Few quarrels in political thought are as durable, or as badly reported, as the quarrel between Marxism and anarchism over the state. The standard summary treats the two traditions as opposites in aim: one side supposedly in love with state power, the other in love with freedom. That summary fails at the first document either tradition produced about its destination. Both traditions name the same destination, a classless and stateless society in which coercive government has been replaced by the free association of producers. They have never disagreed about where they want to arrive. They disagree, with a bitterness that has lasted a century and a half, about the road: about whether a revolutionary movement must pass through a transitional state of its own, or whether building any such state guarantees that the destination will never be reached. Anarchism is named in the opening line of this article because the argument begins there, with the uncomfortable fact that the anarchist side of the dispute made a specific prediction about the Marxist road, wrote it down in 1873, and lived to have it tested.

The comparison is worth doing carefully because it is one of the rare disputes in social theory that evidence can actually grade. Most arguments between schools of political thought dissolve, on inspection, into differences of definition or of temperament, and no conceivable event could settle them. This one contains a component that is different in kind. Mikhail Bakunin, the Russian anarchist who fought Marx for control of the First International, argued in print that if workers took state power, the state they built would produce a new privileged stratum drawn from former workers and professional revolutionaries, a stratum that would govern in the name of the class while ceasing to belong to it. Marx read the claim, annotated it with contempt, and held to the opposite expectation: that a transitional state held by the working class, organised on the model of the Paris Commune, would lose its functions as classes disappeared and would wither rather than entrench. The twentieth century ran the experiment on the Marxist side of the wager. It did not run the anarchist alternative at comparable scale for comparable time. That unevenness is itself part of the finding, and this article will price it honestly rather than pretend the contest was symmetrical. The name for the whole structure is the prediction asymmetry, and the article builds toward it slowly, because the asymmetry is only intelligible once the shared destination, the organisational history, and the actual texts have been set out in each tradition’s own strongest terms.

Anarchism and Marxism: the shared destination

Begin with the agreement, because the agreement is large, specific, and routinely denied by the polemics of both camps. Marxists and anarchists both hold that the state is not a neutral arbiter standing above society but an institution bound up with class division, arising where society splits into groups with opposed economic interests and serving, in its ordinary operation, to hold that split in place. Both hold that capitalism rests on the separation of producers from the means of production, and that political emancipation within a class society leaves the deeper unfreedom untouched. Both hold that the remedy is the abolition of classes themselves, and that when classes go, the state as a separate coercive apparatus standing over society goes with them. The end point of the Marxist tradition is stated in the language of withering: the state is not abolished by decree but dies out as its functions become unnecessary. The end point of the anarchist tradition is stated in the language of federation: free communes and workplaces, linked from below by revocable delegates, coordinating production and distribution without a government of persons. Strip the vocabularies away and the descriptions converge to a degree that embarrasses partisans of both. A society without classes, without a standing apparatus of coercion, administering things rather than ruling people. The phrase about administering things rather than governing men circulated in the socialist movement long before either tradition hardened, and both claimed it.

What is the main difference between Marxism and anarchism?

Both aim at a classless, stateless society. They differ on the route. Marxists hold that a transitional workers’ state is needed to suppress counter-revolution and will later wither. Anarchists hold that any such state entrenches itself and must be abolished immediately.

Everything else in this article is commentary on that short answer, but the commentary matters, because the short answer conceals how much machinery each side built around its half of the disagreement. The Marxist half carries a theory of what the state is, an account of why it cannot simply be smashed and left in pieces, and a sketch, drawn from the Paris Commune of 1871, of the transitional form that supposedly escapes the anarchist objection: officials elected and recallable, paid workmen’s wages, the standing army replaced by the armed population, the functions of government stripped down until administration and coercion begin to merge into ordinary social life. The full theory behind that sketch is treated in the companion study of the marxist theory of the state guide, and the transitional form itself, under its famous and fiercely contested name, is examined in the article on the dictatorship of the proletariat. The anarchist half carries an equally elaborate construction: a theory of power as self perpetuating, an insistence that means determine ends because institutions shape the people who staff them, and a counter sketch of the transition, in which the old state is dissolved at the moment of revolution and replaced from the first day by federated communes, militias, and workplace councils that never consolidate into a separate governing layer. Neither side was naive. Each knew the other’s objection by heart, because they had spent years making those objections to each other in the same organisation, in the same congresses, often in the same rooms.

The agreement on the destination also explains a feature of the quarrel that puzzles outsiders: its intimacy. Marxists and anarchists fight like relatives because they are relatives. They recruit from the same constituencies, read many of the same books, honour some of the same martyrs, and define themselves against the same enemy, the capitalist order and its state. Movements that want different things can coexist or ignore each other. Movements that want the same thing by incompatible means cannot, because every recruit, every union, every insurrection poses the choice between them in practical form. The bitterness is structural before it is personal, which is why the familiar explanation of the split as a clash of personalities, Marx’s authoritarian temperament against Bakunin’s conspiratorial romanticism, explains so little. Personalities supplied the sparks and much of the ugliness. The fuel was a genuine disagreement about what political power does to the people who hold it, and that disagreement would have divided the movement had both men been saints.

One further element of the shared inheritance deserves emphasis before the divergence takes over the story. Both traditions were formed in opposition to two enemies at once: the existing states of Europe, and the milder socialisms that proposed to use those states, by parliamentary pressure or benevolent reform, to soften capitalism into decency. Anarchists and Marxists alike treated that reformist hope as a trap, though for different reasons that foreshadow the split. The Marxist reason was analytical: the state, being bound to the class structure, could redistribute at the margins but could not be the instrument of its own abolition, since abolishing classes would abolish the state’s reason to exist and the interests staffing it. The anarchist reason was moral and institutional at once: any movement that sought power within the state would be absorbed by the state, its leaders converted into administrators, its goals quietly exchanged for the goal of governing. Notice that the anarchist argument against reformism is the same argument Bakunin would turn against the Marxist transition. The state corrupts the project that touches it. For Bakunin that principle admitted no exception for a state flying a red flag and staffed by former workers. For Marx it admitted exactly one exception, a state of a radically different type, held by the immense majority and organised to dissolve the conditions of its own existence. The entire dispute can be read as an argument about whether that exception is real.

One movement, two tempers: the common inheritance

The institutional setting of the quarrel was the International Workingmen’s Association, founded in London in 1864 and known to history as the First International. It gathered trade unionists, exiles, followers of the French socialist Pierre Joseph Proudhon, English labour leaders with no theory beyond their unions, and the circle around Marx, who drafted its inaugural address and its rules. Proudhon, who died in 1865, matters to this story out of proportion to his organisational role, because he had given the anti authoritarian side of socialism its founding vocabulary a generation earlier. He had called himself an anarchist in print, attacked property as theft and as despotism in successive breaths, and proposed a society of independent producers and communes bound by free contract and mutual credit rather than by government. Bakunin absorbed much of this, collectivised it, and carried it into the International when he and his followers joined in the late 1860s, bringing with them organised strength in Switzerland, Spain, Italy, and parts of France. By the time Bakunin’s Alliance of Socialist Democracy sought admission as a body within the International, the association contained two coherent and incompatible theories of its own purpose.

The federalist temper, as the anarchist side came to be called, held that the International itself should prefigure the society it sought: autonomous sections, free federation from below, no central authority capable of imposing a line, resolutions binding only on those who accepted them. The centralist temper, as its opponents named the Marxist side, held that a movement facing centralised states and a coordinated possessing class needed coordination of its own: a general council with real authority, common strategy, and the discipline to fight elections and insurrections as one body rather than as a scatter of local enthusiasms. What looks, in retrospect, like a procedural quarrel about committee powers was the transition dispute in miniature. If you believed that centralised organisation inevitably bred a caste of officials with interests of their own, you could not accept a strong general council even provisionally, because the provisional had a way of becoming permanent. If you believed that dispersed movements were simply defeated in detail by concentrated enemies, you could not accept a federation too loose to act, because defeat was a worse corruption than bureaucracy. Each side’s organisational demand followed from its theory of power, and each side understood the other’s demand as the theory made visible.

Beneath the organisational argument sat a disagreement about the revolutionary agent, which the brief for this comparison rightly lists among the secondary disputes and which deserves its weight. Marx centred the industrial proletariat: the class produced by capitalism itself, concentrated in factories and cities, disciplined by cooperative labour, and positioned to take over production because it already, in a sense, ran it. Bakunin, formed by the politics of the Russian and southern European countryside as much as by the factory districts, gave far greater weight to the peasantry and to what Marxist language called the declassed elements, the students, the ruined artisans, the brigands and rebels of the margins, whom he regarded as less corrupted by the existing order and readier for total revolt. The difference was partly sociological observation and partly temperamental wager. It mattered for the state question because an agent concentrated in industry could plausibly administer a transition through institutions it built in advance, unions and parties and councils, while an agent scattered across countryside and margin could only dissolve the old order and federate what emerged. Strategy followed sociology. A movement of the concentrated could imagine seizing and holding. A movement of the dispersed could only imagine destroying and linking.

There was also a disagreement, quieter at the time and louder in retrospect, about science and authority. Bakunin’s writings return repeatedly to the image of a government of savants, a legislature of scientists and engineers ruling in the name of knowledge, which he treated as among the most hateful forms of domination because it would be domination armed with a claim to competence. His objection was not to knowledge but to its conversion into office. The Marxist tradition, confident that it possessed a science of society, was slower to see the danger in that form, and Bakunin’s prediction about a new ruling stratum, examined in its own section below, has two branches precisely because of this: a branch about former workers elevated into governors, and a branch about the credentialed experts who would govern beside them in the name of the plan. Readers who know only the first branch underestimate the prediction. Bakunin foresaw a stratum defined not simply by origin but by position and by knowledge, the people who administer, coordinate, and know, ruling over the people who produce. That refinement will matter when the twentieth century evidence is priced.

The roots run deeper: Proudhon and the first quarrel

The argument of 1872 had a rehearsal a quarter of a century earlier, and the rehearsal fixed habits of mind on both sides that the later contestants inherited without noticing. In 1846 Proudhon published his System of Economic Contradictions, an attempt to build socialism out of the tensions internal to political economy, and the young Marx, who had corresponded with him and sought his alliance, answered in 1847 with The Poverty of Philosophy, a demolition so thorough and so contemptuous that it became the template for Marxist polemic against the anti authoritarian tradition: the opponent’s categories dissolved as confused, his dialectics exposed as a misunderstanding of Hegel, his practical proposals, mutual credit and the exchange bank, dismissed as petty bourgeois tinkering that left the structure of property intact. Much of the demolition was earned. Proudhon’s economics was genuinely muddled at crucial points, and his social vision, for all its influence on the French labour movement, carried limitations his admirers had to spend decades outgrowing, including a hostility to women’s emancipation and to strikes at odds with the movement he helped create. Yet something else was established in that exchange besides the verdict on one book. Marx’s side learned to treat the anti authoritarian tradition as a pre scientific stage of socialism, a moral protest awaiting replacement by analysis. Proudhon’s side, and after it Bakunin’s, learned to treat Marxist analysis as a bid for intellectual government, the savants organising in advance the hierarchy they would later staff. Each lesson contained a fragment of truth about the other’s weakness. Each also functioned as an inoculation, ensuring that when the two traditions met inside the International they would hear each other’s strongest arguments through a filter installed decades before.

The substantive legacy of the Proudhon episode matters alongside the temperamental one. Proudhon’s federalism, his vision of society as a web of autonomous communes and associations bound by contract rather than command, supplied the architectural drawing from which Bakunin built, adding collectivism about land and the instruments of production where Proudhon had wanted individual possession secured by mutuality. His critique of government as such, the argument in his later writings that every government, whatever its origin, tends toward the absorption of society by the state, supplied the theorem Bakunin would apply to the workers’ state with such effect. And his distance from insurrectionary politics, his preference for building the new society’s institutions inside the shell of the old, marked the point where Bakunin broke with the master: for Bakunin the old state had to be destroyed by revolt before federation could breathe, a difference within the anarchist inheritance that later generations would replay as the tension between revolutionary and gradualist anarchisms. When Marx attacked Proudhon he was therefore attacking, without knowing it, the draftsman of the building his own movement would spend a century arguing about. The building’s designer had one more unsettling quality, from the Marxist point of view. Proudhon distrusted all systems that promised to administer society from a centre of knowledge, including the system being assembled in the British Museum reading room by his young correspondent. The government of savants appears in Bakunin fully armed, but its first sketch is Proudhonist: the fear that socialism, misunderstood, might become the rule of those who understand society, over those who merely live in it.

The quarrel inside the International

The end, when it came, was institutional. Through the congresses of the late 1860s and the first years of the 1870s the two camps contested resolutions on political action, on inheritance, on the powers of the General Council, and the votes tracked a real division in the movement’s geography: the Marxist resolutions carried where industrial organisation was strongest, in England and Germany and the emigrant circles of America, while the federalist resolutions carried the Latin and Swiss sections, where craft workers, peasants, and artisans predominated and where suspicion of central government was a lived tradition rather than a doctrine. The Paris Commune of 1871 sharpened everything. Marx claimed the Commune for his theory in a famous address, reading its measures, the recallable officials and the suppressed standing army, as the discovered form of working class rule. Bakunin’s supporters claimed the same event for theirs, reading its federal structure and its spontaneity as proof that revolution needed no directing party. A single event had supplied both sides with their emblem, which is usually a sign that the event is being asked to settle a question it cannot settle, and the struggle for the International became a struggle over which reading would own the movement’s machinery.

The Hague Congress of September 1872 settled the ownership question by the methods organisations use when argument has failed. Marx attended in person, an unusual step, and his side commanded a majority of mandates, some of them contested. The congress voted to strengthen the General Council, voted resolutions committing the International to political action and the conquest of political power, and then voted the expulsion of Bakunin and his ally James Guillaume, on charges centred on the Alliance, the semi secret organisation Bakunin had maintained within the International after its formal dissolution had been announced. The proceedings had the character of a trial conducted by the victors, and the evidence and counter evidence about the Alliance’s real existence and purposes have been fought over by historians ever since. What is not contested is the sequence and its meaning. The Marxist leadership chose to expel the federalists and move the General Council to New York, beyond the reach of European factional war, rather than let the International continue as a federation containing an organised anti authoritarian opposition. The expelled and their sympathisers met days later at Saint Imier in Switzerland and founded a rival International on federalist principles, declaring that the destruction of all political power was the first duty of the proletariat. The split was thereby formalised: two Internationals, each claiming the inheritance, and within a few years both were shadows, the New York body dissolving in 1876 and the federalist body fading by the end of the decade. The argument, however, had been fixed in its enduring form by the manner of the parting. It would never again be one argument inside one movement. It would be a frontier between movements, patrolled by memory.

