Dialectical method is the most frequently claimed and least frequently taught instrument in the Marxist tradition, and the gap between the claiming and the teaching is the problem this article exists to close. A doctoral student in sociology, geography, education or political economy can read a hundred theses that describe themselves as dialectical and finish the hundredth no closer to knowing what the writer actually did differently on the first morning of the research. The word appears in methodology chapters like a coat of arms on a shield. It announces allegiance. It rarely announces a procedure, and a reader who asks what operations the word names, what decisions it forced, what it ruled out, is usually answered with a definition of dialectics itself, which is a different thing entirely. Definitions belong to the doctrine. This article belongs to the workshop. It treats dialectical method in research as a small number of decisions that can be stated, defended and criticised, and it shows each decision being made, on real material, with its costs made visible.

The central claim can be put before any apparatus. Most disagreements in social research that present themselves as disagreements about facts are, on inspection, disagreements about three settings that researchers adjust without announcing: how widely the object of study is extended in space and time, at what level of generality it is examined, and from what vantage point it is viewed. That claim comes from Bertell Ollman, whose account of the process of abstraction is the most operational version of dialectical method available in print, and it converts an apparently mystical vocabulary into an audit. Where two studies of the same strike, the same housing estate or the same examination system reach opposite conclusions, the dialectical researcher asks first whether the studies were looking at the same object at all, or whether one had bounded the object at the factory gate and the other at the world market, one had pitched its concepts at the level of a single firm and the other at the level of a mode of production, one had viewed the process from the standpoint of the employer and the other from the standpoint of the night shift. If the settings differ, the findings may both be accurate and the quarrel may be an artefact. If the settings match and the findings still differ, then and only then is there a factual dispute worth having.

A second claim disciplines the first. A word earns its place in a research design only when removing it changes something, and this article names that requirement the work test. The word dialectical is doing work in a piece of research only if deleting it changes a claim: if the sentences that surround it survive its removal without loss, the word was decoration, and decoration of that particular kind is the tradition’s characteristic vice. Supervisors who tell students that their use of dialectical is vague are almost always applying the work test informally. Examiners who write that a methodology chapter asserts rather than demonstrates are applying it too. The remedy offered here is not to abandon the word but to put it under load, attaching it to the settings, to a stated unit of analysis, to a contradiction claim that could fail, and to a written record of what happened when claims did fail. Method, in the account that follows, is nothing grander than decisions made explicit enough for another scholar to disagree with them productively.

A third commitment should be declared at the outset, because honesty about it is part of the method’s defence. There is no algorithm. No sequence of steps exists which, followed mechanically, guarantees a dialectical study in the way that a sampling formula guarantees a representative sample, and any writer who promises one is selling a ceremony. What exists instead is a discipline of explicitness about abstraction, a distinction between the order in which material is gathered and the order in which it is presented, a procedure for rising from thin abstractions to a concrete account that holds many determinations together, and a family resemblance to neighbouring approaches, systems thinking, process tracing and critical realism among them, from which dialectical research must be distinguished by what it adds rather than by what it denounces. The pages that follow take those elements in turn, then carry one complete research design through all of them, then state the strongest sceptical objection at full strength and answer it section by section, conceding what has to be conceded.

Dialectical method as a set of decisions

Begin with what the method is not, because the confusions are load bearing. Dialectical method is not dialectical materialism, although the two are routinely fused in textbook summaries and in the minds of students who meet them together. Dialectical materialism names a doctrine about the world, a set of claims concerning matter, motion and the development of nature and society, and its exposition belongs to the series guide to dialectical materialism as a complete doctrine, which owns that definitional ground. Where the method survives inside the working disciplines, discipline by discipline, is mapped by the series survey of dialectics in social science, which owns that territory in turn; this article owns the procedure that a researcher in any of those disciplines can carry away. A method, by contrast, is something a researcher does on a Tuesday with a notebook, a dataset or an archive box. A scholar could in principle accept the doctrine and conduct slovenly research, just as a scholar could reject the doctrine entirely and still find Ollman’s settings useful as an audit of framing decisions. Keeping the two apart is not pedantry. It is what allows the method to be assessed on its performance rather than cheered or booed as a badge, and it is what allows a researcher outside the tradition to borrow its instruments without importing its conclusions, a borrowing the tradition should welcome rather than police.

Nor is the method a technique of data analysis, in the sense in which regression, thematic coding or process tracing are techniques. It operates one level above technique, at the point where the object of study is constituted: where its boundaries are drawn, which relations are treated as internal to it and which as background conditions, which changes count as the object becoming something else and which count as the object merely fluctuating. Two researchers can run identical statistical models on identical data and be conducting different studies, because one has constituted the object as a labour market and the other as a set of individual job seekers, and the models will obediently describe whatever object they were handed. Dialectical method concerns itself with that prior constitution and refuses to let it happen by default. This is why methodology chapters that list techniques and then add the word dialectical as a seasoning satisfy nobody. The word belongs in the section where the object is defined, and if it cannot be made to do work there, it will do no work anywhere else in the document.

The positive account starts from a single observation about abstraction. Every act of research abstracts. To study anything is to lift it out of the dense simultaneity of the world and hold it still under a description, and the lifting is unavoidable: even the most determined empiricist, transcribing an interview or cleaning a spreadsheet, has decided what counts as a case, what counts as the same case continuing through time, and whose description of the situation will structure the record. Abstraction in this sense does not mean vagueness. It means the opposite. An abstraction is a boundary drawn deliberately, and the dialectical tradition’s contribution is to insist that the drawing be treated as a research act with consequences, recorded like any other research act, rather than as a preliminary to be hurried past on the way to the findings. Marx’s own practice exhibits the insistence. The opening of Capital does not begin with factories, strikes or prices as they are met in experience. It begins with the commodity, an abstraction of ferocious thinness, chosen because the relations that organise a whole society can be unfolded from it in a controlled order. Whatever a reader concludes about the economics, the methodological gesture is unmistakable: start where the determinations are fewest, know exactly why you started there, and earn each complication by showing the relation that requires it.

Ollman’s contribution was to take that gesture apart and name its adjustable parts. In Dialectical Investigations, published in 1993, and again in Dance of the Dialectic, published in 2003, he argues that abstraction has three dimensions along which any researcher, knowingly or not, fixes a position. Extension fixes how far the object reaches across space and how long it persists through time. Level of generality fixes the altitude of the concepts used, from the unique event through the historically specific form to the most general features of a kind of society. Vantage point fixes the position from which the object is viewed, the standpoint built into the questions asked and the evidence gathered. Ollman’s further claim, the one that gives the apparatus its bite, is that these settings are usually invisible to the researchers who hold them, so that scholars argue past one another for years, each treating the other’s framing as an error of observation. Make the settings explicit, and a large class of sterile disputes either dissolves or sharpens into a genuine disagreement about which setting suits the question. Either outcome is progress, and both are obtainable by any researcher willing to fill in a worksheet and defend the entries.

It is worth pausing on how modest this apparatus sounds, because the modesty is deceptive and the sceptic’s strongest case rests on it. Settings, boundaries, standpoints: every competent methodologist recognises the territory, and a later section of this article states, at full strength, the objection that dialectical method is ordinary good research wearing ceremonial dress. The reply, when it comes, will not deny the overlap. It will point instead to what the tradition does with the settings that ordinary practice does not: it treats the three dimensions as a single coordinated audit rather than as separate housekeeping items, it adds the doctrine of internal relations, which changes what a boundary means, it adds the ascent toward the concrete as a rule for presentation, and it adds a test, the work test, that can be run over any draft, including this one. Whether those additions amount to a method or to a manner of speaking is exactly the question the objection presses, and the article’s verdict, stated in advance so that the reader can watch it being earned, is that they amount to a method of a specific and limited kind: a method for constituting objects of study and for ordering an exposition, which must be married to ordinary techniques of evidence and which guarantees nothing about conclusions.

One more preliminary distinction will be needed throughout, and it concerns the difference between a contradiction and a difficulty. In ordinary speech the words are interchangeable, and in a great deal of published dialectical writing they are interchangeable too, which is one reason the vocabulary attracts suspicion. In the disciplined usage this article follows, a contradiction is a relation between two aspects of a single object such that each aspect requires the other and simultaneously undermines it, so that the object’s ordinary functioning generates pressure for its own transformation. A difficulty is merely an obstacle. A shortage of funding is a difficulty for a hospital. A hospital funded per procedure while charged with preventing the conditions that produce procedures faces something closer to a contradiction, because the mechanism of its survival and the content of its mission pull against each other by design rather than by accident. The category receives its full treatment in the series article on contradiction in Marxist method; here it matters because a dialectical research design must state its contradiction claim in a form that could fail, and a claim that could not fail is, by the work test, not a claim at all.

The three settings of abstraction

How do you actually use dialectical method in research?

Make three explicit decisions about your object, namely its extension in time and space, its level of generality, and your vantage point, study it in its relations and its becoming rather than as fixed, and separate the order of your inquiry from the order of your presentation.

The worksheet that organises this article appears in full in a later section, with its example carried across every row. Before the table can be used, each setting needs its own examination, because each conceals a characteristic error and each repays a characteristic question. The three settings also interact. A wide extension usually forces a higher level of generality, since a researcher who extends an object across two centuries cannot also track the intentions of every participant. A vantage point chosen at the level of the individual biography makes some extensions natural and others perverse. Doctoral designs most often fail not because one setting is wrong but because the three are inconsistent with one another: a research question framed in terms of an entire mode of production, evidence gathered at the level of a single workplace, and conclusions announced at the level of a national class structure, with the silent escalations between levels doing all the argumentative work. The settings audit exists to catch exactly that slippage while it is still cheap to repair, in the design, rather than in the viva.

What does extension mean in the process of abstraction?

Extension is the reach of the object in space and time: where the researcher draws its boundary, when it begins, and when it ends. A strike studied across one depot and one week is a different object from the same strike studied across an industry and a decade, and the findings differ because the objects differ.

Extension, in fuller terms, is where the researcher draws the object’s outer boundary, when it is taken to begin, and when, if ever, it is taken to end. A study of a strike may be bounded by one depot and one week, or it may run across an entire sector over ten years, and the two studies will report different causes because they are examining different objects that happen to share a name. Neither boundary is the true one in the abstract. Each is a decision, defensible by the question being asked and vulnerable to exactly one test, whether the relations that explain the outcome fall inside or outside the line that was drawn.

The characteristic error at this setting is the boundary that follows the archive rather than the problem. Records survive in administrative units, so studies inherit administrative shapes: the school rather than the catchment, the firm rather than the supply chain, the parliamentary term rather than the policy’s career through several governments. The dialectical habit is to ask, before fieldwork begins, what the object does that crosses the inherited boundary. A school whose intake is determined by a housing market, whose staffing is set by a national labour shortage and whose results are consumed by employers in another region is being studied at the wrong extension if the explanatory relations stop at its gates. Widening the boundary is not free. Each widening multiplies the material, dilutes the detail and tempts the researcher toward the level of generality at which everything connects to everything and nothing can be falsified. The craft lies in extending just far enough to capture the relations that do the explaining, and in writing down why the line was drawn there, so that a critic can propose a better line rather than merely suspecting a worse one.

Temporality deserves separate emphasis, because the temporal boundary is the one researchers set most carelessly. To begin a study of a trade union in the year of its founding charter is to assume that the founding explains the form; to begin it in the year of a famous defeat is to assume the defeat does. Marx’s historical writing habitually violates the tidy start date, reaching back to the enclosures when explaining a factory system or forward to tendencies still embryonic when judging a political form, and the habit is methodological rather than decorative. An object extended through time is an object allowed to become, and becoming is where dialectical claims live. The corresponding danger is the extension that never ends, the study in which every outcome is provisional and every conclusion deferred to a longer run. A design should therefore name not only its start and end points but the event or condition that would count as the object ending or transforming, so that the temporal boundary is a hypothesis about the object’s career rather than a convenience of the calendar.

Why does the level of generality change what a study can see?

Level of generality is the altitude of the concepts in use, running from the unique event, through historically specific forms, up to features of a whole type of society. A study pitched at one rung cannot see what neighbouring rungs show, so the choice of level partly decides the findings before evidence is gathered.

Ollman’s scheme, spelled out, distinguishes a ladder running from the unique, through the historically specific, up to the most general features of a type of society. At the lowest rung stand events that happened once: this negotiation, this eviction, this cohort of trainees. Higher stand forms specific to a period and place: the postwar British council house, the late twentieth century American suburb, the contemporary platform firm considered as a kind of organisation. Higher still stand the concepts that claim relevance across a whole mode of production: wage labour, the commodity form, capital itself. What one altitude discloses, the altitudes on either side of it cannot, much as a map at one scale cannot display both the street layout and the continental watershed, and most charges of blindness traded between scholars are charges that the other party chose the wrong altitude for the question at hand.

The characteristic error is level slippage inside a single argument. A researcher assembles evidence at the level of a handful of interviews, interprets it through concepts pitched at the level of a social formation, and publishes conclusions at the level of a policy recommendation, and each transition is invisible because the same nouns, class, power, resistance, recur at every altitude with different meanings. The noun is doing smuggling work. The settings audit forces the transitions into the open by requiring the design to state, for each major concept, the rung at which it will be used and the observations that belong to that rung. Where a concept must move between levels, the movement becomes an explicit step in the argument, with its own justification, instead of a slide. This single discipline, honestly applied, would retire a substantial fraction of the vagueness for which dialectical writing is mocked, because much of that vagueness is not dialectical at all. It is altitude sickness, and it afflicts every tradition that works with layered concepts, from functionalist sociology to rational choice history.

