Historiography & Historical Methods

A clerk in Berlin accepts a sheaf of numbers

On 16 January 1917 the German Foreign Office handed the United States embassy in Berlin a telegram it could not read. The message was in cipher, which was itself a concession: Washington had opened its diplomatic wire to German traffic in the hope of brokering a negotiated peace, and had asked that anything sent over it be in plain language. President Wilson had relaxed that condition. Ambassador James Gerard accepted the coded groups and they went out over American wires, through Copenhagen, to the German embassy in Washington, and from there by commercial telegraph to Mexico City.

The text, once anyone could read it, instructed the German minister in Mexico to offer Mexico an alliance if the United States entered the war: joint war, joint peace, "ample financial support", and the recovery of what it called her lost territory of New Mexico, Texas and Arizona. It asked that Mexico invite Japan to join. It closed, in the clipped register of the trade, "Acknowledge Receipt", over the signature Zimmerman, with the second n of the Foreign Secretary's name dropped to save a character.

Within a day a cryptanalyst named Nigel de Grey, working in Room 40 at the British Admiralty, had most of it. What happened next is the reason this document, rather than some other, is the one we use to teach source criticism. Reading it was easy compared with the problem of admitting that it had been read.

Four questions, asked in an order that matters

You will meet long checklists for interrogating documents. Most of them are variations on four questions, and the order is not decoration. Each one narrows what the next can sensibly ask.

1. Provenance. Where did this object come from, through whose hands, in what physical form, and how do I know?
2. Purpose. What was it made to do? Not its subject: its job.
3. Audience. Who was meant to read it, who was meant not to, and what could the writer assume the reader already knew?
4. Survival. Why is this in front of me when so much else is not?

Run them on the Zimmermann telegram and watch each one change what the document can be used to prove.

Provenance, or why there is no such thing as the telegram

Ask where the document came from and the single object dissolves into at least five. There is the German Foreign Office file copy in Berlin. There is the enciphered text that travelled the American diplomatic wire, in a newer code. There is the version the German embassy in Washington re-transmitted to Mexico City by commercial telegraph, re-enciphered in the older code numbered 13040, which Room 40 had reconstructed far more completely. There is the copy a British agent known only as Mr H bought from a clerk in the Mexican commercial telegraph office. There is the English translation, typed in London, that an American ambassador carried home.

Those are not copies of one source. They are five sources with different evidential value. The Berlin file copy would show what the Foreign Office intended to say. The Mexico City copy shows what a telegraph clerk in Mexico could be bribed to part with in 1917, which is a fact about the security of commercial cable traffic and not about German war aims. The English translation shows what the British wanted the Americans to read, in words the British chose.

Provenance is also where deliberate deception lives. Admiral William Reginald Hall, who ran Room 40, sat on the decrypt for three weeks. Publishing it would tell Berlin that its diplomatic codes were broken and tell Washington that Britain was reading American diplomatic traffic. Both disclosures were worth more to Germany than the telegram was worth to Britain. So the British manufactured a provenance. They obtained the 13040 version in Mexico, which let them say in public that the text had been stolen there, and they privately handed the American government enough of the code to check the message against the files of the American commercial telegraph companies. When Arthur Balfour gave the papers to Ambassador Walter Hines Page on 23 February 1917, he said the text had been bought in Mexico. It had. That was the half of the truth that could be said out loud.

Key idea: a stated provenance is itself a claim, made by someone, for a reason. Treat it as evidence about the person who stated it, not as a property of the document.

Purpose: what the paper was for

The telegram's subject is a proposed alliance. Its purpose is narrower and more useful to you: it is an instruction, conditional on an event that had not happened yet. Zimmermann was not predicting that Mexico would invade Texas. He was authorising a diplomat to make an offer only if the United States entered the war, which Berlin expected because it was about to resume unrestricted submarine warfare on 1 February.

Get the purpose wrong and you will read the document as a plan. Get it right and you can see what it actually demonstrates: that the German Foreign Office, in January 1917, regarded American belligerency as near certain and was willing to spend money and credibility to slow American shipments of arms. That is a strong finding, and it does not depend on Mexico taking the offer seriously. Mexico did not. President Venustiano Carranza put the proposal to a military commission, which reported that the war was unwinnable, and the offer went nowhere.

Audience: two readers, and a third nobody planned for

The intended reader was Heinrich von Eckardt, the German minister in Mexico, a professional who needed no explanation of what an alliance entailed. That is why the message is so short and so blunt. Documents written for insiders compress; documents written for outsiders explain. When you find a source that carefully defines its own terms, suspect that it was written for people outside the circle, and ask who those people were.

The unplanned reader was Room 40, and after 1 March 1917 the American newspaper public. Watch what happens to the document's meaning when the audience changes. Read by Eckardt, it is a contingency instruction. Read by a reader of the Chicago press, it is an insult aimed at a continent. The Hearst papers and others argued it was a British forgery, and that argument was not stupid: the official provenance was a lie, the release was conveniently timed, and the British had every motive to invent it.

What ended the forgery argument was not a document at all. It was a press conference. On 3 March 1917 Zimmermann told an American journalist, "I cannot deny it. It is true." On 29 March he defended the telegram in the Reichstag. A historian could have argued about handwriting, cable registers and code groups for a decade; the author's admission settled authenticity in a sentence. Authentication often arrives from outside the document, and you should look for it there.

Survival: the question most readers skip

This telegram survives in abundance because four organisations with incompatible interests all kept a copy. Most documents are not like that. Survival is a filter, and the filter has a shape: institutions keep what they need for administration, families keep what flatters them, and whatever is written on cheap paper by people with no institution keeps nothing at all.

Ask the survival question as a counterfactual. If the German Foreign Office had used a one-time pad, Room 40 would have had nothing and the United States might have entered the war some weeks later on the submarine issue alone. If the Americans had held to their cleartext condition, the message would never have gone down the cable that Britain was tapping. The document exists in your hands because of two administrative decisions made for unrelated reasons. Every source you will ever use is downstream of decisions like that, and the ones taken in the other direction left no record of themselves.

The same telegram, sorted by the question you are asking

Now the part that most people have backwards. Primary and secondary are not categories of object. They describe the relation between a source and a question. The same physical page moves between them when you change what you want to know.

Notice the third row. For the question about American opinion, the telegram itself is almost beside the point; what matters is what editors printed and what senators said. Choose your question first and the sorting follows. Choose the label first and you will spend a month reading the wrong pile.

When a work of scholarship becomes a primary source

Take Edward Gibbon's History of the Decline and Fall of the Roman Empire, published from 1776. If your question is about the reign of Julian, Gibbon is a secondary source, and a dated one: he could not use the papyri, inscriptions and coin hoards that have appeared since. If your question is about how educated Britons in the 1770s thought about religion, empire and the causes of decline, Gibbon is a superb primary source, and his footnotes are part of the evidence rather than an apparatus you skip.

