a foundation. Explanatory levels justify themselves through captured regularities and successful intervention, and by that standard structural explanation has earned standing in domains ranging from labour markets to organisational convergence to the patterned formation of desire. Each element of the reply concedes something real to the opposing camp: enactment, mechanism, dependence, and the permanent possibility of correction from below. The concessions are not decorative. They define a structuralism that has absorbed the individualist critique and continues the argument on ground that critique helped clear.

What remains genuinely contested, after concessions on both sides are honoured, is narrower and more interesting than the shouting version of the debate. Contest persists over how often structural regularities are robust enough to support explanation without detailed mechanism, over which mechanisms deserve priority when several fit the evidence, over the ethics of attributing interests that agents disavow, and over the political implications of treating wants as formed rather than sovereign. Those questions admit progress through case work: comparative history, organisational ethnography, formal models with endogenous formation, natural experiments created by policy change, and careful listening to agents whose self understanding is data rather than decoration. A reader who finishes this section with a taste for that case work, and with suspicion toward any method claiming in advance to have settled every dispute of level, has received the intended lesson. Explanation in social life is a layered craft. The layers quarrel, correct, and occasionally illuminate each other, and the craft advances when no layer is permitted to silence the rest by definition.

A final consideration returns to the worker and the capitalist with whom the structural case began, because abstraction should answer to the lives that prompted it. The worker who must sell labour power is not a puppet, and the capitalist who must accumulate is not a villain in a morality play, and structural explanation that turns either figure into a cartoon has failed its subject. Both deliberate, improvise, love, err, and occasionally act against every pressure their position applies, at costs that measure the pressure more eloquently than any treatise. Explaining the pressure is compatible with honouring the deliberation, just as explaining a river in flood is compatible with admiring a swimmer. Reduction to preferences fails, in the end, not because preferences are unreal or unimportant, but because they are downstream of relations that also deserve names, analysis, and, where judgement finds them wanting, transformation. A social science that can hold both truths together, upstream relations and downstream choices in a single account, serves understanding better than either orthodoxy guarding its level against the other, and the debate over micro-foundations is most valuable when it forces that integration rather than awarding a monopoly to one side.

The collective action problem for class

Classical accounts of class politics move with an ease that later scrutiny found difficult to sustain. A group occupies a shared position inside relations of production, develops interests opposed to those of another group, and then, through struggle, acts upon the antagonism that its position created. Each step in that sequence appears plausible when described at a distance. Proximity dissolves the smoothness. Between recognising a grievance and bearing the cost of redressing it lies a gap that interest alone does not close, and the most influential statement of why that gap persists came from outside Marxist theory altogether. Mancur Olson published The Logic of Collective Action in 1965, and the argument developed there has shadowed every subsequent defence of class agency, whether defenders acknowledged the shadow or preferred to work inside it without naming the source.

Olson began from a premise that much political writing had treated as settled. Where people share an objective, organisation follows because advancing the objective serves each person who belongs. A trade association, a professional body, a union, a revolutionary party: all could be read as natural expressions of purposes held in common. His intervention denied the inference. Sharing a purpose supplies a reason to desire an outcome. Supplying effort toward that outcome is a separate decision, governed by a separate calculation, and under identifiable conditions the second decision points away from participation even when the first points strongly toward the goal. The distinction sounds modest. Its consequences reach far, because it separates the sociology of interests from the explanation of action and demands that any theory connecting them specify a mechanism capable of carrying individuals across the interval.

The mechanism problem arises from the character of the goods at stake. Some benefits, once secured, cannot be withheld from people who did nothing to secure them. A cleaner street benefits residents who never joined a clean-up campaign. A change in law benefits citizens who never campaigned for reform. A rise in wages across an industry benefits employees who kept working during a strike exactly as it benefits employees who walked out and lost pay. Olson called attention to this non-excludability and traced its effect on reasoning. If enjoyment of the gain does not depend on personal contribution, then contribution purchases nothing additional for the contributor beyond a share available regardless. Cost, by contrast, remains stubbornly private. Time given to organising cannot be recovered. Money paid in dues leaves a household budget diminished. Risk of dismissal, demotion, blacklisting, or arrest attaches to identifiable persons rather than dispersing across a category. Faced with private cost and shared reward, a rational member has reason to let others bear the burden, hoping to receive the improvement without paying for its production. That posture has a familiar name: the free rider problem.

Group size sharpens the difficulty rather than softening it. Inside a tiny circle, one person’s withdrawal visibly weakens the effort, and others notice absence, remember it, and adjust their regard accordingly. Contribution can plausibly alter the result, so effort retains a perceptible connection to outcome. Scale dissolves both conditions. Among thousands or millions, a single additional participant rarely decides success or failure, and anonymity shelters non-participation from observation. Each member can reason that personal involvement will make no appreciable difference while personal abstention will attract no appreciable notice. When many members reason identically, the aggregate effect is paralysis produced not by apathy, disagreement, or false belief, but by the structure of incentives confronting people who may understand their situation perfectly well. Olson’s unsettling suggestion was that failure to organise can be a product of rationality operating under adverse conditions, rather than evidence that a group lacks interests or awareness.

Applied to class, the argument acquires particular force because a class is among the largest groups political theory has ever asked to act. Higher wages benefit workers who strike and workers who stay at work alike. Shorter hours, once conceded by employers or imposed by legislation, extend to the cautious and the militant without distinction. Political transformation on the scale envisaged by revolutionary traditions would alter conditions for an entire population of wage earners, including those who opposed the transformation, feared it, ignored it, or quietly relied on braver colleagues to attempt it. Class interest, however objectively grounded in the wage relation, does not translate into class action without further explanation. A worker may detest low pay, resent arbitrary management, and wish fervently for a different order, yet still conclude that joining a stoppage, funding an organisation, or confronting the state offers personal loss in exchange for a negligible increment to a collective prospect. Preference and participation part company.

Classical Marxism possessed rich resources for describing exploitation, crisis, ideology, and historical change. What the tradition supplied less fully was an account of individual contribution under conditions where contribution is costly and dispensable. Appeals to historical necessity described direction without explaining motivation at the level of the person deciding whether to attend a meeting after a shift. References to the discipline of factory life described shared experience without showing why shared experience overcomes the incentive to hang back. Expectations that crisis would radicalise populations described moments of heightened feeling without specifying why feeling should convert into sustained, organised, risky effort rather than into private adaptation, emigration, resentment directed sideways, or exhaustion. The gap is not a minor omission at the edge of an otherwise complete edifice. Agency sits near the centre of any politics that expects the exploited to become authors of transformation, so a missing account of agency leaves a central promise unsupported.

For generations, the supposed bridge across that interval was consciousness, and the fuller discussion of class consciousness explained sets out how awareness of position was expected to convert position into purpose. Consciousness mattered because it promised more than information. A worker who understood exploitation as systematic rather than personal might identify with others similarly placed, interpret private hardship as collective injury, and accept sacrifice as meaningful within a shared project. Olson’s challenge survives that promise in a sharpened form. Even granting full clarity about interests, even imagining away every distortion attributed to ideology, the calculation regarding contribution remains. Knowledge tells a person where advantage lies for the group. Knowledge alone does not establish that personal effort is worth personal cost when the fruits of success arrive irrespective of effort. A theory may therefore solve the problem of awareness completely and still leave the problem of action untouched.

That persistence is why the difficulty deserves treatment as a genuine unsolved difficulty for class-based politics regardless of anyone’s methodology. Analytical Marxists pressed the point from inside a sympathetic reconstruction of historical materialism. Rational choice theorists pressed the point from a liberal individualist starting position. Historians of labour movements encountered versions of the puzzle in archives full of failed appeals, lapsed memberships, and strikes that collapsed when support proved thinner than rhetoric suggested. Trade union organisers met the puzzle in doorstep conversations long before encountering formal statements of it. Convergence across such different vantage points indicates that the obstacle belongs to the phenomenon rather than to a fashion in theorising about the phenomenon. Methodological commitments determine how a writer phrases the obstacle, which solutions appear respectable, and which evidence counts as decisive. No methodological commitment dissolves the underlying structure: benefits diffuse across a vast category, costs concentrate on identifiable contributors, and each contributor’s marginal impact shrinks as the category grows.

Plain statement serves honesty better than protective qualification here. The problem remains unsolved. Partial mechanisms exist, documented in workplaces, movements, and organisations across many countries and periods, and later portions of this section examine those mechanisms in detail. Partial mechanisms, however, are not a general solution. A general solution would explain how contribution becomes rational, habitual, or otherwise reliably produced among millions of dispersed people whose individual efforts are neither decisive nor easily observed, without quietly presupposing the solidarity, discipline, or organisational strength whose emergence requires explanation. No candidate answer has yet met that standard to a degree commanding wide assent even among writers sympathetic to class politics. Pretending otherwise, by treating familiar vocabulary as if vocabulary were a mechanism, weakens the very tradition such pretence aims to defend.

Critics sometimes respond that Olson modelled people too narrowly, as if every decision emerged from a ledger of material gain and loss computed in isolation. The objection carries weight against a crude reading, yet less weight against the core argument than its proponents often assume. Broadening motivation to include dignity, anger, loyalty, moral conviction, and desire for respect changes the contents of the calculation without removing the structural feature that generates the puzzle. A person moved by anger still faces danger when confronting an employer backed by law. A person devoted to comrades still sacrifices income during a stoppage while a neighbour who shares none of that devotion receives identical improvements afterward. Unless devotion is distributed widely and reliably enough to sustain action at scale, invoking richer motivation relocates the inquiry toward the distribution of devotion itself. Why do some acquire it while others do not? Why does it endure under defeat in certain settings and evaporate under far milder pressure elsewhere? Those questions are extensions of Olson’s problem, not refutations of it.

Historical evidence complicates any simple dismissal because workers plainly have organised, struck, built parties, and occasionally overturned states. Existence of action, however, does not refute a theory about the obstacles confronting action. Bridges exist despite gravity; their existence directs attention toward engineering rather than toward denial that weight presses downward. Labour history, read through Olson’s lens, becomes a record of devices invented to counteract free riding: dues check-offs that removed repeated decisions about payment, picket lines that made absence visible, social sanctions within tight communities, strike funds that reduced the private cost of participation, and legal frameworks that altered the risks attached to membership. Each device testifies to the reality of the tendency it counteracted. Movements did not flourish by assuming contribution would flow spontaneously from interest. Movements flourished, when movements flourished, by constructing environments where contribution became easier, safer, more visible, or harder to avoid. The historical record thus supports the problem’s importance even where outcomes contradict a prediction of universal inaction, a prediction Olson himself never made in that crude form.

Careful reading also matters because the 1965 argument is often caricatured as a claim that large groups never act. Olson’s actual position distinguished among group types, allowed that certain large organisations plainly existed, and sought to explain their existence through features additional to shared purpose. Latent groups, in his terminology, possessed interests without automatic organisation. Privileged groups contained at least one member for whom the benefit of action exceeded its total cost, giving that member reason to provide the good alone. Intermediate groups were small enough for members to notice one another and bargain. The taxonomy matters for class analysis because a proletariat, defined by position in production rather than by membership lists, fits the latent category uncomfortably well. Nobody receives a membership card from the wage relation itself. Boundaries blur at the edges where supervisors, contractors, small proprietors, and salaried professionals occupy ambiguous positions. Numbers run into tens of millions within a single national economy. If any collectivity should struggle to convert interest into coordinated effort, a class should, and the burden falls on class theory to identify the additional features that historically overcame that struggle in particular places and periods.

What did Olson’s study establish about group action?

Olson showed that shared interest alone cannot explain joint action in large groups, because each person gains the benefit whether personal effort helped secure it or not, while bearing private cost alone, so abstention becomes rational unless contribution is separately rewarded, compelled, or sustained by relations that alter the calculation for participants.