Choosing a level is also a wager about where the explanation lives. Pitch the study of a failing school at the level of the unique and the explanation will be a story of personalities and accidents, true perhaps and ungeneralisable certainly. Pitch it at the level of the education system of a whole society and the explanation will concern funding formulas and labour markets, with the particular school reduced to an illustration. Pitch it at the intermediate level, the historically specific form of schooling in a particular urban regime, and mechanisms become visible that both neighbours miss: the interaction between a local housing market, a selective admissions rule and a national accountability metric. Dialectical method has no formula for the right rung. It has a demand, that the rung be chosen in awareness of the ladder, that the choice be stated, and that the researcher be able to say what the adjacent rungs would have shown and why the question does not need them. A design that can answer that challenge has done something real. A design that cannot has not yet begun, whatever its word count.

What is a vantage point in Ollman’s sense?

A vantage point is the position inside the object from which the researcher views it: the roster office, the night shift, the finance desk. The position partly constitutes the findings, because each standpoint makes some relations visible and hides others, so the design must declare its standpoint and state what that standpoint cannot see.

Vantage point, in fuller terms, is the position built into the act of looking, and it is the setting that researchers are most reluctant to declare, because declaring it feels like confessing a bias when it is actually the precondition of correcting for one. Every object of social study presents different faces to different positions within it. A shift system viewed from the roster office is a puzzle of coverage and cost. The same system viewed from the night bus home is a pattern of broken sleep and missed school plays. Both views are of the same object, both can be documented rigorously, and neither is the view from nowhere. Ollman’s point is not the familiar one that perspective colours interpretation, although it does. His point is stronger and more useful: the vantage point partly constitutes the object, because the relations that appear internal to the object, and the changes that appear as its development, depend on where the viewing begins. A study of a hospital from the vantage point of its finance office will find a contradiction between demand and budget. The same hospital from the vantage point of its porters will find a contradiction between the formal hierarchy and the informal knowledge that actually moves patients through the building. The contradictions are both real. They are not the same finding, and a design that does not know its vantage point will report whichever one its evidence happened to collect, mistaking an accident of access for a property of the world.

Declaring a vantage point has three practical consequences that a design can be held to. The first concerns evidence: each standpoint makes some sources natural and others invisible, so the declaration should be accompanied by an account of what the chosen position cannot see and how, if at all, the study will compensate. The second concerns concepts: categories that are obvious from one position are artefacts from another, and the researcher owes the reader a note on which of the study’s organising categories belong to the participants, which to the administration being studied, and which to the analyst. The third concerns change over time, the dimension in which vantage point does its subtlest work. To view a process from the standpoint of its outcome is to risk reading necessity backward into every earlier stage, the error Marxist historiography at its worst committed wholesale and at its best diagnosed in others. To view it from within an early moment, with the outcome unknown, restores the openness that the participants experienced and that any honest account of their choices requires. A design may need both views, taken in sequence and labelled. What it may not do is alternate between them without signalling, because the reader who cannot tell whether a judgement is retrospective or internal to the moment cannot assess it, and unassessable judgements are the raw material of the filler problem.

A worked miniature makes the three settings tangible before the full design arrives. Consider two published styles of study of the same phenomenon, the growth of food banks in a wealthy country. The first extends the object across one town and two years, works at the level of the unique and the locally specific, and views the scene from the vantage point of the volunteers: its findings concern referral pathways, warehouse logistics and the moral economy of the queue, and they are detailed, humane and silent about why the town needs the warehouse. The second extends across four decades and a national economy, works at the level of the wage relation and the welfare state as historically specific forms, and views the scene from the vantage point of the household budget: its findings concern benefit levels, rent burdens and the price of food relative to wages, and they explain the warehouse while saying nothing about its opening hours. Neither study is the dialectical one by virtue of its scale. The dialectical demand falls on both equally: state the settings, show that the settings fit the question, and mark the place where the findings of the other scale would alter the conclusions if they were admitted. A literature review written to that standard stops being a parade of summaries and becomes an argument about settings, which is the first genuinely useful thing a doctoral student can be taught to build, and a later section returns to the literature review as a dialectical exercise in its own right.

The abstraction settings worksheet

The worksheet below is the findable artifact of this article, and it is deliberately unglamorous. A table cannot do research for anybody. What it can do is force four questions per setting onto a single page, where they can be answered in sentences short enough to be argued with. The example carried across all three rows is a study of casual teaching labour in universities: the employment of teachers on short, fractional or hourly contracts to deliver a growing share of undergraduate teaching. The example is developed fully as a worked design later in the article. Here it serves only to show what filled cells look like, because a worksheet described in the abstract teaches nothing, while a worksheet shown completed, with its trade offs exposed, can be imitated, criticised and improved. Readers building their own version for a different object should keep the row structure and replace every cell, including the example’s quiet assumptions, which are contestable and are meant to be contested.

Setting	Question to ask	Setting A	What A makes visible, and what it hides	Setting B	What B makes visible, and what it hides
Extension	Where does the object begin and end, in space and in time?	One university, one decade	Visible: departmental budgets, hiring decisions, named managers, the lived texture of insecure work. Hidden: the national funding settlement and the labour market that supply the casual pool.	The higher education systems of one country across four decades	Visible: the funding formula, the expansion of enrolment, the long substitution of casual for permanent labour. Hidden: the departmental discretion that decides who is hired and who is dropped.
Level of generality	At what altitude do the organising concepts operate?	The historically specific form of the marketised university	Visible: audit culture, student fee income, league table pressure as mechanisms acting on hiring. Hidden: features shared with casualised labour in other sectors, and any claim about wage labour as such.	Wage labour and its reserve of available workers in general	Visible: the structural pressure that a pool of qualified underemployed graduates exerts on all academic wages. Hidden: everything specific to universities, and any finding a policy maker could act on.
Vantage point	From which position inside the object is it viewed?	The standpoint of the hourly paid teacher	Visible: income volatility, unpaid preparation, the career ladder experienced as a wall, dependence on favour. Hidden: the budgetary constraints managers cite, and the experience of students taught by rotating staff.	The standpoint of the institution as an employer	Visible: flexibility as a survival strategy, enrolment risk transferred downward, the arithmetic of the teaching budget. Hidden: the human cost that the arithmetic distributes, and the teacher’s view of the same flexibility as insecurity.

Three lessons fall out of the completed rows, and each generalises beyond the example. The first is that visibility and concealment are paired in every cell. There is no setting from which everything appears, and a methods section that lists what its framing reveals without listing what it hides is advertising, not methodology. The second is that the settings of a real design are chosen together. The researcher who selects the national extension and the general level has, whether the fact is acknowledged or not, also selected a vantage point far from any participant, and should expect the evidence to consist of series and statutes rather than voices. The third is that the worksheet is a record to be kept, not a ritual to be performed once and discarded. When a study changes its extension midway, as honest studies often do when the archive disappoints or a relation leads outward, the change belongs in the worksheet’s margin and later in the methods account, dated and explained. The audit trail of settings is to dialectical research what the lab notebook is to experimental science: not the finding, but the condition of trusting the finding.

Keeping that record has become easier than it was for the generation that worked in card indexes, and the practical arrangements deserve a sentence in any methods discussion. A researcher can hold the worksheet, the dated setting changes, the contradiction claims and their outcomes, and the reading notes that prompted each revision in a single study file on VaultBook, where the notes for a project sit together, searchable and offline, rather than scattered across a laptop, a library account and a supervisor’s annotated drafts. The tool does not make the decisions. It makes the decisions retrievable, which, three years into a doctorate, amounts to much the same service.

Internal relations and the unit of analysis

Beneath the three settings lies a commitment that distinguishes dialectical research from other kinds of framing talk, and it must be stated plainly because it is the commitment the sceptic is entitled to interrogate. The commitment is the doctrine of internal relations: the claim that some of the relations in which a thing stands are not external connections between already constituted items but are part of what the item is. A landlord and a tenant are the standard illustration. Strip away the tenancy relation and neither party remains what it was: the landlord without tenants is a person with an expensive building, the tenant without a landlord is a person with an income and a housing problem of a different shape. Capital and wage labour are Marx’s own large scale version of the same point. Each term of the pair presupposes the other, and the history of each is unintelligible except as the history of their changing relation. Where relations are internal in this sense, the unit of analysis cannot be either term alone. The unit must be the relation, or the structured whole within which the relation operates, and a research design that takes individuals as its units while making claims about internally related positions has built a contradiction into its foundations, this time in the logician’s sense rather than the dialectician’s.

The practical force of the doctrine shows in a rule that one of this article’s extra questions puts directly: a dialectical study must name its unit of analysis before it names its forces. Forces, pressures, tendencies and contradictions are all predicates of a unit. To say that a contradiction operates between flexibility and security means nothing until the design has said within what whole the two press against each other: within the individual contract, within the department’s staffing model, within the institution’s finances, or within the sector’s labour market. Each choice yields a different contradiction and a different study. Researchers who begin by listing the forces at play in a situation, as brainstorming exercises encourage, are building on air, because the same list rearranges itself completely when the unit changes. The unit first, then the relations internal to it, then the contradiction claim stated as a claim about those relations, then the evidence that could embarrass the claim: that ordering is the closest thing to a procedural skeleton that dialectical method possesses, and a design document that follows it can be examined, faulted and repaired at specific joints.

Internal relations also govern how change is conceived, and here the tradition’s language of becoming earns its keep or fails to. If the terms of a relation are constituted by the relation, then a sufficient change in the relation is a change in the identity of the terms, not merely in their circumstances. The hourly paid teacher who wins a permanent post has not simply improved a situation; in the vocabulary of the object under study, a different position in the structure is now occupied, with different interests attaching to it, and a longitudinal design that follows persons will record continuity where a design that follows positions records a replacement. Neither record is false. They answer different questions, and the settings worksheet is where the choice between following persons and following positions belongs, argued rather than assumed. Much confusion in mobility research, in organisational ethnography and in oral history comes from sliding between the two mid analysis. The dialectical contribution is not a verdict in favour of positions. It is the insistence that the identity of the unit through time is itself a theoretical decision, since an object that is a knot of relations can persist, transform or dissolve, and the design must say in advance what evidence would mark each outcome.

Against this, a caution from inside the tradition deserves equal weight, because the doctrine of internal relations is easily overextended into a holism that explains nothing. If everything is internally related to everything else, then no boundary can be drawn and no claim can be isolated long enough to be tested, and the researcher has purchased profundity at the price of paralysis. Ollman himself is clear that abstraction is precisely the art of provisionally treating relations as external for the purpose of examination, of holding a boundary steady while a mechanism is traced, in full knowledge that the boundary is a decision rather than a discovery. The settings are the record of those provisional cuts. Sayer’s work on method in social science makes the complementary point from the realist side: social science deals in open systems, where mechanisms combine irregularly and constant conjunctions of events are rare, so the researcher’s task is to identify structures and their tendencies and then examine how they play out under specific conditions, not to hunt for laws of the closed form. Read together, the two cautions define the working posture of the method at its best. Draw boundaries boldly, since nothing can be studied without them. Record them honestly, since everything concluded is concluded within them. And treat the boundary itself as one of the findings when the evidence rebels against it, as evidence in a well kept study regularly does.

What separates a relation from a mere connection?

A relation is internal when removing it would change what the connected items are, not merely what happens to them. A landlord without tenants and a tenant without a landlord are different beings, not inconvenienced versions of themselves, so the tenancy relation belongs inside the unit of analysis rather than among its external conditions.

A relation counts as internal, for design purposes, when removing it would change what the connected items are, and not merely what happens to them. The test is applied by counterfactual inspection of the object’s own constitution. A commuter town is connected to a capital city by trains, timetables and property prices, and most of those connections are external: interrupt them and both places persist, inconvenienced and poorer, recognisably themselves. A company town is related to its single employer internally, because the town’s institutions, its housing stock, its politics and its very population were constituted by the employment relation and begin to transform into something else when the relation is withdrawn. The distinction cannot be read off the strength of the connection. It must be argued from the history of the object and from the dependencies that its ordinary functioning reveals, and the argument belongs in the design, where a critic can get at it.

Designs go wrong at this joint in two symmetrical ways. The first inflates every connection into an internal relation, producing studies in which a supermarket’s car park, a minister’s speech and a grandmother’s habits are all moments of one totality, each invoked to explain the others in a circle that no evidence could interrupt. The second deflates every relation into an external connection, producing studies of exquisitely measured items whose definitions smuggle in the relations the analysis claims to discover: the unemployed person defined without reference to employers, the wage defined without reference to the bargain that sets it. A dialectical design steers between the two by making the internality claim specific and revisable. It names the relations it treats as constitutive, accepts the burden of showing what would change if they were removed, and leaves the remainder of the object’s connections in the honest category of influences, conditions and contexts, which can be weighed without being metaphysically promoted. The claim, like every claim in the design, is subject to the work test. If treating a relation as internal changes nothing in the analysis, the designation was ceremonial, and the worksheet should be corrected to say so.

Inquiry and presentation are two different orders

Marx left one methodological text that researchers quote more than they use, the introduction written in 1857 for the work that became the Grundrisse, and its most consequential distinction for research practice is between the method of inquiry and the method of presentation. Inquiry appropriates the material in detail. It is archival, empirical, patient and largely shapeless, a long accumulation of evidence whose order is dictated by availability, by accident, by the sequence of libraries and interviews and datasets. Presentation is another matter entirely. In presentation, the material, now mastered, is reconstructed so that the life of the object appears in the arrangement of the argument itself: the exposition begins from the simplest determinations and builds, relation by relation, toward the concrete whole, and the reader retraces in an afternoon the conceptual labour that took the researcher years of formless gathering. The two orders are not merely different. They are frequently opposite. The archive that supplied the key insight may be the archive visited last. The category that opens the finished book may have been the category the researcher understood last and worst at the start.