The same move works on anything. A 1935 school textbook is a weak secondary source on the French Revolution and an excellent primary source on what a state wanted its teenagers to believe in 1935. A census return is primary for who lived in a house and secondary, at best, for how people described their own occupations, because an enumerator wrote the entry. Each time, the question moved and the label followed.

The point: "Is this a primary source?" is not a question about the source. Rewrite it as "primary for what?" and it becomes answerable.

Common misconceptions

• Primary sources are more reliable than secondary ones. They are closer, which is not the same thing. A forged diary is a primary source. A careful monograph that has weighed a hundred documents may be far more reliable about what happened than any one of them.
• A document written at the time must be honest, because there was no reason to lie. There is almost always a reason. The telegram's own provenance was falsified within weeks by a government that wanted it believed.
• If the archive holds it, someone checked it. Archives preserve; they do not adjudicate. A file can contain a rumour, a draft that was never sent, or a forgery, and its presence certifies only that it was kept.
• The surviving record is a sample of the past. It is a sample of what institutions and families chose to keep, which is a very different population. Treating it as representative is the commonest quantitative error in historical work, and the next lesson is about where it leads.

What to carry forward

You now have a procedure rather than a sentiment. Ask where the object came from and in what physical form, and treat any stated provenance as somebody's claim. Ask what the document was made to do, which is usually narrower than its subject. Ask who was meant to read it, because that sets the level of explanation and tells you what was assumed. Ask why it survived, and name the decisions that could have destroyed it. Then, and only then, decide whether it is primary or secondary, which will depend entirely on the question you have chosen to ask.

Worth holding on to: the Zimmermann telegram convinced Americans of its contents not because it was found, but because its author admitted it. Authentication usually comes from outside the document, and a historian who will not look outside is at the mercy of anyone who can forge.

Sources

1. National Archives and Records Administration. (n.d.). Zimmermann Telegram (1917). Milestone Documents. archives.gov
2. Bloch, M. (1953). The Historian's Craft (P. Putnam, Trans.). Alfred A. Knopf. archive.org
3. Boghardt, T. (2012). The Zimmermann Telegram: Intelligence, Diplomacy, and America's Entry into World War I. Naval Institute Press.
4. Tuchman, B. W. (1958). The Zimmermann Telegram. Viking Press.
5. Tosh, J. (2015). The Pursuit of History: Aims, Methods and New Directions in the Study of Modern History (6th ed.). Routledge.

A palace with a dead man's name on it

Above the town of Milot in northern Haiti stand the ruins of a palace built by Henry Christophe, king of Haiti, between 1810 and 1813. It is called Sans Souci. Every guidebook tells you it was named for Frederick the Great's palace at Potsdam, and the comparison is flattering enough that nobody checks it.

There is another candidate. Colonel Jean-Baptiste Sans Souci was a Congo-born officer who led one of the insurgent bands in the Haitian war of independence and refused to place himself under Christophe's command. In 1803 Christophe had him killed. Seven years later Christophe began building a palace and gave it the dead man's name.

The Haitian anthropologist Michel-Rolph Trouillot opens from this coincidence and refuses to resolve it neatly. His question is not which explanation is correct. It is how one explanation became so obvious that the other stopped being asked. That question is the subject of his 1995 book, and it is the subject of this lesson.

Where silence enters

Trouillot's framework is the most useful thing anyone has written about archival absence, because it refuses to lump every kind of absence together. He identifies four moments at which a silence can be produced. He calls them the making of sources, the making of archives, the making of narratives, and the making of history in the final instance.

The moments are cumulative and they are different in kind. A fact that was never written down is missing for a different reason than a fact that was written, filed, and then thrown out, and both differ from a fact that sits in a box nobody has been given a reason to open. If you treat all three as "the archive is silent", you cannot do anything about any of them.

The first moment: what nobody bothered to record

Records are made by people with a reason to make them, and the reason is nearly always administrative. A plantation ledger records an enslaved woman because she is an asset that has a price, a work rate and a replacement cost. It records her first name, her age within a decade, and sometimes her mother. It does not record her surname, because the system did not recognise one, and it does not record what she thought about anything.

This is not an oversight to be indignant about. It is a structural property of the source, and it is informative. The ledger tells you exactly what the enslaver thought worth writing down, which is a finding. Saidiya Hartman's essay Venus in Two Acts pushes the point further: for many enslaved women the entire documentary trace consists of a line in a ship's log or a coroner's entry, produced by the violence that killed them. The archive preserves the moment of destruction and nothing else.

The same structure operates far from slavery. Illiterate people appear in the record when they are taxed, conscripted, prosecuted or buried. Domestic work, subsistence agriculture and unpaid care leave almost no paper anywhere. Ask of any group: what transaction would have generated a document about these people, and who would have kept it?

Why this matters: the first silence is not in the archive. It is upstream of the archive, in the decision about what counted as worth recording, and no amount of searching will fix it. It has to be worked around by inference.

The second moment: making an archive, and unmaking one

Archives are assembled, which means someone decided what to keep. Sometimes that decision is a fire.

On 30 June 1922, at the opening of the Irish Civil War, the Record Treasury of the Public Record Office of Ireland, housed at the Four Courts in Dublin, was destroyed by explosion and fire. Centuries of court, probate and parish material went with it, including nineteenth-century census returns. Irish social history was reshaped by a single afternoon, and researchers still work around the hole. The Virtual Record Treasury of Ireland project has since rebuilt part of the collection from copies, calendars and transcripts scattered across other repositories, which is a reminder that a destroyed archive is not always a total loss: what matters is whether anyone else had made a copy for their own purposes.

Accident does some of this work. Most of the schedules of the 1890 United States census burned in a Commerce Department fire on 10 January 1921, and the damaged remainder was disposed of afterwards. The effect is a twenty-year blind spot across the peak decades of immigration and internal migration, exactly where a researcher most wants to follow a household from one enumeration to the next.

Policy does the rest. Between the 1950s and the 1970s, British colonial administrations carried out what is now called Operation Legacy: officials were instructed to identify files that might embarrass the government or reveal racial or religious bias, and to burn them or dump them at sea. Destruction certificates were sent to London to confirm the work was done. One administration warned the Foreign Office of the risk of, in its own phrase, "celebrating Independence Day with smoke". Some 8,800 files from more than twenty territories were not destroyed but removed to Britain, where their existence was denied until 2011, when litigation brought by Kenyan claimants over the Mau Mau detention camps forced the admission. They are now the FCO 141 series at Kew.

Notice the methodological gift buried in that story. The destruction was bureaucratic, so it generated its own paperwork. The instruction, the criteria and the certificates survive even where the files do not, which means the shape of the hole is documented. Where an archive has been purged systematically, look for the purge's own records.

The third moment: retrieval, and the catalogue as an argument

A document you cannot find is, for practical purposes, a document that does not exist. Retrieval is governed by finding aids, and finding aids embody the assumptions of whoever wrote them.