Establishing that distinction between interest and action reorganised debates far beyond economics, where the book first found its audience. Political science absorbed the lesson into accounts of lobbying, where concentrated industries routinely out-organised diffuse consumers despite consumers vastly outnumbering producers and despite aggregate consumer losses exceeding aggregate producer gains. Environmental politics absorbed the lesson into explanations of why polluters, few and well resourced, often defeated majorities favouring cleaner air whose individual stakes in any single regulatory battle appeared small. Sociology absorbed the lesson into studies of movements that succeeded only after building infrastructures of meetings, publications, and local chapters capable of converting sympathy into assignments. In each field, the durable contribution was a discipline of questioning: whenever a writer attributed action to interest, a reader could ask what carried particular people from wishing to doing, and the question retained force even when the eventual answer proved complicated.

The establishment of that question also exposed a weakness in functional explanation that later critics of classical Marxism developed at length. Functional explanation attributes the existence of an institution or pattern to beneficial consequences the institution produces for some group or system. Workers need representation, therefore unions arise. Capital requires certain conditions, therefore states supply those conditions. Olson’s reasoning interrupts the slide from need to existence. Needs do not summon their own satisfaction into being. Between a beneficial consequence and the behaviour producing that consequence lies a chain of individual decisions, each vulnerable to the incentive to abstain. A satisfactory explanation must traverse the chain link by link, showing at every step why acting made sense, or became habitual, or was compelled, for the people whose combined behaviour generated the outcome. Where the chain breaks, beneficial consequences remain unrealised, and history fills with interests that never found organisational expression, grievances that never became movements, and majorities that never became forces.

Subsequent scholarship refined the original formulation without overturning its central insight. Experiments in behavioural economics demonstrated that many experimental subjects contribute to shared funds at rates higher than a strict free rider prediction suggests, especially during early rounds of a game, yet contributions typically decay as rounds repeat and participants observe others withholding. Field studies found cooperation flourishing under monitoring, communication, and graduated sanctions, conditions that communities sometimes build deliberately. Network analysis showed that position inside a web of relationships predicts participation better than attitude surveys alone. Each refinement qualifies the starkness of the model while confirming its diagnostic value. Human beings are not relentless maximisers sealed inside private preferences. Human beings are also not automatic collectivists whose shared circumstances generate shared action without mediation. Between those caricatures lies the terrain where explanation must operate, mapping the conditions under which cooperation survives contact with scale, anonymity, and risk.

One further clarification protects the argument from a misunderstanding that recurs in polemical exchanges. Free riding describes a structural position, not a moral character. A worker who declines to strike may be caring for a sick parent, servicing debts that permit no interruption of income, doubting the competence of strike leadership, or judging the specific demand unwinnable. A worker who crosses a picket line may hold views indistinguishable from those of strikers on every question except timing and risk. Olson’s vocabulary tempts moralisation because riding free sounds like parasitism, yet the analytical point concerns incentives confronting anyone placed inside the structure, saint or opportunist alike. Indeed, the problem gains its sting precisely from applying to conscientious people. If only the selfish abstained, exhortation might suffice. When the prudent, the encumbered, and the sceptical also abstain for intelligible reasons, exhortation reaches its limit and institutional design becomes unavoidable.

Thinking about design, however, immediately raises a second-order version of the same difficulty. Building an organisation is itself a collective undertaking. Founders must invest time before any selective benefit exists to reward investment. Early joiners accept risks before protection reaches meaningful strength. A founding meeting can be infiltrated, a founding leaflet can cost a job, a founding fund can be confiscated. Who supplies the initial contribution that creates the machinery capable of sustaining later contributions? Historians answer with particular names, neighbourhoods, chapels, pubs, ethnic associations, craft traditions, and political sects whose prior solidarities seeded labour organisation. Such answers are empirically rich and theoretically revealing, because each seed represents social capital accumulated under conditions different from the class-wide setting whose explanation is sought. The seed explains a beginning. Extending the explanation from seed to scale, from a chapel congregation or craft workshop to a national working class, remains the unfinished labour.

Selective incentives were Olson’s own route out of the impasse his argument created, and the route deserves examination before alternatives enter the discussion. A selective incentive rewards contributors with a benefit withheld from non-contributors, or punishes non-contributors with a penalty contributors escape. Insurance schemes tied to union membership, access to legal representation, training credentials, hiring hall privileges, and social facilities restricted to members all exemplify the positive form. Loss of standing, exclusion from desirable shifts arranged through collective agreements, or fines exemplify the negative form. Where such incentives operate, participation purchases something private alongside the shared gain, and the calculation shifts. A large organisation can then exist as a by-product: the collective good emerges from an institution sustained by private goods, much as a commercial firm produces social effects beyond the wages and profits motivating its participants.

Elegance should not be mistaken for completeness. Selective incentives explain why people join and remain inside organisations capable of dispensing excludable benefits. Class scale strains the explanation at several points. Dispensing benefits requires administrative machinery, funds, staff, and legal standing, all of which presuppose an organisation already functioning at considerable size. A revolutionary transformation, as distinct from a wage settlement, offers no office from which to distribute rewards beforehand and no credible schedule of private payments capable of offsetting the extreme risks of insurrectionary participation. Benefits that motivate routine membership, such as insurance or representation in a grievance hearing, bear little proportion to dangers involved in confronting armed authority. Moreover, the most consequential collective goods sought by class politics, including a reordered economy, cannot be packaged as membership perks. A transformed society, if achieved, arrives for members and non-members together, restoring the original non-excludability at the highest level of ambition. Selective incentives therefore illuminate the maintenance of unions and parties as organisations while leaving the deepest objectives of those organisations only partly explained.

Commercial analogies sometimes conceal that limitation. A person joins an automobile association for roadside assistance and incidentally funds lobbying; a person joins a union for representation and incidentally funds political campaigns. By-product theory handles such layering convincingly where the private service is valuable, deliverable, and separable from the collective purpose. Difficulties multiply when members disagree about the collective purpose funded by their dues, when employers or states can supply rival private services designed to detach workers from independent organisation, or when the organisation must ask members for sacrifices vastly exceeding the value of services received. A strike call tests the by-product structure severely. During a long dispute, insurance and camaraderie may sustain commitment, yet the decisive sacrifices involve wages forgone in pursuit of a shared outcome, returning participants to exactly the terrain where selective rewards shrink relative to collective stakes. Organisations survive such tests regularly enough to prove the rewards matter. Failures under comparable pressure reveal that those rewards fall short of a general account.

Why does class interest not translate into class action?

Wages, hours, and political change, once won, reach strikers and non-strikers alike, so a worker can share the gain without risking dismissal, lost pay, or arrest, and millions of strangers cannot monitor one another, which means interest supplies a reason to welcome victory but not a reason to help produce it.

Translation failure operates at several levels simultaneously, and separating those levels clarifies why remedies aimed at a single level disappoint. At the level of information, a worker may misjudge where interest lies, although Olson’s problem persists even after every informational defect is repaired. At the level of trust, a worker may doubt that enough colleagues will participate to give action a chance, and doubt alone can be self-fulfilling because each person’s hesitation confirms the hesitation of others. At the level of efficacy, a worker may trust colleagues completely yet judge the employer, the market, or the state too powerful for the proposed action to succeed. At the level of priority, a worker may endorse the goal while ranking rent, childcare, health, or migration plans above the meeting scheduled for the evening. A successful mobilisation must clear all four hurdles for enough people at the same moment. Failure at any single hurdle returns the group to inaction, even when every other condition appears favourable. Class interest, stated abstractly, clears none of the hurdles by itself.

Threshold dynamics deepen the analysis in ways associated with later formal work on cascades and tipping points. Individuals differ in how much visible support they require before joining. A small number will act almost regardless of company. Others join once a handful of respected colleagues commit. Others wait until participation looks like a majority practice. Still others join only when victory appears nearly certain, contributing least while arriving in time to share credit. A mobilisation succeeds when early participants trigger successive waves, each wave lowering the perceived risk for the next. A mobilisation fails when the opening wave stalls below the threshold of the second group, leaving pioneers exposed and confirming widespread caution. Outcomes can therefore hinge on sequencing, visibility, and the reputations of early movers, factors poorly captured by aggregating interests across a population. Two workplaces with identical pay, identical grievances, and identical demographics can diverge sharply because in one setting a respected figure moved first and in the other setting a comparable figure hesitated. Contingency of that kind frustrates any theory expecting action to follow mechanically from structure.

Employers and states, for their part, rarely remain passive while thresholds approach. Counter-mobilisation exploits the same logic from the opposite direction. Singling out early organisers for discipline raises the cost borne by pioneers and interrupts cascades at the point of greatest vulnerability. Offering concessions to a strategically chosen segment divides a workforce into groups whose thresholds and interests no longer align. Shifting production across sites weakens beliefs about efficacy by demonstrating that local sacrifice may simply relocate employment elsewhere. Delay exhausts funds and attention, allowing private pressures to reclaim people whose commitment depended on momentum. Legal restrictions on secondary action prevent sympathetic groups from pooling strength, fragmenting a potential class-wide effort into isolated disputes each facing the full disparity of resources alone. None of these tactics requires employers to persuade workers that exploitation is benign. Tactics succeed by manipulating the conditions under which already aggrieved workers decide whether acting is worth the exposure. Persuasion about interests and manipulation of participation costs are different arts, and the second has often proved more effective than the first.

Geography and industrial structure add further fragmentation that abstract class categories flatten. Workers compete for jobs across regions and borders, and that competition is a lived reality rather than a failure of understanding. A wage gain in one plant can accelerate automation or relocation, pitting the long term interests of a local group against gains available elsewhere in a wider labour market. Occupational hierarchies divide manual and supervisory staff, permanent employees and agency staff, citizens and migrants, each division carrying distinct legal exposure and distinct fallback options. Gendered divisions of unpaid labour shape who can attend evening meetings, who absorbs the domestic consequences of lost wages, and whose participation gets recorded as commitment while comparable contributions at home remain invisible. A category as capacious as the working class contains multitudes whose situations intersect without coinciding. Interest at the level of the category is real in the sense that all depend on selling labour power or on a household member doing so. Interest at the level of decision-making groups is fractured, uneven, and sometimes directly opposed across short time horizons. Olson’s framework predicts exactly such fracturing effects, because smaller subgroups with distinct stakes organise more readily than the encompassing category, even when subgroup organisation obstructs class-wide goals. Craft exclusiveness, sectional wage bargaining, and protection of national employment against foreign competition all illustrate subgroup rationality defeating encompassing solidarity.

Generational experience introduces still another layer that interest-based accounts handle awkwardly. Cohorts entering employment during expansion encounter unions as established features of the landscape, inherit expectations of gradual improvement, and may treat organisation as a service rather than a cause. Cohorts entering during defeat encounter closed workplaces, weakened protections, and narratives portraying collective effort as futile or dangerous. Identical objective positions thus arrive wrapped in divergent inherited expectations about what action can achieve and what action typically costs. Expectations, once formed, guide the interpretation of fresh events: a lost strike confirms futility for one cohort while teaching tactical lessons to another. Path dependence of this sort means that the capacity for action at any moment reflects a history of previous attempts, victories, betrayals, and settlements, layered over the structural position that class theory foregrounds. Structure sets the stage repeatedly. History decides which actors believe the play can end differently.