The distinction matters practically because its neglect produces two recognisable diseases of academic writing. The first is the thesis structured as a diary, in which chapters follow the order of discovery and the reader is dragged through every false start at full length, on the apparent theory that labour undergone is knowledge conveyed. The second is the opposite fraud, the thesis that presents its conclusions as if they had descended in their finished order, concealing the inquiry so completely that the reader cannot judge how the evidence was won or what it cost the argument when recalcitrant material was set aside. Dialectical presentation is not the second of these. When Marx distinguishes the modes, he warns explicitly that the reconstructed exposition can create the illusion that the object was generated by thought itself, an illusion he names in order to refuse it, and the refusal imposes a duty: the presentation must keep its debt to inquiry visible, through the evidence it cites, the settings it declares and the failures it records. A presentation that hides its inquiry has not become dialectical. It has become idealist in the precise sense Marx criticised in Hegel, which is why the comparison between the two dialectics, treated fully in the series article on Hegel’s dialectic and Marx’s dialectic, is a live methodological issue and not a courtesy to the history of philosophy.

Applied to the design of a study, the distinction yields concrete instructions. Keep the research record in the order of inquiry, unedited, because its mess is evidence about the path taken and about the paths not taken. Build the written study in the order of presentation, which is the order of dependence among the concepts: what must be understood first for the next element to be intelligible. Expect the two documents to diverge, and do not apologise for the divergence in the methods chapter; explain it, with one honest paragraph about how the inquiry actually proceeded and one about why the presentation departs from that sequence. Supervisors and examiners reward the explanation, because it demonstrates the distinction in operation rather than in citation. The same discipline governs the smaller units of a study. A journal article that opens with its thinnest abstraction, shows the relation that complicates it, and only then admits the institutional detail that a chronological account would have put first, is practising the distinction at paragraph scale. A literature review that proceeds by dependence rather than by date, gathering positions according to the settings they share rather than the years they appeared, practises it too, and such a review does argumentative work that a chronological parade cannot.

What is the difference between the method of inquiry and the method of presentation?

The method of inquiry is the order in which material is actually gathered, following access, accident and availability. The method of presentation is the order in which the finished study rebuilds that material for a reader, following dependence among concepts from the simplest determination toward the concrete whole. The two orders rarely coincide, and neither should imitate the other.

To be exact about the contrast: inquiry proceeds through archives visited as they open, interviews conducted as access allows, data cleaned in the sequence problems appear, while presentation begins with the simplest determination and adds relations as the argument requires them. Inquiry follows opportunity and evidence. Presentation follows dependence among concepts, so the two sequences rarely coincide, and confusing them produces theses structured as diaries and articles whose logic only their author can retrace.

There is a temptation, especially among beginning researchers, to treat the distinction as a licence, as if dialectical presentation authorised the writer to arrange material in whatever sequence flatters the conclusion. The tradition’s own exemplars refuse the licence. Marx’s presentation in Capital is ruthless about sequence, yet every stage of the ascent is anchored in material that the inquiry sections, the historical chapters on the working day, on cooperation, on primitive accumulation, supply in documentary bulk. The order of the concepts governs the exposition, while the weight of the exposition rests on appropriated fact. A design can be checked against that standard with a simple question: for each step in the presented argument, could the researcher produce the inquiry record, the notes, the sources, the dead ends, from which the step was built? Where the answer is no, presentation has floated free of inquiry and the study is exposed, however elegant its architecture. Where the answer is yes, the rearrangement of material is not a distortion but the whole point of having a method of presentation at all, and the methods chapter that maps steps to records, briefly and without vanity, gives the reader the means to verify the claim.

The distinction also rescues researchers from a false shame about how knowledge actually arrives. Insight in real inquiry is opportunistic. The decisive document surfaces in the wrong archive, the interviewee who changes the argument is met by introduction at a conference dinner, the dataset that settles a question is compiled for another purpose entirely. A methods culture that demands the written study mirror that contingency would be demanding worse books in the name of honesty. The dialectical position is more exact and more forgiving at once: the inquiry is what it is, opportunistic and cumulative, and honesty consists in recording it truthfully, while the presentation is a construction and honesty consists in making its logic examinable. Confusing the two standards, judging a presentation by its resemblance to the inquiry or an inquiry by its elegance, is the beginner’s error the 1857 introduction exists to correct.

The ascent from the abstract to the concrete

The second move Marx names in the same introduction is the rise from the abstract to the concrete, and it is the element of the method most often quoted and least often performed. The concrete, in this vocabulary, is not the tangible. A handful of soil is tangible and, for most research purposes, abstract, because almost nothing about it has been determined. The concrete is the concentration of many determinations, the point at which an object stands reconstructed with its relations, its history and its tendencies held together in a single account. A named street in a named town becomes concrete in the relevant sense when the study can hold together its housing tenure, its price history, the employment that sustains its households, the policies that zoned it, the credit that built it and the direction in which each of those is moving. The everyday research failure that the vocabulary diagnoses is the study that mistakes the starting chaos of impressions, Marx’s phrase for the raw given is a chaotic conception of the whole, for the concrete, and reports the chaos at length. Interview transcripts piled without determinations are chaos, however many pages they fill. So are descriptive statistics unconnected to any structure that would explain them. The ascent is the discipline of refusing to stop there, and equally of refusing to stop at the opposite pole, the thin abstraction, the model or definition so purified that nothing in the world answers to it any longer.

Performing the ascent in a design means scheduling it. The study begins from abstractions chosen for their thinness and fertility, states them exactly, and then earns concreteness in planned stages, each stage adding a determination that the previous stage showed to be necessary. Marx’s example is economic: value, then money, then capital, then wage labour, then the state and the world market, each category presupposing and complicating its predecessors. A study of casual academic labour can be scheduled in the same form. Begin with the employment contract in its thinnest description, a sale of teaching hours. Add the determination of the labour market that supplies more qualified teachers than there are permanent posts, and the contract changes character. Add the institution’s funding model, in which teaching income arrives per enrolled student and fluctuates, and the contract appears as a device for transferring that fluctuation downward. Add the career structure, in which casual years are simultaneously a probation and a filter, and the contract acquires its hold over the people who accept it. Add the state, its audit regimes and its fee policies, and the concrete object finally stands: not a contract, not a grievance, but a node where a funding settlement, a labour market, an institutional strategy and thousands of individual hopes intersect and press against one another. The study that can narrate that sequence, with evidence at every stage, has performed the ascent. The study that announces the conclusion at stage five while evidencing only stage one has performed a leap, and leaps are what examiners are trained to notice.

Two practical corollaries follow. The first is that the abstract starting point must be defended as a starting point, not smuggled in as a finding. Marx begins with the commodity because he can show that the commodity form is the cell form of the society under study, the simplest unit in which its characteristic relations are already present in embryo. A researcher beginning with the contract, the household, the classroom or the firm owes an analogous argument: this unit, and not a neighbouring one, contains the relations the study needs in their least developed form. The second corollary concerns stopping. The ascent has no natural terminus, since determinations can be added indefinitely, and a study that cannot stop cannot be written. The stopping rule is pragmatic and should be declared: the ascent ends when further determinations cease to alter the contradiction claim at the study’s chosen level of generality. At that point additional richness is decoration, and the researcher who continues gathering it is avoiding the exposure of writing, a recognisable academic condition that no methodology can cure but a declared stopping rule can at least name.

How does the ascent organise the stages of a study?

The ascent guides fieldwork by turning the stages of the argument into stages of gathering. Early fieldwork tests the thin starting abstraction against documents and first interviews. Later fieldwork hunts the determinations the argument has shown to be necessary, the funding rule, the market condition, the career incentive, so gathering stays targeted instead of encyclopaedic.

Fieldwork organised by the ascent knows what it is looking for at each stage and what would count as the stage failing, which keeps a long project from dissolving into general accumulation.

It is equally important to say what the ascent is not, because the phrase attracts a devotional reading in which rising to the concrete becomes a synonym for thinking harder or for adding context until the prose feels rich. Context added without a determination doing work is precisely what the method exists to prevent. Each added element must pass a simple interrogation: what does this determination change about the object as previously described? If the honest answer is that it changes nothing, that it sits alongside the account as scenery, the element belongs in a footnote or nowhere. The interrogation is severe and it is meant to be. It is the same severity the work test applies to vocabulary, applied here to evidence, and a study that survives both interrogations has earned the right to call its conclusion concrete. A study that has merely accumulated will feel the difference in its final chapters, where conclusions have to be assembled from material that was never selected to support them, and the familiar experience of a thesis that is long everywhere and decisive nowhere is usually the experience of an ascent that was announced in the introduction and abandoned by the second month of fieldwork.

Contradiction claims and the work test

Every element so far has been preparation for the point at which a dialectical design either becomes research or remains rhetoric: the statement of its contradiction claim. The claim is the design’s engine. It identifies, within the named unit of analysis, two aspects that require each other and undermine each other, and it predicts that the object’s ordinary functioning will generate pressure, symptoms and attempted repairs that a study can go and look for. Stating the claim well is a craft with rules. The two aspects must both be internal to the unit, not one aspect inside and an obstacle outside. The dependence must run in both directions, since a one way pressure is a constraint, not a contradiction. The undermining must be produced by the object’s normal operation, not by an accident or an enemy, since a contradiction that requires sabotage to appear is a plot, not a structure. And the claim must be accompanied, in the design document, by its failure conditions: the observations that would show the two aspects coexisting without strain, or the strain resolving without transformation, or the pressure arising from a different mechanism altogether. A claim accompanied by its failure conditions can lose, and only claims that can lose can win.

The work test governs the whole apparatus at the level of language. Apply it by deletion. Take the draft, remove every occurrence of the word dialectical and its cognates, and read what remains. Where a sentence collapses, the word was carrying weight: it marked a setting decision, a claim about internal relations, a stage in the ascent. Where nothing changes, the word was upholstery, and the honest options are to delete it permanently or to do the work it was pretending to do. The test generalises, and researchers do well to run it over the other ceremonial vocabulary of the field: process, dynamic, complex, holistic, relational. Each of those words has honest employment. Each also has a career as filler, signalling that the writer has noticed the world is complicated and would like credit for the noticing. The test takes an hour on a long draft and returns a map of the places where the argument is thinner than its prose. Few methodological instruments offer so much for so little, and none of them requires agreeing with Marx about anything else at all.

Recording failures is the part of the discipline that researchers most often skip, and the skip is understandable, because a failed claim feels like a failed project. The dialectical tradition, properly understood, treats the record differently. A contradiction claim that fails its work test in the field, where the two aspects turn out to coexist stably, or where the predicted strain is traced to an external shock, has produced knowledge of exactly the kind the design was built to produce: it has located the boundary of a mechanism. The record of such a failure should state what was claimed, what was observed, which setting or relation was revised in consequence, and what the revised claim now predicts. Kept in that form, the failure record becomes one of the most citable parts of a study, because other researchers working the same territory need the map of dead ends as much as they need the route that survived. A design that hides its failures forces every successor to repeat them at full price. There is also a colder, professional reason to keep the record. Examiners and referees trust a study whose author can narrate a defeated expectation precisely, because the narration proves the expectations were real enough to be defeated. Vague methods survive everything and convince nobody.

What makes a contradiction claim more than a metaphor?

A contradiction claim is more than a metaphor when its two aspects are internal to one named unit, when each requires and undermines the other through the unit’s ordinary functioning, and when the claim states in advance the observations that would defeat it. A claim that can absorb every outcome describes the researcher’s vocabulary, not the object under study.

Consider the difference in practice. A draft announces that a city’s housing system is riven by the contradiction between housing as home and housing as asset. As it stands, the sentence is a mood. It becomes a claim when the design names the unit, say the private rented sector of one city across two decades, specifies the mutual requirement, in that landlords’ returns depend on tenants’ capacity to pay from wages while asset appreciation depends on scarcity that wages cannot match, and lists the symptoms the mechanism should produce if real: rising arrears alongside rising prices, lengthening tenancies alongside rising evictions, political pressure for controls arriving from constituencies that also hold housing wealth. It becomes research when the design also states what would embarrass it: stable affordability alongside appreciation, borne by rising real wages, which would show the two aspects coexisting without the asserted strain, or strain traced wholly to a planning constraint, which would relocate the mechanism. None of this guarantees the claim is true. It guarantees the claim is a claim, which is the precondition of finding out.

The same discipline exposes the most common counterfeit, the contradiction between a thing and its opposite number imported from outside. A hospital is not in contradiction with a budget cut; it is damaged by one. A university is not in contradiction with a government; it may be cut, regulated or captured by one. External blows belong in a design as conditions, and conditions matter enormously, but the dialectical claim is reserved for the tensions the object generates by being what it is. The reservation is what keeps the concept discriminating. A vocabulary in which every difficulty is a contradiction can no longer distinguish the hospital that must be redesigned from the hospital that must merely be refinanced, and the practical stakes of the distinction, for researchers who advise, teach or organise, are not small.

A worked research design carried through the settings

Abstractions about method persuade nobody for long, so this section carries one complete design through every element of the apparatus, in prose, at the length a real proposal would require. The design is a design. It is presented as a plan with expectations and failure conditions, not as a set of findings, and nothing in it should be read as a report of completed research. The topic is the casualisation of university teaching in one country: the long shift by which a growing share of undergraduate teaching came to be delivered by staff on hourly, fractional or fixed term contracts rather than by permanent post holders. The topic is well chosen for demonstration precisely because it tempts every failure mode the method exists to prevent. It invites moral denunciation in place of mechanism. It invites altitude slippage between the individual teacher’s experience and the structure of a national system. It invites the word dialectical to be sprinkled over a conventional labour process study. Watch, instead, what the apparatus does to it.