Records of enslaved people are commonly catalogued under the name of the enslaving family, because the papers came in as a family deposit. A researcher looking for a person must therefore already know who claimed to own her. Married women's records are filed under husbands' surnames. Colonial material is indexed by the administrative district the colonial power invented, so a search by the name of the people who lived there returns nothing. None of this requires bad faith. It requires only that the catalogue was written by someone whose questions were not yours.

The practical response is to stop searching by subject and start searching by process. Ask which office would have generated paper about your subject as a by-product of its ordinary work, then read that office's series in full. People invisible in a subject index turn up in petitions, court depositions, poor relief registers, ships' manifests, hospital admissions and disciplinary files, because those bodies had reasons of their own to write.

The fourth moment: what a period will let you think

Trouillot's hardest chapter argues that the Haitian Revolution was, in the phrase he uses, an unthinkable history: not merely unknown to Europeans but outside the categories with which they thought. Contemporaries who could see enslaved people winning battles reached for other explanations, from foreign agitators to British money, because a self-liberating black republic did not fit the available account of the world. Silencing at this moment is not about missing documents. The documents were plentiful. It is about what a reader could do with them.

The same mechanism operates on happier material. For a century after 1492 the arrival of Europeans in the Caribbean was narrated as discovery, and that word decided in advance whose experience counted as the event. Trouillot's point is that this last moment is where most of the work of silencing gets done in the present, in textbooks and anniversaries rather than in archives.

What to actually do

Silence is a finding, not an excuse. Four moves make it productive.

• Name the moment. Say explicitly whether the gap is non-recording, destruction, cataloguing or narration. Each has a different remedy and a different implication for your argument.
• Find the counter-archive. Where one institution destroyed, another may have copied. Missionary societies, insurance companies, newspapers, foreign consulates and opposing armies all keep records about people they do not employ.
• Measure the hole. If a series survives for eleven of fifteen years, say so, and say what the four missing years would have had to contain to change your conclusion. That is a testable statement rather than a disclaimer.
• Write the silence into the argument. A sentence such as "the wage books record the men by name and the women as numbers" tells the reader something real about the firm, and it is honest about the limits of what follows.

Common misconceptions

• Archival silence means the historian can say nothing. It means one route is blocked. Structural absence is itself evidence about the institution that produced the record, and other institutions often wrote about the same people for their own reasons.
• Silences are accidents of preservation. Some are. Operation Legacy shows that others were planned, carried out on written instructions and certified on completion.
• Digitisation removes the problem. Digitisation reproduces the catalogue. If the finding aid never named the people you want, full-text search of that finding aid will not find them either, and the confidence a search box inspires makes the gap harder to see.
• Trouillot is arguing that history is only power. He argues that power enters at four identifiable points in the production of history. The point of specifying the points is that they can be examined one at a time.

Where this leaves us

The archive is not a neutral deposit of the past that historians happen to visit. It is the residue of a long series of decisions, some administrative, some accidental, some deliberate and documented. Trouillot's four moments give you a way of asking which decision produced the particular absence in front of you, and therefore what to do next: infer around it, look for a copy made elsewhere, read a different series, or say plainly in your text that the record was made by people who did not think your subject worth a line.

The upshot: when you find nothing, do not write that the sources are silent. Write which of the four moments silenced them, and what that tells you about the people who kept the books.

Sources

1. Trouillot, M.-R. (1995). Silencing the Past: Power and the Production of History. Beacon Press. archive.org
2. Virtual Record Treasury of Ireland. (n.d.). Reconstructing the Record Treasury destroyed in 1922. Trinity College Dublin. virtualtreasury.ie
3. Hartman, S. (2008). Venus in two acts. Small Axe, 12(2), 1-14.
4. Farge, A. (2013). The Allure of the Archives (T. Scott-Railton, Trans.). Yale University Press.
5. Fuentes, M. J. (2016). Dispossessed Lives: Enslaved Women, Violence, and the Archive. University of Pennsylvania Press.

The miller answers a question about the creation

In February 1584, in the Friulian town of Portogruaro, a fifty-two-year-old miller stood before the Holy Office and was asked what he believed about the origin of the world. He did not hedge. All was chaos, he said, and out of that mass a bulk formed, as cheese is made out of milk, and worms appeared in it, and those worms were the angels.

His name was Domenico Scandella, known to everyone in the village of Montereale as Menocchio. He had already told his neighbours that the pope's authority was a human arrangement, that Christ was born of a man, and that the only real sin was harming your neighbour. He would abjure in 1584, spend some twenty months in prison, be released in 1586 under house arrest with a cross sewn on his clothing, go on saying the same things, be arrested again in 1598, and be burned in 1599 as a lapsed heretic.

Roughly twelve hundred pages of testimony survive from his two trials. We know more about the cosmology of this particular illiterate-class miller than about that of any duke of the same generation, and we know it for a reason that ought to make you uncomfortable.

Why this record exists at all

The trial dossier survives because an institution with the power to kill him wanted to know precisely what he thought. Inquisitors were trained to press for detail: not merely whether the accused denied purgatory but in what words, learned from whom, said to which neighbours, on what day. That procedural appetite produced, as an accident, the most detailed record of one sixteenth-century peasant's mental world that anyone possesses.

This is the central paradox of most evidence about people without power. It was produced by the institutions that policed them: courts, prisons, poor law boards, missions, plantations, welfare offices, asylums. The record is hostile, coerced, and shaped by the interrogator's agenda, and it is also the only place those voices survive at all. Refusing to use it means agreeing that the poor left no trace. Using it naively means writing history from the interrogator's chair.

Remember: the sources richest in the words of the powerless are usually the products of coercion. That is a reason to read them carefully, not a reason to leave them alone.

What Menocchio had read, and what happened to it in his head

Carlo Ginzburg found a summary of the case in 1963 in an inquisitor's index while working on Friulian witchcraft trials, returned to the documents in 1970, and published Il formaggio e i vermi in 1976. Menocchio had read: a vernacular Bible, a popular compendium of biblical stories, a chronicle, Mandeville's Travels, Boccaccio's Decameron, and probably an Italian version of the Koran. None of these books says that the world congealed like cheese.

Ginzburg's decisive move is to stop treating the books as sources of Menocchio's ideas and start treating them as raw material for an operation performed on them. Between the printed page and the miller stood a screen, made of the oral culture of his village, and that screen did the choosing. Menocchio read Mandeville's account of distant peoples with strange religions and came away with the conclusion that no single faith could be the true one. That conclusion is not in Mandeville. It is what a man already inclined to think it does with Mandeville.

So the historical finding is not "peasants absorbed elite books". It is that a peasant reader was an active filter with a culture of his own, one strong enough to bend printed authority into shapes its authors would not have recognised. Ginzburg proposed that this oral peasant culture had deep roots, an argument that remains contested; what almost nobody contests is the demonstration that reading was transformation rather than transmission.

The operation, step by step

Reading against the grain sounds like an attitude. It is a procedure, and it can be written down.