Small groups offer a contrasting setting where several of the obstacles weaken together, and workplace organisation has always drawn strength from that contrast. Face to face settings change the calculation because absence becomes conspicuous, contributions become attributable, and relationships accumulate obligations that extend beyond a single dispute. A team sharing a shift observes who volunteers, who hangs back, who can be trusted under pressure, and who performs enthusiasm while avoiding exposure. Reputation within such a circle carries tangible value: cooperation in future emergencies, inclusion in informal exchanges of favours and information, protection against isolation when a supervisor applies pressure. Refusal also carries tangible cost, registered in cooled greetings, withheld assistance, and diminished standing among people encountered daily. Under those conditions, participation may be sustained by the ordinary fabric of social life rather than by extraordinary ideological commitment. Many durable movements grew from exactly such fabrics, rooting formal organisation in chapels, clubs, extended families, and neighbourhood networks where people already monitored and valued one another.

Scale, however, refuses to disappear merely because small units function well. A class is not a small group, and federation among small groups reproduces collective action difficulties at a higher level. A local branch may discipline its own members effectively while facing incentives to let other branches carry the burden of a national campaign. A union in a profitable sector may settle separately rather than prolong sacrifice for workers in a weaker sector whose victory would benefit the movement broadly. A national leadership may invoke solidarity while regional bodies calculate the local price of compliance. Each level of federation presents fresh opportunities for abstention dressed as prudence, autonomy, or fidelity to immediate members. Historical movements devised congresses, levies, and binding mandates to manage inter-unit free riding, with uneven success and frequent conflict over whether central discipline represented solidarity or bureaucratic domination. The problem therefore scales fractally: solving it inside a workshop leaves analogous problems between workshops, between industries, and between countries, each demanding additional institutional invention.

Workplace concentration, celebrated in classical writing as the proletariat’s school of cooperation, deserves a balanced assessment under this light. Factories gathered large numbers of workers into shared spaces, synchronised their routines, and made common conditions unmistakable. Gathering also placed workers under surveillance, subjected gatherings to managerial permission, and tied livelihoods tightly to a single employer’s goodwill. Concentration facilitated communication and intimidation simultaneously. Dispersed forms of employment, from domestic service to agriculture to platform-mediated delivery, remove even the communicative advantage while retaining vulnerability. Gig arrangements radicalise the Olsonian condition: workers may never meet, may compete through algorithmic allocation for each task, may be classified as independent contractors, and may face deactivation without any visible adversary to confront collectively. Where classical theory expected capitalist development itself to assemble its own gravediggers into ever larger battalions, segments of contemporary employment assemble isolation instead. A theory of class action adequate to dispersed labour must explain coordination among people whom production never brings into contact, a harder assignment than explaining solidarity inside a mill.

Coercion presents itself as an answer whenever voluntary mechanisms falter, and labour history supplies abundant instances. Closed shops made union membership a condition of employment, converting an individual choice into a structural requirement. Union security agreements deducted dues automatically, removing the recurring decision to pay. Picket lines applied physical and social pressure against entering a struck workplace. States sympathetic to organisation sometimes compelled recognition, mandated bargaining, and protected membership from employer retaliation. Each arrangement demonstrably raised participation and stabilised organisations that voluntary recruitment alone might never have sustained. Explanation by compulsion, however, relocates the question rather than resolving the underlying puzzle. Who established the closed shop, and through what prior mobilisation was that establishment achieved? Who enforces the rule when enforcement itself requires effort and exposes enforcers to retaliation? Compulsion presupposes a coalition already strong enough to impose and police compulsion, so citing compulsion as the foundation of strength argues in a circle at the point where foundations matter most.

Circularity aside, coercion carries costs that movements ignore at their peril. Participation secured by penalty differs in quality from participation secured by conviction, and the difference surfaces during crises demanding initiative rather than mere compliance. A member retained by a closed shop may pay dues reliably while declining every voluntary task, withholding the discretionary effort on which campaigns depend. Resentment against compulsion can be harvested by employers or rival organisations promising liberation from union discipline, turning the machinery of solidarity into evidence for hostile narratives. Democratic legitimacy suffers when members experience the organisation as another authority extracting payments and issuing instructions, mirroring the workplace hierarchy the organisation exists to contest. None of these costs proves coercion unjustifiable; many defensible institutions, from taxation to conscription in wartime, rest on compulsion justified by the impossibility of voluntary provision. Costs do prove that coercion is a technology with trade-offs requiring political judgement, not a theoretical solvent dissolving the free rider difficulty without residue.

Repeated interaction offers a subtler mechanism, grounded in the logic of ongoing relationships rather than in formal penalties. Where people expect to deal with one another indefinitely, defection on one occasion invites retaliation at a later encounter, and the shadow of the future disciplines conduct in the present. Game theorists formalised this insight through iterated dilemmas in which conditional cooperation can stabilise once participants value future encounters sufficiently. Workplaces approximate iterated settings because colleagues meet across years, depend on mutual cover, and exchange assistance whose withdrawal would hurt. Reputation aggregates the history of those exchanges into a portable judgement: reliable people receive trust, invitations, and support, while unreliable people find doors closing without any formal proceeding. Such informal enforcement can sustain contributions that one-shot calculations would predict to vanish, and ethnographies of factories, docks, and mines document elaborate reputational economies governing effort, loyalty, and mutual aid.

Limits appear when horizons shorten or membership churns. Casualised employment truncates the future that disciplines the present, because a worker expecting dismissal or transfer within months gains little from investing in local reputation. High turnover dissolves the shared memory on which reputational judgements depend, allowing defection to escape record. Mobility between employers, regions, and occupations weakens the reach of any single community’s sanctions. Anonymity returns through the side door whenever scale or churn prevents stable recognition. Furthermore, reputational systems enforce conformity to whatever norms a group happens to hold, which may include norms against organising as readily as norms favouring it. A workplace culture prizing individual advancement, deference to management, or suspicion of outsiders will deploy reputation against activists with the same efficiency that a militant culture deploys reputation against strikebreakers. Repeated interaction explains how norms, once established, perpetuate themselves. Establishment itself, and the direction norms take, require separate explanation rooted in history, leadership, and circumstance.

Social norms widen the discussion from enforcement toward internalised obligation, and no treatment of the subject can avoid Jon Elster, whose work traversed the territory between rational choice and norm-guided behaviour with unusual candour about difficulties encountered along the way. Elster, a Norwegian philosopher and social theorist sympathetic to Marxism’s explanatory ambitions while critical of its functionalist habits, examined collective action in works including Ulysses and the Sirens, The Cement of Society, and his contributions to analytical Marxism during the nineteen eighties. Norms, in his account, are shared injunctions sustained by emotional responses such as shame and contempt, operating partly independently of outcome calculations. A person may contribute because abstention feels disgraceful in the eyes of relevant others and in the person’s own eyes, even where material reasoning counsels withdrawal. Norms of fairness prove especially relevant: many people will punish free riders at personal cost, reject grossly unequal divisions even when rejection leaves them worse off, and contribute conditionally when assured that others contribute likewise.

Conditional cooperation deserves emphasis because survey and experimental evidence repeatedly identify it as a widespread disposition. Large numbers of people describe willingness to do their share provided others do theirs, alongside refusal to be exploited as the sole contributor while neighbours idle. Mobilisation, on this picture, is less about manufacturing generosity from nothing than about providing credible assurance that generosity will be matched. Public commitments, visible turnout, transparent counts of pledges, and rituals of mutual recognition all function as assurance devices. Failure, conversely, often reflects assurance collapse: rumours of low turnout, defections by prominent figures, or memories of previous efforts abandoned midway convince conditional cooperators to protect themselves by staying home. Leaders intuitively grasp assurance dynamics when they inflate expected attendance, stage early demonstrations of strength, or secure endorsements from trusted intermediaries before asking ordinary members for exposure. The craft of organising consists substantially in managing beliefs about what others will do, because those beliefs mediate between disposition and deed.

Elster’s later doubts about norms give this section of the debate its distinctive honesty. Having deployed norms to explain behaviour that rational choice models missed, he grew sceptical about the ease with which theorists invoke norms as residual explanations whenever calculation fails. A norm cited without independent evidence risks becoming a label for the behaviour it purports to explain: people contributed because a contribution norm existed, and the norm’s existence is inferred from the contribution. Circularity of that kind mirrors the solidarity problem examined below. Elster also questioned how norms originate, how competing norms within the same person get adjudicated, and how norm-guided action interacts with strategic reasoning when actors anticipate how displays of norm adherence will be read by audiences. His scepticism never amounted to denial that norms move people. Scepticism amounted to a demand for mechanisms: specify the emotions involved, the audiences whose judgement matters, the sanctions applied, and the conditions under which the norm yields to fear, fatigue, or countervailing obligation. Without such specification, norm talk decorates an explanation rather than completing one.

Does an appeal to solidarity answer the problem?

An appeal to solidarity names the willingness to act for others despite personal cost, yet the puzzle asks why such willingness arises and persists at scale, so invoking the disposition explains action only by assuming beforehand the very disposition whose origin, strength, and durability the collective action problem calls into question.

The standard correction stated in that answer paragraph applies across a wide family of responses that share its logical shape. Solidarity is the name for the disposition whose existence the problem questions. Treating the name as an explanation commits the error of answering a question by restating the question in warmer vocabulary. If workers act together because workers feel solidarity, the next inquiry asks why feeling reached sufficient intensity among sufficient numbers at the decisive moment, why feeling survived the first defeat, why feeling extended beyond the familiar faces of a single department to strangers in distant towns, and why feeling outweighed competing attachments to family security, religious counsel, national loyalty, or personal ambition. Each sub-question is empirical and historical, answerable in particular cases through narrative and evidence. None is answered by the word itself. Vocabulary that conceals unanswered questions behind resonant affirmation may serve morale at a rally. Vocabulary of that kind fails the standards of a theory claiming to explain historical agency.

Distinguishing levels of solidarity helps expose where genuine explanatory work begins. Solidarity within a family or friendship circle rests on affection and lifelong reciprocity, resources rarely available across a class. Solidarity within an occupation rests on shared craft pride and daily interdependence, resources available in pockets. Solidarity within a movement rests on constructed identity, common rituals, shared enemies, and remembered victories, resources that organisations labour to create and maintain. Solidarity within a class as such would rest on recognition of common position among people divided by occupation, region, language, faith, and citizenship status, the most demanding construction of all. Sliding between these levels, citing the undeniable warmth of the first as evidence for the feasibility of the last, is a rhetorical habit that careful analysis should refuse. Each level requires distinct supports, faces distinct rivals, and breaks under distinct pressures. A history of dockers sustaining one another through a lockout demonstrates something real about dockers in a port community. Generalising from that history to the capacities of wage earners everywhere demands argument about transferability, and transferability is exactly where Olson’s scale argument regains its footing.

None of this denies that solidarity, however generated, changes outcomes when present in strength. Strikes held together by mutual commitment have endured hardships that incentive models would predict to shatter them within days. Communities have hidden organisers, fed strikers’ children, and ostracised strikebreakers across generations, weaving collective action into the moral fabric of a place until participation became almost unthinkable to refuse. Such cases demonstrate the power of the disposition. Demonstration of power is not explanation of origin. A chemist who observes that a compound explodes has learned something important without yet knowing the compound’s structure. Social theory similarly owes an account of the structure: the institutions, narratives, networks, and repeated practices from which strong solidarity precipitates, together with the solvents, from repression to prosperity to internal scandal, under which the precipitate dissolves. Demanding that account is not pedantry directed against a noble sentiment. Demanding that account is how respect for the sentiment’s real historical effects gets converted into knowledge usable by people hoping to reproduce those effects under less favourable conditions.