The unit of analysis is named first, before any forces are listed. The unit is the employment relation between a university and the teachers it engages to deliver undergraduate teaching, considered as a structured relation with two poles, the institution that must deliver a fluctuating volume of teaching from a fluctuating income, and the qualified teacher who must assemble a livelihood from engagements that no single institution guarantees. The unit is not the individual casual teacher, whose biography would pull the study toward the level of the unique, and not the national academic workforce, whose aggregation would dissolve the relation into stocks and flows. The relation is the unit because the claims to come are claims about what each pole requires of the other and does to the other, and only a relational unit can host them. Around the central relation the design places its conditions: the funding settlement for higher education, the supply of doctoral graduates, the student demand cycle, the regulatory audit of teaching quality. Conditions are studied as conditions, weighed and evidenced, but the contradiction claim is not allowed to hide among them.

The settings follow, and each is set against its live alternative. Extension is set at the higher education system of one country across roughly four decades, from the period when casual teaching was a marginal supplement to the period when it became a structural feature, with the endpoint defined not by the calendar but by a condition: the point, if it arrives, at which the relation stabilises into a recognised two tier profession with its own career structure, which would mark the object’s transformation into something else. The alternative extension, one university across one decade, was rejected for a stated reason: the mechanism under investigation operates through a national labour market and a national funding rule, and a single institution study would record the mechanism’s local weather while mistaking it for climate. Level of generality is set at the historically specific form, the marketised mass university of the late twentieth and early twenty first centuries, rather than at wage labour in general, because the design wants mechanisms a policy could touch, while a standing instruction in the analysis plan requires every general concept used, reserve of labour, flexibility, insecurity, to be cashed out at the specific level before it enters a conclusion. Vantage point is set primarily at the pole of the hourly paid teacher, for a declared reason: the relation’s costs are least visible from the institution’s accounts, where they appear as savings, and most visible in the household budgets where they land. The design compensates for the standpoint’s blindness by building the institution’s view into the evidence plan as a structured counterweight: finance office documents, workload models and managers interviewed explicitly as occupants of the other pole, their accounts treated as evidence about the relation rather than as the truth or the enemy of it.

The contradiction claim can now be stated in its disciplined form. The claim: within the employment relation as constituted, the institution requires a teaching workforce that is simultaneously disposable and committed. Disposable, because teaching volume fluctuates with enrolment and income, and the financial survival of the institution under its funding model requires that teaching capacity be expandable and contractible at short notice without fixed cost. Committed, because the same institution sells, and is audited on, educational quality that depends on teachers who know the students, develop the curriculum across years, and invest unpaid effort in preparation and feedback, investments that rational workers make only in futures they expect to share. Each requirement is generated by the institution’s ordinary functioning under its conditions, fee income, audit, expansion. Each requirement undermines the conditions of the other: disposability, pursued consistently, destroys the continuity on which the audited quality rests, while commitment, honoured consistently, recreates the fixed cost structure the funding model cannot carry. The claim predicts observable symptoms. It predicts a permanent management literature about engaging and valuing the casual workforce, produced by the same offices that engineer its disposability. It predicts quality assurance systems that measure the traces of commitment, feedback speed, availability, documentation, with increasing precision as the substance becomes harder to secure. It predicts periodic crises in which the relation’s strain becomes public: a marking boycott, a recruitment failure in a shortage subject, an accreditation warning. And it predicts attempted repairs that relocate rather than resolve the tension: teaching only contracts that formalise disposability while renaming it, or fractional posts that extend commitment to a fraction of the workforce while deepening the two tier structure the claim describes.

The failure conditions are stated with the same care, because the claim was built to be able to lose. The claim fails, first, if the evidence shows disposability and commitment coexisting without strain across a full enrolment cycle: departments sustaining audited quality through stable teams of long serving fractional staff whose security in practice matches their formal insecurity not at all, which would show the formal contract to be a poor guide to the relation and the contradiction to be an artefact of paperwork. It fails, second, if the strain, where found, traces wholly to a removable condition, a single funding formula whose reform abolishes the tension while leaving the relation otherwise intact, which would demote the claim from a contradiction of the relation to a critique of a policy. It fails, third, if the direction of dependence proves one way: if institutions are shown to secure commitment without cost by non contractual means, prestige, vocation, the hope of permanence operating as a complete substitute for security across whole careers, then the mutual undermining is broken and what remains is exploitation of hope, a finding worth having and a different study. Each failure, notice, leaves the researcher with a result. That is the mark of a design whose claims were real.

The ascent schedules the study’s construction. Stage one works with the thinnest abstraction, the contract itself, analysed from documents: the standard forms, the handbooks, the stated rates and the hours they imagine. Stage two adds the labour market determination: the flow of doctoral graduates relative to permanent posts, reconstructed from published series, which converts the contract from a legal form into a market outcome. Stage three adds the institution’s income model, the dependence of teaching budgets on enrolment, which converts the market outcome into a managerial strategy. Stage four adds the audit determination: the quality regimes that make commitment measurable and therefore make its absence expensive, which is where the contradiction claim acquires its second pole in documentary form. Stage five admits the state in its several roles, funder, regulator, sponsor of expansion, and stage six, the concrete, reconstructs named departments in named institutions across the period, with the relation’s career traced through their staffing decisions, their crises and their repairs. Fieldwork is distributed by stage. The early interviews test stage one descriptions against lived contracts. The later interviews, with managers, union officers and the teachers who crossed from one pole to the other by winning permanence, are aimed at the determinations, and the interview schedules differ by stage because the questions the ascent requires differ. The stopping rule is declared: the ascent ends when additional determinations stop altering the contradiction claim, and the design names its expectation that departmental culture, however fascinating, will be the first candidate for that cut.

Presentation is planned separately from inquiry, and the plan is stated so a reader can later check the performance against it. The inquiry will proceed as opportunity dictates, archives and access first, elegance nowhere. The written study will open with the contract, rise through the determinations in the order of the ascent, state the contradiction claim only when both of its poles have been built, and give the failure record a section of its own rather than a confessional footnote. The literature will be reviewed by settings, grouping existing studies according to their extension, level and vantage point, so that the review demonstrates the article’s central contention, that apparent factual disagreements about casualisation largely encode framing decisions, on the body of the field itself. A design that survives its own worksheet, and a review that performs its own method, have between them done something no invocation of dialectics could do: they have made the method visible enough to be refused, which is the only form in which it deserves to be accepted.

A doctoral researcher adapting this design to another object, a hospital’s nursing establishment, a city’s private rented sector, a platform’s courier fleet, should steal the sequence and not the content. Name the relation before the forces. Set the three settings against live alternatives and record the rejections with their reasons. State the contradiction claim with its symptoms and its failure conditions in the same paragraph. Schedule the ascent and its stopping rule. Plan the presentation as a separate act with its own obligations. Then run the work test over the proposal, deleting the ceremonial vocabulary and repairing whatever collapses. The proposal that emerges will be shorter than the one that went in, and it will be the first version that an examiner, a referee or a funder can actually argue with, which is to say the first version that is research rather than positioning.

Retroduction and the critical realist inheritance

If the settings audit is the method’s discipline of framing, retroduction is its discipline of inference, and it arrives from the critical realism associated with Roy Bhaskar. The reasoning pattern can be stated without ceremony. A researcher encounters a phenomenon that is regular enough to need explaining and puzzling enough that surface description will not explain it: a department that cannot retain staff despite paying at the market rate, a reform that improves every measured indicator while the service visibly deteriorates, a workforce that reports satisfaction in surveys and strikes in the same season. Retroduction asks what the world would have to be like, what structures and mechanisms would have to exist and be operating, for the phenomenon to be possible. The researcher constructs a model of the underlying structure, derives from the model what else should be observable if the structure is real, and returns to the evidence to look. The movement is neither deduction from theory nor induction from data. It is the disciplined construction of a candidate depth, followed by the exposure of that candidate to whatever the surface can be made to disclose.

Bhaskar’s schema, first set out in A Realist Theory of Science in 1975 and extended to the social sciences in The Possibility of Naturalism in 1979, distinguishes the real, the structures and mechanisms that exist whether or not they are observed, from the actual, the events that occur when mechanisms combine, and from the empirical, the subset of events anyone experiences or records. The distinctions matter to a dialectical researcher for a practical reason: in open systems, where many mechanisms operate together and interfere, a real mechanism may produce no regular empirical pattern at all, because its effects are cancelled, deflected or masked by others. The absence of a correlation is therefore not, by itself, the absence of a mechanism, and the presence of a correlation is not, by itself, the presence of one. This is the philosophical licence, carefully limited, for the dialectical habit of looking beneath patterns rather than worshipping them. The limit matters as much as the licence. Retroduction does not certify the structures it imagines. It proposes them, and the proposal then owes the ordinary debts of research: independent evidence for the structure, evidence that the structure was operating in the case, and an account of the rival structures that would produce the same surface. Bhaskar’s later Dialectic: The Pulse of Freedom, published in 1993, attempts to rebuild dialectics itself on realist foundations, and whatever a reader makes of that large construction, the earlier schema remains the most serviceable bridge between dialectical talk of depth and the evidential conscience of working social science.

Andrew Sayer’s Method in Social Science, first published in 1984 and revised in 1992, performs the same bridging work from within methodology, and for many researchers it is the single most useful book named in this article. Sayer insists on the difference between intensive and extensive research designs: extensive designs map patterns across many cases and answer questions about distribution and regularity, while intensive designs examine mechanisms within few cases and answer questions about how a process works in context. Neither is superior. They answer different questions, and the chronic methodological quarrels of social science often reduce to one design being graded by the other’s standards. A dialectical study, in the terms developed here, is typically intensive in its core and extensive in its supporting evidence, and Sayer gives the researcher the language to say so without apology and the standards to do both halves properly. His account of abstraction, distinguishing rational abstraction, which isolates a real and significant relation, from chaotic abstraction, which lumps unlike things under one label, belongs on the same shelf as Ollman’s settings, and the two accounts correct each other usefully: Ollman supplies the adjustable dimensions, Sayer supplies the quality standard for the concepts being adjusted.

David Harvey’s propositions about dialectics, set out in Justice, Nature and the Geography of Difference in 1996, supply a third translation, aimed at researchers working on place, environment and urban process. Harvey writes as a practitioner reporting on his own craft, and his propositions emphasise process over object, the internality of relations, the presence of the whole in each part and the reality of change driven by tensions within a process rather than by impacts from outside it. Read alongside the settings, Harvey’s list functions as a set of reminders for fieldwork in open systems: expect the object to be a moment in a process rather than a thing, expect boundaries to leak, expect the case under study to carry the marks of the larger whole that contains it, a neighbourhood bearing the imprint of a national housing settlement, a classroom bearing the imprint of a funding formula. The reminders are not a procedure, and Harvey does not pretend otherwise. Their value is diagnostic. A draft in which nothing leaks, nothing develops and no part shows the pressure of the whole is a draft whose dialectical vocabulary is very likely decorative, and Harvey’s propositions give a supervisor a vocabulary for saying exactly that, in a form a student can act on.

The three translations share a trait that should be made explicit, because it defines the honest boundary of the whole method. None of them generates evidence. Retroduction proposes mechanisms, the settings discipline framing, the propositions police drafts, and not one of them interviews a witness, cleans a dataset or dates a document. Dialectical method, on the account given here, is a method of design and exposition layered over the ordinary crafts of inquiry, and it neither replaces those crafts nor excuses their absence. A study can be impeccably dialectical in architecture and worthless in evidence, just as a study can be evidentially rich and architecturally inert. The tradition’s rhetoric sometimes obscures the layering, suggesting that the method itself confers rigour. It confers, at most, a set of questions that rigour must answer. Researchers who internalise that limit will find the method a demanding servant. Researchers who expect more of it will produce the kind of work the sceptic, in the objection section below, is waiting to quote.

Neighbours: systems thinking, process tracing and numbers

A method earns its outline by contrast with its neighbours, and dialectical research has three neighbours close enough to be confused with it in methods chapters and in referees’ reports. The first is systems thinking. A systems approach and a dialectical approach share a suspicion of isolated variables, a taste for feedback and an insistence that wholes behave in ways their parts do not predict. The differences are nevertheless real, and a researcher should be able to state them without caricaturing either side. Systems thinking characteristically seeks the structure of a system in order to model, stabilise or optimise it, and its contradictions appear as malfunctions, bottlenecks or unintended consequences to be engineered away. Dialectical research treats the tension as potentially constitutive, as the form the object’s development takes, and asks not only how the system might be steadied but whether its steadiness is purchased by exporting the tension somewhere the model does not look. The second difference concerns history. A system can be modelled at a moment, elegantly and legitimately. A dialectical object must be given a career, because its identity is bound up with its becoming. The third difference concerns the researcher. Systems practice often aspires to the position of the engineer outside the apparatus. The dialectical tradition builds the vantage point into the account and expects the researcher’s own position, funder, institution, discipline, to be among the determinations declared. None of this makes systems work inferior. It makes it a different instrument, and the design that borrows feedback diagrams while claiming dialectical depth owes its reader an account of which instrument is actually in use.