1. Separate the question from the answer. Inquisitorial records are dialogues. Mark which content originated with the interrogator's formula and which the witness volunteered. Anything volunteered, especially anything the interrogator then had to chase, is the strongest material you have.
2. Find the friction. Look for the moments where the transcript stops running smoothly: where the scribe records confusion, where a question is repeated, where the accused says something that fits no available heresy. Menocchio's cheese is exactly such a moment, which is why it survives so vividly.
3. Read the form, not just the content. The categories of the record are evidence about the recorder. Which heresies the court had names for tells you the shape of official expectation, and therefore what the accused had to be forced into.
4. Establish what the speaker could not have been coached to say. A confession that matches the manual proves the manual. A statement the court plainly did not want and could not use is far better evidence about the speaker.
5. Reconstruct the pressure. Torture, the threat of it, the duration of imprisonment and the presence of neighbours as witnesses all shape what gets said. Write the pressure into the analysis rather than into a footnote.

Clues: the evidential paradigm

In 1979 Ginzburg wrote an essay that generalised the method. He noticed that three nineteenth-century practices shared a logic: Giovanni Morelli attributed paintings by studying earlobes and fingernails rather than composition, because a forger copies what is conspicuous; Sigmund Freud read slips and gestures; Arthur Conan Doyle's detective read cigar ash. All three treat the marginal, involuntary detail as the place where truth leaks, precisely because nobody thought to control it.

Ginzburg called this the evidential or conjectural paradigm and contrasted it with the paradigm of the physical sciences, which seeks repeatable general laws. History, medical diagnosis and detection are all conjectural: individual, non-repeatable, and reasoned from traces. The consequence for you is a rule of attention. In any document, the parts written on purpose are where the author exercised control; the parts written without thinking are where you learn the most.

Applied to an ordinary source, this means the letterhead, the file reference, the routing slip, the crossings-out, the form of address, the units of measurement and the things listed without comment. A shipping manifest controls its cargo column. It does not control what it treats as unremarkable.

The objection that has to be met

The strongest criticism of this kind of work is that the historian ends up sitting where the inquisitor sat. The questions are the court's questions. The categories are the court's categories. Menocchio speaks only in answer, in a room where the alternative to speaking is worse, and every word reaches us through a scribe writing in a language and a formula that were not the miller's.

Ginzburg accepts the difficulty and answers it in two ways. First, the dialogue is unequal but it is a dialogue: the record contains resistance, evasion and initiative that a purely dictated confession would not. Second, he insists that the reader be shown the seam. His book quotes the transcript at length and repeatedly marks where he is inferring rather than reporting, which lets a sceptical reader disagree with him using his own evidence.

That is the standard to take away. You are permitted to read a coerced source for what it discloses about the coerced, provided you show the reader exactly where the coercion sat and exactly which of your claims depend on inference. The alternative is not neutrality; the alternative is silence about anybody who never held a pen.

The exceptional and the normal

One more problem, and it is a real one. Menocchio was burned because he was unusual. Most millers did not lecture their priest on cosmology. So what can one strange man tell you about anybody else?

The microhistorians' answer is that an exceptional case is often the only place a normal structure becomes visible. Menocchio's beliefs were his own; the world he had to defend them in, the books available in a Friulian village, the level of literacy, the reach of the Counter-Reformation into a mountain parish, the way neighbours reported one another, was completely ordinary. The rare event lights up the ordinary machinery around it, in the way that a fault reveals a system in normal operation. Historians call this the normal exception, and it is the standard defence of working at very small scale, which the module on the Annales and microhistory examines directly.

Common misconceptions

• Reading against the grain means assuming the document is lying. It means separating what the document reports from what its existence, form and friction demonstrate. A tax roll can be entirely accurate and still tell you most about the state that wanted the numbers.
• Menocchio proves that ordinary people rejected the Church. He proves that one man did, in a documented setting. What generalises is the mechanism of active, transforming reading, not the content of his beliefs.
• Ginzburg reconstructed peasant culture from Menocchio. He argued for a deep oral substratum, and that argument has been disputed for thirty years. The demonstration that survives criticism is narrower and more durable.
• Coerced testimony is worthless as evidence. Coercion changes what a source can prove, not whether it can prove anything. What it cannot support is a claim that the speaker was free to choose the subject.

The short version

A miller's cosmology reaches you because a tribunal wanted evidence for a capital charge, which is how most testimony from people without power reaches anyone. Reading it against the grain is a sequence you can carry out: mark what was volunteered, hunt the friction, read the form as evidence about the recorder, prefer what the court could not use, and state the pressure. Ginzburg's wider claim, that history reasons from involuntary traces rather than from repeatable law, tells you where in any document to spend your attention.

The core of it: the author controls the message and leaves the margins unguarded. Read the margins, and say out loud which of your conclusions come from them.

Sources

1. Ginzburg, C. (1980). The Cheese and the Worms: The Cosmos of a Sixteenth-Century Miller (J. Tedeschi & A. Tedeschi, Trans.). Johns Hopkins University Press. archive.org
2. Ginzburg, C. (1989). Clues, Myths, and the Historical Method (J. Tedeschi & A. Tedeschi, Trans.). Johns Hopkins University Press. archive.org
3. Ginzburg, C. (1993). Microhistory: Two or three things that I know about it. Critical Inquiry, 20(1), 10-35.
4. Levi, G. (1991). On microhistory. In P. Burke (Ed.), New Perspectives on Historical Writing (pp. 93-113). Polity Press.
5. Muir, E., & Ruggiero, G. (Eds.). (1991). Microhistory and the Lost Peoples of Europe. Johns Hopkins University Press.

A question the Dutch colonial state could not answer

The Nationaal Archief in The Hague and the Arsip Nasional in Jakarta hold a long series of Dutch official papers in which administrators of the Netherlands Indies asked one another, decade after decade, which children of European fathers and Asian mothers counted as European. Commissions were appointed. Memoranda were written. Criteria were proposed, adopted, found unworkable and proposed again: paternal acknowledgement, language spoken at home, schooling, dress, household arrangements, the smell of the kitchen. The question was never settled, because it could not be.

Here is the problem that makes this lesson necessary. Suppose you go into that series looking for evidence of colonial racism. You will find it on every page, and you will come out with a finding you already had. Suppose instead you go looking for resistance, reading against the grain as the previous lesson taught. You will find traces, but the series is not mainly about resisting subjects at all. It is about officials who did not know what they were doing and kept writing to each other about it.

So what do you do with an archive whose most striking property is the anxiety of the people who made it?

From archive-as-source to archive-as-subject

Ann Laura Stoler has spent a career in the colonial archives of the Netherlands Indies, and her 2009 book Along the Archival Grain makes a proposal that sounds small and is not. Stop treating the archive as a warehouse you extract facts from, and treat it as an object with its own properties, produced by a process you can study. She describes the shift as one from archive-as-source to archive-as-subject, and she calls the practice an ethnography of the archive.