Attempts to ground solidarity in identity deserve separate scrutiny because identity language dominates much contemporary discussion of mobilisation. Shared identity can lower the psychological cost of contribution by reframing sacrifice as self-expression: a person acts not for strangers but for people recognised as part of an extended self. Framing of that power is well documented in movements organised around nation, faith, race, and gender, as well as class. Yet identity is neither singular nor automatically ranked. A worker is also a parent, a tenant, a believer or unbeliever, a supporter of a football club, a speaker of a language, a citizen of a state that may be at war with the state employing fellow workers abroad. Which identity governs a decision depends on salience, and salience is contested terrain where employers, states, media, and movements compete. Class identity has historically lost that competition as often as class identity has won, including catastrophic losses during August nineteen fourteen, when international working class identification yielded within days to national mobilisation across Europe. A theory resting on identity must therefore explain salience contests and their outcomes, returning once more to mechanisms rather than resting in the invocation of belonging.

State action reshapes the entire field in ways that class theory sometimes treats as background scenery rather than as a decisive variable. Legal recognition of unions transforms the cost structure of membership by protecting joiners against dismissal for joining. Welfare provision reduces dependence on a single employer and thereby lowers the risk of industrial action, since losing a job no longer entails immediate destitution for a family. Conversely, legal prohibition, surveillance, and imprisonment raise participation costs beyond what most selective incentives or norms can offset, which explains why labour movements under authoritarian regimes often survive as small heroic networks rather than mass organisations. Public employment creates workplaces where striking carries different risks than striking against a private firm able to relocate. Electoral systems determine whether labour-based parties can plausibly promise legislative gains that reward mobilisation with policy victories. Each of these institutional facts mediates between interest and action. Two countries with comparable industrial structures can therefore sustain wildly different levels of class organisation, a comparative fact that pure interest theory cannot digest and that Olson’s framework, supplemented institutionally, handles with relative ease.

Comparative variation also disciplines romantic generalisations drawn from single national stories. Scandinavian labour movements achieved density and durability through Ghent system arrangements linking unemployment insurance to union administration, a selective incentive embedded in public policy. American labour faced a legal environment oscillating between protection and hostility, employer opposition of exceptional intensity, and racial divisions employers exploited deliberately across generations. German co-determination channelled worker representation into works councils and supervisory boards, producing influence alongside integration into managerial decision-making. British voluntarism prized autonomy from the state while leaving organisation exposed to legal reversals after electoral defeats. Each national pattern represents a distinct settlement among incentives, coercion, norms, and state frameworks. None represents the natural unfolding of class interest in unmediated form. Comparative study thus reinforces the section’s central claim from another angle: wherever class action reached scale and durability, identifiable scaffolding held the structure up, and scaffolding varied because builders confronted different materials, opponents, and weather.

What partial answers exist and what do they leave open?

Selective rewards, small workplace groups, compulsion, repeated dealing, reputation, and social norms each move some people in some settings, yet each either presupposes an organisation already built, works only locally, shifts the puzzle elsewhere, or depends on dispositions requiring explanation, leaving contribution across an entire class still incompletely explained.

Inventorying partial answers clarifies both the progress achieved since nineteen sixty-five and the distance remaining. Progress is real. Nobody writing carefully about mobilisation any longer treats interest as self-executing, and the best empirical scholarship specifies mechanisms with a precision earlier traditions rarely attempted. Researchers can explain why a particular hospital’s nurses organised while a neighbouring hospital’s nurses did not, why a strike held in one coalfield and collapsed in another, why a party rooted itself in one region and remained a sect elsewhere. Explanations of that granularity draw on leadership, networks, resources, framing, political opportunity, and employer strategy in combinations tailored to cases. What remains elusive is a unified account travelling from case-level mechanisms to class-level agency without losing rigour along the way. The encompassing claim of classical Marxism concerned not this union or that strike but the historical role of a class as a whole. Partial answers vindicate local agency repeatedly. Vindication of local agency is compatible with continued scepticism about agency at the encompassing level, and holding both judgements together is the intellectually honest position the evidence supports.

Aggregation is the technical name for the unresolved passage from local to general. Suppose every workplace contained groups capable of cooperation under favourable conditions. Class action requires synchronisation across workplaces whose favourable conditions rarely coincide, whose demands differ in timing and content, and whose members may regard one another’s priorities as distractions or threats. Synchronisation demands coordinating bodies, and coordinating bodies demand resources, authority, and trust from units jealous of autonomy. Victories at coordination level, such as general strikes, have occurred, yet their rarity relative to local disputes measures the height of the aggregation barrier. General strikes also tend to be defensive, brief, and triggered by provocations supplying unusual unity, rather than offensive instruments wielded at moments of a movement’s choosing. A weapon usable mainly in emergencies, under conditions supplied by opponents, differs considerably from the sustained historical agency classical theory envisaged. Aggregation analysis therefore tempers celebration of local successes without diminishing their genuine human significance for participants whose conditions improved because a local fight was won.

Free riding also reappears inside organisations built to defeat free riding, a recursive quality that organisational sociology documents unsparingly. Members delegate to officials, officials develop interests in organisational survival distinct from members’ goals, and monitoring officials becomes a fresh collective good that ordinary members have incentives to neglect. Robert Michels formulated an early version of this worry as an iron law of oligarchy, observing tendencies toward leadership entrenchment across parties formally committed to democracy. Later scholars qualified the law’s inevitability while confirming the tendencies. Rank and file apathy, often lamented as betrayal or false consciousness, fits Olson’s logic comfortably: attending meetings, scrutinising accounts, and contesting elections are costly activities whose benefits diffuse across the membership, so rational members abstain and then deplore the oligarchy abstention permits. The recursive pattern does not make organisation futile. Recursion does mean that solving the problem once, at founding, purchases no permanent exemption. Maintenance of internal democracy is itself a collective action achievement requiring continuous inputs of attention that the theory predicts will be under-supplied.

Generosity toward the problem also requires acknowledging what Olson’s framework does not attempt. The framework takes preferences and interests as given and asks what follows for action. Marxist theory at its strongest asks how preferences form, how people come to understand situations, and how participation itself transforms participants. That transformative claim carries genuine insight. People who join a strike frequently report changed self-understanding, new capacities, and altered loyalties persisting long after the dispute ends. Action generates the solidarity that later action draws upon, reversing the causal direction assumed when solidarity is treated purely as a precondition. Bootstrapping of that kind softens the puzzle’s logical edge: a movement need not possess in advance the full disposition it eventually exhibits, because early participation, even if triggered by narrow or accidental causes, can manufacture commitment through experience. Yet bootstrapping still requires an account of the early participation and of why transformative experiences generalise in some trajectories while curdling into disillusionment in others. Defeat transforms people too, often toward withdrawal. Experience is a teacher whose lessons depend heavily on outcomes nobody controls.

An adequate answer would therefore have to explain contribution at scale without assuming the disposition it needs. Explanation of that kind must show, step by step, how dispersed strangers acquire reasons, assurances, protections, and habits strong enough to sustain costly effort, how early cooperation gets built before organisations exist to reward it, how local cooperation aggregates into coordination across workplaces and regions, and how the resulting commitment survives defeat, turnover, and counter-mobilisation. Until an account performs that work, class action remains something history occasionally achieves through fragile scaffolding rather than something class interest reliably produces, and Olson’s question retains its unsettling force for any politics resting upon collective agency.

Where a mechanism may stop, and what the debate changed

Resolution in this dispute does not require victory for one camp over the other in every respect. Instead, resolution arrives through separation of demands that earlier discussion had fused into a single test of legitimacy. One demand asks for mechanisms: a worked account showing how a pattern comes about through identifiable steps rather than through invocation of a large tendency. A second demand asks for reduction, meaning translation of every legitimate explanation into statements about individual wants and beliefs. Treating those demands as identical made disagreement look total. Pulling them apart shows agreement to be wider than polemic suggested, while disagreement narrows to a sharper and more interesting question about explanatory priority. A mechanism may satisfy the first demand without satisfying the second. Whether satisfaction of the first demand suffices for good explanation is precisely the point at issue, and the answer defended across this section is affirmative, subject to a stated defeater and to an honest concession about a thin premise borrowed from the individual level.

Consider what termination means in practice. Every explanation stops somewhere. Even accounts celebrated for depth rely on background conditions taken as given for purposes of a particular inquiry. A chemist explaining combustion does not pause to derive atomic structure. A demographer explaining migration flows does not pause to reconstruct household psychology in full. Stopping is therefore not a vice peculiar to structural explanation. Rather, stopping becomes a problem only when the chosen terminus hides the very process that needed illumination, or when the terminus smuggles in assumptions stronger than evidence permits. Judged by that standard, structural positions can serve as legitimate termini under specifiable conditions. Those conditions concern narrowness of options, stability of constraints, and transparency about premises retained. Where options are broad and constraints weak, stopping at position explains little. Where options are tightly bounded and constraints bite hard, stopping at position can explain a great deal, because variation in private motivation makes little difference to the aggregate outcome.

Compulsion supplies the clearest illustration. Under a wage relation, a person lacking independent access to means of production faces a restricted menu. Selling labour power to an employer, seeking support from kin, relying on public provision where such provision exists, or enduring destitution exhaust the realistic alternatives for most occupants of that position in most historical settings examined by Marxist writers. Employers likewise face discipline from competition, credit conditions, and the need to secure a surplus sufficient for continued operation. Neither side needs a rich psychology for the central pattern to follow. Workers appear at workplaces, accept terms they would reject under roomier circumstances, and return after disputes that end without transformation of ownership. Capitalists press for longer hours, lower costs, or intensified effort even when personal inclination favours ease. Behaviour converges across temperaments because the position punishes deviation. Generosity in an employer that raises costs beyond what rivals bear invites exit from the market. Refusal in a worker that rejects all available employment invites hunger, eviction, or dependence. Convergence under punishment is the explanatory core, and convergence is visible without peering inside each mind.

Minimal assumption does the remaining work, and its modesty matters. The compulsion account assumes only that people generally take the less ruinous option open to them when ruin is recognisable and alternatives are understood in outline. Such an assumption says nothing about whether a worker prefers leisure to income, status to security, solidarity to advancement, or obedience to risk. Preferences across those dimensions may vary wildly while conduct remains similar, because the option set filters which preferences can be acted upon at acceptable cost. A worker who loves craft and a worker who hates the trade both clock in when rent falls due. An employer devoted to paternalist welfare and an employer indifferent to welfare both resist wage increases that threaten solvency. Explanation succeeds by showing why divergent wants produce convergent deeds under shared constraint. Insisting on a full inventory of wants before granting explanatory status would discard that achievement for no gain in predictive power over the pattern under discussion.

Distinction between explaining action and explaining motivation clarifies the point further. Structural accounts aim chiefly at patterned action under constraint: hiring, firing, striking, migrating, investing, defaulting, complying. Motivation enters as background, held at a thin level, while foreground attention falls on relations that distribute options unevenly. By contrast, accounts centred on preferences aim chiefly at why particular agents selected one course among several genuinely open courses. Both aims are legitimate. Confusion arises when legitimacy of the second aim is treated as a requirement governing the first. A railway timetable explains arrival patterns without explaining why passengers wished to travel, although wishes plainly exist and plainly matter for other questions. Likewise, an account of the wage relation explains labour market participation without cataloguing why each participant values income, although valuation plainly occurs. Silence about a factor is not denial of that factor. Silence reflects a judgement about which factors carry explanatory weight for a specified explanandum, and such judgements are unavoidable in every science that has progressed beyond chronicle.