Why was the split enacted at a congress rather than settled in print?

Congresses controlled mandates, funds, affiliations, and the authority to speak for the International, so the quarrel over doctrine became a contest for machinery. Expulsion protected the Marxist majority’s resolutions from an organised federalist opposition, while the federalists, once expelled, could found a rival body on their own principles immediately.

That the settlement was institutional rather than intellectual is the single most clarifying fact in the whole history, and the brief for this article is right to put it near the front. No congress refuted Bakunin’s argument about transitional states, and no pamphlet refuted Marx’s argument about the necessity of political power, because neither argument was of a kind that the other side’s best efforts could simply demolish. They were wagers about institutional behaviour over time, and wagers are settled by events, not by votes. The Hague voted on who would hold the International. History was left to vote on who had been right, and history’s ballot took most of a century to count. Meanwhile the institutional settlement had a distorting effect on the intellectual one, because each tradition thereafter developed in partial isolation, training its members on its own canon, nursing its own martyrology, and refining its answer to an opponent it increasingly encountered as caricature. The Marxist tradition produced shelves of refutations of an anarchism few of its writers had studied at first hand. The anarchist tradition produced an equally confident literature about a Marxism reduced to its most authoritarian custodians. Serious comparison has to tunnel back under that accumulated polemic to the moment when the two positions were stated by their ablest holders in full knowledge of each other.

The federalist case had been stated in its most quotable form a year before the Hague, in the Sonvillier Circular of November 1871, issued by the Jura Federation, the watchmaking communities of the Swiss mountains whose sections formed Bakunin’s most disciplined base. The circular protested against a conference held in London that autumn, which had asserted the General Council’s authority and committed the International to forming working class political parties in every country, decisions the Jura socialists regarded as the substitution of a directing centre for the free federation the International’s founding rules had promised. Their counter statement deserves its fame, because it formulates the anarchist theory of organisation as a principle of prefiguration with complete clarity: the society of the future, the circular argues in substance, cannot be built by an organisation that contradicts it in the present, and the International, as the embryo of the coming society, must therefore refuse every institution, a government, a priesthood of doctrine, a permanent directing caste, that the coming society is meant to abolish. A free society can only be created by free organisation, in the circular’s logic, so the means are not a ladder to be kicked away but a mould that shapes the casting. Set that paragraph beside Bakunin’s prediction and the two documents illuminate each other: the circular states the principle for movements, the book applies it to states, and in both the claim is that forms are not neutral vessels but active ingredients, that what an organisation is, it tends to make more of. The Marxist reply, at the Hague and after, held that an embryo is not the organism, that a fighting organisation facing states is not yet the society it fights for, and that confusing the two disarms the fight in the name of a purity the enemy does not share. The exchange has never been improved upon as a statement of the organisational dispute, and seminars still begin from the circular because its authors managed, in a few pages of provincial polemic, to say the thing their whole tradition exists to say.

The charges exchanged at the Hague also deserve a sober word, because they prefigure the methods each tradition would later attribute to the other. The Marxist side accused Bakunin of maintaining a secret hierarchy inside an ostensibly open movement, an invisible dictatorship of initiates, and the accusation was not groundless in form even where it was exaggerated in detail: Bakunin did build revolutionary brotherhoods with graded membership, and his organisational practice sat uneasily beside his federalist doctrine. The anarchist side accused Marx of manipulating mandates, packing the congress, and ruling the General Council as a personal instrument, and that accusation too had its measure of substance: the Marxist leadership fought the factional war with every procedural weapon available, and was not fastidious about the difference between defeating an argument and disabling its holders. A neutral reader should register the symmetry. Each camp, in the heat of an internal struggle, behaved in ways its own theory should have predicted and condemned, the centralists centralising beyond their stated purpose, the anti authoritarians constructing authority without admitting it. The episode is a small, early exhibit in the larger inquiry this article conducts: whether organisational forms generate their predicted pathologies regardless of the intentions of the people inside them. On the evidence of the International itself, both sides have something to answer for, and any honest account must say so before proceeding to the asymmetric part of the story.

Bakunin’s prediction: a state of former workers

Bakunin’s Statism and Anarchy, written in 1873 in the aftermath of the expulsion and published after his death, contains the most consequential passage in the anarchist canon for this comparison. The book is a sprawling polemic against the German state tradition, against Marx, and against the very idea of a people’s state, and in the middle of it Bakunin poses a question to the Marxists with a precision that lifts it out of polemic altogether. If the proletariat is to be the ruling class, he asks, over whom will it rule? His answer works through the composition of any actual workers’ government. The millions who labour cannot themselves sit in government. They will be represented, and their representatives, however proletarian their origins, will cease to work, will take up the standpoint of the state, and will come to view the world of labour from the height of administration. Around them will gather the second element of the new stratum, the men of science and credentialed expertise, whose knowledge the government of a complex economy cannot do without and whose authority will wear the legitimating dress of competence. The result, in Bakunin’s forecast, is a new privileged class, a red aristocracy of former workers and professional knowers, governing in the name of the class it has left, and defending its position with the instruments every state possesses. He was equally blunt about duration. A dictatorship, he argued, whatever its founding purpose, can have no other aim than to perpetuate itself, because the people who hold emergency power acquire interests in the emergency.

Three features of the prediction give it its unusual scientific standing in a field of vague prophecy. First, it is specific about mechanism: representation at scale, the separation of governing from working, the standpoint shift that position induces, the alliance between elevated ex workers and credentialed experts. A prediction with a mechanism can be checked piece by piece, and the pieces can be separately confirmed or disconfirmed. Second, it is unconditional in form. Bakunin did not say that a workers’ state would entrench itself if its leaders were corrupt, or if conditions were unfavourable, or if the revolution occurred in the wrong country. He said that the outcome followed from the structure of state power itself, from what a state is, and would therefore follow in the most favourable case run by the most sincere revolutionaries. Third, it was made in advance, in print, dated 1873, before any workers’ state existed anywhere, by a named author arguing against the named programme that would later be tried. Those three features are what the namable claim of this article, the prediction asymmetry, is built from. Hold them in mind, because the Marxist counter expectation, the withering of the transitional state, possesses none of the three in the same degree, and that difference in logical shape, more than any difference in sincerity or insight, is what the twentieth century was able to grade.

What exactly did Bakunin predict a workers’ state would produce?

He predicted a new privileged stratum: representatives drawn from the working class who would stop working, adopt the standpoint of government, and rule alongside credentialed experts whose authority rested on knowledge. The stratum would govern in the class’s name, defend its own position, and convert emergency power into permanent rule.

Bakunin’s reasoning about the state went deeper than a forecast about one kind of government, and the deeper layer explains why no Marxist reassurance about the transitional state’s special character could touch it. In his analysis the state is not a tool that happens, under capitalism, to be held by the wrong hands. It is a form of organisation with its own logic: a minority set apart to coerce a majority, necessarily developing interests distinct from the majority’s, necessarily reproducing the division between rulers and ruled that it was supposed to administer. Domination, for Bakunin, is not a policy a state pursues but the substance of what a state is, so the phrase workers’ state contains, on his account, a contradiction that time will expose. The workers in it will either cease to be workers, becoming the new rulers, or cease to hold the state, in which case others will hold it over them. He applied the same analysis without favour to his own camp’s temptations, warning against revolutionary dictatorships of every label, including any that might be erected in anarchism’s name, and against the invisible government of revolutionary elites manipulating movements from behind the scenes, a warning his own organisational practice did not always honour but which his doctrine stated without ambiguity.

The prediction of 1873 was not an improvisation of the quarrel’s final phase. Bakunin had been building toward it for years, in the federalist writings of the late 1860s where the architecture of free communes first takes shape in his work, and in his polemics against the idea, which he attributed to the German social democrats of his day, that the conquest of state power by a workers’ party could be a stage on the road to a stateless society rather than a detour into a new servitude. Statism and Anarchy, left unfinished at his death and assembled from manuscripts, gave the argument its fullest form because it was written with the Hague behind him and the German Empire in front of him, the most powerful new state in Europe presenting itself to his mind as the shape of the future his opponents proposed to capture. Reading the book’s polemic against Bismarck’s Germany beside its polemic against Marx is instructive: the same analysis serves both, because for Bakunin the two were one subject, the state as such in its modern, centralised, scientifically administered form, and the workers’ state differed from the Kaiser’s, in his account, chiefly in the biographies of its personnel. That refusal to grant the revolutionary state a separate analytical category is the whole dispute in one gesture, and Marx’s conspectus, which labours precisely to establish the separate category, class state against class state, transitional form against permanent form, is the reply in one gesture. Every later document in the quarrel is a variation on those two gestures, performed by people who mostly did not know they were performing them.

The prediction found a second great carrier in the next anarchist generation, and the carrying changed its emphasis in ways worth recording. Peter Kropotkin, the geographer and naturalist who became anarchism’s most systematic theorist after Bakunin, restated the analysis of the state as a historical institution with a birthday and, in principle, a deathday: his unfinished study of the state argued that the modern centralised state was a comparatively recent construction, layered over older federative and communal forms that it had spent centuries destroying, and that what history had assembled, history could disassemble. Where Bakunin’s forecast is a warning delivered in the heat of a schism, Kropotkin’s version is a research programme, an invitation to examine how states actually grow, by absorbing the functions of communes, guilds, and mutual associations and converting free cooperation into administration. The difference of register matters for the prediction’s afterlife. In Bakunin’s form it is a prophecy to be graded against the transitional states. In Kropotkin’s form it becomes a standing hypothesis about institutional growth everywhere, applicable to empires and republics as much as to revolutionary governments, and it quietly entered the common sense of later historiography and sociology, often without its author’s name attached. Traditions should be graded partly by what they succeed in making obvious to people who reject them, and by that measure the anarchist analysis of the state has had a stranger and larger victory than its political history suggests: the suspicion that apparatuses pursue their own preservation, once a factional slogan, is now the ordinary premise of half the social sciences, while the withering expectation survives mainly as a period phrase. That inversion of fortunes, the losing movement’s idea becoming the common property, is a pattern the history of thought knows well, and it should caution any reader against grading traditions by their organisational survival alone.

His positive alternative completes the picture and must be stated as strongly as the prediction, because the comparison fails if anarchism appears as pure negation. The old state, in Bakunin’s programme, is to be abolished in the revolution itself, not captured and not inherited, and in its place society organises from below: workplaces and communes governing their own affairs, federating through mandated and revocable delegates for purposes that exceed local scale, defence conducted by the armed population organised federally rather than by a standing army answering to a centre. The transition, on this account, is not a phase between the old order and the new but the new order in the act of forming, since any separate transitional apparatus would congeal into the disease it claimed to cure. Whether that alternative could survive contact with a determined counter revolution is the great unpriced question of the comparison, and this article will return to it when the strongest Marxist objection is stated in full. For now the point is structural. Bakunin had a theory of transition. It was a theory in which the transition and the destination are the same process viewed at different moments, and its elegance is inseparable from its risk: it permits no fallback position, no temporary machinery to be dismantled later if the pure form falters, because the pure form is all there is.

Marx’s answer: the conspectus and the Commune model

Marx read Statism and Anarchy and filled his notebooks with marginal comments in 1874 and 1875, a private working document not published in his lifetime and known to scholarship as the conspectus. The tone is frequently scornful, and the scorn is worth noting because it has misled casual readers into treating the exchange as abuse rather than argument. Beneath the marginalia there is a real argument, and it has three moving parts. The first is a refusal of Bakunin’s framing question. Asked over whom the ruling proletariat will rule, Marx replies, in substance, that the question dissolves once the terms are made concrete: so long as other classes survive, above all the capitalist class and the peasantry in transition, the proletariat must use coercive measures against them, but it rules as a class organised for that suppression, not as a government caste standing over itself. His mockery fastens on the image of a class simultaneously ruling and being ruled in the same respect, which he treats as a verbal puzzle generated by abstract nouns. Classes, in his analysis, are not persons, and a class organised as the state is the class itself in its political form, not a committee that has escaped from it.

The second part of the reply concerns the machinery by which the danger Bakunin named is to be prevented, and here Marx leans on the Paris Commune, which he had analysed in The Civil War in France in 1871 and which had become, for the Marxist tradition, the empirical template of the transitional form. His most quoted sentence on the subject runs to a dozen words: the working class, he wrote, cannot simply lay hold of the ready made state machinery. The conclusion he drew was not that workers should abstain from state power but that they must build a different kind: a working body rather than a parliamentary one, executive and legislative fused, officials elected, recallable at any time, and paid no more than workmen’s wages, the standing army abolished in favour of the armed people, the police stripped of political functions. Each Commune measure reads, in this light, as a targeted answer to a clause of Bakunin’s prediction. Recall answers the separation of representatives from represented. Workmen’s wages answer the material privilege of office. The armed population answers the monopoly of coercion by a separate body. Whether the answers suffice is the empirical question of the next century. That they were designed against exactly the danger Bakunin later codified, and designed before Statism and Anarchy appeared, is a fact the anarchist tradition has never found comfortable, since it shows the Marxist founder treating the pathology of office as real and engineering against it, rather than waving it away.

The third part of the reply is the least quotable and the most important. Marx insists, against Bakunin, that the transition is governed by material conditions rather than by institutional choice alone. A revolution inherits a level of production, a geography, a peasantry, a set of enemies with armies, and the forms of its transition are hammered out under those constraints rather than selected from a catalogue of principles. Where Bakunin treats the transitional state as a free choice that a wise movement declines, Marx treats some apparatus of coercion and coordination as a necessity imposed by the survival of hostile classes and by the sheer administrative weight of a complex economy in upheaval, a necessity that can be shaped toward dissolution but cannot be wished away on the first day. The withering expectation belongs to this part of the argument: as classes and class antagonisms disappear, public power loses its political character, coercion becomes administration, and the state, in Engels’s later formulation, is not abolished but dies out. The full anatomy of that expectation, including its internal tensions, belongs to the companion article on the withering away of the state, and is deliberately not duplicated here. What matters for the comparison is the expectation’s logical shape. It is conditional, gradual, and tied to a process, the disappearance of classes, whose timetable no text specifies and whose completion no observer could certify. Compare that shape with the shape of Bakunin’s forecast and the prediction asymmetry is already visible in embryo, a generation before either claim met its evidence.