The phrase in her title is deliberately set against the familiar slogan. Reading against the grain asks what the record conceals. Reading along the grain asks what the record is, how it was made, in what genres, at what tempo, with what gaps between the handwritten note and the printed report, and what its makers took so thoroughly for granted that they never wrote it down. Her term for that unwritten layer is colonial common sense: the assumptions officials shared so completely that they appear in the files only as the things nobody argues about.

What matters here: against the grain and along the grain are not rivals. Against the grain recovers the people the record was used on. Along the grain recovers how the record, and therefore the rule it carried out, actually operated. A serious project needs both, and should say which it is doing in any given paragraph.

Epistemic anxiety, and why it is evidence

Colonial rule is often written as if it were confident: a machine of classification that sorted populations and acted. Stoler's archives show something else. They are full of hesitation. Questions are referred upward and come back unresolved. Categories are invented, then qualified, then abandoned. Officials request guidance on cases the rules do not cover, which turns out to be most cases. She calls this epistemic anxiety, and her argument is that it is not noise around the real business of governing. It is the business of governing.

Take the pauperism question that worried the Indies administration around the turn of the twentieth century: what to do about Europeans who were poor. A confident racial order would have had no such category, because European status was supposed to carry prosperity with it. The existence of commissions on the subject is evidence that the order did not work as advertised, and the terms in which the commissioners argued tell you which parts of it they believed and which they were improvising.

This is why the form of a file matters so much. Consider what each genre can tell you:

Read downward through that table and the character of the evidence changes. The annual report is the most polished and least informative about how decisions were reached. The marginalia are scrappy, undated, often unsigned, and frequently decisive.

The counterpart from the other side: Guha on counter-insurgency prose

Stoler's project has a close relative in Indian historiography. In 1983 Ranajit Guha published an essay called The Prose of Counter-Insurgency, about the documentary record of peasant uprisings in colonial India. Almost everything written about those uprisings was produced by the apparatus that put them down: district officers' reports, police files, commission findings, and later the nationalist histories that absorbed those sources.

Guha sorted this writing into levels. Primary discourse is written by officials at the time and for official use. Secondary discourse is written later, often by the same establishment, digesting the first. Tertiary discourse is written by historians, including sympathetic ones, who take their framework from the first two without noticing. His point is that a nationalist or Marxist historian can adopt the causal structure of a police report simply by relying on it for the facts, so that peasant action keeps appearing as a reaction triggered by outside agitators rather than as action with its own reasoning.

Put Stoler and Guha together and you have both halves of the instruction. The record is not merely biased, as though a correction factor could be applied. It has a structure, and the structure will organise your argument for you unless you study it deliberately.

Doing it: six questions to ask of a series rather than a document

The unit of analysis changes. You are no longer interrogating one page; you are characterising a run of paper.

1. What generated this series, and when did it start and stop? A series that begins in 1870 and ends in 1889 brackets a problem the state thought it had during those years.
2. What is the rhythm? Steady monthly returns mean routine. Clusters of urgent correspondence mean crisis. Long gaps mean either calm or a filing practice you have not yet understood.
3. Which categories are stable and which keep being redefined? A category redefined four times in twenty years is the site of a real problem, and it will be the most interesting thing in the series.
4. What do officials explain to one another? Explanation marks the boundary of shared assumption. Whatever is never explained is the common sense.
5. Where is the friction between genres? Compare what the printed report says with what the handwritten note beside it says. The distance between them is a measurement.
6. What did the series make it impossible to record? A form with no field for a thing renders that thing administratively invisible, which has consequences in the world and not only on paper.

The danger in this method

There is a cost to turning the archive into the subject, and honest practitioners of the approach say so. If every project becomes a study of how colonial officials felt, the people those officials ruled can disappear from the history a second time, now in the name of methodological sophistication. Attending to a district officer's uncertainty is not a substitute for finding out what happened in the district.

The discipline is to keep both questions live and to be explicit about which one a given chapter answers. Stoler's own books work on the colonial state; her earlier study of the Sumatra plantation belt worked on labour and its organisation. The lesson is not that one replaces the other but that a historian who never asks how the record was made will import its assumptions without noticing, which is exactly Guha's charge against tertiary discourse.

Common misconceptions

• Reading along the grain means accepting the document's account. It means studying the document's production. You can conclude that a report is self-serving and still learn a great deal from the fact that it had to be written at all.
• Colonial administrations knew what they were doing. The files show repeated commissions, unresolved referrals and categories abandoned and reinvented. Confidence was a presentation, and the presentation is one genre among several in the same file.
• Bias in a source can be corrected for, like a known measurement error. Bias here is structural: it determines what was recorded at all, which categories existed, and what questions a later historian will find it natural to ask.
• Guha's objection applies only to colonial administrators. His sharpest target is tertiary discourse, written by historians sympathetic to the insurgents, who inherit the official framework along with the official facts.

Pulling it together

An archive is a made thing. Its series have start dates, rhythms, genres and silences, and all of those are evidence about the institution that produced them. Stoler's shift from source to subject turns the apparent obstacle, a record written entirely by the powerful, into a research object: the hesitation in the files tells you that rule was improvised, and the things never explained tell you what everyone involved took for granted. Guha's levels of discourse warn you that the framework travels with the facts, so a historian who borrows one borrows the other.

Bottom line: before asking what a series says, characterise what kind of paper it is and what problem brought it into existence. Then decide which of your questions it can answer, and say so in the text.

Sources

1. Stoler, A. L. (2009). Along the Archival Grain: Epistemic Anxieties and Colonial Common Sense. Princeton University Press. archive.org
2. Guha, R. (1983). Elementary Aspects of Peasant Insurgency in Colonial India. Oxford University Press. archive.org
3. Stoler, A. L. (2002). Colonial archives and the arts of governance. Archival Science, 2(1-2), 87-109.
4. Guha, R. (1983). The prose of counter-insurgency. In R. Guha (Ed.), Subaltern Studies II: Writings on South Asian History and Society (pp. 1-42). Oxford University Press.
5. Burton, A. (Ed.). (2005). Archive Stories: Facts, Fictions, and the Writing of History. Duke University Press.

Before you read: hold one prediction

Suppose you interview forty people about an event and half of them place it in the wrong year, and the wrong year is the same wrong year in every case. Write down, in one sentence, what you would conclude. Keep the sentence. You will be asked to compare it with Portelli's conclusion, and the difference is the whole lesson.

A shooting outside the gates

On 17 March 1949 the police shot and killed Luigi Trastulli, a twenty-one-year-old steelworker, outside the works at Terni in Umbria. He had walked out with his colleagues to a rally against Italy joining the North Atlantic Treaty, signed in Washington twelve days later. The facts of the killing are documented in press, police and union records, and nobody in Terni disputes that he died.

Three or four years later the same works dismissed workers by the thousand. The layoffs of 1952 and 1953 broke the strength of a plant that had defined the town, and in October 1953 there was fighting at the gates with the police. That struggle was lost.