Historical writing inside the Marxist tradition often proceeded in exactly this fashion long before methodological controversy gave the practice a name. Analyses of enclosure treated loss of common access as a transformation of options, after which former commoners sought wages not because tastes had altered but because subsistence outside wage labour had been closed off. Analyses of factory discipline treated bells, fines, and supervision as devices narrowing discretion during the working day, so output regularity followed from institutional design rather than from a sudden spread of industrious character. Analyses of crisis treated competition as a coercive law imposed on each producer regardless of intention, so accumulation and cost cutting followed from market position rather than from greed considered as a personal trait. Greed, ambition, fear, and habit were never absent from human affairs in those accounts. Rather, those accounts located causal leverage in rearrangements of access and penalty. Later methodological debate made explicit a logic already present in strong historical practice, then argued over whether explicitness required a different and deeper terminus.

Readers seeking orientation in the broader controversy will find the analytical marxism complete guide useful for placing this stopping question alongside disputes about functional explanation, rational choice modelling, and the status of class. Placement matters because the stopping question is easily caricatured when isolated. Defenders of structural termini were not defending vagueness, historical laws immune to evidence, or collective entities endowed with minds. Critics demanding mechanisms were not demanding that sociology dissolve into psychology, nor that every macro pattern be rebuilt from scratch in each study. Serious positions on both sides accepted evidence, accepted causal reasoning, and accepted that individuals act. Disagreement concerned how much about acting individuals an explanation must state before the explanation counts as complete enough for its purpose. Framed that way, the dispute becomes assessable. Claims about completeness can be tested against cases. Cases can be compared for leverage gained or lost when additional detail is supplied. Assessment replaces denunciation, and the literature improves accordingly.

Where can a mechanism legitimately stop?

A mechanism may stop at structural positions when those positions define an option set narrow enough that action follows from a minimal assumption about avoiding ruin. Explanation ends there because further detail about wants adds no leverage over the pattern being explained, given the constraints already specified in full.

Elaboration of the criterion helps prevent misuse. Narrowness is a matter of degree, established by examining alternatives actually available rather than alternatives imaginable in abstraction. An option counts as available only if pursuing that option leaves the agent able to meet basic requirements of subsistence, shelter, and social standing sufficient for continued participation. Options carrying catastrophic penalties are available in a formal sense yet unavailable in the sense relevant to explanation of regular conduct. A starving person can refuse food offered on degrading terms, and occasional refusals occur and deserve study as exceptions illuminating the rule. Regular conduct, however, tracks penalties. Explanation of regularity may therefore treat catastrophic options as effectively closed without denying logical possibility. Precision about effective closure distinguishes compulsion from determinism. Determinism would claim that position dictates a single act for every occupant on every occasion. Compulsion claims something weaker and more defensible: position raises the cost of some acts so sharply that aggregate behaviour clusters, individual exceptions persist, and clustering itself becomes the fact demanding explanation.

Stability forms a second part of the criterion. Positions explain best when relations defining those positions endure across the period under study, so occupants face similar option sets despite turnover in personnel. Labour markets, property regimes, and credit systems often display such endurance. Individuals enter and exit, yet the slots remain, and behaviour associated with slots recurs. Recurrence across different occupants is strong evidence that position rather than personality drives the pattern. If a workplace sees successive managers all press for speedups under unchanged competitive pressure, attributing each episode to separate quirks of character multiplies explanations without improving understanding. If successive cohorts of tenants all fall into arrears after a rent increase under stagnant wages, attributing arrears to improvidence cohort by cohort misses the shared arithmetic. Slot based reasoning captures what person based reasoning fragments. That capture is a genuine explanatory gain, purchased without any claim that persons are interchangeable in all respects or that biography never matters.

Transparency forms a third part. A structural terminus earns legitimacy by declaring premises openly, especially the thin premise about avoiding ruin, so readers can judge whether the premise holds in the case at hand. Concealed premises invite suspicion that structural language hides individual assumptions too controversial to state. Declared premises invite testing. Does evidence suggest occupants understood alternatives in outline? Did penalties for deviation actually fall as described? Were there support networks, communal resources, or illicit economies widening options beyond the formal picture? Such questions are empirical, answerable in principle, and fatal to particular applications when answers go the wrong way. A compulsion account of a particular labour market fails if investigation reveals ample alternative livelihoods that occupants ignored for reasons rooted in honour, faith, or political commitment. Failure in an application does not, however, invalidate the form. Forms survive failed applications when other applications succeed and when failure conditions are themselves informative about scope.

Scope conditions deserve emphasis because overextension discredited structural reasoning in some quarters. Not every pattern arises under tight constraint. Consumption choices among affluent households, voting choices among cross pressured citizens, and cultural choices among artists with multiple patrons all involve roomy option sets where wants and beliefs differentiate outcomes powerfully. Stopping at position in those domains would leave the interesting variation untouched. A defender of structural termini need not defend stopping everywhere, only stopping somewhere, under stated conditions, for specified patterns. Modesty about scope strengthens rather than weakens the resolution. Grand claims that class position explains everything invited justified resistance. Limited claims that position explains labour market participation, competitive cost cutting, and similar compelled regularities rest on firmer ground. Resistance to grand claims sometimes slid into resistance to limited claims as well, and recovering the limited claims from the wreckage of the grand ones is a central achievement of the resolution defended here.

Another clarification concerns levels. Stopping at position does not posit a collective agent acting above individuals. Positions are relational facts about individuals: who owns what, who owes what, who may exclude whom, who must seek permission from whom. Statements about positions are statements about arrangements among individuals, not statements about a super individual. A landlord tenant relation involves at least two persons plus enforceable rules. A creditor debtor relation involves parties plus courts, customs, or coercion sustaining repayment. Describing such relations is describing the social world individuals inhabit, not inventing a separate world floating free. Individualist critics sometimes wrote as if any vocabulary above preference and belief must refer to mystical entities. That inference fails. Relational vocabulary refers to patterned connections among individuals, and patterned connections are as real as the individuals connected. A bridge is not a mystical entity above its components, yet engineering explanations stop at properties of the bridge that no inventory of molecular preferences, if molecules had any, could replace. Analogy has limits, since social relations depend on recognition and enforcement in ways physical structures do not, but the limited point stands: higher level description is not automatically occult description.

Enforcement deserves separate treatment because enforcement seems to reintroduce individual action at every joint. Rules persist because officials enforce them, neighbours observe them, and participants anticipate enforcement. All true, and all compatible with structural stopping for many purposes. An explanation of why tenants pay rent may stop at lease, income, and eviction risk without reconstructing the psychology of bailiffs, judges, and landlords in each transaction. Enforcement background can be taken as given when enforcement is routine and uncontested during the period studied. When enforcement itself becomes the explanandum, because a regime is collapsing or a rule is contested, explanation must descend to the actions sustaining or undermining enforcement. Descent is then required by the question asked, not by a universal methodological law. Questions dictate depth. A universal law dictating maximum depth for every question would paralyse inquiry, since every answer can be met with a further why. Productive inquiry accepts provisional depth governed by the pattern under investigation and by leverage gained from descending further.

Leverage provides a practical test. Descend when descent changes predictions, alters intervention points, or resolves anomalies that the higher level account cannot handle. Stop when descent merely redescribes the same pattern in more granular vocabulary without changing any of those things. Granularity alone is not a virtue. Maps at maximum detail cease to function as maps. Models that reproduce every idiosyncrasy forfeit the generality that made modelling worthwhile. Social science needs both detailed case reconstruction and portable accounts of recurring mechanisms. Portable accounts travel across cases precisely because detail varying between cases is omitted. Omission is therefore a feature when omitted detail is causally inert for the pattern, and a defect when omitted detail drives the pattern. Sorting inert detail from driving detail is empirical work, case by case, and no a priori rule about preferred levels can perform that sorting in advance. The resolution thus replaces a level rule with a leverage rule. Retain the level that carries leverage for the question at hand, declare premises, specify scope, and remain ready to descend when anomalies demand descent.

Strongest objection must be stated without softening, because a resolution surviving only a weak version of opposition settles little. Individualist proponents, with Jon Elster above all, pressed a regress argument against structural starting points. Whatever structural position an explanation starts from, the question why people in that position act as they do arises immediately. A landlord collects rent, a worker accepts a shift, a manager orders a speedup, a tenant vacates after notice. Each verb describes human doing. Doings invite reasons. Reasons involve what agents want and what agents believe about means for securing desired ends. Answering the immediate question therefore requires reference to wants and beliefs, whether the analyst wishes to supply that reference or not. On this view, structural starting points defer the individual level rather than dispensing with the individual level. Deferral masquerades as independence. Strip away the masquerade, and every structural explanation rests on unstated individual premises that intellectual honesty should drag into the open and that explanatory rigour should develop in full.

Force of the objection grows when hidden premises are examined in concrete cases. Suppose an account claims that competition compels cost cutting. Competition among whom, enacted how? Rivals lower prices, buyers switch suppliers, lenders tighten terms, shareholders sell holdings. Each link involves persons perceiving signals and responding according to expectations and aims. Suppose an account claims that propertylessness compels wage seeking. Propertylessness bites because persons need food, landlords demand payment, and employers offer income conditional on obedience to workplace rules. Needs, demands, and offers are intelligible only against background assumptions about purposive conduct. Suppose an account claims that a fiscal crisis compels austerity. Ministers choose cuts fearing market reaction, officials draft budgets anticipating resistance, citizens protest or acquiesce depending on judgements about alternatives. At every joint, purposive conduct carries the causal chain. Elster and allied writers concluded that mechanisms worthy of the name must specify those joints in individual terms, ideally with the precision of rational choice modelling where feasible, and at minimum with explicit desires, beliefs, and opportunities as the canonical triad. Anything shorter was, on the strongest version, a sketch awaiting completion or a black box awaiting opening.

Appeal of the objection is easy to understand. Social science had seen too many explanations invoking tendencies, functions, and necessities without showing how any of those abstractions could move a single person to do a single thing. Hegemonic ideology explained compliance, except when compliance failed, whereupon false consciousness was said to be cracking. Capital logic explained state policy, except when policy diverged, whereupon relative autonomy absorbed the exception. Such elasticity protected theories from refutation while draining them of content. Demanding mechanisms at the level of acting persons promised discipline. A writer forced to say who does what, in response to which signal, under which incentive, loses the freedom to let abstractions act in place of people. That discipline improved Marxist writing where it was absorbed, and even writers rejecting reduction absorbed much of the discipline. Acknowledging the improvement is part of steelmanning honestly. The objection was not pedantry about vocabulary. The objection targeted a real pathology, diagnosed the pathology accurately in many instances, and proposed a remedy whose successes in particular studies are undeniable.

Further strength comes from the problem of divergent response. Occupants of the same position sometimes act differently, and differences matter historically. Some workers organise unions while others inform on organisers. Some employers pioneer welfare provision while others squeeze relentlessly under similar market pressure. Some tenants resist eviction collectively while others leave quietly one by one. Structural accounts predicting clustering must also accommodate dispersion, and dispersion seems to point back toward wants and beliefs as differentiating causes. If position alone explained conduct, dispersion would be noise. Yet dispersion occasionally becomes the main story, as when a minority response spreads and transforms the situation for everyone. Revolutions, strikes, and bank runs all involve thresholds where minority conduct cascades. Explaining cascades seems to require heterogeneity in thresholds, information, and commitment, all properties of individuals. The individualist therefore argues that even successful structural accounts borrow explanatory power from unexamined individual variation, drawing on a reservoir that structural vocabulary cannot replenish. When the reservoir runs dry, when variation rather than regularity demands explanation, structural stopping fails conspicuously.