The same months produced the other canonical Marx text on transition, the Critique of the Gotha Programme, written in 1875 as a private letter of objections to the programme of the newly unified German workers’ party and published only after Marx’s death, against the wishes of the party leadership it criticised. Its contribution to this comparison is a single austere passage. Between capitalist society and the classless society, Marx argues, lies a period of revolutionary transformation, and the state corresponding to that period can be nothing, in his deliberately unadorned phrase, but the revolutionary dictatorship of the proletariat. Almost everything in the sentence has been fought over since: the noun, with its Roman echo of emergency office; the absence of institutional specification, which allowed later regimes to pour their own machinery into the phrase and allowed later libertarians to point out that the Commune writings supply a specification the phrase omits; and the framing of the transition as a period, a stretch of history with its own laws, rather than as a moment. For the comparison the passage matters chiefly as evidence of how Marx himself held the two ends of his position together. The destination in the Gotha critique is emphatically stateless, a society that has outgrown the narrow horizon of bourgeois right and distributes according to need, and the road runs unapologetically through a state. Bakunin, reading over the same years, could find in the passage the whole error in one sentence: a period, governed by a state, whose length no one specifies, entrusted to a class whose representatives no mechanism restrains, and expected to end in the state’s own extinction by the logic of a destination the state itself will interpret. The conspectus and the Gotha critique, taken together, are the complete Marxist answer as its author left it, and later Marxism’s history with the question is largely the history of which of the two texts, the Commune’s safeguards or the Gotha period’s open cheque, its custodians chose to emphasise when their turn came to hold the state the texts describe.

A fair account must also register what Marx’s reply does not contain. He does not predict that the transitional state will entrench itself, obviously, but neither does he offer a mechanism by which its dissolution is guaranteed against the interests of the people staffing it, other than the Commune devices and the underlying transformation of class relations. His confidence rests on the proposition that a state held by the immense majority, in a society where the material basis of class privilege is being dismantled, lacks the social foundation from which a new ruling stratum could grow. Bakunin’s prediction is precisely the denial of that proposition: position, he claims, manufactures interest regardless of foundation. The dispute, reduced to its core, is about whether a governing stratum needs a basis in property to consolidate itself, or whether control of the apparatus is basis enough. That is a beautifully precise disagreement, and the twentieth century spoke to it with unusual directness, because the regimes that claimed Marx’s inheritance abolished private ownership of the major means of production while concentrating control of the apparatus to a degree Marx never sketched. If a privileged and self perpetuating stratum can arise on the basis of control alone, without property, Bakunin’s mechanism is confirmed in its strongest form. If those regimes are disqualified as not genuinely transitional, not genuinely of the Commune type, not genuinely socialist in their foundations, the withering expectation survives every test by declining each one. Both moves were made, at length, by serious people, and section by section this article will keep them distinct.

Engels on authority: coordination is not the state

The sharpest short document on the Marxist side of the quarrel is not by Marx but by Engels: the essay On Authority, written in 1872 for an Italian socialist almanac and aimed at the anti authoritarians of the International in the immediate aftermath of the Hague. It is barely a few pages long, and generations of readers have misfiled it as an argument for the state. Its actual target is narrower and its actual claim more interesting. Engels argues that authority, the imposition of one will upon another in coordinated activity, is a feature of material organisation that no revolution and no change of ownership can remove. A factory running on powered machinery imposes the authority of the process itself: hours fixed by the interdependence of tasks, decisions taken by those charged with them, compliance enforced because the alternative is the machine stopping for everyone. A railway cannot be run by perpetual assembly. A ship in a storm cannot vote on each order. Engels’s examples are chosen from production and transport rather than from politics, and the choice is the argument. If authority of this functional kind survives every change of regime, then the anarchist slogan of abolishing authority as such is, in his word, an absurdity, and the honest dispute is about which authorities, exercised by whom, for what purposes, and subject to what revocation.

His closing move is the one that entered the tradition’s bloodstream. Having argued that authority is unavoidable, he turns the anti authoritarians’ charge around: the people who demand the abolition of all authority before the social revolution have the sequence backwards, because a revolution is itself, as he puts it, certainly the most authoritarian thing there is, an act in which one part of the population imposes its will on the other by force of arms. The sentence is quoted wherever this dispute is rehearsed, and its function in Engels’s essay is dialectical rather than celebratory. He is not praising authoritarianism. He is denying that the anarchists occupy ground outside the problem. Their revolution, too, will coerce: it will expropriate the expropriators, suppress armed resistance, compel compliance with decisions taken under fire. The difference, he concludes, is that the Marxist side admits the coercion, names it, organises it as a transitional state, and argues about its limits, while the anarchist side performs the same coercion under federal and spontaneous descriptions that conceal its character and thereby exempt it from scrutiny. Coercion acknowledged can be watched. Coercion denied is answerable to no one. That is the strongest form of the Marxist case on authority, and it deserves to be felt at full weight before any reply is entertained.

How does Engels separate authority from the state?

He treats authority as a function of coordinated material activity, present wherever interdependent work must be directed, while the state is a specific institution: a public force separated from the population and organised for class rule. Abolishing the state, on his account, removes the institution; it cannot remove the function, which any complex production process regenerates daily.

The anarchist replies, stated at matching strength, begin by granting the factory and disputing the inference. No serious anarchist proposed that a railway timetable be settled by acclamation or that a ship’s crew deliberate while the ship founders. The federalist tradition distinguishes, with as much care as Engels could wish, between authority that is technical, temporary, and answerable, the authority of the engineer over the bridge design or the delegate carrying a mandate, and authority that is governmental, permanent, and self extending, the authority of a body that commands because it commands. Bakunin himself wrote that in matters of boots he deferred to the authority of the bootmaker, while reserving his reason against any bootmaker who wished to rule him in matters beyond boots. The distinction between competence and command is the anarchist answer to On Authority, and it is a real distinction, though Engels would reply, and did in effect reply, that under revolutionary conditions the distinction collapses at exactly the moments that matter, when the competent body must compel the unwilling and the temporary emergency refuses to end. The second anarchist reply is the institutional one that runs through this whole comparison: whatever authority production requires can be lodged in federated, recallable, rotating bodies arising from the workplaces themselves, acquiring no separate corporate existence, whereas a transitional state concentrates the same functions in a body that does acquire such an existence and then defends it. The essay settles none of this. What it settles, permanently, is the terms: after 1872 no participant in the dispute could honestly pretend that the choice was between authority and its absence. The choice was between forms of authority, differently housed and differently watched, and the argument had to be conducted at that level or not at all. A neighbouring article takes up the charge that follows naturally from this one, whether Marxism’s account of the state leaves room for democracy at all, under the heading of the marxism and democracy critique.

The transition comparison

Arguments of this size need a fixed point of reference, a place where the two positions stand side by side on identical questions so that the reader can check any later claim against the original alignment. That is the office of the transition comparison, the findable artifact of this article. It sets out six questions: the end state, the transitional form, the agent, the organisational principle, the view of authority, and the prediction each tradition makes about a workers’ state, with a final column recording what the historical record shows where the record is capable of showing anything. The last column is deliberately uneven. On some rows history has delivered volumes of evidence. On others it has delivered almost none, because the anarchist alternative was never operated at national scale across decades, and the table marks those rows as untested rather than allowing either tradition’s partisans to fill the gap with confidence. A comparison that grades one side on evidence and the other on intention is rigged, and the rigging is usually invisible. Making the empty cells visible is the table’s chief virtue.

Question	Marxist position	Anarchist position	What the record shows
End state	A classless society in which the state has withered as classes disappear	A classless society organised as a free federation of communes and workplaces	Agreement on the destination; no society of either description has existed at scale
Transitional form	A workers’ state on the Commune model, holding power to suppress counter revolution and reorganise production	No transitional state; the old state is abolished in the revolution and federation begins at once	Transitional states were built repeatedly under Marxist banners; none withered
The agent	The industrial proletariat, organised and conscious of its historical role	Workers together with peasants and the declassed, rising in spontaneous revolt	Industrial workers were central in some revolutions, marginal in others; no single pattern holds
Organisational principle	A disciplined party and centralised coordination under democratic forms	Free federation from below, mandated and recallable delegates, no permanent centre	Centralised parties proved durable and effective at seizing power; federations proved fragile under military pressure
View of authority	Authority is inherent in coordinated production and must be organised openly and transitionally	Technical authority is accepted; governmental authority is rejected as self perpetuating	Both traditions exercised coercion in power; only one tradition’s theory expected to
Prediction about a workers’ state	It will wither as its class basis disappears, provided it takes the Commune form	It will breed a new ruling stratum of former workers and experts and perpetuate itself	Bakunin’s predicted stratum appeared in recognisable form; withering did not occur in any case

Read the table from the bottom up and the article’s argument is in front of you in miniature. The first row records an agreement so complete that the quarrel cannot be about ends. The middle rows record disagreements of design, each defensible, each carrying costs the other side enumerated in advance. The last row records an asymmetry of a different order from everything above it: two predictions, only one of which events were able to seize and grade. Before that grading is examined, the middle rows deserve their own fuller treatment, because the secondary disagreements are not decorative. Each of them fed back into the state question and hardened it. The agent disagreement decided whose institutions would be trusted with a transition. The organisational disagreement decided what a movement would build before the revolution, and movements fight their revolutions with the organisations they have, not the organisations their theory prefers in the abstract. The authority disagreement decided what each side could admit about its own conduct, and a tradition that cannot admit its coercion cannot correct it.

The secondary disagreements: agent, organisation, authority

Consider the agent question with the seriousness both sides gave it. Marx’s proletariat was not a sentimental preference for factory workers over other sufferers. It followed from his analysis of where power over production actually sits. The class that works the mines, mills, railways, and docks can, by combining, stop the economy, and can, by the same combination extended, run it; its members are gathered by capital itself into large cooperative units where the habits of collective action are learned as a by product of employment. A revolution made by that class carries its successor institutions inside itself, unions becoming organs of distribution, factory committees becoming organs of management. Bakunin’s wider and wilder agent followed from an opposite reading of the same Europe. In the countries he knew best, Russia, Spain, Italy, the industrial class was a small island in a peasant ocean, and a theory that waited for the island to become the continent postponed revolution indefinitely. The peasantry, land hungry and rebellious by tradition, and the declassed elements of the towns, students without prospects, artisans ruined by the factory, discharged soldiers, had in his judgement nothing to conserve and therefore the purest revolutionary instinct. Marxists replied that instinct without position produces jacqueries, magnificent and sterile, burned out in a season and leaving the structures untouched. Anarchists replied that position without instinct produces administrators, that the most organised sections of the working class were precisely those with the most to lose, and that a revolution captained by the labour aristocracy of the advanced countries would stop at the frontier of its own comfort. Each reply landed on something real, which is why the argument outlived both men and reappears inside every later revolutionary movement as a dispute about whom to organise first.

How do the traditions differ on who makes the revolution?

Marxists centre the industrial proletariat, gathered and disciplined by production itself and able to run the economy it stops. Bakunin gave decisive weight to peasants and the declassed, judging them less integrated into the existing order and therefore readier for its total overthrow rather than its negotiated reform.

Organisation is where the secondary disagreements stop being secondary. The Marxist commitment to the party followed from the transition theory in a straight line: if the revolution must hold state power against armed counter revolution, expropriate an economy, and administer a society through a dangerous interval, then the instrument of that work must exist before the interval begins, must be disciplined enough to act as one body under fire, and must be capable of becoming the skeleton of the transitional state. The anarchist rejection of the party followed from the anti state theory with equal directness of its own: an organisation built to seize and hold power will find reasons to hold it, its leading layer will practise governing before the revolution and will not unlearn the practice after, and the party is therefore the embryonic form of the very stratum Bakunin predicted. Federation was not, for the anarchists, a sentimental preference for looseness. It was a prophylactic. Structures that cannot command cannot tyrannise; delegates who can be recalled and whose mandates expire cannot consolidate; a movement with no permanent centre has no palace to storm and no throne to inherit. The Marxist counter argument writes itself and was written, often: structures that cannot command cannot win, and a revolution that cannot win delivers its participants to the prisons and gallows of the old order, which is a high price for prophylaxis. The record of the following century gave evidence to both lines of argument in different theatres, disciplined parties surviving and triumphing where federations were scattered, and triumphant parties exhibiting, in power, the pathologies the federalists had forecast for them. The two findings are compatible, and holding them together without flinching is most of what maturity in this subject consists in.

Authority, the third secondary disagreement, has already been introduced through Engels’s essay, and its remaining dimension is the question of what each side’s theory permits it to see in itself. A tradition that expects coercion to continue through a transition builds, at least in principle, instruments for watching the coercers: recall, rotation, publicity, the Commune devices. A tradition that defines its own arrangements as non authoritarian by construction lacks a vocabulary for its own exercises of power, and its exercises of power, where they occurred, militia discipline, revolutionary tribunals improvised under siege, the suppression of rival organisations, tended to be carried out without the self scrutiny the other tradition’s theory at least demanded. This is not a debating point invented for symmetry. It is the anarchist movement’s own internal literature, its post mortems on Spain especially, that records the discovery, made in government offices and at fronts, that refusing the name of power had not refused the thing. Conversely the Marxist tradition’s vocabulary for its coercion proved, in the great test, to be no protection either: the devices existed on paper in constitutions while the apparatus answered elsewhere. A form of words in a founding document, whether the word is federation or withering, settles nothing by itself. What settles outcomes, if the evidence of both traditions is allowed to accumulate, is whether the people exercising power can be removed by the people over whom it is exercised, through channels that work under emergency conditions and not only in calm ones. Neither nineteenth century tradition put that criterion at the centre of its theory, though the Commune model gestures toward it and the federalist model assumes it, and the omission in both is one of the places where a modern reader can see further than either founder, using their own materials.

The test arrives: the record of the revolutionary century

The generations after the split did what generations do with an inheritance: they spent it under conditions the founders never faced. Revolutionary movements claiming the Marxist succession took and held state power across a vast arc of the twentieth century, beginning with Russia in 1917 and extending through eastern Europe, China, and a series of postcolonial states. The full record of that experience belongs to the companion history of soviet marxism complete guide and is deliberately summarised rather than narrated here, because this article’s question is not what those regimes did in general but what they show about one prediction. Summarised at the level of the prediction, the record has a stark shape. In every case a governing stratum consolidated: a party and state apparatus whose members enjoyed privileged access to housing, goods, education for their children, and security, whose tenure did not depend on the producers they administered, and which reproduced itself across generations without owning the means of production in the classical sense. Control, not property, was the basis of the privilege, exactly the possibility Marx’s reply to Bakunin had treated as dissolved by the class character of the state and that Bakunin’s mechanism had placed at the centre of the forecast. The stratum’s members were, in the founding generations, disproportionately people of working class and peasant origin elevated by the revolution, joined, as the economies grew more complex, by the credentialed administrators and technical experts whose parallel rise Bakunin had also specified. No transitional state of the Marxist type withered. Several grew more coercive across decades. One, the Soviet Union itself, eventually collapsed, and its apparatus stratum converted control into property in the aftermath with a fluency that read, to many observers, like the prediction’s final footnote.