When Alessandro Portelli interviewed Terni workers in the 1970s and 1980s, he found that a great many of them put Trastulli's death in 1953, during the fight over the dismissals. Not scattered randomly across the decade. In the same wrong place, again and again.

The wrong answer, stated properly

Here is the conclusion most people reach, and it is not stupid. Memory decays. Thirty years after the event, witnesses conflate two episodes of violence at the same gates. The error is systematic because both events are emotionally similar and because the community has told the story to itself so often that the versions have contaminated one another. Therefore the oral testimony is unreliable on chronology and should be checked against documents, and where it conflicts with documents the documents win. Use the interviews for colour and the archives for facts.

Every step in that reasoning is defensible. The conclusion is still wrong, and it is worth being precise about where it fails.

Where the reasoning breaks

It fails at the word "therefore". The argument establishes that the testimony is a poor instrument for dating an event. It then treats that as a verdict on the testimony as a whole, as though the only thing an interview could be evidence for were the date.

Look at what the error is doing. Moving the death from 1949 to 1953 moves it from a protest about foreign policy to a struggle over jobs. In 1949 the men walked out over a treaty; a young worker died in a cause that, from the vantage point of a ruined town, could look abstract. In 1953 they were fighting for the plant itself and lost. Relocating the death attaches the community's one martyr to the defeat that actually shaped its history. It supplies a reason for the loss, and it makes the death mean something a foreign-policy protest could not.

That is not decay. Decay is random. This is a consistent, directional revision, which means it has a cause, and the cause is the thing you wanted to know about in the first place: what the events meant to the people who lived through them.

So what?: a factual error in testimony disqualifies that testimony as evidence about the fact and simultaneously qualifies it as evidence about the meaning. Portelli's formulation is that errors and inventions lead us through the facts to what they meant.

What oral sources are good for

Portelli's methodological essay, written at the end of the 1970s, sets out the differences that follow. They are worth taking one at a time, because each changes what you should do with a recording.

The fourth row is the one beginners underestimate. An interview is not a deposit being withdrawn; it is a conversation created by two people, one of whom asked. If a worker explains what a shift system was, he has judged that you do not know, and that judgement is data. If he refuses a topic, the refusal is on the recording. Portelli's practice is to keep his own questions in the published transcript, which is the oral historian's version of showing the seam.

Also weigh: what oral history cannot do

Being clear about the strength requires being blunt about the limits. Oral testimony is a bad instrument for chronology, for numbers, for anything the speaker did not personally witness, and for the internal workings of institutions the speaker did not sit in. It cannot tell you how many workers were dismissed at Terni; the company's records can. It cannot establish what a minister decided in Rome.

It is also bounded by mortality, which is a hard constraint people forget. Oral history of the 1930s is now impossible to conduct. Every project has a closing window, and choosing what to record is a decision about what future historians will be able to ask.

There is a further trap. Because interviews feel like access to authentic experience, they invite the historian to treat a vivid speaker as representative. Forty interviews in one town are forty interviews in one town, weighted towards those who survived, stayed, and agreed to talk. Sampling logic still applies, and the survivors of a defeat are a particular population.

Working a transcript

A practical sequence, once you have a recording.

1. Check the checkable. Establish dates, names and numbers against documents first, so you know precisely where the account departs from the record. The departures are your material.
2. Characterise the departures. Are they random, or do they all move in one direction? Directional error is a finding; scattered error is usually just time passing.
3. Attend to form. Where does the speaker shift from I to we, from past tense to present, from summary to scene? A sudden switch into the present tense usually marks the part of the story that still matters most.
4. Keep your own voice in. Note which questions produced elaboration and which produced a sentence and a pause.
5. Compare across speakers. A shared error is a community's interpretation. An idiosyncratic one is an individual's.
6. Deal with consent and rights properly. Recorded testimony belongs in part to the person who gave it. The Oral History Association's published principles cover informed consent, deposit and reuse, and they are not optional extras.

Common misconceptions

• Oral history is what you do when there are no documents. It answers different questions. Even where records are abundant, no document tells you what a person thought they were doing, and few record what happened afterwards to the meaning of an event.
• A good interviewer stays silent so as not to influence the account. The account is co-produced whatever you do. The remedy is to record and publish your own questions, not to pretend you were absent.
• Consistent testimony is more reliable than inconsistent testimony. Consistency across many speakers can mean a shared public version has replaced individual recollection. That is itself important, but it is not corroboration.
• Memory error is noise to be filtered out. Random error is noise. Systematic, directional error is signal, and it points at exactly the interpretive question the project usually exists to answer.

What to remember

Terni gave Portelli a community that shifted a killing four years forward in time, together, and in a direction. The naive reading discards the testimony for being wrong about the date. The correct reading notices that the error is patterned, asks what the pattern accomplishes, and finds an answer about how a defeated town made sense of its defeat. Oral sources are weak evidence for chronology and unique evidence for meaning, and the discipline consists of checking the first so that you can trust yourself about the second.

In short: compare what people say with what the documents show, and then treat the gap between them as the object of study rather than as an obstacle to it.

Sources

1. Portelli, A. (1991). The Death of Luigi Trastulli and Other Stories: Form and Meaning in Oral History. State University of New York Press. archive.org
2. Oral History Association. (n.d.). Principles and Best Practices. oralhistory.org
3. Passerini, L. (1987). Fascism in Popular Memory: The Cultural Experience of the Turin Working Class. Cambridge University Press.
4. Thompson, P., & Bornat, J. (2017). The Voice of the Past: Oral History (4th ed.). Oxford University Press.
5. Abrams, L. (2016). Oral History Theory (2nd ed.). Routledge.

A plantation ledger, two years, one hundred and sixty entries

Bennet H. Barrow, a cotton planter in West Feliciana Parish, Louisiana, kept a record of punishments on his plantation. For the two years from December 1840 it lists roughly one hundred and sixty whippings. In 1974 two economists divided that total by the number of hands and by the number of years and published the result: about seven tenths of a whipping per hand per year.

The calculation is arithmetically unimpeachable. The argument about what it means has run for fifty years, and it is the best case in the discipline for learning what quantitative evidence can and cannot do.

The book, stated at its strongest

Robert Fogel and Stanley Engerman published Time on the Cross in two volumes, one of narrative and one of evidence and methods. It was a work of cliometrics, the application of economic theory and statistical inference to historical records, and it reached a set of conclusions that contradicted almost everything the standard account said.

Their claims, in the form they would recognise:

• Slavery was profitable and expanding, not a decaying institution that the market would have ended. Slave prices were rising and would have continued to rise. This extended an argument Alfred Conrad and John Meyer had made in 1958.
• Southern agriculture using the gang system was substantially more productive than Northern family farming, by an index that put large plantations roughly a third ahead.
• The enslaved worked hard and skilfully, and a considerable number held positions of responsibility. Fogel and Engerman argued this against a racist inheritance shared by abolitionists and slaveholders alike, which had portrayed black workers as incapable.
• Material conditions in food, housing and medical care were roughly comparable to those of free urban labourers.
• Expropriation was smaller than assumed. In their words, over a lifetime a field hand "received approximately ninety percent of the income produced".
• Systematic slave breeding was essentially a myth, family formation was encouraged, and punishment was infrequent.