Process tracing is the second neighbour, and the kinship is closer. The process tracer follows a mechanism through a single case, testing the observable implications of each link, and the discipline of deriving expectations in advance and hunting for evidence that discriminates between rival mechanisms is one any dialectical researcher should admire and imitate. The difference lies in what each approach takes a mechanism to be attached to. Process tracing, in its standard form, works with variables and sequences, holding the identities of the actors and the setting steady while the causal chain unfolds. Dialectical research reserves the right to find that the unfolding changed the actors: that the union local which began a dispute as a wage claimant ended it as a political organisation with different interests, that the department which adopted a reform as a compliance exercise became, in the adopting, a different kind of department. Where the identities are stable, the two approaches converge and the dialectical vocabulary adds nothing, a verdict the work test will deliver honestly. Where the identities move, process tracing must either freeze them by fiat or borrow the dialectical account of internal relations to say what happened. The honest summary is that the approaches are complements with different default settings, and the design that knows which defaults it has accepted is stronger than the design that pledges allegiance to either flag.

The third neighbour is quantitative research itself, and the supposed incompatibility between dialectical method and numbers is a misreading this article is charged with correcting. Nothing in the settings, the ascent or retroduction forbids counting. What the method forbids is counting under unexamined categories, and since every statistic rests on categories, the prohibition amounts to a demand that the categories be treated as part of the research rather than as part of the furniture. A dialectical use of statistics begins by auditing the categories in the series, asking whose administrative purposes formed them, what extension and level they silently encode, which relations they render invisible by construction. It proceeds by treating breaks in a series as evidence in their own right, since a reclassified category is often the footprint of the very transformation under study. And it reports distributions alongside the mechanisms that plausibly generate them, refusing both the fetish of the coefficient and the romantic refusal to count at all. Marx’s own practice was omnivorous in this respect: the chapters of Capital lean on factory inspectors’ reports, parliamentary papers and mortality tables with an appetite any empirical researcher would recognise, and the tradition that claims him while disdaining his evidence base has some explaining of its own to do.

How can dialectical research use statistics without abandoning its method?

Dialectical research uses statistics by treating the categories inside the numbers as research objects rather than as furniture. The researcher audits who formed a category and for what administrative purpose, reads breaks and reclassifications in a series as evidence about the object itself, and reports distributions together with the mechanisms that plausibly produce them.

Worked briefly, the point about categories becomes concrete in any study of unemployment. The headline rate is not a fact of nature. It is the residue of decisions about who counts as seeking work, over what reference period, with what availability, decisions revised by statistical offices as labour markets change, so that the series contains the history of its own definitions folded inside the history it purports to record. A study that plots the rate across four decades without excavating the definitions has extended its object across a boundary that its own evidence does not recognise. The dialectical researcher does the excavation, reports the seams, and where the seams make long comparison unsafe, says so and works at the extension the evidence supports. This is slower than downloading the series. It is also the difference between using statistics and being used by them, and it generalises to every administrative category a social researcher inherits, from truancy to productivity to the casual contract whose shifting legal definition the worked design above had to track as part of its object.

There remains the question of scale, which quantitative neighbours press with justice. A method of design that specialises in intensive study must say how its findings travel. The dialectical answer is that findings travel by mechanism and by setting, not by sample. A mechanism established in one setting, with its conditions specified, licenses a disciplined expectation elsewhere: where the conditions recur, look for the mechanism, and where the mechanism is found, expect its characteristic pressures, with the expectation stated before the looking. That is a weaker warrant than statistical generalisation and a stronger one than anecdote, and its honesty consists in naming its own weakness. The settings worksheet is what makes the warrant usable, because the worksheet of the original study tells the travelling researcher exactly which features of the new setting must match and which may differ. Generalisation, in this tradition, is a comparison of worksheets, and a field that kept its worksheets could accumulate knowledge across intensive studies far faster than a field that keeps only its conclusions.

The sceptical objection, at full strength

The objection must now be given the floor without interruption, in the strongest form a fair sceptic can give it, because a method article that answers a weakened version of the charge against it has answered nothing. The sceptic speaks as follows. A method, properly so called, is a procedure that different competent practitioners can apply to the same material and be brought, by the procedure itself, into confrontation when they diverge. It can be taught as steps. It can be misapplied, and the misapplication can be shown. Dialectical method, as actually encountered in journals and theses, has none of these properties. There is no sequence of operations to teach, only a repertoire of gestures: relate, historicise, find the tension, rise to the concrete. There is no way to perform it wrongly, because any outcome can be redescribed in its vocabulary after the fact. A finding of stability becomes a temporarily contained contradiction. A finding of change becomes a contradiction unfolding. A finding that fits neither becomes a mediation. A vocabulary that accommodates every result explains no result, and the history of the tradition’s empirical claims gives the sceptic no reason to relent, since its confident predictions have been revised, deferred and reinterpreted with a suppleness that a falsifiable programme would not have survived. What the vocabulary does reliably accomplish is social rather than cognitive. It signals membership. It marks a thesis, a department, a journal submission as belonging to a tradition, with shared ancestors and shared enemies, and it performs that marking work precisely in proportion as it fails to perform any methodological work. On the sceptic’s account, then, dialectical method is ordinary good research, framing awareness, mechanism hunting, historical care, wearing ceremonial dress, plus a loyalty badge, and the ceremony would be harmless if it did not consume the time of students who could be learning the ordinary good research directly.

Parts of that indictment are simply true, and the reply begins by conceding them in the open, since a defence that starts by denying the evidence will not be believed when it reaches the evidence’s limits. It is true that no reproducible procedure exists, in the strong sense of a sequence whose faithful execution by independent practitioners converges on the same design. It is true that the vocabulary can absorb any finding, and that in a depressing proportion of published work it does exactly that, which the tradition’s own internal critics have said for decades. It is true that the word functions as a badge, in hiring, in journal cultures and in the tribal geography of conferences, and that badging is one of the things a young researcher is actually purchasing when the word is adopted early and understood late. And it is true that much of what this article has presented, the suspicion of unexamined categories, the care with framing, the hunting of mechanisms, is ordinary good research that other traditions teach under other names. A reply that cannot absorb those four concessions is not a reply. It is a recruitment pamphlet.

What remains, after the concessions, is a narrower and more defensible claim, and it can be tested section by section against the apparatus this article has built. Take the settings first. Framing awareness is indeed ordinary good research, but in ordinary practice it is dispersed across a dozen unconnected habits: a paragraph on scope in the introduction, a reflexive passage in the methods chapter, a limitation acknowledged in the discussion. The settings audit gathers those fragments into one instrument with a fixed form, three dimensions and a demand for stated alternatives and recorded changes, and an instrument with a fixed form does work that scattered awareness does not, in the same way that a checklist does work that conscientiousness does not. The sceptic may reply that a checklist is a modest thing to found a tradition on. The reply is that the tradition, at its best, founds itself on nothing grander, and that the worksheet’s modesty is the point: here, at last, is a dialectical practice that can be misapplied, since a worksheet can be filled lazily, inconsistently or not at all, and a supervisor can show the failure on one page.

Take internal relations next. Here the addition over ordinary practice is a genuine piece of theory with methodological consequences, not a slogan. The claim that some relations constitute their terms tells the researcher where the unit of analysis must sit, why individuals and their attributes cannot always be the atoms of a design, and what kind of change counts as transformation rather than variation. Ordinary good research has no equivalent commitment; it treats the unit as a pragmatic choice among levels, micro, meso, macro, and is regularly ambushed by phenomena whose logic is relational through and through. The sceptic is entitled to demand cases where the commitment changed a finding, and the tradition can supply them in the study of classes, of households, of firms in supply chains, of professions and their clients, wherever the identity of the parties is unintelligible apart from the relation the study exists to examine. Where the commitment changes nothing, the work test deletes it, and the deletion is part of the method rather than an embarrassment to it.

Take the ascent and the presentation distinction together. Ordinary research training teaches writing as the reporting of results, in the order the results arrived or the order a journal’s template imposes. The dialectical tradition teaches exposition as a second intellectual act with its own logic, the order of dependence, and the practical consequence is visible in the difference between theses that must be excavated and theses that can be followed. The sceptic may call that rhetoric rather than method. The distinction is thinner than the sceptic needs. A rule that governs the construction of the account, that can be followed well or badly, that changes what readers can check and therefore what claims can survive, is doing methodological work by any definition that is not rigged in advance to exclude it.

Take, finally, the contradiction claim under its failure conditions, because this is where the unfalsifiability charge must be met head on or surrendered. The charge is fatal to the vocabulary used loosely, and the loose use is the statistical norm in the literature, which the reply concedes without flinching. But the charge does not survive contact with the claim as this article has specified it: two internal aspects, mutual requirement, mutual undermining through ordinary functioning, predicted symptoms, and stated observations that would defeat it. Such a claim can lose, in specified ways, on evidence a researcher can actually gather, and the worked design above was constructed to display the losing conditions as prominently as the predictions. The sceptic’s strongest response at this point is to note that the specification, not the tradition, did the work, that any mechanism claim in any tradition can be disciplined in the same way. That response is correct, and it is also the verdict this article announced at the start: the method, disciplined, is a method of a narrow and defensible kind, a discipline for constituting objects, stating relational claims and ordering expositions, which must marry ordinary evidential craft to live. What the tradition adds to that craft is the audit, the relational unit, the ascent and the test. What it does not add is exemption. A researcher who needs the exemption should be refused it by referees, and a researcher who does not need it will find the additions, modest as they are, worth the price of the vocabulary.

One further concession completes the reply, concerning the guarantee the method refuses to give. Dialectical framing tilts a study toward expecting tension, development and transformation, and a tilt is a bias when the object in fact sits still. Some objects do sit still, for long periods, for reasons that are themselves interesting: settlements locked in by law, professions protected by closure, routines sustained by the sheer cost of changing them. A method that cannot see stability except as frozen motion will misdescribe such objects, and the settings audit contains no automatic correction for the tilt. The correction has to be supplied by the researcher’s honesty in the failure record, where a claim of unfolding contradiction that met a decade of quiet must be written down as met and quiet. Traditions are not responsible for their slogans, but they are responsible for their characteristic temptations, and this one, the temptation to find drama, belongs in the methods chapter of every study the tradition produces, named as plainly as any other limitation.

Failure modes, and how the method appears in a finished study

The characteristic failures of dialectical writing are worth cataloguing in diagnostic form, because a researcher who can recognise them in a draft can repair them while repair is still cheap. The first is the vocabulary transfer, in which dialectical, contradiction and mediation replace the ordinary words complex, problem and connection throughout a text whose research operations are entirely conventional. The work test exposes it in an hour. The second is the eternal return of the same paragraph, in which every section of a study concludes that the phenomenon is contradictory, historically produced and internally related, in rotating synonyms, so that the findings chapters add bulk without adding information. The repetition scan of the author’s own conscience catches it: if the conclusions of successive chapters can be exchanged without loss, the chapters were not findings. The third is the unowned setting, in which a design drifts between extensions or vantage points as the evidence dictates, always arriving where the material is richest and never recording the journey, so that the study’s object quietly changes identity between chapters. The worksheet, kept and dated, is the specific remedy. The fourth is the contradiction of convenience, in which the two poles of an alleged contradiction are selected because they are the two things the researcher dislikes most about the object, rather than because each demonstrably requires the other. The mutual requirement clause is the remedy, and it is a hard clause to satisfy dishonestly, since demonstrating that a funding model requires both disposability and commitment, in the worked example, took evidence about audits, enrolment and income rather than adjectives. The fifth failure is the missing failure, the study in which everything predicted occurred and nothing surprised, which experienced researchers read the way auditors read accounts without a single correction: as evidence that the record has been tidied. A study with no recorded surprise has usually not had expectations definite enough to be surprised by, and the failure record is the proof of definiteness.

Set against the failures, the marks of the method in a finished study are equally recognisable, and a reader who wants to learn the craft should learn to spot them in the work of others before attempting them. The settings are visible, stated early and used later, so that the conclusions carry their framing on their face. The unit of analysis is a relation or a structured whole where the argument needs one, and the text can say why. The contradiction claim, where one is made, arrives with its evidence and its scars, the conditions under which it would have been abandoned and the places where it nearly was. The exposition ascends: the early chapters are visibly thinner than the later ones, and the concreteness of the ending feels earned rather than asserted. The literature is organised by the framings of the studies reviewed, so that the review is itself an analysis. And the prose uses the tradition’s vocabulary sparingly, at the joints of the argument, where the deletion test would draw blood. A study with those marks could delete the word dialectical from its title and lose nothing of its method, which is the deepest sense in which the work test can be passed.

The literature review deserves its own paragraph, because it is where doctoral researchers most visibly practise or fail to practise the method, and because a dialectical review is a teachable artefact. The conventional review proceeds by summary: one study per paragraph, findings recited, a gap announced at the end like a vacant plot. The settings based review proceeds by grouping. Studies are sorted by their extensions, and the review shows how the field’s object has expanded or contracted across its history. They are sorted by level, and the review shows which altitudes are crowded and which are empty. They are sorted by vantage point, and the review shows whose experience the field has treated as the object’s face and whose it has treated as noise. The gap, when it is announced, is then a specific absence: no study at this extension from that vantage point, no mechanism traced at the intermediate level, and the candidate’s own design enters as the filling of a located absence rather than as a claim that nobody has studied the topic, a claim that is false of every topic worth a doctorate. Reviewers and examiners respond to the located gap because it demonstrates command of the field’s structure rather than of its contents page. Students preparing the groundwork for such a review often assemble the reading that feeds it from earlier examination material and topic guides, and there is a practical sense in practising with previous-year questions on ReportMedic as a way of discovering which framings examiners reward, before the review commits its author to one. The instrument is an examination aid, but the habit it trains, reading a question for the settings it presupposes, is the same habit the review requires.