Against that stark shape stand the qualifications, and a serious article must give them their full size rather than a dismissive sentence. The first qualification is the Commune objection: the regimes in question departed, usually early and usually under the pressure of civil war and invasion, from the specific institutional devices, recallable officials on workmen’s wages, an armed population in place of a standing army and a separate bureaucracy, that Marx had identified as the transitional form’s safeguards. A defender of the withering expectation can therefore argue that the experiment tested a different machine from the one specified, a centralised party state of a type Marx never described and would not have recognised. The objection is not frivolous. It is, however, double edged, because Bakunin’s prediction was explicitly about what would happen to any workers’ state under real conditions, and an experiment that can only be run in perfect conditions is an experiment that cannot be run. The second qualification is the encirclement argument: every transitional state faced hostile powers, embargoes, invasions, and the permanent emergency of backwardness, and emergency is the classic solvent of liberties and the classic nourishment of apparatuses. Perhaps, under favourable conditions, the withering would have begun. Perhaps. The argument is unfalsifiable in exactly the way that matters for the prediction asymmetry, since favourable conditions never arrived anywhere and the expectation therefore never faced a clean test. The third qualification runs the other way and belongs to the anarchists’ debit column, and it concerns their own record when history handed them their chance.

The anarchist movement’s great test came in Spain in 1936, where the strongest anarcho syndicalist movement in the world, the CNT, with its federation the FAI, confronted military rebellion and civil war in a country where it organised millions. The Spanish case is named here as the brief for this comparison requires, and left at summary level, since detailed treatment sits outside this article’s territory. The summary is chastening for the anarchist side in a specific and instructive way. Faced with war, the movement’s organisations built the very instruments their theory proscribed: militia columns with command structures, then incorporation into a regular army; representatives entering the Catalan government and then, in November 1936, the government of the Spanish Republic itself, anarchist ministers sitting in a state their movement existed to abolish. The reasons were the reasons states always give: coordination, supply, diplomacy, the need to speak to foreign powers and to run a war economy. The compromise did not save the revolution, which was ground down between its fascist, republican, and communist enemies, and anarchist participation in government was revoked by defeat rather than by withering. Spain therefore does not test the anarchist alternative in the way Russia tests the Marxist one, because the anarchist alternative, a federated war effort and a socialised economy without a state, was abandoned by its own custodians within months under the pressure of events. But Spain tests something else: the claim that a movement immunised by doctrine against state power will decline that power when declining it becomes expensive. On that claim the record is plain, and it belongs in any honest ledger. The earlier Russian episode of Kronstadt in 1921, where sailors who had been heroes of 1917 rose against the party state and were suppressed by it, tests the other side’s ledger in a single emblematic event: a workers’ government using organised force against workers demanding soviet democracy, the prediction’s mechanism and the revolution’s tragedy compressed into one garrison town.

The prediction asymmetry

Name the claim of this article plainly, in the form in which it can be argued with: the prediction asymmetry. The dispute between Marxism and anarchism contains many elements, differences of analysis, of temperament, of organisational taste, that no evidence could grade. It also contains one element of a rarer kind. In 1873 a participant in the dispute published a specific, unconditional, mechanism bearing prediction about what his opponents’ programme would produce if it were ever tried: a workers’ state would generate a new ruling stratum of former workers and credentialed experts, would govern in the name of the class while separating itself from the class, and would perpetuate the emergency that justified it. The programme was tried, repeatedly, across most of the following century, in countries at every level of development, under leaders of every temperament, some cynical, some sincerely devoted to the withering ideal. The predicted stratum appeared in recognisable form in case after case. The counter expectation, that a transitional state of the right type would progressively lose its functions and die out as classes disappeared, was confirmed in no case, and was protected from disconfirmation by its conditional and gradual form: the type was never quite right, the classes had never quite disappeared, the conditions were never quite favourable. One side’s central empirical claim survived contact with the century in good order. The other side’s survived by being unsuited to contact. That difference is the prediction asymmetry, and it makes this dispute, almost uniquely among the great quarrels of social theory, partially decidable.

What makes one side’s claim testable and the other’s elastic?

Bakunin specified a mechanism, named the stratum’s composition in advance, and made the forecast unconditional, so any durable workers’ state either exhibits the stratum or refutes him. The withering expectation is conditional on the right form, the disappearance of classes, and favourable circumstances, so every failure can be charged to a missing condition.

Precision about what the asymmetry does not establish matters as much as the claim itself, because both traditions have overdrawn it in their pamphlets. The asymmetry does not establish that anarchism is true. A prediction about the failure of a rival programme is not evidence for the success of one’s own, and Bakunin’s alternative, the immediate abolition of the state and its replacement by federation, generated predictions of its own, about defence, coordination, and abundance, that history left almost entirely untested for reasons the next section examines. Nor does the asymmetry establish that Marx’s analysis of the state was worthless: the theory of the state as bound to class relations, treated fully in the companion guide already linked, retains its explanatory force about ordinary states under capitalism whatever one concludes about transitional ones, and the Commune devices remain the most concrete attempt ever made to design a coercive apparatus that cannot consolidate. Nor, finally, does the asymmetry settle the question of what anyone now living should do, since both traditions are here being examined as historical theories with testable components, not as recruitment literature. What the asymmetry establishes is narrower and, for that reason, harder to evade. On the specific question the founders argued about in 1873, what a workers’ state does over time, the anarchist was right and the Marxist expectation was wrong or, at the most charitable grading, permanently unproven. A tradition’s maturity is shown by whether it can absorb a finding of that shape without either denying the evidence or surrendering everything else it knows. The Marxist tradition’s internal critics, the council communists and dissident currents treated below, are best read as attempts at exactly that absorption from within.

Why does the asymmetry exist at all? Part of the answer is temperamental, but the deeper part is structural, and it reflects the different logical characters of the two theories. Bakunin’s theory of the state is a theory about institutions as such: any separated coercive apparatus generates the predicted behaviour, so the prediction follows from the theory’s core and cannot be detached from it. Marx’s theory is a theory about history as a process: outcomes depend on the development of the productive forces, the maturity of classes, the international setting, so its expectations are inherently staged and conditional, and each staging supplies a fresh explanation when events run otherwise. Process theories are not thereby false; most true theories of society have that character, and a theory that predicted social outcomes unconditionally would for that reason alone be suspect. But the difference in character has a consequence for intellectual honesty. The holder of a conditional theory owes the world a statement of what would count against it, a specification of the conditions under which continued failure would be admitted as refutation rather than as further delay. That debt was rarely paid by the Marxist tradition’s custodians across the twentieth century, and the refusal to pay it, more than any single doctrine, is what turned a scientific disagreement into a theological one in the eyes of outside observers. Bakunin’s heirs owe a symmetrical debt, which the strongest objection will now collect.

The strongest objection: an untested alternative

Now the Marxist objection must be stated at the full strength the brief for this article demands, because the prediction asymmetry, left unqualified, flatters the anarchist tradition in a way the evidence does not license. The objection runs as follows. Bakunin’s prediction was correct about the transitional state, but the comparison the asymmetry invites, between a programme that failed in practice and a programme vindicated by that failure, is rigged by history. Anarchism has never sustained a large scale alternative for long enough to be tested. Its federated communes have never had to run a continental economy through a decade of scarcity, defend a long frontier against a modern army, or coordinate the rebuilding of a shattered industrial base, for years rather than months, with their institutions intact at the end. Every anarchist experiment at scale either collapsed quickly under external force, as the Ukrainian free territory associated with Nestor Makhno was ground between White and Red armies, or compromised its way into the state it opposed, as in Spain. The Marxist programme, for all its catastrophic deformations, at least demonstrated that its instrument could seize power, hold it, industrialise backward countries, and survive total war, which is why serious people kept choosing it. A test in which one contestant’s machine breaks down on the road and the other contestant’s machine has never left the garage does not crown the garage. That is the objection, and it is not a debater’s trick. It identifies a real asymmetry of evidence operating in the opposite direction from the prediction asymmetry, and any honest treatment must hold both asymmetries in view at once.

The answer begins with a concession, because the objection is largely sound. The asymmetry of testing cuts both ways. The failure of the transitional state to wither is evidence against the withering expectation, but the absence of a durable anarchist society is not evidence for the federated alternative; it is an absence, and absences confirm nothing. Bakunin’s predictions about his own construction, that federation could defend itself through the armed people, that coordination without a centre could provision cities and run railways and factories through a revolutionary war, remain in exactly the epistemic position the Marxist custodians claimed for the withering: untested under the conditions that would decide them, with the Spanish evidence pointing, where it points anywhere, in a discouraging direction, since the federation’s own leaders concluded within months that the war could not be run on federal principles. An anarchist who celebrates the prediction asymmetry while declining to notice the testing asymmetry is doing with Bakunin’s vindication what the orthodox did with Marx’s postponements, converting a research result into a loyalty badge. The second part of the answer, however, restores a measure of balance, because the two asymmetries are not of equal weight on the specific question this article examines. The dispute of 1872 was not, at its core, about whether anarchist federation could run a war economy, though that question matters enormously to any practical judgement. It was about what a workers’ state would do. On that question one side predicted and the century answered, and the answer does not stop being an answer because a different question, the viability of the alternative, remains open. A doctor can be right that a treatment will poison the patient without possessing a cure. The finding licenses a negative conclusion with confidence and a positive conclusion not at all, and intellectual honesty consists in keeping the proportions exactly there.

There is a third consideration, harder to state and impossible to omit, about the selection of tests themselves. The transitional state was tried so often partly because it worked at the task of taking and holding power, and movements in desperate circumstances rationally reached for the instrument that worked, whatever its deferred costs. The anarchist alternative was tried so seldom partly because its conditions of success are more demanding: it requires a population already organised in workplaces and communes dense enough to federate, a military situation survivable without central command, and neighbours who permit the experiment to proceed. History rarely supplies that combination, and where it nearly did, in Spain, the international setting was poisoned by fascist intervention on one side and by the calculated non intervention of the democracies and the hostile patronage of the Soviet Union on the other. A fair tribunal would note that neither tradition was tested under laboratory conditions and that the Marxist tests, run under siege conditions almost without exception, were run under precisely the conditions its own theory flagged as most favourable to the pathology Bakunin predicted. Emergency is not an alibi external to the theory. It is the normal weather of revolutions, and a theory of transition that only works in fair weather has been refuted for practical purposes even where its logic remains intact. Bakunin’s prediction, note, made no exception for weather. That is why it could be tested, and that is why it passed.

The overlap: libertarian Marxism and the traditions between

If the two traditions were the sealed and hostile systems their polemics describe, the story would end with the asymmetries. It does not, because the border between them has always leaked people and ideas in both directions, and some of the most interesting political thought of the century after the split was produced exactly on the border. Council communism, arising from the German and Dutch left after the First World War and associated with figures such as Anton Pannekoek, rejected the party state model from within Marxism, held that the working class could rule only through its own councils and never through a party ruling in its name, and arrived by that route at positions the anarchists recognised as neighbouring, while continuing to reject anarchism’s economics and its account of the state. Rosa Luxemburg, orthodox in her Marxism and merciless toward anarchism as a doctrine, nonetheless warned her own tradition, in writing, that freedom reserved for supporters of the government is no freedom, and that the elimination of democracy would leave public life to a bureaucracy, warnings that read now as the prediction asymmetry registering inside the Marxist conscience before the evidence was complete. The young Marx of the Paris manuscripts and the Marx of the Commune address have been quarried by generations of libertarian readers for a Marx against the Marxists, while Bakunin’s insistence on material organisation, his contempt for purely moral revolt, and his collectivism about production have been quarried from the other side. Autonomism in later twentieth century Italy and the syndicalist currents inside several labour movements worked the same seam. The article on the new left marxism explained traces how these convergences widened after 1956 and 1968, when disillusion with the existing workers’ states drove Marxist dissidents toward councilist and libertarian themes and drove anarchists toward the class analysis they had once left to their rivals.

Can a Marxist reject the transitional state and remain a Marxist?

The council communists answered yes in practice: they kept the analysis of capital, the centrality of the working class, and the goal of a classless society, while denying that a party state could be the road. The orthodox answered no, treating the transitional state as definitional. The border dispute itself shows the traditions are spectra, not fortresses.

The practical question buried in the overlap deserves direct treatment, since it is among the most common forms in which this comparison is actually asked: whether the two traditions can work together. The historical answer is that they have, repeatedly, and that the collaborations exhibit a pattern. Marxists and anarchists have cooperated effectively against common enemies and in common institutions: in unions where both organised the same workplaces, in anti fascist resistance across occupied Europe, in the streets of May 1968 in Paris, in countless local campaigns where the doctrinal frontier dissolved in the solvent of a shared immediate task. The pattern is that cooperation flourishes at the level of resistance and strains at the level of construction, because construction poses the transition question inescapably: what is to be built the morning after, and who commands it. Alliances have repeatedly broken on that morning, sometimes honourably, sometimes in blood, as in Russia after 1917 when anarchist organisations that had fought beside the Bolsheviks against the old order were raided and suppressed by the new one within a few years. A sober summary would run: the traditions are compatible as opposition and incompatible as government, unless the government question is settled in advance by a shared commitment to council and federal forms that neither orthodoxy, left to itself, reliably supplies. That sentence could serve as the whole modern history of the relationship in miniature.

The defence problem: the anarchist transition under fire

Every comparison of the two traditions eventually arrives at a single concrete scene, and it is fairer to stage it explicitly than to let it haunt the margins. A revolution has succeeded in a country. The old army has dissolved, the possessing classes are dispossessed on paper, and across the border, and usually inside it, armed forces are assembling to restore them. The factories must be fed with raw material, the cities with food, the front with shells and boots and medical supplies, and all of it must move on railways whose timetables were written by the old order’s ministries. Who decides, on a Tuesday, whether the steel goes to ploughs or to guns, and whose decision binds the factory that disagrees? The Marxist answer is the transitional state: a centre empowered to decide, staffed by the revolution’s organisations, legitimated by the councils or assemblies that sustain it, and pledged to dissolve when the emergency passes. The anarchist answer is federation under arms: the militias elect their delegates, the communes and unions federate their supply organs, decisions of war are taken by congresses of mandated representatives and executed by a coordinated staff that remains the servant, recallable and rotated, of the bodies that created it. Stated in the abstract, the anarchist answer is not absurd; federal armies have existed, and some have fought well. The question the record presses is what happens to the federal staff in the third year of the war, when continuity of command has become a military necessity, when the supply organs have acquired permanent officials because rotation lost the knowledge each rotation discarded, when the congresses meet rarely because the front consumes the delegates. The Spanish evidence suggests the federal forms bend toward the central ones under sustained pressure, and bend fast. The Marxist tradition reads that bending as the refutation of anarchism by reality itself. The anarchist tradition reads it as the proof of its own theorem in reverse: the moment the revolution builds a centre, the centre begins the career Bakunin described, and Spain’s tragedy was not that federation failed but that it was traded away.