Micro detail also promises unification, and unification is a genuine scientific good. If disparate macro patterns can be derived from a shared account of purposive action under constraints, social science gains the kind of integration that physics prizes. Game theory appeared to offer a common language for conflict, cooperation, bargaining, and coalition formation. Rational choice models of collective action recast familiar historical puzzles in tractable form. Olsonian logic about free riding explained why large latent groups often fail to organise despite shared interests, a result with obvious bearing on class organisation. Schelling style models showed how mild individual tendencies could generate sharp aggregate segregation, demonstrating that macro patterns need not mirror macro intentions. Such demonstrations dazzled for good reason. Surprising aggregate results derived from transparent individual assumptions teach something real about social processes. Enthusiasm for derivation is understandable, and defenders of structural stopping who dismiss derivation wholesale cut themselves off from powerful tools. The resolution defended here does not dismiss derivation. Rather, the resolution denies that derivability in principle settles legitimacy in practice, and denies that every good explanation must exhibit its derivation before earning trust.

Answer begins by distinguishing presupposition from explanatory requirement. Structural accounts presuppose that individuals act, perceive, and respond, just as demographic accounts presuppose births and deaths without supplying obstetric or pathological detail. Presupposition in that thin sense is universal and harmless. Explanatory requirement is stronger: a factor is required when omitting that factor leaves the pattern unexplained or mispredicted. Wants and beliefs in full specificity are presupposed as existing, yet not required for compelled regularities, because the option set does the differentiating work. Vary the wants across plausible ranges and the pattern persists. Vary the option set and the pattern shifts. Counterfactual sensitivity identifies where explanation lives. If workers given independent plots cease seeking wages while workers with unchanged wants but no plots continue seeking wages, access rather than desire explains participation. If employers shielded from competition relax cost pressure while employers with unchanged temperament but exposed to rivals intensify pressure, exposure rather than temperament explains intensity. Sensitivity tests of that kind are the bread and butter of causal reasoning, and those tests vindicate structural termini in the compelled cases.

Option set explanation also answers the immediate why question, albeit at a level the individualist finds too coarse. Why did a particular worker accept a particular shift? Because refusal risked dismissal, dismissal risked destitution, and the shift, however unwelcome, preserved income, housing, and standing. That answer cites position, penalties, and a minimal assumption about ruin avoidance. The answer is not empty, not circular, and not a refusal to explain. Further questions remain available: why did the worker rank housing above rest on that occasion, why did beliefs about dismissal risk take that shape, why did family obligations weigh as heavily as they did. Availability of further questions does not invalidate the answer already given, any more than availability of molecular questions invalidates a thermodynamic answer about pressure and volume. Every answer generates further questions. Inquiry must still distinguish between answers sufficient for a stated purpose and answers insufficient for that purpose. Sufficiency is purpose relative, and purposes in social science include comparison across cases, identification of intervention points, and explanation of aggregate regularities. For those purposes, option set answers often suffice, and additional psychological depth, while interesting, functions as a different inquiry rather than as completion of the original one.

Honest concession follows, and the concession should not be minimised. The minimal assumption is itself an assumption about individuals. Stating the premise attributes a disposition to persons, not to positions: occupants, on the premise, steer away from ruin when a safer course lies within reach. Positions do not take options. Persons do. A structural account therefore borrows a thin individual premise rather than operating with no individual premise at all. Purity claims on the structural side fail, and defenders do well to abandon purity. Borrowed thinness, however, differs enormously from the thick specification demanded by reduction. A thin premise covers almost all occupants, requires no ranking of substantive ends, survives wide cultural variation, and can be stated in a sentence. Thick specification requires modelling particular desire orderings, the formation of beliefs, the costs of information, and strategic expectations case by case. Treating thin borrowing as vindication of thick reduction commits a slide from some individual content to maximal individual content, as if any mention of persons licensed a demand for complete psychology. The slide is a fallacy of degree. Dosage matters in explanation as in medicine. A premise everyone accepts because evidence for ruin avoidance under recognised threat is overwhelming cannot bear the weight of controversial modelling choices that thick reduction must defend case by case.

Analogy with other sciences reinforces the dosage point without pretending social life resembles gases or fluids in substance. Epidemiologists explain outbreak patterns through contact structures and transmission probabilities without modelling each carrier in psychological depth, although transmission occurs through human acts saturated with intention and belief. Traffic engineers explain congestion through road topology and vehicle volumes without reconstructing commuter biographies, although each vehicle is steered by a purposive driver. Economists themselves, including many sympathetic to individualism, explain market clearing through supply and demand schedules that aggregate away most detail about why suppliers hold the schedules they hold. Practice across fields accepts layered explanation, with each layer stopping at the factors carrying leverage for its questions. Social explanation centred on class relations claims no exemption from a standard that would also condemn epidemiology, traffic engineering, and price theory. Rather, class centred explanation claims the ordinary right to layer, subject to ordinary duties of declaration, scope specification, and openness to descent when anomalies accumulate.

Divergent response, pressed earlier as strength for the objection, can be accommodated without surrendering the resolution. Clustering claims are probabilistic and aggregate. Dispersion around a cluster is expected whenever additional factors vary, and additional factors plainly include wants, beliefs, networks, and biographies. Accommodation becomes evasion only if dispersion is invoked ad hoc to rescue every failed prediction. Discipline against evasion comes from specifying in advance which patterns the structural account claims to explain and which patterns lie outside its scope. Labour market participation under propertylessness falls inside scope. Choice between two unions offering different strategies falls outside scope, or at the boundary where position underdetermines selection and finer factors govern. Cascades and thresholds fall partly outside scope as well, because tipping dynamics depend on distributions of commitment that position alone does not fix. Conceding those domains to finer grained analysis is not defeat. Conceding those domains maps the boundary of a tool, and tools with mapped boundaries are more trustworthy than tools claiming universality. The individualist objection proves too much if boundary mapping counts as capitulation, since every successful theory in every field has a mapped boundary.

Verdict can therefore be stated plainly, together with the condition that would overturn the verdict, so readers are not asked to accept an unfalsifiable preference for one level over another. Judgement defended here is that mechanisms need not terminate in preferences. Mechanisms may terminate in structural positions and in the options those positions make available, provided option sets are narrow, premises are declared, and scope is respected. Grounds for the judgement are cumulative: compelled regularities display counterfactual sensitivity to option sets rather than to desire profiles, slot based patterns recur across different occupants, thin premises suffice where thick premises add no leverage, and layered stopping is standard practice in neighbouring fields. None of those grounds denies that preferences exist, influence conduct at boundaries, or deserve study for other questions. Denial would be absurd. Affirmation is narrower: preference specification is not a universal tollgate through which every mechanism must pass before counting as an explanation.

What would defeat the judgement defended here?

The judgement would fall if someone demonstrated that every structural starting point secretly presupposes a fuller account of why occupants act as they do. Such a demonstration would show that option sets explain nothing until preferences and beliefs are specified independently, leaving no legitimate stopping point above the individual level.

Defeater deserves unpacking because promissory arguments and completed demonstrations are easily confused, and confusion on that point has prolonged the dispute beyond its useful life. A completed demonstration would need to show, for the compelled cases at the heart of the resolution, that option set reasoning collapses without independent specification of wants and beliefs. Showing collapse would require more than pointing out that persons act, more than noting that further questions can be asked, and more than exhibiting cases where finer detail helps. Pointing, noting, and exhibiting are compatible with the resolution, which already grants all three. Demonstration would need to establish necessity: that apparent explanatory success of option set accounts is illusory, borrowed entirely from tacit preference assumptions doing the real work, such that varying those tacit assumptions across plausible ranges destroys the pattern while varying option sets leaves the pattern intact. Establishing necessity would reverse the sensitivity results described earlier. Nothing in the literature accomplishes that reversal for the core compelled cases. Sensitivity in those cases runs the other way, repeatedly, across labour markets, rental markets, and competitive product markets studied by historians and sociologists with divergent theoretical commitments.

Plain statement is owed on the status of the individualist reply in light of that standard. The reply sketched in the objection, for all its force as a challenge and for all its success in exposing lazy structural writing, is a promissory argument rather than that demonstration. A promissory argument announces that a deeper account must exist and must be required, while deferring delivery of the proof that requirement holds universally. Promissory arguments have value. Promissory arguments direct research, motivate model building, and keep complacency at bay. Value should not be confused with settlement. Decades of sophisticated work inspired by the individualist programme produced illuminating models of particular processes, yet production of illuminating models in some domains does not demonstrate necessity in all domains. Existence of a deeper account, even where achieved, shows possibility of descent, not obligation to descend. Obligation requires the stronger showing described above, and the stronger showing remains undelivered. Until delivery occurs, the verdict stands, not as dogma, but as the judgement best supported by comparative explanatory performance in the cases both sides treat as central.

Burden allocation follows from that assessment. Defenders of structural stopping have discharged a burden by producing accounts that predict clustering, survive occupant turnover, identify intervention points such as changes in access or penalty, and declare premises for inspection. Critics alleging hidden dependence bear a burden to expose the dependence in a form that damages predictions or reveals circularity. Exposure attempts have succeeded against particular weak accounts, and success against weak accounts should be celebrated rather than resented, since pruning weak work strengthens a tradition. Success against weak accounts does not generalise automatically to strong accounts with narrow option sets and declared premises. Generalisation needs argument pitched at the strong cases. Much individualist polemic pitched argument at weak cases, at functionalist excess, at historicist prophecy, at dialectical obscurity, then claimed victory over structural explanation as such. Victory over excess is real but limited. Limited victories leave the resolution intact, because the resolution was constructed precisely by discarding excess and retaining only the compelled core that polemic never directly engaged at full strength.

Comparison invited by the analytical vs classical marxism contrast helps locate the verdict historically. Classical writing often combined profound historical insight with methodological looseness, moving between levels without signalling transitions, invoking necessity where tendency was all evidence supported, and treating class categories as self explanatory. Analytical writing reacted by imposing clarity, sometimes at the cost of narrowing the questions deemed respectable. Verdict defended here sides with analytical clarity about premises and scope while resisting analytical narrowing about permissible termini. Retention of structural termini under clarity conditions is not nostalgia for looseness. Retention is a claim that clarity, properly understood, permits multiple legitimate stopping points and evaluates stopping points by leverage rather than by allegiance to a favoured level. Such a claim preserves what was best in the classical historical imagination, namely attention to how relations distribute life chances before any individual deliberation begins, while accepting what was best in the analytical correction, namely refusal to let large words substitute for identifiable processes.

Influence beyond the originating dispute is extensive, and tracing that influence changes how the episode should be remembered. Argument among Marxists about mechanisms, reduction, and collective action helped propel a general turn toward mechanism based explanation across the social sciences. Philosophers of social science took up the vocabulary of mechanisms to escape sterile opposition between covering laws and interpretive understanding. Sociologists took up the vocabulary to escape sterile opposition between grand theory and narrow empiricism. Political scientists took up the vocabulary to connect institutional rules with strategic conduct without reducing institutions to mere arenas. Economists, despite a home discipline already committed to optimisation models, took up mechanism language in institutional and behavioural subfields where optimisation alone explained form poorly. A style of explanation once associated with a faction inside Marxism became a shared methodological idiom. Shared idioms rarely retain labels recording their origins, and mechanism talk is no exception. Textbooks present mechanisms as a general advance in social scientific reasoning, often without mentioning Cohen, Elster, Roemer, Wright, or the journals and conferences where Marxists fought over the meaning of historical materialism.

Analytical sociology represents the most direct institutional descendant. Peter Hedström founded and named a research programme explicitly building on the mechanisms tradition, codifying a triad of wants, convictions, and chances as standard components for explaining social outcomes through action and interaction. Hedström acknowledged intellectual debts openly, tracing lines back through the analytical Marxist discussions and through earlier mechanism oriented sociology. Practitioners gathered under the analytical sociology banner include researchers with no connection to Marxism, many of whom would not describe their work in political terms at all, working on segregation, diffusion, networks, education, and organisational behaviour. Migration of a programme from a Marxist debate into general sociology involved transformation as well as transmission. Emphasis shifted toward formal modelling, simulation, and network data. Class receded as the privileged structural category, replaced by a plural inventory of positions and relations. Political stakes receded as methodological identity came forward. Transformation should not obscure lineage. Lineage matters for intellectual history, and lineage also matters substantively, because questions analytical sociology inherited, about how macro patterns arise from situated action without dissolving structure, were sharpened first under pressure of defending or criticising Marxist claims.