The research question itself can be anatomised in the same terms, since students ask, reasonably, what a dialectical research question looks like as opposed to an ordinary one wearing the word. The honest answer is that the grammar differs less than the commitments behind it. An ordinary question asks what factors influence an outcome, and it imagines the world as items and influences. A dialectical question, in the disciplined sense, names a relation and asks about its career: how the relation between an institution and its casual workforce reproduces itself, what pressures its ordinary functioning generates, and what forms its transformation might take. The factors question can be answered by a list with weights. The relational question can only be answered by an account, with a unit, a mechanism and a history, and its sub questions fall out of the apparatus: one for each setting, one for each pole of the contradiction claim, one for the failure conditions. A supervisor who is handed the relational question should still ask the factors question’s rude little sibling, namely what evidence could possibly be gathered, by this student, in the time available, that would bear on the answer. Dialectical ambition has wrecked more doctorates through scale than through error, and the settings worksheet, honestly filled, is an early warning system for scale, since the extension cell and the fieldwork plan sit on facing pages and cannot both be fantasies.

Publication raises the last practical question, and it can be answered without romance. Dialectical research is published in mainstream journals whenever its evidential craft meets the journal’s standards and its framing decisions are stated in terms the referees can engage, which increasingly means stating them as decisions about scope, level and standpoint, exactly the vocabulary the settings supply, with the tradition named as provenance rather than brandished as authority. What referees resist, and resist rationally, is the unaccompanied vocabulary: the submission whose contribution is said to be a dialectical perspective, whose method section performs the allegiance and whose findings would be unchanged under any other flag. The remedy is the one this article has applied throughout. Lead with the decisions and the claims. Let the tradition appear in the citations and in the one paragraph where its specific commitments did specific work. Journals reward findings that travel, and a mechanism with stated conditions travels first class.

The tradition practised: inquiry before it was theorised

The method’s components were practised before they were named, and two episodes from Marx’s own research life repay study by anyone who suspects the apparatus of being a retrospective invention. The first is the questionnaire he composed in 1880 for a French workers’ journal, a hundred questions addressed to workers themselves about the concrete determinations of their working lives: hours, wages, the division of the day, the fines, the meals, the state of their health, the behaviour of overseers. The document is the method of inquiry in its purest surviving form. It does not ask workers what they think about capitalism in general. It appropriates the material at the lowest level of generality, in the categories of the working day as lived, on the stated theory that the concrete knowledge of the relation must be built from its determinations and cannot be deduced from its concept. A researcher designing an interview schedule can still learn from its architecture: begin where the respondent’s knowledge is exact and daily, ascend toward the relations the respondent may never have had occasion to name, and let the schedule’s order mirror the ascent the analysis will later perform. The second episode is the construction of Capital itself, whose inquiry filled decades of notebooks with excerpted statistics, reports and histories, and whose presentation discarded the order of that accumulation entirely, opening instead with the commodity. The gap between the notebooks and the book is the largest worked example in the tradition of the distinction between inquiry and presentation, and the notebooks’ survival is a standing argument for keeping the inquiry record intact: without them, later scholarship could not have audited the presentation, revised its judgments or understood which of its claims rested on evidence its author had gathered and which on evidence its author had hoped to find.

The episodes also correct a flattering error about the tradition’s history, the error of supposing that the method descended from philosophy into research as an application. The traffic ran the other way. The research practice, voracious, archival, quantitative where quantities existed, came first and kept correcting the philosophy, and the methodological texts, the 1857 introduction above all, read as reflections by a practitioner on his own unfinished work rather than as instructions issued from a system. This is why the strongest pages in the secondary literature on method are so often written by practitioners reporting on a craft, Harvey on geography, Sayer on the design of social inquiry, Ollman on the teaching of a course, rather than by commentators on the doctrine. A doctoral student deciding what to read should weight the shelf accordingly, and the reading route below is built on that weighting. It is also why the neighbouring series article on researching Marxism through its sources complements this one: that article owns the question of how the tradition itself is researched, its editions, archives and apparatus, while this article owns the transferable procedure, and a researcher working on any topic at all needs the procedure more than the provenance.

There is a third episode, quieter than the other two, that belongs in a methods article because it concerns the revision of settings under the pressure of evidence. Marx’s late studies, the notebooks on communal landholding and the drafts of his reply to Vera Zasulich in 1881, show a researcher widening the extension of his object and lowering the confidence of its most general claims in response to material that did not fit the earlier framing: the Russian commune, examined closely, refused to behave as a mere survival awaiting dissolution, and the drafts multiply qualifications accordingly. Whatever position a scholar takes on the substance, the methodological spectacle is the one this article has been recommending throughout. Settings were changed, in writing, because the evidence rebelled, and the changes were recorded in drafts that later readers could inspect. The worksheet is nothing but that spectacle, routinised and given a table.

Keeping the record: a protocol for the settings

Because the apparatus lives or dies in paperwork, it is worth stating the paperwork as a protocol a researcher can adopt on the first day of a project and maintain in an hour a month. The protocol has five artefacts. The first is the worksheet itself, dated at its first completion and amended in dated layers thereafter, never silently overwritten, so that the history of the framing survives as data about the research. The second is a unit statement, one page at most, naming the unit of analysis, the relations treated as internal to it with the argument for each, and the relations demoted to conditions, with the demotions argued too, since a demotion is as much a decision as a promotion. The third is the claim sheet: the contradiction claim where one exists, written with its predicted symptoms and its failure conditions, and beneath it, in the same document, any rival mechanism that would produce the same symptoms, because a claim without its rival on the page invites the researcher to gather only friendly evidence. The fourth is the failure record, which begins empty by design and is expected to fill: every defeated expectation, every archive that refused the framing, every interview that embarrassed a setting, entered briefly and without self pity at the time it happens. The fifth is the presentation plan, drafted early and revised late, which states the order of dependence among the study’s concepts and therefore the order of the chapters, so that the difference between the plan and the eventual book is itself a finding about what the inquiry taught.

Maintaining five artefacts sounds like bureaucracy until the alternative is priced. The alternative is the third year discovery that the thesis’s chapters presuppose three different objects, the supervisor’s question in a progress review that cannot be answered from any notebook, the referee who asks why the study’s population changed definition between the second and fifth chapters. Each artefact answers one of those disasters in advance. The artefacts also have a social function that should not be underestimated. A doctoral researcher who can hand a supervisor a worksheet, a unit statement and a claim sheet has converted a relationship of general encouragement into a relationship of specific criticism, and specific criticism is the scarcest resource in postgraduate education. Supervisors cannot fault what they cannot see. The protocol makes the design visible at the stage when faulting it is a kindness. The same visibility serves the researcher in funding applications, in ethics review, where the vantage point declaration does real work about whose experience is being gathered and why, and in collaboration, where most interdisciplinary friction is, on inspection, friction between unannounced settings: the economist’s extension against the ethnographer’s, the historian’s level against the sociologist’s, each party experiencing the other as careless when both are merely set differently and in silence.

Two habits keep the protocol alive across a long project. The first is the monthly settings review, half an hour in which the researcher rereads the worksheet against the month’s evidence and writes one paragraph, kept, on whether the settings still fit. Most months the paragraph records no change, and its value is the record of the check rather than the discovery of a problem. The second habit is the deletion pass, run over any substantial draft before it leaves the desk: the work test applied to the ceremonial vocabulary, the sentence openings scanned for the chapter concluding formula, the conclusions of successive sections compared to ensure they could not be exchanged without loss. Neither habit requires talent. Both require the belief, which the tradition at its best embodies, that method is mostly the refusal to let decisions happen by default, repeated at intervals until the project ends.

What to read, and in what order

The reading route for this method is short by design, because the literature that teaches operations is small and the literature that celebrates vocabulary is endless. Begin with the primary methodological text, the introduction Marx wrote in 1857, available as the opening of the Grundrisse, and read it slowly for the two moves this article has built on: the separation of gathering from exposition, and the climb from thin abstractions to the concrete whole, noting how compressed the original is compared with the edifices later raised on it. Read Ollman next, Dialectical Investigations for the philosophy of internal relations and Dance of the Dialectic for the process of abstraction with its three settings, since the settings are most fully presented there, worked through teaching examples. Read Sayer’s Method in Social Science alongside, as the realist conscience of the enterprise, for intensive and extensive design, for rational as against chaotic abstraction, and for the discipline of mechanisms in open systems. Read Bhaskar selectively: the early realist schema of the real, the actual and the empirical, and the account of retroduction, before attempting the later dialectical constructions, which divide readers and are not needed to use the inference pattern. Read Harvey’s propositions in Justice, Nature and the Geography of Difference as a practitioner’s coda, a list to pin above the desk rather than a text to summarise. Levins and Lewontin’s The Dialectical Biologist, published in 1985, deserves its place as the worked example from another discipline entirely, a demonstration that the method’s commitments about parts, wholes and becoming can organise a research programme in evolutionary biology, written by practitioners who had to survive referees in a field with no patience for ceremony. End with a return to a piece of finished research in the reader’s own field, chosen for its reputation, and audit it against the worksheet: find its settings, its unit, its claim and its failure conditions, stated or not. The audit of another scholar’s masterpiece teaches more in a week than a month of methodological exhortation, not least because it reveals how rarely even the masters state their settings, and therefore how much room a new researcher has to be clearer than the canon.

Citation practice for this literature carries one warning that experience justifies. The tradition’s key terms are translation sensitive and edition sensitive, and a researcher citing the 1857 introduction or the abstraction literature should cite by the structural location, the work and the section, as well as by the page of the edition used, since pagination varies across the editions in circulation and a reader following a bare page number into a different edition will arrive at the wrong passage and conclude, unfairly, that the citation was decorative. Decorative citation is, in this field, a real and common scholarly fault, and the habit of citing locations that survive translation is the simplest protection against committing it. The same care applies to the secondary canon: Harvey’s propositions and Sayer’s distinctions are frequently cited at second hand in forms their authors would not recognise, and the researcher who reads the originals acquires, as a by product, the ability to spot the counterfeits in other people’s bibliographies, a small skill with a long professional half life.

Case selection, sampling and the settings

Design questions that sound purely technical, which cases to study, how many, chosen how, change their meaning when they are asked downstream of the settings, and dialectical research has answers to them that differ from the textbook defaults in instructive ways. Case selection in an extensive design seeks representativeness, so that patterns in the sample can be projected onto a population. Case selection in a dialectical design seeks something else: the case in which the relation under study is most fully developed, or the pair of cases whose settings differ in one dimension while matching in others, so that the effect of the differing setting can be watched. Marx’s choice of England as the classic ground for studying the capitalist mode of production was a case selection of exactly this kind, made on the argument that the relation appeared there in its most developed and least disguised form, with the explicit warning that the findings would arrive elsewhere in locally altered shapes. The logic is defensible and its condition is honesty: a case chosen for development must not then be described as typical, and the worksheet should record which logic governed the choice. The deviant case has a special role in this scheme, because a case that refuses the design’s framing is evidence about the settings in a way that a confirming case can never be. Where an extensive design treats the outlier as noise to be trimmed, the dialectical design budgets time for it in advance, on the expectation that the object is most likely to disclose its real boundaries at the point where the researcher’s boundaries fail.

Sampling inside the case follows the unit rather than the population. If the unit is a relation, the sample must cover both poles of it and the mediations between them, which is why the worked design interviewed managers as occupants of the institution’s pole rather than as informants about the teachers’ world. A sample drawn from one pole and generalised to the relation is the single most common empirical fault in relational research, and it survives every statistical sophistication applied to it afterwards. Saturation, the qualitative researcher’s stopping signal, also changes its sense under the settings: a dialectical study is saturated not when interviews stop adding themes but when they stop adding determinations that alter the central claim, which ties the fieldwork stopping rule to the ascent’s stopping rule and gives both a rationale that can be written into a methods chapter without embarrassment. Time, the scarcest resource in any project, is allocated differently under this logic. Less of it goes to enlarging the sample in the directions the claim does not need. More of it goes to the second pole, the hostile archive, the deviant case and the failure conditions, the places where the design can lose. A researcher defending that allocation to a funder or a supervisor is not indulging a taste for difficulty. The allocation is the method, expressed as a calendar.

The audit applied to a field: reading other people’s settings

The settings apparatus does its most immediate service not on the researcher’s own design but on the literature the researcher inherits, because a field, seen through the worksheet, resolves from a crowd of competing findings into a structured space of framing decisions, many of them never announced by their authors. Performing the audit on a field means reconstructing, for each significant study, the extension it actually used, which is often narrower than the extension its title advertises, the level at which its operative concepts worked, which is often lower than the level of its conclusions, and the vantage point its evidence embodies, which is often the vantage point of whichever institution kept the records. The reconstruction is an interpretive act and should be flagged as one, with the evidence for each attribution given. Done fairly, it transforms the researcher’s relation to the canon. Studies that appeared to refute one another are revealed as studies of differently constituted objects, each accurate within its settings, and the field’s apparent chaos becomes a map on which the unoccupied positions are visible. Studies that genuinely conflict, settings matched and findings opposed, stand out with a clarity they never have in a summary based review, and those are the conflicts a new project should attach itself to, because they are the ones evidence can settle.