Between those two readings sits a body of theory neither founder lived to write, the theory of emergency institutions, and a fair comparison should acknowledge how much it owes to hindsight the founders lacked. Both traditions were theorising before the age of total war, before economies were so interdependent that a decision about steel in one region starved furnaces in another within weeks, before the administrative state had demonstrated how much of modern life runs through permanent specialised machinery. Marx’s Commune model was drawn from a city under siege governing itself for two months in 1871, an inspiration and a warning in equal measure, since the Commune’s own story includes improvised centralisation, a Committee of Public Safety revived from the earlier revolution, and defeat. Bakunin’s federation was drawn from the communes and cantons of an earlier Europe and from the insurgent traditions of peasant countries, worlds in which the productive economy was local enough for federation to grip it. Neither man confronted the transition problem at the scale of a modern integrated economy under modern military pressure, and each side’s modern heirs have had to improvise theory their founders never supplied: Marxists wrestling with planning, markets, and the information problem, anarchists wrestling with coordination at scale and the tyranny, named by their own later writers, of structurelessness, the informal and unaccountable hierarchy that grows inside organisations that forbid formal hierarchy. That last concept, articulated within the feminist and anarchist movements of the later twentieth century, deserves its place in this comparison, because it is Bakunin’s theorem turned inward with admirable honesty: power does not vanish when it is denied a name and an office; it migrates to the charismatic, the experienced, and the well connected, where it cannot be recalled because it officially does not exist. A tradition capable of producing that critique of itself is not intellectually dead, whatever its untested status at the level of states.

What withering would have to look like

Because the withering expectation is so easily protected by vagueness, the comparison benefits from an unfashionable exercise: specifying, in advance and in concrete terms, what evidence would have counted in its favour, had the twentieth century supplied it. A transitional state on the road to withering would show identifiable symptoms across its life. The proportion of the population employed in administration and coercion would fall across decades rather than rise. The material privileges of officials, in pay, housing, access, and legal immunity, would converge toward the average of the administered rather than diverging from it. Tenure would rotate; the same names would not govern across a generation; officials would return to ordinary work and be replaced by others who had done ordinary work. Decisions of increasing weight would migrate downward and outward, to workplaces and localities, with the centre’s interventions becoming rarer and more easily reversed. The coercive organs would shrink in personnel, budget, and legal remit, and their contact with ordinary citizens would become less frequent and less feared. Above all, the apparatus would lose functions to society rather than accumulating them, surrendering the management of culture, of science, of internal movement, of opinion, until the residue was the thin administrative layer both traditions describe as the end point. None of this is utopian bookkeeping. Each symptom is measurable, and contemporaries did measure several of them informally, in the jokes citizens told about their rulers and in the emigration statistics of the skilled. Measured against such a schedule, the historical transitional states moved, with brief and contested exceptions in their earliest years, in the opposite direction on nearly every line, and moved furthest in the direction of accumulation precisely in the decades when, on the withering timetable, the disappearance of classes should have been emptying the state of purpose. The exercise matters because it removes the last shelter of the expectation, the claim that no one can say what its fulfilment would look like. One can say. It looks like the opposite of what happened.

Set beside that schedule, the Commune devices can be assessed as engineering rather than as scripture. Recall failed historically not because it was tried and defeated but because it was among the first devices surrendered: elections to the councils that were to exercise it became formalities within a few years in the Russian case, and the councils themselves were subordinated to the party apparatus whose officials no electorate could recall. Rotation failed where it met the genuine problem of expertise, and there the anarchist prediction about the savants returns with force: a complex economy requires continuity of knowledge, continuity of knowledge creates a stratum, and the stratum defends the scarcity that makes it valuable. Workmen’s wages for officials failed almost immediately everywhere, replaced first by supplements, then by the elaborate systems of closed distribution through which the apparatus rewarded itself while preserving the fiction of equality in the published salary scales. The lesson the comparison extracts is uncomfortable for a particular style of Marxist apologetic that treats the devices as a spell that was never properly cast. Devices are not spells. They are rules that require enforcement by people with the power to break them, and the only enforcement ever proposed was the vigilance of the class in whose name the state ruled, exercised through the very councils the apparatus learned to manage. Bakunin’s deepest point was never that revolutionaries are wicked. It was that a design whose safeguards depend on the permanent virtue and attention of the governed, against a permanent and organised interest of the governors, has the asymmetry of effort built into it. The governors attend to their position every day. The governed have factories to run and children to feed. On the evidence, that mundane asymmetry of attention did more work than any ideology.

Misreadings that keep the quarrel stupid

Four misreadings account for most of the confusion in popular treatments of this comparison, and clearing them is a service to both traditions, since each misreading is somebody’s caricature of the other side. The first holds that anarchists want chaos, disorder, or the absence of organisation. The historical anarchist movement was, in fact, organisationally prolific to a fault: unions, federations, schools, publishing houses, cooperatives, militias, all built by people whose objection was never to organisation but to organisation that hardens into a separate power over its members. Anarchism’s canonical texts argue about the design of institutions at obsessive length. A doctrine whose central practical problem is how to coordinate without commanding is many things, but a wish for chaos is not among them, and the dictionary sense of anarchy as disorder describes the condition anarchists are accused of causing by opponents who find the accusation cheaper than the argument. The second misreading holds that Marxists want a permanent, all embracing state, beloved for its own sake. The founding texts say the opposite with monotonous consistency: the state is an evil inherited by the revolution, to be cut down to its smallest possible compass and destined for extinction, and the Marxist politicians who built permanent states were, at the level of doctrine, heretics against their own canon even when they quoted it most fluently. A reader who cannot distinguish a tradition’s texts from its custodians will misread every tradition in this series, and this one most of all.

The third misreading reduces the split to personality: Marx the authoritarian schemer, Bakunin the romantic conspirator, two egos colliding in a small organisation. The personalities were real and the collision was genuinely ugly, marked by ethnic slurs in private letters, by manoeuvres neither side’s admirers enjoy reading, and by the shabby episode of the Hague trial. But the proof that personality explains little is that the same split reproduced itself, along the same theoretical seam, in movement after movement whose participants had never heard the founders’ insults: syndicalists against parties in France, councilists against Bolsheviks in Germany, libertarian against authoritarian currents in every revolutionary crisis of the century that followed. Wherever revolutionaries have had to decide whether to build a state or to refuse one, the argument of 1872 has reassembled itself from the situation, wearing local clothes. Structural disagreements behave that way. Personal ones die with the persons. The fourth misreading concerns Engels’s essay and has been corrected in its own section above: On Authority is an argument about coordination in production, conscripted afterwards into service as a general defence of state power, a service its text does not perform. Readers who meet the essay only through quotation in polemics inherit the conscription rather than the argument, and the same warning applies to Bakunin’s prediction, which circulates in anarchist literature as a trophy, a paragraph quoted without its surrounding analysis of science, representation, and the German state tradition that gives the forecast its mechanism. Both traditions are ill served by their own quotation habits. The texts are short. The Hague documents, the essay on authority, the conspectus, and the relevant books of Statism and Anarchy can be read in a week, and this article’s comparison is best used as a map for that reading rather than as a substitute for it.

The documentary record: how the split was written down

The afterlife of the dispute was shaped by who kept the papers, and a section on the evidence itself is owed to any reader who wants to know why the standard accounts lean the way they do. The Marxist tradition became, early and thoroughly, a tradition of collected works: Marx’s and Engels’s writings edited, translated, and canonised, the conspectus on Bakunin eventually published within that canon, framed by editors who knew which side they were on. The anarchist record is scattered by comparison: Bakunin died in 1876 leaving manuscripts in several languages and states of completion, his works assembled piecemeal by devoted editors across decades, much of Statism and Anarchy circulating in partial translations that flattened its argument into a prophecy with footnotes. Guillaume, the expelled ally of the Hague, became the federalists’ historian, and his multi volume history of the International remains indispensable and partisan in equal measure, a combination it shares with the Marxist tradition’s official histories. The result is an archive with a slope. The Marxist side of the quarrel is documented at the level of the marginal note; the anarchist side must often be reconstructed from correspondence, congress reports, and the publications of movements that were illegal in most of the countries where they operated. Historians of the split have spent a century correcting for the slope, and their findings have generally moved in one direction: the closer the documentary work gets to the congresses and the texts, the less the split looks like a morality play with a villain, and the more it looks like two coherent theories of power colliding inside an organisation too small to hold them. That convergence of serious scholarship, across scholars of every sympathy, is itself a finding worth reporting in an article committed to grading claims by evidence.

One document deserves a paragraph of its own because it condenses the whole problem of the archive into a single artefact: the report on the Alliance presented at the Hague, the evidentiary basis for Bakunin’s expulsion, compiled by a commission whose members were aligned with the prosecuting side and drawing in part on materials supplied by interested parties. Later scholarship has picked the report apart, finding genuine material about Bakunin’s network building interwoven with inference and exaggeration, and the anarchist tradition has its own counter archive alleging forged mandates and procedural coups by the Marxist majority. A reader need not adjudicate every allegation to extract the durable point. The split was executed by methods that illustrated both theories of organisational pathology simultaneously, the majority behaving as Bakunin said majorities with machinery behave, the minority having built, as Marx charged, structures its own doctrine could not acknowledge. When the founders’ organisations are the first exhibits in the trial of the founders’ theories, and both exhibits cut against their owners, the sensible conclusion is not that both theories are worthless. It is that the phenomenon both theories point at, the tendency of power to consolidate in the hands of those who administer it, is stronger than either movement’s immunity to it, which is precisely the proposition the twentieth century would test at a scale neither man could have imagined and at a cost neither would have wished on his worst enemy’s country.

Later schools inherited the dispute and sorted themselves along its seam with instructive regularity. The Marxism that governed states read the split as settled: anarchism a petty bourgeois deviation, refuted by its own marginality, its literature untranslated and its name a term of abuse in the official lexicon. The Marxism that opposed those states from the left read the split as unfinished business, recovering the Commune writings, the council tradition, and Luxemburg’s warnings, and treating Bakunin, when it dared, as an unwelcome relative whose question had never been answered. Anarchism, for its part, divided between readings of Bakunin as prophet vindicated and readings that quietly rebuilt his federalism with modern materials, syndicalist, platformist, and communalist variants arguing among themselves about how much coordination a free society needs, which is the On Authority argument continuing inside the anarchist house. Even the academic study of organisations, innocent of the old banners, reproduced the seam in its own vocabulary, in the iron law of oligarchy formulated by Robert Michels from his study of the German socialist party itself: organisation, Michels concluded, implies the tendency to oligarchy, whoever the organisers and whatever their democratic creed. Michels had read the same Europe the founders had fought over and reached, by sociological road, a generalised form of Bakunin’s theorem, this time aimed at parties rather than states. That a theorem about the Marxist movement’s own organisations, stated by a sociologist, should converge with the anarchist’s prediction about the Marxist state suggests the prediction was tracking something real about large organisations under pressure, something neither tradition owns and both must answer.

The argument before the test: the decades of separation

Between the deaths of the founders and the revolutions that tested their theories, the two traditions lived separate lives, and the separation did intellectual damage to both that the comparison should record. Anarchism, expelled from the mainstream of organised socialism, developed for some decades at the movement’s edge, and its development was marked by a tragic detour into individual acts of violence, the assassinations and bombings of the 1890s and the years before the First World War, carried out by a minority whose deeds fixed the word anarchist in the public mind as a synonym for the bomb. The movement’s serious theorists and its syndicalist organisers repudiated or outgrew the detour, but the damage to serious comparative discussion was durable: a tradition could be dismissed by pointing at its worst publicists, and Marxist polemic availed itself of the shortcut for a century. Meanwhile the same decades showed anarchism’s constructive strength in the one arena where its organisational theory fitted the terrain: the revolutionary syndicalism of France, Spain, and Latin America, where anarchists built unions on federal principles that organised hundreds of thousands of workers without a party holding their reins, and demonstrated, at the scale of industries and cities, that coordination without a political centre was not a fantasy. The unions’ weakness, visible in the same decades, was the obverse of their strength: without a political instrument, their victories remained industrial, reversible by any government willing to use troops, and the question of the state returned through the factory gate every time a general strike met an army.

Marxism in the same decades built the opposite profile. In Germany its party became the largest in the country, a state within the state, with newspapers, unions, cooperatives, and a parliamentary fraction of growing weight, and the very completeness of that apparatus revived the old argument inside the Marxist house: whether the machine built to conquer power was itself becoming an interest, its officials a stratum with salaries and pensions and a horror of risk. Michels formulated his iron law from observation of exactly that party, and the revisionist controversy, treated elsewhere in this series, asked in electoral language the question Bakunin had asked in institutional language. When 1914 came and the great parties of the International voted, in their several parliaments, for the war credits of their own governments, the anarchist tradition read the collapse as the theorem’s confirmation in yet another register: organisations integrated into the state, even in opposition, even with revolutionary programmes on paper, behaved at the crisis as the state required. The Marxist tradition’s honest historians have never fully answered that reading, and the revolutionary Marxists who rebuilt their movement after 1917 did so in explicit reaction against it, determined that their organisations would never be absorbed. The determination produced parties of a new hardness, and the hardness produced, in power, the most complete examples of the stratum Bakunin had predicted. The wheel of the argument turned full circle within fifty years of the Hague, each tradition generating, from its own attempt to escape the other’s critique, fresh evidence for it. That recursiveness is the deepest reason the comparison remains alive. It is not a museum dispute between dead men. It is the record of a problem that revolutionary politics has repeatedly failed to engineer its way out of, stated first by the two men who saw it earliest and clearest from opposite sides.