How did a Marxist dispute reshape social science method?

The dispute pushed mechanism based explanation into the mainstream of social science, supplied the acknowledged foundation for analytical sociology, and made collective action a central problem in political science and economics. Its influence spread far beyond Marxism, even though histories of method rarely credit the Marxist debate that generated it.

Collective action supplies a second channel of influence, arguably broader than the first. Before the analytical Marxist engagement with Olson, class organisation was often treated within Marxist writing as a natural expression of shared position, obstructed mainly by ideology or repression. After that engagement, organisation became a problem requiring explanation in its own right: why individuals bearing costs for benefits shared with non contributors would participate at all. Political science absorbed the problem as foundational for interest groups, social movements, parties, and revolutions. Economics absorbed the problem through public goods theory, mechanism design, and studies of cooperation under varying institutional rules. Solutions multiplied: selective incentives, repeated interaction, network embeddedness, identity commitments, entrepreneurial leadership, and institutional guarantees each received extensive development. Development occurred largely outside Marxist venues, in journals and departments where Marxism held little sway. Yet the sharpened form of the problem, applied to class as the paradigmatic latent group, owes much to writers who took Olson seriously precisely because class organisation was the case their tradition could least afford to leave unexplained. A challenge absorbed under maximum pressure generated tools of maximum portability.

Philosophy of social science absorbed a third channel. Debates about whether mechanisms must terminate in preferences matured into general accounts of explanatory pluralism, levels of analysis, and interlevel relations. Positions developed in response to Elster and his critics reappear, often with examples changed, in discussions of neuroscience, cognitive science, and biology, where reduction debates have their own long history. Social case proved instructive for general philosophy because social levels are conspicuously entangled: institutions depend on actions, actions depend on institutions, and neither dependence settles priority by itself. Entanglement forced philosophers to distinguish dependence, which is mutual and ontological, from explanatory priority, which is question relative and pragmatic. That distinction, standard across methodological writing, was hammered out in arguments whose original stakes were whether historical materialism could survive methodological scrutiny. Survival of the distinction after fading of the original stakes is a mark of genuine philosophical progress, the kind that outlives the controversy producing the progress.

Rarity of credit invites reflection without resentment. Several factors explain why influence travelled without attribution. Cold War academic politics made Marxist origins a liability for ideas seeking general audiences, so adopters had incentives to present mechanisms as neutral method. Generational turnover replaced participants who knew the origins with students who encountered mechanisms through textbooks scrubbed of genealogy. Success itself erases origins, because a victorious idiom comes to seem like common sense that nobody in particular invented. Disciplinary boundaries fragmented memory further, with political scientists crediting Olson, sociologists crediting Merton or Hedström, and philosophers crediting general movements in the philosophy of science, each lineage partial and each defensible within its own canon. Partial lineages combine into a collective forgetting of the Marxist debate as a generative site. Recovering that site is not an exercise in sectarian pride. Recovery corrects the record about where methodological innovation can originate, namely inside substantive traditions wrestling with hard cases, rather than only inside methodology considered as a separate specialism. Innovation under substantive pressure has a distinctive fertility, because concepts forged to solve a pressing explanatory problem carry marks of the problem that make the concepts sharp.

Settlement, finally, must be described with care, because different components of the original dispute stand in different conditions and lumping those conditions together reproduces the conflation that caused so much confusion. One component is dissolved. Conflation of the mechanism demand with the reduction demand no longer survives careful statement, and few participants in later literature defend the conflation explicitly. Dissolution is a real achievement, comparable to philosophical progress elsewhere where a question disappears because distinctions reveal that the question rested on equivocation. A second component is absorbed. Demand for mechanisms, understood as demand for identifiable processes linking conditions to outcomes, has been absorbed into ordinary standards of good work across traditions that once resisted the demand or never encountered the demand in Marxist form. Absorption means the demand no longer needs a faction to champion the demand. Referees enforce the demand. Graduate training transmits the demand. Textbooks codify the demand. Victory through absorption is quiet, and quietness can make victory hard to see, yet absorption is arguably the most complete form a methodological proposal can achieve.

A third component remains open, and openness should be stated without embarrassment. Reduction demand, separated from the mechanism demand, survives as a substantive dispute about whether preferences are prior to relations. Priority here means more than temporal precedence in a given episode. Priority means explanatory fundamentality: whether wants and beliefs should be treated as the bedrock from which relations are built, or whether relations should be treated as shaping the wants and beliefs that agents bring to any episode. Evidence and theory both bear on that dispute. Evidence about preference formation shows wants adapting to circumstances, aspirations adjusting to perceived possibility, and beliefs tracking informational environments structured by position. Such evidence pressures any picture treating preferences as fully formed inputs arriving from outside social relations. Theory about strategic interaction shows, in the opposite direction, how given wants under given rules generate outcomes nobody intended, a result supporting the fruitfulness of taking wants as starting points for particular models. Both bodies of work are genuine. Neither body closes the dispute, because the dispute concerns relative priority across domains rather than exclusive truth in a single domain. Relations shape preferences over long horizons. Preferences drive choices within relations over short horizons. Timescale relativity may eventually transform the dispute into a set of domain specific findings, at which point the grand form of the dispute will look as dated as the conflation does. Transformation of that kind would itself be a settlement of sorts, achieved by decomposition rather than by coronation of one level.

Character of the remaining dispute matters for how research should proceed. Treating the remaining dispute as methodological, as if a rule could settle priority in advance of inquiry, invites sterile restatement of allegiances. Treating the remaining dispute as empirical and theoretical invites studies comparing how much variance in wants is explained by position, how stable wants remain after position changes, how quickly beliefs update after informational environments shift, and which interventions alter outcomes more reliably: interventions targeting option sets or interventions targeting convictions. Migration studies, natural experiments following plant closures or benefit reforms, and longitudinal surveys tracking occupational transitions all supply relevant evidence. Evidence accumulates slowly and points in mixed directions across domains, which is exactly what a domain relative view predicts. Mixed evidence frustrates partisans seeking a universal verdict, yet mixed evidence is information, and information of that kind gradually replaces slogans with maps. Maps show where structural stopping works, where preference specification is indispensable, and where feedback between levels is so tight that any single level account misleads. Cartography of that sort is less dramatic than foundational victory, and more useful.

State of the question, then, is neither triumph nor stalemate described in simple terms. Conflation is dissolved, mechanism demand is absorbed, reduction demand persists as a live dispute about priority that empirical and theoretical work continues to reshape domain by domain. Marxism as a living political and intellectual movement played a smaller role in later stages of that reshaping than in early stages, as the tools migrated and the originating questions were reformulated in vocabularies owing little to historical materialism. Migration and reformulation do not diminish the originating episode. Rather, migration testifies that a debate conducted with unusual clarity about unusually hard problems, the explanation of class conduct, the logic of collective action, the status of functional claims, produced distinctions sturdy enough to survive transplantation. Whether future social science recovers the genealogy or continues to use the tools without the genealogy, the tools themselves carry the imprint of the problems that forged the tools. A mechanism may stop at a position, under declared conditions, with a thin premise openly borrowed. That sentence, modest as the sentence sounds after so much argument, marks the ground the debate cleared, and the ground remains available for any tradition willing to work under the discipline the clearing imposed. One further implication deserves recording before the account closes. Teaching reflects the settlement described above more faithfully than polemical memory does. Students encountering social explanation through textbooks meet mechanisms early, meet levels treated as complementary rather than as rivals, and meet collective action presented as a puzzle rather than as an embarrassment for group based reasoning. Encounters of that kind normalise the resolution without naming the resolution. Normalisation completes a long arc: a demand born in controversy, refined under objection, qualified by concession, and tested against compelled cases has become part of the ordinary equipment with which new researchers approach factories, markets, movements, and states. Equipment used daily seldom prompts questions about provenance, yet provenance, once recovered, enriches understanding of why the equipment has the shape it has and where its edges remain rough.

Standards borrowed across disciplines, and the misreadings corrected

One reason the quarrel lasted so long is that its participants were often enforcing standards native to different disciplines while speaking as if a single tribunal existed. Economics had developed its own usage of microfoundations in the same decades, where the term named the project of grounding aggregate relations in optimising conduct, and that usage lent the Marxist dispute a borrowed air of settled practice. Sociology carried a different inheritance, in which the relation between structure and agency was the organising problem and in which purely individual starting points had been criticised since Durkheim insisted that social facts be treated as things with their own mode of constraint. History worked to yet another standard, accepting narrative sequences of situated action as explanation without demanding a general model beneath them. None of these standards is self-evidently the correct one for the others. Carrying a standard across a disciplinary border is a step that needs its own defence, because what counts as a completed explanation is partly a function of the questions a field exists to answer. Much of the heat in the microfoundations debate came from unargued imports: a standard at home in one faculty was wielded in another as if its passport needed no inspection.

Why do borrowed standards need justification?

Because a standard of explanation is fitted to the questions a discipline asks, and a requirement that disciplines one field can deform another. Economics may reasonably demand optimising foundations for aggregate relations, while history reasonably accepts situated narrative, and neither verdict transfers without argument about what the receiving field is trying to explain.

The vocabulary of the dispute encouraged the confusion by offering paired slogans in place of positions. Holism and individualism were traded as if each named a single doctrine, when holism covered everything from the modest claim that relations constrain conduct to the extravagant claim that societies think and intend, while individualism ranged from the modest claim that only persons act to the extravagant claim that social vocabulary is disposable shorthand. Emergence supplied a more careful term for what the modest holist meant: properties of organised wholes that depend on their parts without being predictable from the parts in isolation, as the properties of water depend on hydrogen and oxygen without being a sum of them. A debate conducted at the level of the modest claims would have been short and largely consensual. It was conducted instead at the level of the extravagant ones, each side refuting the other’s extreme and reporting victory over the whole position. The separable demands cut through this by refusing the package deals: they identify the exact requirement under discussion and grade that requirement alone.

Five misreadings recur wherever the debate is summarised, and each is worth correcting in the plainest terms. The first treats microfoundations and methodological individualism as the same demand, which is the original conflation and the source of the rest. The second holds that accepting mechanisms commits one to individualism, when in fact a mechanism is a form of account, not a level of account, and mechanisms can be specified between positions, institutions, and relations as readily as between persons. The third supposes that structural explanation is inherently unrigorous, a supposition that survives only by ignoring the structural explanations that do specify their transmission processes in detail. The fourth treats the collective action problem as answered by an appeal to solidarity, which restates the problem, since solidarity names the disposition whose reliable existence the problem questions. The fifth declares the debate settled, in either direction, when the honest position is narrower: the methodological quarrel dissolved under separation, while the substantive dispute about the priority of preferences and relations continues as ordinary theoretical and empirical work. A reader who can decline all five misreadings possesses the debate’s usable residue, which is the point at which an old controversy becomes equipment rather than allegiance.