The authors were explicit that they were not defending slavery. Their stated aim was to destroy a set of myths about black incapacity by showing, in the slaveholders' own accounts, that the labour force was efficient and skilled. The book was reviewed in Newsweek and Time; its authors appeared on television. Fogel later shared the 1993 Nobel Memorial Prize in Economic Sciences with Douglass North.

The reply, stated at its strongest

Herbert Gutman published Slavery and the Numbers Game in 1975, and in 1976 Paul David, Gutman, Richard Sutch, Peter Temin and Gavin Wright published Reckoning with Slavery. Their case was technical, not moral, which is what made it devastating.

On the whipping figure, the objection is about what the arithmetic mean can represent. The same Barrow record yields a whipping roughly every four and a half days on a single plantation, in front of everyone. A large share of the hands were whipped at least once in the year. And an average punishment rate says nothing at all about the mechanism the punishment served, because the point of a public whipping is its effect on those watching. A system of terror does not require frequent application; it requires credible application. Dividing by the number of hands measures the wrong thing.

On the efficiency index, the criticism cut to the construction of the measure. Total factor productivity compares output value to weighted inputs. Southern and Northern farms grew different crops on different land in different climates, so an index comparing them is heavily sensitive to how outputs are priced and how land and labour are weighted. Change plausible assumptions and the Southern advantage shrinks or reverses. Worse, a higher measured output per worker is compatible with slaves being driven harder for longer, which the index records as efficiency and a worker would experience as something else.

On the ninety percent figure, the objection is that the number depends on choices about how to value what the enslaved received, over what lifetime, and net of what. Different but equally defensible accounting conventions move the result substantially. A number that swings with the convention is not a finding; it is a model output.

Gutman pressed a further point about assumptions hidden inside economic models. Fogel and Engerman's account of incentives assumed workers who responded to rewards and penalties much as a wage labourer would. Gutman argued that this quietly imported a nineteenth-century work ethic into people whose relation to the crop and the whip was not that of employees at all.

And the critics noted an omission with large consequences: the forced migration of roughly a million people from the Upper South to the cotton frontier, which broke families on a scale that the book's treatment of family stability could not accommodate.

Key idea: none of these objections says the arithmetic was wrong. They say the numbers do not measure the thing the argument needs them to measure. That is the commonest failure mode in quantitative history, and it is invisible to anyone checking only the calculations.

What survived

A dispute is only useful if you can say what it settled. Sort the claims:

Fogel returned to the subject in Without Consent or Contract in 1989, keeping the economic viability argument and rebuilding the moral case around the theft of liberty rather than the level of consumption. That is the right lesson to draw about scholarly disputes: they rarely end in one side leaving the field. They end with a redistribution of which claims still stand.

The general problem this case exposes

Strip away the subject and four transferable errors remain, each of which you will meet again in any quantitative work.

1. The measure is not the concept. Whippings per hand per year is a measure. Brutality is a concept. When you substitute one for the other without argument, you are asserting a theory of how the institution worked, and you should defend it.
2. The mean hides the distribution and the mechanism. A low average is compatible with a system organised around the certainty that punishment will arrive. Ask who is in the tail, and ask what the observers of a single event do afterwards.
3. Index numbers carry assumptions. Any comparison of unlike things requires weights, and the weights encode a judgement. Report how the result moves when the weights are varied, or you have reported a preference.
4. Behavioural models smuggle in a subject. Assuming that people respond to incentives in a particular way is an empirical claim about those people, and in historical work it is usually the claim most in need of evidence.

Common misconceptions

• The critics objected because the conclusions were offensive. The published critiques are dense with recomputation, sensitivity analysis and source criticism. The moral stakes explain the intensity of the reception, not the content of the objections.
• Time on the Cross was refuted and quantitative history discredited with it. Several of its findings stand, and the reply itself was quantitative. What the episode discredited was arithmetic presented without the assumptions that make it mean anything.
• The problem was bad data. The sources were mostly the same ones everybody used, including the Barrow record and the census. The dispute was about construction and interpretation, which is where most quantitative disagreements actually live.
• Fogel abandoned his position. He kept the economic argument, revised specific findings, and changed the moral frame. Distinguishing those three moves is exactly the skill this lesson is teaching.

The takeaway

One ledger, one division, one number, and fifty years of argument about whether the number measures the thing anybody cared about. Fogel and Engerman were right that slavery was profitable and that the enslaved were skilled workers, and those findings changed the field. They were wrong to think that a punishment rate could stand in for how a coercive system worked, and wrong to treat an efficiency index as a fact rather than as a construction resting on choices.

The upshot: before you accept a historical statistic, ask what concept it is standing in for and what would have to be true for the substitution to hold. Checking the arithmetic is the easy half and almost never the half that is wrong.

Sources

1. Fogel, R. W., & Engerman, S. L. (1974). Time on the Cross: The Economics of American Negro Slavery. Little, Brown. archive.org
2. Gutman, H. G. (1975). Slavery and the Numbers Game: A Critique of Time on the Cross. University of Illinois Press. archive.org
3. David, P. A., Gutman, H. G., Sutch, R., Temin, P., & Wright, G. (1976). Reckoning with Slavery: A Critical Study in the Quantitative History of American Negro Slavery. Oxford University Press.
4. Conrad, A. H., & Meyer, J. R. (1958). The economics of slavery in the ante bellum South. Journal of Political Economy, 66(2), 95-130.
5. Fogel, R. W. (1989). Without Consent or Contract: The Rise and Fall of American Slavery. W. W. Norton.

A clerk writes down a christening in 1538

In September 1538 Thomas Cromwell instructed every parish in England to keep a book and to enter in it every wedding, every christening and every burial. Compliance was patchy at first and then general. Four centuries later, in 1981, E. A. Wrigley and R. S. Schofield used a sample of 404 of those parish registers to reconstruct the population of England from 1541 to 1871, year by year, before any census existed.

That reconstruction is one of the great achievements of historical demography, and everything difficult about historical statistics is visible in it. The registers do not record births; they record baptisms, which happened days or weeks later, and only for children who lived that long and whose parents belonged to the established church. They do not record deaths; they record burials in consecrated ground. The 404 parishes were not a random sample of English parishes: they were the ones whose registers were complete and legible enough to use, which is a selection on record quality, and record quality correlates with parish wealth and clerical diligence.

Wrigley and Schofield knew all of this. Their method, back projection, works forwards and backwards from later census totals to constrain the earlier estimates, and it carries explicit correction factors for under-registration. The numbers are defensible precisely because the corrections are stated. What you must never do is copy the resulting figure into your own work as though somebody had counted.