The government of savants: the second branch of the prediction

Bakunin’s fear of the credentialed deserves a section of its own, because it is the part of his forecast that has travelled furthest beyond the original dispute and because it complicates any simple reading of him as a tribune of spontaneous revolt. His writings return, with unusual insistence for a nineteenth century revolutionary, to the danger of knowledge converted into office. A government of scientists and engineers, ruling society as an object of expertise, struck him as potentially the most oppressive of all governments, because it would experience its own commands as technical necessities and its subjects’ resistance as ignorance, and because the ruled, lacking the credentials, would be disarmed in argument before they were disarmed in fact. He extended the same suspicion to the intellectuals of his own movement, warning that socialist men of science who expected to tutor the revolution would end by ruling it. The point is frequently misread as hostility to science, and Bakunin’s texts contain the correction in his own hand: he bowed, he wrote, before the authority of knowledge in its proper sphere, and consulted specialists as a man consults a bootmaker about boots; what he refused was the translation of that authority across spheres, from competence in a subject to command over a society. The distinction between the authority of the bootmaker and the rule of the bootmakers’ guild, generalised, is the whole anarchist theory of expertise in one image, and it maps with embarrassing precision onto the later history of planned economies, in which the planning apparatus, armed with the claim that its directives were science rather than politics, achieved an immunity from ordinary political challenge that the cruder dictatorships of property never enjoyed.

The Marxist tradition was not blind to the problem of expertise; it was, however, structurally slower to theorise it, for a reason internal to its own achievement. A movement that had built its identity on possessing a science of society, historical materialism as the discovered law of social development, could not easily treat the custodians of that science as a candidate ruling stratum without turning its suspicion on itself. It did produce the instruments for doing so. The analysis of bureaucracy in the dissident Marxist literature, the theory of the new class formulated most famously by Milovan Djilas from inside the Yugoslav leadership, and the various accounts of a managerial or technocratic stratum in both capitalist and post capitalist societies, all wrestle with Bakunin’s second branch using Marxist tools. Djilas’s case is emblematic and uncomfortable: a vice president of a workers’ state concluding, from the inside, that a new class had formed on the basis of administrative control rather than ownership, and being imprisoned by that state for the conclusion. The episode belongs in this article because it closes a circuit. The prediction of 1873 was framed against Marx’s programme by an anarchist. Its most famous twentieth century confirmation from within a ruling party was framed by a Marxist heretic against his own government, using the vocabulary of class he had learned from the tradition Bakunin opposed. Ideas that can migrate across the frontier of a schism and do work in the enemy’s camp are usually tracking something real. The second branch of the prediction, the government of savants, may in the end prove the more durable of the two, since the fusion of expertise with office has continued to develop in societies of every label, long after the specific regimes of the revolutionary century disappeared.

Is anarchism simply Marxism without the state?

No, though the formula captures the loudest disagreement. The traditions also differ on the revolutionary agent, on organisation, on the status of expertise, and on whether the new society is built by federation from the first day or reached through a transition. Subtracting the state from Marxism leaves those differences standing, and several have independent histories.

The difference restated at three levels

Comparisons of this dispute go wrong most often by conducting the whole argument at one level when the dispute lives at three, and separating the levels is a useful discipline to carry away from the table and the history. At the level of ends, there is no dispute worth the name: classless, stateless, self governing association, described in different vocabularies by both traditions and recognisably the same country. At the level of diagnosis, there is a real but narrower dispute than the polemics suggest. Both sides analyse the existing state as an instrument bound to class power. They differ on the depth of the binding. The Marxist analysis treats the state as derivative, its behaviour explained by the class structure it serves, so that a change in which class holds it changes its nature. The anarchist analysis treats the state as additionally autonomous, a form with its own appetite, so that changing the holders changes the personnel of domination more reliably than its fact. Much of the empirical argument of the last century can be organised as evidence about that single difference: how far does the character of a state follow the class that staffs it, and how far does it follow the form itself? Readers who keep the question in that shape will find evidence on both sides in unexpected places, including the behaviour of states in capitalist societies whose forms outlast every change of governing class. At the level of strategy, finally, the dispute is total and probably irreconcilable, because the strategies are logical contraries: seize and transform the state, or abolish it in the act of revolution; build the instrument of transition in advance, or refuse all instruments that resemble the destination’s opposite. A person can split the difference in diagnosis. Nobody can split it in strategy on the morning of an insurrection, which is why the argument, so often pronounced dead by academics, reassembles itself with undiminished force wherever insurrections actually occur.

Restating the levels also clarifies what each tradition’s strongest modern defenders actually claim, as opposed to what their founders’ custodians claimed for them. The strongest contemporary forms of the Marxist position tend to concede the diagnosis halfway: the transitional states of the past are acknowledged as catastrophes of consolidation, and the theoretical work goes into specifying the institutional conditions, plural and competing centres of working class organisation, radical publicity, rotation with teeth, independent unions and press even against a revolutionary government, under which a transition might avoid the predicted fate. That is, in substance, the Commune programme rebuilt with the twentieth century as its tutor, and its proponents are entitled to note that no such specified attempt has yet been made, so the experiment, in its careful form, remains unrun. The strongest contemporary forms of the anarchist position make a mirror concession: federation is acknowledged to face unsolved problems of scale, defence, and coordination, and the theoretical work goes into the design of federal institutions with real capacity, confederal planning, mandated administration, defensive organisation, intended to answer Engels’s factory and the general staff’s map. Both research programmes are, in effect, attempts to repair the exact failure the record exposed in their own tradition, using the other tradition’s critique as the specification. Viewed that way, the dispute’s modern form is less a war than a division of critical labour, each side serving as the other’s most informed building inspector. The inspection reports, compiled above, stand as this article’s findings: the Marxist building failed its inspection on consolidation, repeatedly and at terrible cost; the anarchist building has yet to be erected at a scale where inspection bites, and its drawings contain known unresolved stresses at exactly the joints the Marxist inspectors identified in 1872.

A method for judging disputes of this kind

Something general can be salvaged from the wreckage, and an article in a series devoted to method as much as to doctrine should salvage it explicitly. Theoretical disputes in politics are usually graded, when they are graded at all, by eloquence, by loyalty, or by the later respectability of the disputants. The Marxism and anarchism case suggests a sterner set of questions, applicable far beyond it. Ask first what each side predicted, in writing, before the events, and how specific the prediction was: a dated, unconditional, mechanism bearing forecast is worth more than a shelf of retrospective wisdom, and movements should be held to their founders’ published expectations rather than to their publicists’ later reconstructions. Ask second what each side’s theory permits it to admit: a theory with no imaginable falsifier is not thereby false, but its holders owe a specification of what evidence would change their minds, and the specification, or its absence, is itself evidence about the theory’s health. Ask third whether the dispute’s terms survived contact with events: the withering of the state did not survive as a description of anything that happened, while the consolidation of a governing stratum survived as a description of a great deal, and survival of terms is a rough but real index of contact with the world. Ask finally what each side does with its own failures: the traditions in this article each produced, from inside, critics who named the failure in the tradition’s own language, Luxemburg and the councilists on one side, the Spanish post mortems and the theorists of structurelessness on the other, and a tradition’s capacity to generate such critics may matter more, across a century, than any single doctrine it defends. Graded by those questions, this dispute yields a verdict with edges rather than a trophy: a clear anarchist advantage on the fate of transitional states, a clear Marxist advantage on the demonstrated capacity to act at scale in history, no verdict available on the anarchist alternative, and a shared debit, on both sides, for the methods used at the Hague and the habits of caricature learned there.

What test would settle the remaining question between the traditions?

A revolutionary transition conducted under the specified Commune conditions, or a federated society surviving at national scale across a generation, would each test what the record leaves open. Until some society runs either experiment under observation, the honest verdict keeps the asymmetry: one prediction confirmed, one expectation unfulfilled, one alternative untested.

That closing balance sheet is the point at which comparison must stop, and it is worth a sentence about why stopping there is a strength rather than an evasion. Both traditions contain resources this article has not needed and judgements it has not attempted: philosophies of history, theories of value, accounts of alienation and of freedom, on which the state dispute sheds only oblique light. To convert a graded finding on one question, the fate of transitional states, into a total ranking of two civilisations of thought would repeat, at the level of scholarship, the error the Hague committed at the level of organisation: settling by scope what can only be settled point by point. The reader who wants the wider rankings will find the materials across this series, in the studies of the state theory, of the Commune, of the Soviet record, and of the traditions that grew from the split. This article’s narrower claim has the advantage of being checkable. In 1873 one side wrote down what the other’s road would build. The road was built, many times, and what it built resembled the description. Everything else is argument. That sentence, with its uncomfortable distribution of credit, is as close as honest comparison gets to a verdict here.

The Paris Commune: the emblem both traditions claim

No event sits closer to the centre of this comparison than the Paris Commune of 1871, and none illustrates its difficulties better, because the Commune is the one historical case both traditions cite as the portrait of their own theory in action. For seventy two days, between the proclamation of March 1871 and the slaughter of the final week in May, Paris was governed by an elected council after the national government withdrew to Versailles and the city refused to surrender its arms. The measures the Commune enacted in that interval became scripture for the Marxist account of transition: delegates elected by universal suffrage and recallable at any time, officials paid at workmen’s rates, the standing army suppressed in favour of the armed population organised as a National Guard, the police stripped of political character, church and state separated, abandoned workshops handed to the workers who had staffed them. Marx, watching from London, wrote The Civil War in France as the events unfolded and immediately after, and read the Commune as the political form at last discovered under which to work out the economic emancipation of labour. His tradition has read it that way ever since: not as a blueprint copied from theory but as a form invented by practice, which theory then recognised and generalised. The transitional state, on this reading, is not a speculation. It existed, briefly, in Paris, and its features are the checklist against which later claims to the Commune form must be measured.

Bakunin read the same seventy two days and found a different portrait. The Commune, for him and for the federalists who took heart from it, was a revolt of a city against the centralised French state, an assertion that Paris would govern itself and that France should reconstitute itself as a federation of such communes, each autonomous, linked by free pact rather than commanded from a capital. He noted, correctly, that the Commune’s own instincts ran against its governmental inheritance: it hesitated before the Bank of France, respected forms of property its programme should have dissolved, and reproduced, in its Committee of Public Safety, the Jacobin machinery of the earlier revolution, a machinery Bakunin regarded as the old world reasserting itself inside the new. The anarchist reading treats the Commune’s governmental residues as the cause of its fatal hesitations and its federal impulse as its glory. The Marxist reading treats the same residues as proof that even the freest revolutionary government remains a government, exercising coercion amid war, and that the honest course is to name it a state of a new type rather than to pretend otherwise. Both readings can quote the record, because the Commune, improvised under siege by men of several doctrines, Blanquists, Jacobins, Proudhonists, members of the International, genuinely contained both. It was a council of recallable delegates and a wartime government issuing decrees. It was a federation in aspiration and a besieged city in fact. An event that contains both traditions in embryo cannot referee between them, and the scholarly use of the Commune in this dispute should be limited accordingly: it proves that the transitional form Marx described is not a fantasy, since something recognisably like it was assembled by workers in a crisis, and it proves that the form’s libertarian features coexisted, from the first week, with emergency centralisation of exactly the kind Bakunin predicted centralisation follows from. What seventy two days cannot prove is the trajectory across years, and trajectory, not birth, is what the whole argument is about.

The Commune’s suppression supplies a final, grim datum that both sides have used and neither should use cheaply. The Versailles government retook Paris street by street, and the killings of the final week, together with the deportations and executions that followed, ran into the tens of thousands by most serious estimates, making the Commune’s fall one of the bloodiest episodes in the history of European labour. Marxists have cited the massacre as the standing refutation of any politics that disarms before its enemies: this, they argue, is what counter revolution does when it wins, and the transitional state’s severity must be judged against the severity it exists to prevent. Anarchists have cited the same massacre as proof that the old state’s violence needs no provocation beyond the existence of an alternative, and that the Commune fell not because it lacked a state of its own but because it hesitated to destroy the enemy’s power while destroying its own scruples, leaving the Bank untouched and Versailles unharrassed. The two morals cannot both be drawn in full, and the choice between them is the transition dispute in its most concrete historical dress. What the evidence does not permit is the comfortable conclusion, heard in both camps’ folklore, that the Commune’s fate settles the argument for one’s own side. It is the argument’s founding exhibit precisely because it settles nothing except the stakes.

Federation as a positive doctrine

Because the anarchist side of this comparison is so often represented by its negation, the refusal of the state, its positive institutional design deserves a fuller statement than it usually receives in Marxist summaries, and the brief for this article requires both traditions at full strength. The design, assembled across Bakunin, the syndicalists, and later federalist writers, has load bearing parts. The base units are the workplace and the commune, self governing assemblies of the people who work and live in them, deciding their internal affairs directly. Beyond the base, coordination is achieved by federation: congresses and councils composed of delegates carrying imperative mandates, instructions from their assemblies that bind them, subject to recall at any moment and rotating before habit can harden into office. Federal bodies administer; they do not govern, in the doctrine’s careful distinction, because they possess no power separate from the mandates that compose them, no permanent bureaucracy, no coercive apparatus of their own, defence and enforcement resting with the armed population and the federated militias. Economic coordination across regions is handled by federated chambers of labour and statistical bureaux that inform decisions without taking them, the plan emerging from the congresses of producers rather than descending from a ministry. Later anarchist currents argued, sometimes bitterly, about whether this design contains enough centre. The organisational platform proposed by exiled Russian and Ukrainian anarchists in the 1920s, associated with Makhno and Piotr Arshinov, demanded theoretical unity, tactical unity, and collective responsibility within the anarchist movement itself, a proposal its critics inside the tradition denounced as the embryo of an anarchist party on the road to becoming what it beheld. Errico Malatesta, the Italian anarchist whose career spanned the whole period from the First International’s aftermath to the rise of fascism, spent decades refining the middle ground: organisation as dense as reality requires, authority as thin as honesty permits, and perpetual suspicion of the movement’s own functionaries.

Stated at that fullness, the federal design is a serious answer to Engels, and the Marxist critique of it must be equally serious to count. The critique has three prongs, each of which the record illustrates. The first is speed: imperative mandates and recallable delegates deliberate at the pace of assemblies, and wars, famines, and epidemics do not. Federal systems under pressure historically grow standing committees, and standing committees grow staffs, and the design’s safeguards are then tested at exactly the point of maximum stress, as the Spanish sections of this article record. The second prong is the boundary problem: a federation of communes must decide who speaks for the whole in dealings with hostile states, who apportions scarce resources between a rich region and a poor one when both mandates conflict, and who compels a commune that refuses its share of a common burden, and every answer to those questions is a small state in the act of being born. The third prong is the one Bakunin himself would have recognised as his own theorem returning by another door: administration without the name of government still concentrates knowledge, contacts, and continuity in the people who staff the federal bureaux, and the savants of the federation, however mandated, acquire the quiet power of those who draft the options among which assemblies choose. The anarchist tradition’s best writers did not dodge these prongs. They answered that every safeguard is a practice rather than a mechanism, that rotation and recall work only where the base assemblies are alive, and that the design’s admitted fragility is the price of refusing the one failure mode they regarded as fatal, a centre that can no longer be dissolved by the people beneath it. A federation can fail by dissolving into impotence. A state, on their account, fails by succeeding at self preservation. Between a design that risks collapse and a design that risks consolidation, the traditions chose opposite risks, and the twentieth century, which ran the consolidation risk to its end many times and the collapse risk hardly once at scale, cannot be said to have graded both papers.