A reader approaching the primary literature benefits from knowing the order in which the blows landed, because later texts answer earlier ones and lose their point when read in isolation. The functionalist defence came first, in the 1978 reconstruction of historical materialism that gave the tradition its most rigorous statement and thereby its most exposed target. The individualist critique followed within a few years, aimed less at Marx than at the explanatory habits of his defenders, and its author had himself written sympathetically on the tradition before turning examiner. Replies then multiplied along two tracks that the separable demands help to sort: some supplied the missing transmission processes and thereby conceded the first demand while vindicating the tradition’s conclusions, while others contested the destination requirement and developed the vocabulary of supervenience, emergence, and positional compulsion. A fourth group changed the subject productively, rebuilding class analysis and exploitation theory with the new tools and letting the methodological quarrel cool behind them. Reading any single contribution without that sequence invites the characteristic error of the secondary literature, which is to grade a text against a demand its author was never addressing. The quarrel also left a stylistic legacy worth naming. Marxist writing that passed through the exchange tends to mark its explanatory claims more carefully, distinguishing what a passage asserts from what it merely suggests, and that discipline survived the school that imposed it. Habits of precision, once learned under hostile examination, are rarely unlearned when the examiners depart.

For students and exam candidates

The single most valuable habit this debate teaches is separation. When a question asks whether Marxism needs microfoundations, the strong answer refuses the question’s packaging and grades its two contents separately: the mechanism demand, which is reasonable and which the tradition met, and the reduction demand, which is a substantive commitment that may be refused on argument. Markers reward the candidate who can state each demand in one sentence, attribute it to a named proponent, and give the collective action problem as the point where the individualist challenge genuinely bites. The comparison with the classical tradition is set out in the companion study of analytical and classical Marxism, and a worked card on the two demands, with the table reproduced for revision, is kept in the VaultBook study notes. Practise the defeater move as well: state the judgement, then state what demonstration would overturn it. That second sentence is what separates an answer that reports the debate from an answer that has understood it.

For teachers and seminar leaders

The collective action problem is the compact teaching instrument in this territory, because it fits on one board and resists every quick reply a seminar will offer. Set Olson’s argument from 1965, let the group apply it to a strike or a revolutionary organisation, and then work through the attempted escapes one at a time: selective incentives, small groups, coercion, norms, solidarity. Each escape teaches something real about organisation, and each leaves a residue the group can be made to name. The two demands table serves a different exercise: distribute anonymised passages from both camps and ask students to sort every sentence into the mechanism row or the reduction row. Passages that fit neither row are the most instructive of all, since they are usually the rhetorical ones. Close by asking what evidence or argument would change each side’s mind, which returns the room to the defeater standard and away from team loyalty.

For researchers, historians, and journalists

For the researcher, the durable inheritance of this dispute is the mechanism standard itself, which migrated from a quarrel inside Marxism into the general philosophy of social science and into analytical sociology, where it is practised with little memory of its origin. Tracing that migration is a genuine historiographical opportunity: the demand was sharpened on Marxist texts, above all in the exchange over functional explanation, before it was generalised. The individualist position in its strongest form is documented in the companion article on Jon Elster and methodological individualism, which should be read alongside any use of his name as a slogan. Journalists covering disputes about whether a social science finding is properly explained will find the separation useful as a diagnostic: ask whether the critic is demanding a mechanism, which is a request for work, or demanding a reduction, which is a request for allegiance to one ontology. The two requests deserve different receptions, and conflating them in print reproduces the original confusion for a wider audience.

Frequently Asked Questions

Q: Does Marxism need microfoundations?

It depends which of two demands the word is carrying, and separating them is the article’s central move. Marxism needs mechanisms: when it claims that a relation or institution produces an outcome, it owes the reader the steps through which the producing happens, and meeting that obligation, under pressure from Jon Elster and others, made the tradition’s explanations better. It does not owe the further step of ending every such account with individual wants and beliefs. That further step is a doctrine about the furniture of the social world, and doctrines stand or fall by argument. Holding the first answer and the second together is the position this article names the separable demands.

Q: What are the two demands behind microfoundations?

The two requirements differ in kind. One asks how a cause does its work: name the process, the steps, the bearers, so that the connection between cause and effect is shown rather than asserted. The other asks where the showing must stop: at persons, with the beliefs they hold and the ends they pursue treated as the final vocabulary of explanation. The first requirement is a standard of completeness that every science recognises in some form. The second is a theory of what the social world contains, advanced under the first requirement’s protection. They were pressed together, by the same authors in the same works, and were refused together, which is why the argument generated decades of cross purposes before the separation became available.

Q: Can structures explain without individuals?

Structures explain without being reducible to individuals, which is a different claim from explaining without individuals existing. The wage relation holds whatever any participant believes about it: the worker who despises it must still sell labour power, and the employer who regrets competition must still accumulate or fail. A mechanism may therefore rest at the level of positions and the option sets they impose, given a minimal assumption about how people respond to penalties and openings. What structural explanation cannot do is dispense with conduct altogether, since relations persist only through the actions they constrain. The dispute is about the legitimate resting place of an explanation, not about whether persons are involved in social life.

Q: What is supervenience in this debate?

Supervenience is the middle position several participants accepted as the defensible core of the individualist demand. It says that the social level rests on the individual level in one thin respect: alter nothing about persons, their relations, and their circumstances, and nothing social can alter either. Crucially, that dependence does not deliver reducibility. Chemistry supervenes on physics in the same thin sense, yet chemical explanation proceeds at its own level with its own regularities. A Marxist can therefore concede that every change in class structure is accompanied by changes at the level of persons, while denying that class explanation must be rewritten as a story about preferences before it counts as explanation.

Q: Does class interest produce class action?

Not automatically, and the gap between the two is one of the debate’s permanent results. A class shares an interest when its members would benefit from a common outcome, but benefit does not organise itself. Mancur Olson’s argument of 1965 shows why: where the benefit reaches members whether or not they contribute, each member has reason to leave the cost to others. Interest explains what a class would gain from acting; it does not explain the acting. Bridging the gap requires an account of organisation, incentives, norms, or compulsion, and supplying that account is work the classical tradition often assumed away by treating interest as a motive already installed in its bearers.

Q: What is the collective action problem for class?

It is the difficulty that a benefit shared by a whole class gives no individual member a sufficient reason to help win it. A strike’s gains in wages or hours reach the worker who picketed and the worker who stayed at the bench alike, while the costs of striking fall only on those who strike. Olson’s book of 1965, The Logic of Collective Action, developed the general argument, and its application to class is direct: the larger and more dispersed the group, the weaker each member’s incentive to contribute. Selective incentives, small-group organisation, coercion, and norms each answer part of the problem. Stated honestly, the problem remains unsolved at the scale of a class, and any class-based politics must reckon with it rather than rename it.

Q: Is the demand for mechanisms reasonable?

Yes, and this article treats that answer as settled in the demand’s favour. A mechanism requirement asks only that an explanation show its working: how an alleged cause brings its effect about, the feedback by which a beneficial institution is maintained, the steps by which a structural pressure becomes an event. Functional explanation in Marxism was genuinely vulnerable here, since showing that an institution serves capital does not explain why it exists or persists. Meeting the demand forced better history and better theory, and its standards migrated into the philosophy of social science generally. Reasonableness ends, however, where the demand quietly changes into the second one, and a request for processes becomes a requirement that processes end at preferences.

Q: What did this debate influence outside Marxism?

Its largest influence was the turn toward mechanism-based explanation throughout the social sciences. The demand that explanations identify the processes linking cause and effect, sharpened in argument over Marxist functional explanation, became a standard position in the philosophy of social science. Analytical sociology, founded by Peter Hedström, took the mechanism requirement as its explicit basis and built a research programme on it, practised by scholars with no connection to Marxism and often with no knowledge of the debate that supplied its method. The collective action problem, the debate’s other export, became central to political science and economics. A dispute conducted in the margins of one tradition thus set methodological terms for disciplines that rarely credit it.

Q: Must mechanisms terminate in preferences?

No, and the claim that they need not is the article’s defended judgement. A mechanism is allowed to end with positions inside a structure and with the choices those positions leave open, as a compulsion account of the wage relation illustrates: conduct follows from the option set and its penalties, granted only the thin premise that occupants usually pick the opening that ruins them least. The judgement carries a stated defeater. It would be defeated by a demonstration that any structural starting point presupposes an account of why individuals occupying those positions act as they do. That demonstration has been promised by the individualist reply, which observes that the question of motivation always arises, but a promissory argument is not the completed demonstration.

Q: Is the microfoundations question settled?

The conflation is settled and the substance is not. As a methodological quarrel, the debate dissolved once its two demands were separated: the mechanism demand was absorbed into general standards of explanation, and nobody serious defends explanation by bare function any longer. As a substantive dispute, the underlying question remains open, because it concerns whether preferences are prior to social relations or formed by them, and that is an empirical and theoretical matter on which evidence and argument continue to accumulate. The question is therefore best described as reframed rather than answered. What would settle it is identified in this article: a completed demonstration that structural starting points presuppose individual-level accounts, or a structural research programme that keeps explaining without needing one.

Q: How do the separable demands grade the microfoundations debate?

They grade it by splitting its verdict. On the mechanism demand, the critics were right and the tradition’s defenders who resisted were wrong: explanation by function alone was a real defect, and repairing it improved Marxist work measurably. On the reduction demand, the grading reverses: the critics presented a substantive ontology as a standard of rigour, and defenders were entitled to refuse it, though many refused it in the wrong way, by rejecting rigour along with reduction. A single verdict on the debate, favourable or hostile, is therefore always a grading error. The separable demands replace it with two verdicts of opposite sign, each with its own evidence, which is why the distinction does more work than any summary judgement of who won.

Q: What should a student separate between demanding a mechanism and demanding a reduction?

A student should separate the request for work from the request for allegiance. Demanding a mechanism asks the writer to show a process: how an alleged cause brings about the outcome attributed to it, through what steps, borne by which actors in which positions. That request can be met inside almost any theoretical framework, and meeting it is a skill worth drilling. Demanding a reduction asks the writer to accept that only individual beliefs and preferences finally explain, which is a doctrine to be argued for, not a standard to be assumed. In an examination, the separation is worth performing explicitly in the opening paragraph, since it converts a slogan question into two answerable ones and signals that the candidate knows where methodology ends and theory begins.

Q: Why did one methodological demand split into two?

Because the two requirements were delivered as a package and had to be unpacked by argument. They arrived together in the same critical works, above all in the assault on functional explanation, where the call for mechanisms and the commitment to methodological individualism were advanced by the same authors as parts of one programme. Defenders responded to the package as a package, refusing both demands or swallowing both. Only slower work, including concessions by participants on each side, separated the general standard from the particular ontology it had travelled with. The split was therefore not present at the debate’s start as a distinction anyone drew cleanly; it is the debate’s product, visible in retrospect once supervenience and structural mechanisms gave each demand its own address.

Q: How does the two demands table order the requirements?

The table orders them by ascending commitment. Its first row asks for the working: show the steps connecting an alleged cause to its effect, because an unexplained connection is a postulate wearing the dress of an explanation; Marxism can meet that row, and did, to its benefit; accepting it commits one only to diligence. Its second row asks for a destination: every mechanism must end with what individuals believe and prefer; it is made because only individuals act; meeting it costs the tradition its claim that relations shape the wants people bring to them; accepting it commits one to methodological individualism as an ontology. The ordering matters because conceding the rows in sequence is the debate’s characteristic trap: agreement earned cheaply on the first row is spent as if it had been given on the second.

Q: What separates a structural explanation with mechanisms from one without?

The presence of a transmission story. A structural explanation without mechanisms states that a position or relation produces an outcome and leaves the connection as a postulate: the state serves accumulation, the school reproduces labour power, the relation compels the result. A structural explanation with mechanisms keeps the same structural starting point but shows the working: the penalties that close off alternatives, the option set a position imposes, the feedback that maintains an arrangement because of what it does. Both explanations are structural, and neither reduces to preferences. What separates them is not ontology but completeness. The first asserts a power; the second demonstrates one, which is precisely the improvement the mechanism demand was entitled to exact from the tradition.