The audit also protects against the most seductive error in interdisciplinary work, the importation of a finding across a settings boundary without noticing the crossing. A result established at the level of individual psychology, about how workers experience insecurity, does not transfer unaltered to the level of the labour market, where insecurity functions as a discipline on workers who never experience it directly. A result established within one country’s extension, under one funding settlement, does not transfer to another’s without an argument about which determinations carried it. Journals are full of discussions in which a finding is criticised for failing in a setting it never claimed, and of defences that quietly relocate the claim to whichever setting the criticism cannot reach. A researcher trained to keep worksheets, their own and other people’s, becomes difficult to fool in either direction, and develops the rarer skill of stating, for any admired study, exactly the territory over which its authority runs. That skill, more than any vocabulary, is what senior scholars mean when they describe a junior colleague as having methodological maturity, and it can be acquired deliberately, study by study, with the worksheet as the training instrument.

Historians will recognise a version of the audit in their own discipline’s practice, where the criticism of sources has always involved reconstructing the position from which a document was produced and the extension of the world its author could see. The dialectical contribution is to generalise that source criticism into a design criticism, applied to modern studies and to one’s own project with the same severity. Ethnographers will recognise the vantage point setting in their discipline’s reflexive turn, and can use the worksheet to give reflexivity a forward looking form: not a confession appended to the findings, but a setting declared in advance, with its compensations designed rather than improvised. Interview based researchers will recognise the ascent in the discipline of the schedule that moves from the daily and exact toward the relational and abstract. The method, in this light, is less an import into the established crafts than a common grammar for decisions the crafts already make separately, and its claim on a researcher’s loyalty should be assessed in exactly those terms: not as a worldview competing with the crafts, but as the grammar that lets their decisions be compared, criticised and improved in one another’s company.

A design clinic: three flawed proposals repaired

Method is learned fastest from repairs, so this section takes three flawed proposals of the kind supervisors meet every term and shows the apparatus working on each. The flaws are composites, drawn from the standard failure modes rather than from any identifiable student, and the repairs are deliberately unglamorous, because the distance between a weak design and a strong one in this tradition is usually a page of decisions, not a new worldview.

The first proposal announces a dialectical study of resistance in a logistics warehouse. Its evidence plan is interviews with workers, its unit is never named, and the word resistance does duty for everything from organised strikes to jokes in the canteen. The repair begins with the unit. Set the unit as the relation between the warehouse’s labour process, with its measured pace and its algorithmic allocation of tasks, and the workforce engaged to perform it, and resistance acquires a definition with edges: conduct that contests the terms of that relation, distinguished from coping, which accommodates them, and from exit, which abandons them. The settings follow. Extension is set at the warehouse and the agency network that supplies it, since the relation runs through the agency’s door. Vantage point is declared at the pole of the pickers, with the supervisors interviewed as the other pole rather than as commentators. The contradiction claim can then be attempted: the labour process requires a pace that the human body, its other indispensable input, cannot sustain indefinitely, so the process generates the fatigue, the turnover and the informal pace negotiations that management then polices as misconduct. The claim has symptoms to hunt and failure conditions to fear, among them the possibility that the pace is comfortably sustainable and the discontent traces entirely to pay, which would make the study a wage study and a better one for knowing it. The proposal that returns from the repair is recognisably the same project, with its topic, its site and its sympathies intact. What has changed is that every noun in it can now be interrogated, and the project can lose.

The second proposal is an archive based history of a strike wave, framed at the level of the working class of an entire nation, evidenced from two local newspapers and a union journal, and concluding in advance that the strikes expressed a rising class consciousness. The flaws are altitude and verdict. The evidence lives at the level of two towns, the conclusions at the level of a class, and the central interpretive term, consciousness, has been decided before the archive is opened, so the inquiry can only illustrate it. The repair lowers the claims to the evidence or raises the evidence to the claims, and given a student’s calendar it lowers the claims. Extension is reset to the two towns and the industries that connect them. The level is reset to the historically specific form of those industries in that decade. Consciousness is demoted from conclusion to question, and given observable proxies that the sources can actually bear: the language of the strike leaflets as between economic and political demands, the durability of the organisations the strikes left behind, the behaviour of the strikers at the subsequent election. The contradiction claim, if one survives the demotion, concerns the local unions, organisations built to negotiate within an industry and repeatedly required by their members to act against it. The repaired history is smaller and vastly more defensible, and its author has learned the transferable lesson, that a claim’s level is a budget constraint, set by the evidence that can be gathered, and not by the importance of the topic.

The third proposal is quantitative: a study of whether union membership protects workers from wage stagnation, using a national survey, framed in a methodology chapter that calls the design dialectical because work is a social relation. The framing sentence is decoration and the deletion test removes it without trace, which the repair accepts as the starting point. What the apparatus can genuinely add to this design is work at the categories and the settings. Union membership in the survey is a binary inherited from administrative convenience, while the relation the researcher cares about has determinations the binary flattens: membership in a workplace with bargaining coverage, membership without it, former membership, membership in a union in dispute. The repair rebuilds the variable from the survey’s own sub questions where the data allow, states the level of the claim as the labour market of one country in one period rather than employment in general, and adds the mechanism the distribution will be asked to testify about, so that the results, when they arrive, are evidence in an argument about how protection works, where it works, rather than a coefficient in search of a paragraph. The word dialectical may or may not return to the methodology chapter at the end of the repair. If it returns, it returns attached to the category work and the settings, where the deletion test can find it doing something. If it does not return, the study has lost nothing but a badge, and the researcher has learned the work test in the only way it is ever learned, on a draft of their own.

Working an archive under the settings

Archives are where the settings stop being a proposal’s furniture and become daily decisions, because an archive is itself a constituted object, assembled by an institution for its own purposes, and the researcher who forgets that will mistake the archive’s vantage point for the past’s. The extension question arrives first and in disguise. A record series begins and ends for administrative reasons, a department created, a form redesigned, a retention schedule enforced, and the researcher who adopts the series’ boundaries as the object’s boundaries has let a filing decision set the temporal frame of the study. The dialectical habit is to ask what the object was doing before the series opens and after it closes, and to budget the awkward work of bridging the seams, through parallel series, through the records of neighbouring institutions, through the trade press and the memoir, rather than allowing the seam to masquerade as a turning point. Many a celebrated discontinuity in the historiography of an industry or a movement is, on excavation, a change in record keeping wearing the costume of an event, and the settings audit, which demands that the extension be chosen rather than inherited, is the specific protection against manufacturing such events.

The vantage point question arrives in the archive wearing a different disguise, as the catalogue itself. Records survive in the categories of the bodies that produced them, minuted in their language, filed under their headings, and a researcher who works the catalogue uncritically will reproduce the producing institution’s view of the object with documentary authority. The counterweight is a deliberate search for the records of the other pole: the inspected as against the inspectors, the applicants as against the committee, the workforce as against the board, and where the other pole kept no archive, as subordinated parties often did not, the absence belongs in the study as a finding about the relation rather than as a gap to be apologised for in a footnote. Reading along the grain of an archive tells the researcher what an institution thought it was doing. Reading against the grain, in the margins, the appendices, the disciplinary files and the rejected applications, tells the researcher what the institution’s categories could not hold. A dialectical archive practice schedules both readings and keeps them distinguishable in the notes, because the eventual presentation will need to know which of its materials testify to the relation and which testify to the relation’s official self image, and confusing the two testimonies is how institutional propaganda survives peer review disguised as evidence.

Level of generality does its archival work at the moment of note taking. The temptation of the reading room is accumulation at the level of the striking instance: the vivid minute, the quotable letter, the outrageous statistic, each copied with delight and none connected to the rungs above. The ascent supplies the corrective discipline. Notes taken under a stage of the ascent know what they are for, since each stage has its determinations to evidence, and the researcher can feel the difference between a document that adds a determination and a document that adds colour. Colour has its place in a finished study, in the presentation, where a concrete instance can carry a stage of the argument for the reader. It has no place governing the inquiry, where every hour is priced against the determinations still missing. The practical rule that falls out of this is austere and liberating at once: before opening a new box, the researcher should be able to say which determination the box is expected to evidence and what in it would count as the determination failing. Boxes opened on that rule are rarely wasted. Boxes opened on the hope of treasure produce the familiar thesis that is rich in anecdotes and poor in argument, however many years it took to assemble.

The methods chapter and the viva

Everything in this article converges, for the doctoral researcher who is its most likely user, on two performances: the methods chapter and the examination that follows it. The methods chapter written to the standard set here has a recognisable shape. It opens with the object and its settings, stated against their alternatives, in the first pages rather than the last. It names the unit and argues the internal relations. It states the central claim with its symptoms, its rival and the observations that would defeat it. It describes the inquiry honestly and briefly, including the changes of setting and the failures, and it explains the presentation as a separate construction governed by dependence among the concepts. It closes with the limits, the tilt toward expecting transformation named alongside the access limits and the extension the evidence could not support. Such a chapter is shorter than the defensive sprawl it replaces, because it does not need to justify the project against every methodology ever published. It needs to expose the project’s decisions to view, and exposure, unlike defence, can be done in a few thousand words.

The viva then becomes a different event. Examiners probe what they cannot see, and a candidate whose decisions are visible spends the examination defending choices rather than excavating them, which is a contest the candidate can prepare for and often enjoy. The questions that destroy candidates in this tradition are predictable in form. Why this period and not the longer one your conclusion gestures toward? Your second chapter studies the institution and your fourth studies the workers, so which is the object? You call this a contradiction, so what would the world have looked like if it had not been one? Each question is a settings question, a unit question or a failure conditions question, and each has been answered in writing, in advance, by the researcher who kept the protocol. The candidate who has not kept it must improvise metaphysics under pressure, and the improvisation is what gives dialectical theses their reputation in examination folklore. There is no mystery about the reputation’s origin and no excuse for adding to the stock of anecdotes. The protocol costs an hour a month. The folklore costs a career’s opening years, in corrections, in reputation, in the quiet decision of examiners never to cite the work. Doctoral researchers are sometimes told that methodology is the tax a project pays to the academy. The account given here suggests a better metaphor. Methodology of this kind is the project’s memory, kept in a form other people can inspect, and a project with a memory can learn from its own first year instead of repeating it in its third.

Journalists and professional researchers outside the academy face the same convergence in a compressed form, the pitch and the editorial conference, where a framing defended in two sentences survives and a framing that exists only as a feeling dies. The settings give the pitch its two sentences. This is the object, bounded here for this reason. This is the relation inside it that generates the story, and this is what would make us drop it. An editor who hears the second sentence knows the proposal is research rather than advocacy, because advocacy never names its dropping conditions. The discipline travels across every setting in which inquiry is practised for an audience, which is the final reason for presenting it here as method rather than as doctrine: doctrines stay home, while disciplines go to work wherever they are carried.

When the method should not be used

A method article that cannot name the conditions under which its method is the wrong instrument is an advertisement, so the limits belong in the body of the text rather than in a critic’s mouth. Dialectical design is a poor instrument where the object is genuinely decomposable: where the parts interact weakly, the identities are stable, and the research question is a question of distribution or frequency. How many, how often, with what spread, across which groups: those questions want extensive designs, clean categories and samples that represent, and wrapping them in relational vocabulary slows the work and inflates the prose. It is a poor instrument where the timescale of the process dwarfs the timescale of the study and no archive can bridge the gap, because the becoming that the method wants to trace will be supplied by imagination instead, and imagination in a methods chapter is called speculation. It is a poor instrument where the researcher cannot gain access to more than one pole of the alleged relation, since a relational claim evidenced from a single pole is a monologue with a dialectical accent. And it is a dangerous instrument in the hands of a researcher who has already decided what the transformation ought to be, because the vocabulary of contradiction and becoming flatters a foregone conclusion with the appearance of a discovery procedure. In each case the honest move is the same: use the ordinary instrument the question calls for, and let the settings audit, which costs an afternoon and improves almost any design, stand as the tradition’s whole contribution to that project.

There is a corresponding limit on the tradition’s side of the exchange that deserves stating, because researchers meet it in the literature and mistake it for depth. Some questions that present themselves as methodological are in fact questions of evidence that no framing decision can answer. Whether a strike was caused by a wage cut or by an organiser’s arrival is not settled by any setting, level or vantage point. It is settled, if it is settled, by documents, sequences and testimony, weighed by the ordinary crafts. The method’s role in such a dispute is confined to ensuring that the disputants are examining the same object, that each side’s claim is stated in a form evidence could defeat, and that the exposition of the verdict ascends from the established determinations rather than leaping to the politically convenient whole. Researchers who ask more of it will be disappointed, and will sometimes blame the disappointment on insufficient faith. The disappointment is better blamed on a category error, the error of expecting a grammar to write the sentence. Grammar matters, sentences fail for want of it every day, and no grammar ever wrote a true sentence by itself.

The second study: replication as a comparison of settings

A method shows its worth in what it lets a field do next, and the settings apparatus has a consequence for the design of follow up research that deserves stating before this article closes its argument. Replication in the extensive tradition means repeating a measurement on a new sample and comparing the numbers. Replication in the dialectical tradition means something richer and more argumentative: repeating a study with one setting deliberately altered and the others held, so that the pair of studies together map the territory over which the original claim runs. A study of casual teaching labour conducted at the extension of one country invites its successor at the extension of another, under a different funding settlement, not in the hope of the same findings but in the expectation that the differences will locate which determinations were carrying the mechanism. A study conducted from the vantage point of the hourly paid teacher invites its mirror from the institution’s side of the relation, conducted by a different researcher with different sympathies, and the confrontation between the two accounts is itself a datum about the object, since a relation that presents genuinely different faces to its two poles is telling the field something no single account could.