The Marxist replies across the tradition

The anarchist prediction did not go unanswered inside Marxism after the founders, and the sequence of replies is itself instructive, because the replies change character as the evidence accumulates. The first great restatement is Lenin’s The State and Revolution, written in hiding in the summer of 1917, weeks before its author’s party took power. As a reading of Marx and Engels it is the most libertarian document ever produced by a successful revolutionary leader: the state machine is to be smashed, not captured; officials are to be elective, recallable, and paid workmen’s wages so that, in Lenin’s phrase, every cook may learn to govern; the standing army and bureaucracy are to be abolished outright; the withering of the state begins, he insists, immediately upon the revolution. Bakunin is not named as the opponent on most pages, but the book reads throughout as the conspectus continued by other means, an attempt to prove that the Marxist transition already contains every safeguard the anarchists demand, minus only the refusal of organised coercion against the dispossessed classes. The historical irony is too well known to need decoration: within a few years the state built under that book’s authority had restored a standing army with former imperial officers, a political police of growing reach, one party monopoly, and officials whose privileges were becoming hereditary in fact. Defenders reply, with evidence, that civil war, invasion, famine, and the death or dispersal of the very working class in whose name the safeguards were designed destroyed the conditions of the book’s programme within months. Critics reply that the sequence is the prediction, that emergency is the normal climate of revolutions rather than an accident befalling this one, and that a programme whose safeguards dissolve under its expected operating conditions has been answered by events in Bakunin’s favour. The argument between those replies is the twentieth century’s central argument about itself, and this article does not pretend to close it. It reports the shape: the strongest Marxist text conceded the anarchist design’s demands in theory, and the strongest anarchist text predicted the practice that followed in fact.

The later replies are more interesting still, because they come from men administering the predicted stratum and, in the honest cases, knowing it. Trotsky, who had organised the Red Army and defended revolutionary terror as a necessity of civil war, spent his exile analysing the Soviet bureaucracy as a parasitic caste resting on the nationalised economy, a phenomenon he insisted was a deformation rather than a new class, since the bureaucracy owned nothing and could be removed by a political revolution that left the social foundations intact. The distinction between caste and class in that analysis carries the whole weight of the conspectus against Bakunin: a stratum that cannot bequeathe its position as property, the argument runs, has not consolidated as a ruling class, and the withering, though monstrously delayed, remains the trajectory of a system whose property forms point the other way. Djilas, arriving at the opposite conclusion from inside another ruling party, replied that control exercised collectively and exclusively by a party apparatus is ownership in every sense that matters to the people administered. Between Trotsky’s caste and Djilas’s class lies the exact seam where Marx’s reply and Bakunin’s prediction meet, and the fact that the seam was located by practitioners rather than by professors suggests the dispute had by then become a problem in the management of real states rather than a quarrel of pamphlets. The anarchist tradition, watching its prediction debated in its rivals’ vocabulary, can be forgiven a certain grim satisfaction. Satisfaction, however, is not a programme, and the same decades that vindicated the warning left the warner’s alternative exactly where 1872 had left it: designed, argued, beloved, and unbuilt at the scale where its own predictions would finally have to answer for themselves.

Two theories of power underneath the dispute

Underneath the argument about institutions runs an argument about what power is, and surfacing it explains several features of the dispute that the institutional level leaves mysterious, including why the two sides so often talk past each other while using the same words. In the Marxist analysis, power is fundamentally relational and derivative: it flows from position within the relations of production, so that the capitalist’s power over the worker, and the state’s power over both, are expressions of an underlying economic structure that assigns control of the means of life to one class and dependence to another. Change the structure and the power changes its nature, which is why a state can, in principle, be an instrument of emancipation in the hands of the class whose interest is the structure’s abolition. Domination, on this account, has a root, and the root is class. In the anarchist analysis, power is additionally, and irreducibly, a phenomenon of hierarchy as such: the relation of command and obedience, wherever it appears, in the state, in the church, in the army, in the school, in the party, tends to reproduce and extend itself and to form the people inside it into rulers and ruled with opposed interests. Class is one species of that genus, the most important species in the economic sphere, but not the whole of it, which is why abolishing class while preserving the hierarchical form leaves, on the anarchist account, the genus alive to grow a new species. Bakunin’s writings pair the state and the church as twin institutions of the same principle, authority consecrated and obeyed, and his polemics reserve a special contempt for the idea that hierarchy becomes harmless when its personnel are revolutionaries or its doctrines scientific. Each theory sees a danger the other is built to miss. A movement armed only with the class theory will be surprised, repeatedly, by the behaviour of its own hierarchies. A movement armed only with the hierarchy theory will be surprised by the resilience of economic power, which survives the dissolution of governments and reconstitutes command through property, credit, and hunger long after the palaces have been emptied.

The difference in theories of power also explains the two traditions’ divergent attitudes toward knowledge, already touched in the section on the savants, and toward gradual construction in the present. If power is derivative from economic structure, then building alternative institutions inside the old society, cooperatives, unions, communes, is preparatory work of real but limited significance, since the structure will bend every enclave to its logic until the structure itself is confronted at the level of the state. If power is hierarchy as such, then the enclaves are the revolution in its only honest form, because a society’s character is nothing but the sum of the relations practised inside it, and means that rehearse domination cannot rehearse freedom out of existence. Syndicalism and the later anarchist emphasis on prefiguration descend from the second premise; the Marxist insistence on the priority of political power descends from the first. Modern readers will notice that the two theories of power are not, strictly, contradictories: an economic structure could generate hierarchies which then acquire autonomy and outlive their generator, which is close to a summary of what the twentieth century appeared to show in the transitional states, where a stratum born of revolution administered an economy it did not own and obeyed incentives no founder had chosen for it. The synthesis, if it is one, was available in neither founder’s lifetime vocabulary, though Bakunin’s mechanism describes the autonomy and Marx’s analysis describes the birth. Serious comparison should say plainly that the dispute’s deepest level has moved toward the anarchist premise on the autonomy of power while retaining the Marxist premise on its origins, and that both traditions’ living theorists, whatever their banners, tend in practice to work with some version of that compound view.

Scale, selection, and the honest use of the record

A comparison that rests on the historical record owes its readers a short audit of how that record can mislead, because the record in this dispute is not a neutral sample and both traditions have mined it selectively. Begin with the crudest selection effect. The transitional state was attempted in many countries across most of a century, under conditions ranging from industrial Germany’s near miss to agrarian Russia and China, and the sample is large enough that the regularity of the outcome, consolidation without withering, carries genuine evidential weight: a prediction confirmed across varied conditions by varied personnel is not easily explained away as a run of bad leadership or bad luck. The anarchist alternative, by contrast, has a sample that can be listed in a breath: the Commune’s federal impulse inside a single besieged city, the Ukrainian free territory amid a multi sided civil war, the Spanish collectives and militias inside a republic at war, and a scatter of smaller experiments in communes and workplaces that never faced the problems of scale at all. Comparing a large sample with a tiny and uniformly besieged one is legitimate only if the verdict’s confidence is adjusted to the samples, which is what this article’s asymmetry language is designed to do: confirmation where the sample is large, suspension where it is small. Polemic in both traditions characteristically skips the adjustment, anarchists grading Marxism by its full record and anarchism by its intentions, Marxists grading anarchism by its shortest experiments and Marxism by its unattempted ideal form. The symmetry of the cheating is worth more attention than either cheat, because it reveals the gravitational pull that loyalty exerts on evidence, the very pull a comparison article exists to resist.

Duration deserves its own caution. The anarchist experiments were short, and brevity cuts in both directions at once: their institutions never faced the slow tests, succession, generational turnover, the conversion of emergency arrangements into customs, that destroyed the withering expectation elsewhere, so their purity is partly the purity of the untested, while their suppression by enemies stronger than themselves proves little about their internal viability, since the transitional states were also born under fire and survived it. What brevity does settle is the military question in its narrowest form: no federation of the anarchist type has yet defended a revolution against a modern state determined to crush it, across years, without either centralising toward the state form or perishing. That is a finding about the world as it has been, not about the world as it must be, and honest statement requires the qualifier, since military technology, international conditions, and the social density of federal organisation have all changed since the experiments were run. Symmetrically, the long durations of the transitional states settle the consolidation question in its narrowest form: across decades, in every observed case, the apparatus grew rather than withered, and the qualifier honesty requires here runs the other way, since it remains possible that the specified Commune form, never in fact maintained, would have behaved differently. Possible is doing heavy work in that sentence, and the article leaves it there deliberately. A finding and a hope are both entitled to their names, provided they are not exchanged for each other in the dark. The founders, at their best, observed that discipline: Bakunin published his prediction where events could grade it, and Marx published his Commune analysis where later movements could check their states against it. Their followers have graded and checked with varying honesty ever since, and the reader who finishes this comparison equipped to do both, to every tradition including the reader’s own, has taken from it the skill its subject most rewards.

The safeguards debated in earnest

Suppose the dispute is treated, for one section, as an engineering problem bequeathed to the living rather than as a inheritance to be defended. The record then reads as a specification of failure modes, and the interesting question becomes what institutional safeguards each failure mode demands, whoever’s banner flies over them. Consolidation through the separation of representatives from represented demands recall that functions under emergency conditions, not recall that exists in constitutions while elections are managed, and the record suggests that recall functions only where independent organisations outside the state, unions, associations, a press, can publicise and organise a recall campaign without asking the recalled official’s apparatus for permission. Consolidation through material privilege demands transparency with teeth: published incomes and perquisites for officials, audited by bodies the officials do not appoint, since every historical apparatus preserved formal salary equality while building the real distribution system in closed channels, and the closed channel, not the salary scale, is where the stratum actually lived. Consolidation through the monopoly of knowledge demands rotation joined to the systematic diffusion of administrative skill, the unglamorous work of training many people to do each governing job, because a function performed by the only people who understand it is a function owned by them whatever the constitution says. Consolidation through emergency demands sunset mechanisms with external enforcement, powers that lapse by default and must be affirmatively renewed by assemblies the emergency government does not control, since every historical case shows emergency powers renewed by the body that wielded them, on the body’s own certification that the emergency continued. And consolidation through the fusion of party, state, and economy in a single hierarchy demands, above all, plurality of centres: more than one organisation entitled to speak for the producers, so that the governed possess somewhere to stand from which the governors can be seen.

The value of the exercise is that it translates both founders into a common specification language, and the translation is not forced. Marx’s Commune devices are safeguards of exactly this kind, aimed at separation, privilege, and the standing coercive body. Bakunin’s federalism is the same safeguards radicalised, plurality and recall built into the foundations rather than bolted onto a state. Read as engineers, the two men differ on one design decision: whether a centre strong enough to coerce the old classes can be fitted with safeguards strong enough to dissolve it afterwards, or whether the fitting must be done by refusing the centre and distributing its functions beyond recall by any single hand. Read as engineers, they also agree on the test protocol: watch what the institution does to the people inside it across time, and judge by trajectories rather than by founding speeches. That they would both have despised the reduction of their dispute to design specifications is beside the point, and in a way it is the point. The century that tested their theories was not interested in their intentions. It graded the designs. A comparison that ends by stating the shared specification, the points of divergence, and the observed trajectories has done everything comparison can honestly do, and has earned the right to stop before prophecy begins again in modern dress.

The anarchist self warning

Honesty about the anarchist tradition requires one more exhibit, which its own historians supply and its folklore tends to skip: the tradition’s recurring discovery that it, too, generates the pathologies it diagnoses, in miniature and without the excuse of state power. Bakunin’s own organisational practice, the graded brotherhoods and the invisible guidance of movements by initiated revolutionaries, was criticised in his lifetime by federalists who smelled the government of savants wearing a libertarian mask, and the criticism was not wrong merely because his enemies also made it. The platform controversy of the 1920s replayed the dilemma in print, with one wing of the exiled movement demanding unity and discipline indistinguishable in form from a party programme, and the other wing answering that an anarchist organisation which cannot tolerate internal dissent has already become the thing it exists to prevent. In Spain the self warning arrived as tragedy rather than debate, when organisations built to abolish government supplied ministers to one, and the movement’s own post mortems, written by participants who had opposed the entry and by some who had accepted it, circle the same admission: the refusal to prepare an answer to the power question did not abolish power, it merely ensured that the answer would be improvised by the least scrupulous hands at the worst moment. Later generations of anarchists codified the lesson in the critique of structurelessness and in renewed attention to formal, accountable, federal organisation, which is to say they moved, on the organisational question, a measurable distance toward the concerns their founders’ opponents had pressed in 1872. The movement across the seam runs in both directions, and has done since the beginning: Marxists in power rediscovering Bakunin’s warning, anarchists in crisis rediscovering Engels’s factory. A reader who grasps that double movement possesses the real conclusion of the comparison, of which the prediction asymmetry is the sharpest single component. The dispute is not a contest between a tradition that understands power and a tradition that does not. It is the permanent argument, inside any politics that aims at a free society, between the danger of building the instrument of liberation too strong to dissolve and the danger of refusing to build it strongly enough to survive. The founders gave that argument its classic statements because they lived at the moment when the instrument was first being designed. Their successors have been re enacting the argument, with fresh evidence and undiminished conviction, ever since, and the evidence assembled in this article suggests that the argument, rather than either of its parties, is what deserves to survive.