Doctoral programmes and research teams can exploit this deliberately. A cluster of projects designed as variations on one worksheet, sharing a unit and a claim while differing by extension, level or standpoint, accumulates knowledge at a rate that unconnected projects cannot match, because every new result arrives already located in the space the earlier results defined. The failures accumulate usefully too. Where a mechanism found at one extension dissolves at another, the boundary of the mechanism has been surveyed, and the field’s map gains a coastline. Where a claim survives every variation the cluster can afford, it has earned a confidence no single study, however large, could give it. This is the closest the tradition comes to a collective research programme with a progressive problem shift, and it requires no agreement on doctrine among the participants. It requires only that they keep their worksheets in a common form and publish the settings alongside the findings, so that the next variation can be designed against a visible background. Fields that adopt the habit will find their literature reviews writing themselves as maps. Fields that do not will continue to rediscover, at intervals of a generation, that their classic disputes were about framings nobody had written down.

For students and exam candidates

Your need is narrower than the researcher’s and the method serves it directly. In an essay, dialectical framing earns its place only when it changes what the essay can say, so build one relational claim, hold it at one level of generality, and let your evaluation turn on whether the claim survives the strongest counterexample you can find. When a question invites you to discuss a thinker, resist narrating the doctrine and instead show a decision the method forced: a boundary drawn, a unit chosen, a contradiction claim stated tightly enough to be wrong. Practise spotting the filler use in your own drafts by deleting the word and reading what survives. The examiners who reward this material reward exactly that discipline, and the fuller writing consequences are developed in the series guide to writing a Marxism essay.

For teachers and seminar leaders

Your version of the problem arrives as a pile of essays in which dialectical functions as a compliment students pay a topic before describing it conventionally. Teach against the pile with the instruments in this article rather than with exhortation. Set the worksheet as a one page exercise on a topic the group knows, and let the disagreements about settings surface in discussion, where students discover that their quarrels about findings were quarrels about extension and vantage point. Set the deletion test as an editing task on a published paragraph. And when a student produces the rare contradiction claim with its failure conditions attached, make the class examine it properly, because one live example of a claim that can lose teaches the method faster than a term of definitions.

For researchers, historians, and journalists

Your layer is this article, so this section is an index rather than a summary. Set the worksheet before fieldwork and keep it dated. Name the unit before the forces. State the contradiction claim with its symptoms, its rival and its failure conditions, and keep the failure record from the first month rather than reconstructing it for the methods chapter. Separate the inquiry record from the presentation plan, and let the exposition ascend. Run the deletion test over every draft that leaves your desk. For the journalist working to a deadline, the compressed version of the same discipline is a single question applied to any story that arrives wearing the vocabulary of crisis and transformation: what is the unit here, the firm, the sector, the workforce, and what would the story look like set one level up or one vantage point over? The question takes ninety seconds and regularly changes the story’s first three paragraphs, which is where framing does its damage or its service.

A method summary to work from

The method, compressed to the size of a desk card, runs as follows. Constitute the object deliberately: set its extension, its level of generality and your vantage point against live alternatives, and record the settings and their changes. Name the unit of analysis before naming any forces, preferring the relation or the structured whole where the terms constitute one another. State the central claim, where a contradiction is alleged, in the disciplined form: two internal aspects, each requiring and undermining the other through ordinary functioning, with predicted symptoms and stated failure conditions. Gather evidence in the order inquiry imposes, and keep the record of that order, including its failures. Present in the order of dependence, ascending from the thinnest abstraction that contains the object’s relations in embryo to the concrete whole of many determinations, and stop the ascent when further determinations stop changing the claim. Test the vocabulary by deletion at every draft stage. And hold the whole apparatus to the standard the sceptic set: if the decisions cannot be stated, defended and shown to have been capable of being wrong, the word dialectical has not yet been earned, whatever the bibliography says.

A last word about temperament, because methods are practised by people and this one asks something specific of the people who practise it. It asks for patience with the object’s refusal to sit still, and for an equal patience with the occasions when it sits still for years and the honest finding is that nothing is unfolding. It asks for a willingness to be caught out in writing, since the failure record is a standing invitation to one’s future critics, and for the rarer willingness to catch out one’s teachers, whose settings, reconstructed respectfully and in print, will sometimes turn out to have been the limits of their conclusions. And it asks for proportion about the vocabulary itself. The word dialectical, on the account defended here, is a small tool with a long history, useful for the jobs named in this article and useless as a substitute for them. A researcher who finishes a project with fewer ceremonial words and more recorded decisions than the project began with has understood the tradition better than one who finishes with the vocabulary perfected and the decisions unmade. The worksheet, the claim sheet and the failure record will not appear in any hall of fame. They are, nevertheless, what the method looks like when it is working, and a field that keeps them will argue better, accumulate faster and waste less of its students’ one irreplaceable resource, which is the years in which they were willing to learn how research is actually done.

Frequently Asked Questions

Q: How do you actually use dialectical method in research?

You use it as a discipline of decisions rather than as a doctrine. Set the three settings of abstraction for your object, its reach in space and time, the altitude of its concepts and the standpoint you occupy, and record each setting alongside the alternatives you rejected. Name the unit of analysis before naming any forces, preferring a relation where the terms constitute one another. State any contradiction claim with its predicted symptoms and its failure conditions, gather evidence in the order inquiry imposes, and present the results in the order of dependence, ascending from the thinnest abstraction toward the concrete. Finally, run the deletion test over the draft, removing the word dialectical wherever its removal costs nothing.

Q: What is Ollman’s process of abstraction?

It is Bertell Ollman’s account of how researchers constitute their objects of study, developed in Dialectical Investigations and Dance of the Dialectic. Ollman argues that abstraction, the unavoidable act of lifting an object out of the world’s simultaneity for examination, has three adjustable dimensions: extension, which fixes the object’s reach in space and time; level of generality, which fixes the altitude of the concepts in use, from the singular event up to the traits of an entire form of society; and vantage point, the standpoint the viewing starts from. His practical claim is that most scholarly disagreements are disagreements about these settings rather than about facts, so making the settings explicit converts sterile disputes into arguable decisions.

Q: What does a dialectical research question look like?

It names a relation and asks about its career, rather than asking which factors influence an outcome. Instead of asking what causes job insecurity in universities, the dialectical form asks how the employment relation between institutions and casual teachers reproduces itself, what pressures its ordinary functioning generates, and what would count as its transformation. The grammar matters less than the commitments behind it: the question presupposes a unit that is a relation, promises a mechanism rather than a list of weighted factors, and generates its sub questions from the apparatus, one for each setting, one for each pole of the central claim, and one for the conditions under which the claim would fail.

Q: How is dialectics different from systems thinking?

The two share a suspicion of isolated variables and a taste for wholes, feedback and unintended consequences. They differ in three ways. Systems thinking characteristically models a system in order to stabilise or optimise it, so its tensions appear as malfunctions to be engineered away, while dialectical research treats tension as potentially constitutive, the form an object’s development takes. Systems models can legitimately describe a system at a moment, while a dialectical object must be given a history, because its identity is bound up with its becoming. And systems practice often places the analyst outside the apparatus, while the dialectical tradition builds the analyst’s vantage point into the account as a declared setting.

Q: What is critical realism and is it dialectical?

Critical realism is the philosophy of science associated with Roy Bhaskar, set out in A Realist Theory of Science and The Possibility of Naturalism. It distinguishes the real, the structures and mechanisms that exist unobserved, from the actual events they produce in combination and from the empirical events anyone records, and it licenses retroduction, the inference from a phenomenon to the structures that must exist for it to be possible. Its kinship with dialectical method is close: both insist on depth beneath patterns and on mechanisms in open systems. Bhaskar’s later work attempts an explicitly dialectical construction. Whether the label applies is disputed, but the practical inheritance, retroduction and the stratified ontology, is available to dialectical researchers whatever the label’s fate.

Q: How do you stop dialectical becoming a filler word?

Apply the work test. Delete every occurrence of the word and its cognates from the draft and read what remains. Where a sentence collapses, the word was carrying weight: it marked a setting decision, a relation treated as internal, or a stage in the ascent. Where nothing changes, the word was decoration, and the choice is to delete it permanently or to do the work it was pretending to do, by stating the setting, naming the unit or specifying the claim it adorned. Run the same test over the neighbouring ceremonial vocabulary, process, dynamic, holistic and relational. An hour with the test on a long draft returns a map of exactly where the argument is thinner than its prose.

Q: Can dialectical research be published in mainstream journals?

Yes, whenever the evidential craft meets the journal’s standards and the framing decisions are stated in terms referees can engage: scope, level, standpoint, mechanism and conditions. Mainstream referees are rarely hostile to the tradition’s substance when it arrives as decisions and claims with evidence attached. What they resist, rationally, is the unaccompanied vocabulary, the submission whose contribution is said to be a dialectical perspective, whose methods section performs an allegiance and whose findings would be unchanged under any other flag. The practical remedy is to lead with the decisions and the mechanism, let the tradition appear in the citations and in the paragraph where its commitments did specific work, and state the failure conditions before a referee has to ask for them.

Q: What should you read to learn dialectical method?

Begin with the primary text, the introduction Marx wrote in 1857, printed at the opening of the Grundrisse, for the two orders of inquiry and presentation and for the rise toward the concrete. Read Ollman’s Dialectical Investigations and Dance of the Dialectic for internal relations and the three settings of abstraction. Read Andrew Sayer’s Method in Social Science as the realist conscience of the enterprise, for intensive and extensive design and for rational as against chaotic abstraction. Read Bhaskar selectively, the early schema and retroduction before the later dialectical constructions. Read Harvey’s 1996 propositions as a practitioner’s coda, and Levins and Lewontin’s The Dialectical Biologist as the method practised inside another discipline.

Q: Is there a reproducible dialectical procedure?

No, and the honest answer matters more than the comfortable one. There is no sequence of operations whose faithful execution by independent researchers converges on the same design, in the way a sampling formula converges on a sample. What exists is a discipline that can be taught, checked and misapplied: the settings audit, the unit named before the forces, the contradiction claim with its failure conditions, the separation of inquiry from presentation, the ascent with its stopping rule and the deletion test for vocabulary. Each element can be done badly in ways a supervisor can point to on a page. That is less than an algorithm and considerably more than a style, and designs should claim exactly that much, no more.

Q: What are the limits of dialectical method?

The method constitutes objects and orders expositions; it generates no evidence and guarantees no conclusions. It is a poor instrument where objects are genuinely decomposable, where the question is one of distribution or frequency, where the process outruns every accessible archive, or where only one pole of an alleged relation can be reached. Its vocabulary can absorb any finding when used loosely, which makes it vulnerable to the charge of unfalsifiability, and it tilts its users toward expecting tension and transformation in objects that may in fact be sitting still. It must be married to ordinary evidential craft, statistics, archives, interviews and documents, and a study strong in the method and weak in the craft is worthless in a well organised way.

Q: How do you set the level of abstraction for a dialectical research design?

State the rung your organising concepts will occupy on the ladder that climbs from a singular occurrence, through forms specific to a period, up to the traits of an entire social order, and state it for each major concept separately, since concepts in one design often sit at different rungs. Choose the rung by asking where the mechanism you are hunting operates and what your evidence can actually reach, then record what the rungs above and below would have revealed and why the question can do without them. Guard against level slippage, the slide in which interview evidence gathered at one rung is interpreted through concepts pitched at another and published as conclusions at a third, with the shared nouns doing the smuggling.

Q: What should a researcher record when a contradiction claim fails its work test?

Record four things, briefly and at the time. First, the claim as originally stated, with its predicted symptoms, so the failure is measured against a real expectation rather than a reconstructed one. Second, what was actually observed: the two aspects coexisting without strain, the strain resolving without transformation, or the pressure traced to a different mechanism. Third, which element of the design was revised in consequence, a setting, the unit, the claim itself, and what the revised version now predicts. Fourth, the implication for other researchers, since a mapped dead end is a citable result. Kept in that form, the failure record is evidence that the study’s expectations were definite enough to be defeated, which is what examiners and referees are looking for.

Q: Why must a dialectical study name its unit of analysis before its forces?

Because forces, pressures, tendencies and contradictions are predicates of a unit, and the same list of forces rearranges itself completely when the unit changes. A tension between flexibility and security means one thing inside an individual contract, another inside a department’s staffing model and a third inside a sector’s labour market, and a design that lists its forces first has built on air. Naming the unit first also forces the decision the doctrine of internal relations requires: whether the unit is an item, a relation or a structured whole, and which relations are constitutive of it. Once the unit is fixed, the claims can be stated as claims about it, with symptoms and failure conditions that evidence can reach.

Q: How does the ascent from the abstract to the concrete guide fieldwork?

It converts the stages of the argument into stages of evidence gathering, each with its own question and its own way of failing. Early fieldwork tests the thin starting abstraction: does the contract, the category or the unit behave as the starting description says, in the documents and the first encounters? Later fieldwork hunts the determinations the argument has shown to be necessary, the funding rule, the market condition, the career incentive, and can therefore be targeted rather than encyclopaedic. Interview schedules, archive requests and data collection differ by stage because the ascent’s questions differ by stage. The stopping rule travels with it: gathering ends at a stage when further determinations stop altering the central claim, which protects a long project from accumulation without end.

Q: What separates dialectical research from ordinary process tracing?

Process tracing follows a mechanism through a case, deriving observable implications link by link and hunting for evidence that discriminates between rival mechanisms, a discipline dialectical researchers should imitate. The difference lies in the identities of the actors. Standard process tracing holds the actors and the setting steady while the causal chain unfolds. Dialectical research reserves the right to find that the unfolding changed the actors: that the organisation which entered a dispute as a wage claimant left it as a political body with different interests, constituted differently by the process itself. Where identities are stable, the approaches converge and the dialectical vocabulary adds nothing, which the deletion test will show. Where identities move, the relational account is needed to say what happened.