The formula and its career: a state that dies out

The withering expectation reached posterity in a single compressed formula, and the formula’s career inside the Marxist tradition is a study in how a conditional expectation hardens into a slogan. Engels, writing Anti Dühring in the late 1870s, gave the tradition its canonical sentence in substance: the state is not abolished, it dies out, in the German image of a process that happens to an institution as the ground beneath it is removed, the interference of state power in social relations becoming superfluous in one domain after another until the whole apparatus can be consigned, in his famous destination for it, to the museum of antiquities beside the spinning wheel and the bronze axe. The image is worth dwelling on because it is doing argumentative work. A thing placed in a museum is not defeated in battle. It is outlived, rendered quaint by the arrival of a world that no longer needs it, and the image therefore presupposes exactly what the anarchist dispute puts in question: that the apparatus will stand idle, unmissed and undefended, while society routes around it. Nothing in the image accommodates officials who experience the routing as a threat to their existence and act accordingly. The formula, repeated by generations of textbooks as the tradition’s settled teaching on the end of the state, performed a quiet transformation on the expectation it carried. In Marx’s Commune writings the expectation is attached to machinery, recall, wages, the armed people, a design for keeping the transitional state too weak and too watched to consolidate. In the textbook formula the machinery drops away and the expectation floats free as a law of history, something that will happen because of what states are in the theory, rather than something that must be made to happen against the resistance of what states are in the world. Bakunin’s prediction can be read, from this angle, as an argument about which of those two registers is the honest one. His claim was never that the withering formula was insincere. His claim was that formulae do not dissolve institutions, that institutions are dissolved, if at all, by people with the power and the will, against other people with the power and the opposite will, and that a theory which assigns the dissolution to history itself has disarmed its own side in advance.

The anarchist side had its own formula and its own hardening, and fairness requires the parallel to be drawn. Abolition, the demand that the state be destroyed in the revolution’s first act, functions in anarchist folklore the way withering functions in Marxist folklore: as a compressed certainty that conceals the machinery it presupposes. The machinery, in the anarchist case, is the federation described in its own section above, and it presupposes a population already organised densely enough to federate, a military balance survivable without a centre, and neighbours willing to let the experiment live. Where those presuppositions failed, in Spain most consequentially, the formula was retained and the failure assigned to betrayal or to circumstance, a bookkeeping indistinguishable in form from the Marxist assignment of failed withering to incorrect conditions. The symmetry is the deepest finding of this comparison, deeper even than the asymmetry that headlines it, and the two findings must be held together to be believed. On the specific, dated, unconditional prediction about transitional states, the anarchist was right and the record is not close. On the general intellectual virtue of refusing to let a formula outrun its machinery, neither tradition distinguished itself, and both built orthodoxies that graded the other side’s failures at full weight and their own at a discount. The founders, read closely and against each other, are better than their orthodoxies on precisely this point, which is why this article has insisted on returning to the texts of 1871 to 1875, where the arguments still carry their mechanisms on their backs and the mechanisms can be inspected, rather than to the slogans under which the traditions later marched them into a century neither man survived to see.

For students and exam candidates

The transition comparison is the essay structure this topic rewards, and answers should be built on it directly. Open with the shared end point, since examiners credit the candidate who refuses the caricature of opposite aims, then stage the disagreement as a dispute about the route: seizure and withering against abolition and federation. Bakunin’s prediction should be quoted in substance with its mechanism, the representatives who stop working, the experts who join them, the emergency that perpetuates itself, because the mechanism is what makes the prediction an argument rather than a slogan. Marx’s reply deserves its Commune safeguards in full, and Engels’s essay should be presented as an argument about coordination, not as a defence of the state, which is the distinction weaker answers always miss. Close with the asymmetry and its limits: one prediction tested and largely borne out, one expectation untested in its specified form, one alternative untested at scale. Keep the primary texts and the comparison side by side in a working file; the study notes at VaultBook are built for exactly that kind of structured revision, with the six rows of the table as headings and the evidence filed beneath each.

For teachers and seminar leaders

This comparison teaches best as a document exercise rather than a lecture, because the texts are short and the students’ prior picture is usually the caricature this article works against. Set the relevant pages of Statism and Anarchy beside the conspectus extracts and the essay on authority, and let the seminar discover that the founders argued about mechanisms, not about freedom in the abstract. The Hague makes a strong second session: give students the expulsion documents and Guillaume’s response and ask them to judge the proceedings by each tradition’s own organisational principles, which produces the uncomfortable symmetry that both sides violated theirs. A third session on the prediction, run as an evidence audit, works well: groups grade Bakunin’s forecast clause by clause against the record, then turn the same audit on the withering expectation and on the anarchist alternative, learning to distinguish a confirmed prediction from an untested one. The aim throughout is a transferable skill, the grading of political theories by their dated, specific, falsifiable content, and this dispute is the cleanest historical case for teaching it.

For researchers, historians, and journalists

The documentary base for this comparison is unusually tractable and unusually sloped, and researchers should know both facts before citing it. The tractable part: the core texts are short, datable, and available in reliable editions, the Hague proceedings are published, and the conspectus gives a rare view of a founder annotating his rival’s argument in private, which allows the published polemics to be checked against the working notes. The sloped part: the Marxist archive was canonised early by institutions with a verdict to defend, the anarchist archive scattered by illegality and by its founder’s death amid unfinished manuscripts, and the standard narratives in both traditions descend from partisan histories, Guillaume on one side and the official movement histories on the other, that must be read as exhibits as well as sources. Three questions in this field remain genuinely open to research. The first concerns the Alliance, whose real structure and membership are still reconstructed from hostile evidence. The second concerns reception: how the prediction travelled, through which translations and movements, into the self criticism of later Marxist dissidents. The third concerns the Spanish archives, where the anarchist movement’s governmental episode is documented by its own participants in recriminating detail that historians have not finished mining. Journalists covering the periodic revivals of both traditions will serve readers best by reporting the shared end point and the asymmetry together: the movements quarrel because they agree about the destination, and the one finding the record supports is narrower, stranger, and more interesting than the folklore of total opposition in which the split is usually retold. Precision about which claims have been tested, and which merely repeated, is the difference between reporting this dispute and re enacting it.

Frequently Asked Questions

Q: What is the main difference between Marxism and anarchism?

Both traditions aim at a classless, stateless society, and the difference lies in the route. Marxists hold that a transitional workers’ state is necessary to suppress counter revolution, reorganise production, and hold power during a dangerous interval, after which the state withers as classes disappear. Anarchists hold that any state, whatever its founding purpose, entrenches itself, breeds a governing stratum, and must therefore be abolished in the revolution itself, with federation beginning at once. Organisation, the revolutionary agent, and the status of authority divide them further, but the transition is the master disagreement from which the others follow.

Q: Why did Marx and Bakunin fall out?

The fall out was doctrinal, organisational, and personal in layers. Doctrinally, they disagreed about the state: Marx defended a transitional workers’ state and political action, Bakunin demanded the state’s immediate abolition and federal organisation. Organisationally, Bakunin’s Alliance operated inside the International as a semi separate body, which the Marxist leadership treated as a rival centre, while the federalist sections resisted the General Council’s growing powers as the embryo of the very domination they opposed. Personally, the two men conducted the struggle with slurs, manoeuvres, and a trial like expulsion at the Hague Congress, which poisoned the memory of the split for both traditions. The doctrine would have divided the movement without the personalities, since the same split recurred wherever later revolutionaries faced the same choice.

Q: Do anarchists and Marxists want the same end state?

Yes, in the respects that define the end state. Both want a classless society without a state: no separate apparatus of coercion standing over society, no class living from the labour of another, production organised by the associated producers themselves. The descriptions differ in vocabulary, withering and administration of things on one side, federation of free communes and workplaces on the other, but the content converges closely. The agreement is precisely why the quarrel is so bitter. Movements with different destinations can ignore each other, while movements sharing a destination and disagreeing about the road compete for the same recruits, the same unions, and the same revolutions, and each reads the other’s road as a betrayal of the common goal rather than as a mere error.

Q: What did Engels argue in On Authority?

Engels argued that authority is a feature of coordinated material activity and cannot be abolished by political decision. Large scale production, railways, and ships at sea impose decisions, timetables, and compliance because interdependent work requires them, whoever owns the enterprise. The anarchist demand to abolish authority as such is therefore, in his argument, a confusion: what can be abolished is the state, a specific institution of class rule, while authority in production survives any revolution and must be organised openly. His closing turn adds that a revolution is itself the most authoritarian act there is, so anti authoritarians practise coercion too, merely without admitting it. Anarchists reply that he conflates technical authority, which they accept, with governmental authority, which is their actual target.

Q: Was Bakunin’s prediction about a new ruling class correct?

Substantially, on the evidence available. Bakunin predicted in 1873 that a workers’ state would produce a privileged stratum of former workers who had stopped working, allied with credentialed experts, governing in the class’s name and perpetuating its own power. The states built under Marxist banners in the following century exhibited that stratum in recognisable form: privilege based on control of the apparatus rather than on property, self reproduction across generations, and no withering of the state in any case. Qualifications matter: those regimes departed from the Commune safeguards Marx specified, and all faced emergency conditions, so defenders argue the specified experiment was never run. But the prediction was unconditional and mechanism based, and case after case matched it, which is why it retains its force.

Q: Do anarchists have a theory of transition?

Yes, though it is a theory in which transition and destination are the same process. The old state is destroyed in the revolution rather than captured, and society reorganises from below at once: workplaces and communes govern their own affairs, federate through mandated and recallable delegates, defend themselves through the armed population, and extend federation as order is rebuilt. There is no transitional apparatus because any such apparatus is held to congeal into a new state. Critics object that the theory underestimates defence and coordination under fire, and the Spanish experience, where anarchist organisations entered government within months of the revolution’s outbreak, is the standing evidence for the objection. Supporters answer that Spain abandoned the theory rather than tested it.

Q: What happened at the Hague Congress?

At the Hague Congress of September 1872 the Marxist majority in the First International voted to strengthen the General Council, committed the association to political action, and expelled Bakunin and James Guillaume, chiefly over Bakunin’s Alliance, the semi secret organisation maintained inside the International. The federalist sections regarded the proceedings as a managed trial, and within days the expelled current met at Saint Imier to found a rival, anti authoritarian International. The parent body was transferred to New York, in part to remove it from the factional war, and dissolved within a few years, while the federalist body also faded. The institutional split endured far longer than either organisation, fixing the frontier between the traditions.

Q: Are there libertarian forms of Marxism?

Yes. Council communism, associated with Anton Pannekoek and the German and Dutch left after the First World War, kept Marx’s analysis of capital and the goal of a classless society while rejecting the party state, insisting that workers can rule only through their own councils and never through a party governing in their name. Rosa Luxemburg, from inside orthodox Marxism, warned that suppressing democracy leaves public life to bureaucracy. Autonomist and syndicalist currents later worked the same territory. These currents occupy the overlap between the traditions: Marxist in economics and in their account of class, anarchist in their suspicion of any state or party claiming to embody the class from above.

Q: Which tradition has the better historical record?

The records are asymmetric and must be graded question by question. On the fate of transitional states, the anarchist prediction was borne out and the Marxist expectation of withering was not fulfilled anywhere. On the demonstrated capacity to seize and hold power, industrialise, and act at scale in history, the Marxist tradition’s organisations were incomparably more consequential. On the anarchist alternative, a durable federated society at national scale, the record is essentially empty, since the experiments were brief, besieged, or compromised by their own leaders’ entry into government. Neither tradition can claim a better record overall without smuggling in a choice about which question matters most, which is a political judgement rather than a historical finding.

Q: Can Marxists and anarchists work together?

They have done so often, and the pattern of success and failure is consistent. Cooperation works at the level of resistance: shared unions, anti fascist struggle, common campaigns, and moments of revolt where an immediate enemy concentrates both traditions on the same task. It strains at the level of construction, because building anything durable poses the transition question, what institution holds power the morning after, in a form that cannot be fudged. Alliances have broken on that question repeatedly, in Russia after 1917 most fatefully. Durable cooperation seems to require settling the government question in advance, usually on council and federal terms, which neither tradition’s orthodoxy supplies on its own. Absent that settlement, joint work is real but provisional, and wise participants plan for the morning after as carefully as for the barricade.

Q: How does the transition comparison handle defence against counter-revolution?

The comparison treats defence as the hardest test of both positions rather than as a point for either. The Marxist row answers defence with the transitional state: a centre able to conscript, supply, and decide, whose coercive powers are accepted as the price of survival. The anarchist row answers with federation under arms: militias and communes coordinating through mandated delegates, on the claim that defence organised by a separate state reproduces the disease the revolution exists to cure. The record shows centralised defence winning wars and consolidating apparatuses, and federal defence bending toward central forms under sustained pressure, as in Spain. The comparison therefore records defence as the question on which the anarchist design is least tested and the Marxist design most costly.

Q: What should a student separate in Engels’s argument about authority?

A student should separate the two things Engels’s essay is constantly made to blur. The first is authority as a function of coordinated production, the discipline imposed by machinery, timetables, and interdependent work, which any society must organise somehow. The second is the state, a specific institution: a public force separated from the population and bound to class rule. Engels argues from the first to a conclusion about political honesty, that revolutionaries who coerce should admit and organise their coercion, not to the conclusion that a permanent state is desirable, which his wider writings explicitly deny. Keeping the function and the institution apart converts a sloganised essay into an argument that can be answered, which is what the anarchist distinction between technical and governmental authority attempts.

Q: Why did Bakunin expect a workers’ state to breed a new elite?

His expectation followed from his theory of what a state is. A state means a minority set apart to govern a majority, and those who govern, whatever class they came from, cease to do the work of the governed, acquire the standpoint and interests of administration, and defend their position with the apparatus in their hands. Applied to a workers’ government, the logic yields representatives who stop being workers in practice, joined by the scientists and experts a complex economy requires, whose knowledge becomes a second title to rule. Emergency powers, in his analysis, have no natural expiry date, because the stratum holding them acquires an interest in the emergency. The elite is therefore not a betrayal of the design from outside but the design’s predictable product.

Q: Where do the two traditions cooperate despite the split?

Cooperation recurs wherever an immediate task overwhelms the transition question. Both traditions built and shared trade unions across Europe and the Americas, sometimes in open rivalry and sometimes in the same locals. Anarchists and Marxists fought in the same anti fascist resistances, shared exile networks, and stood in the same crowds in the revolts of 1968 and similar moments. Joint publishing ventures, defence campaigns for each other’s prisoners, and united fronts against particular wars or laws fill out a long record. The cooperation is genuine but patterned: it flourishes in opposition and thins as soon as a shared victory raises the question of what to build and who commands it, which is the split reasserting itself at the level of practice.

Q: What test would compare their predictions about revolutionary governments?

The fair test begins from the predictions as published, not as later amended. Bakunin’s forecast is graded by examining any durable workers’ state for the predicted stratum: representatives separated from production, privilege resting on control, experts sharing rule, emergency powers persisting. The withering expectation is graded against a schedule of what withering would visibly require: falling administrative employment, converging official privilege, rotation in office, functions migrating to society, shrinking coercive organs. The anarchist alternative is graded only where a federated society actually operated at scale, and the honest finding there is that the sample barely exists. Comparing the three grades, rather than crowning a tradition, is the test this article has applied throughout, and it yields a confirmed prediction, an unfulfilled expectation, and an untested alternative.