Five questions to put to any historical number

Run these in order on any statistic you propose to cite. They take ten minutes and they will stop most of the errors you would otherwise make.

1. Who counted, and what were they trying to accomplish? Nobody collects data for the fun of it. Tax rolls exist to assess, muster rolls to conscript, censuses to apportion representation or plan rationing. The purpose determines what was counted carefully and what was estimated.
2. What was the unit, and did the unit change? A household is not a family. A ton may be long, short or metric. A pound sterling in 1780 is not a pound in 1850. Series that look continuous often contain a redefinition, and the break usually shows up as a step in the graph that gets explained as an event.
3. What is the denominator, and is it from the same source? Rates require two numbers. If the numerator comes from court records and the denominator from a census taken in a different year on different boundaries, the rate can move without anything in the world moving.
4. Who was excluded by definition? Every category has an edge written by someone. Unemployment figures exclude those who stopped looking; crime figures exclude what was not reported; a slave schedule that records age, sex and colour but no names has excluded personhood by design.
5. What incentive did the counter and the counted have? Undercounting to reduce tax, overcounting to increase a grant, and rounding to the nearest convenient figure all leave signatures. Heaping of reported ages on numbers ending in 0 and 5 is the classic one, and it is easy to test for.

Remember: a historical statistic is the output of an administrative process. Study the process and the number becomes interpretable; skip the process and you are quoting a rumour with decimal places.

The same data, one changed input, an opposite conclusion

Committals to trial in England rise steeply through the first half of the nineteenth century. Read as a crime rate, the series describes a society coming apart under industrialisation. Now change one input: notice that the machinery for detecting and prosecuting crime was being built during exactly those decades. County forces were authorised in 1839 and made compulsory in 1856, prosecution costs were increasingly reimbursed from public funds, and summary jurisdiction expanded so that offences previously ignored were processed.

The same series now describes a state acquiring the capacity to notice. Which reading is right is a real historical question that has occupied specialists for decades, and the answer is contested and probably partial in both directions. The methodological point is not contested at all: a count of recorded events measures recording as well as events, and where the recording apparatus changed during the period covered, the two cannot be separated by looking at the series alone.

This generalises brutally. Rising diagnoses of a disease may be rising incidence or a new diagnostic category. Rising literacy measured by signatures on marriage registers may be rising literacy or a change in who was asked to sign. Before you explain a trend, establish that the measuring instrument held still.

What changes when there are a million documents

The digitised corpus is a genuinely new instrument, and it changes the kind of question you can ask rather than merely the speed of answering old ones. Franco Moretti's term for the approach is distant reading: instead of interpreting individual texts, you count features across a corpus too large to read, and the object of study becomes the shape of the whole.

Here is the worked case. The Proceedings of the Old Bailey, published for the London central criminal court, cover roughly two centuries from 1674 and record about two hundred thousand trials. They have been transcribed and made searchable, which means the value of the goods in every larceny indictment can be counted.

Do that counting and a spike appears at thirty-nine shillings. It is not a natural feature of what thieves stole. Under an eighteenth-century statute, theft of goods worth forty shillings or more from a dwelling house was a capital felony without benefit of clergy. Juries who did not want to hang a defendant valued the goods at thirty-nine shillings, and the same pattern appears just below the other capital thresholds. William Blackstone called the practice a kind of pious perjury. A comparable spike sits just under the five-shilling line for theft from a shop.

Now notice what kind of finding this is. No single trial record shows it; each valuation looks like an ordinary assessment. It becomes visible only in aggregate, and once visible it is decisive evidence that juries were systematically nullifying a statute they would not enforce. That is the case for distant reading in one example: some historical facts exist only at the scale of the corpus.

Where it goes wrong

Four failure modes, each of which has produced published nonsense.

The second row deserves a concrete example. Printing before about 1800 uses the long s, a character that optical character recognition frequently reads as f, so that success becomes fuccefs. If your search term contains an s in the middle of a word, your recall in early material may be far below what you assume, and the shortfall will be concentrated in the earliest decades, producing a fake upward trend.

The fourth row is the one that catches experienced people. The Old Bailey Proceedings grew far longer over time as the publication became more official and more nearly complete. A graph of average words per trial therefore charts an editorial regime, not the behaviour of courts, unless you control for it.

What matters here: the digital methods do not replace the five questions. They make the five questions harder to ask, because the interface hides the collection process behind a search box that answers instantly and confidently.

Common misconceptions

• Bigger data means fewer assumptions. Scale removes sampling error and does nothing whatever about selection bias, definitional drift or measurement validity. A million badly defined observations are still badly defined.
• Historical statistics are either reliable or useless. Almost all of them are usable within a stated range for a stated purpose. The skill lies in specifying the range, not in accepting or rejecting the number.
• Full-text search finds everything about a topic. It finds documents containing a string. Contemporaries used different words, and the words you would use may not have existed. Build your search terms from the sources, not from your own vocabulary.
• A striking pattern in a corpus is a finding. A striking pattern is a hypothesis about the corpus. It becomes a finding about the past only after you have excluded changes in what was collected, transcribed and published.

Putting it together

Parish registers record baptisms and not births, and a great reconstruction of English population history is trustworthy because its authors said so and corrected for it. Committals to trial rose while the apparatus for producing committals was being built, so the series measures both at once. Two hundred thousand trials reveal a spike at thirty-nine shillings that no individual trial could show, and that spike is real evidence about juries nullifying capital statutes. The same discipline runs through all three: identify the process that generated the number, then say what the number can carry.

The upshot: ask who counted and why, what the unit was, where the denominator came from, who was excluded and who had an incentive to shade the total. Then, at any scale, plot how much data you have per year before you plot anything else.

Sources

1. Wrigley, E. A., & Schofield, R. S. (1981). The Population History of England, 1541-1871: A Reconstruction. Edward Arnold. archive.org
2. Moretti, F. (2005). Graphs, Maps, Trees: Abstract Models for a Literary History. Verso. archive.org
3. Hitchcock, T., & Shoemaker, R. (2015). London Lives: Poverty, Crime and the Making of a Modern City, 1690-1800. Cambridge University Press.
4. Blackstone, W. (1769). Commentaries on the Laws of England, Book the Fourth: Of Public Wrongs. Clarendon Press.
5. Gatrell, V. A. C., & Hadden, T. B. (1972). Criminal statistics and their interpretation. In E. A. Wrigley (Ed.), Nineteenth-Century Society (pp. 336-396). Cambridge University Press.

A first book with a second book attached

In 1824 a twenty-eight-year-old schoolmaster from Thuringia published a history of the Latin and Germanic nations between 1494 and 1514. Bound with it was a second, shorter work most readers now skip, called A Critique of Modern Historical Writers. In it Leopold Ranke took the most respected narrative of the Italian wars, by Francesco Guicciardini, set it beside the documents, and showed passage by passage where it had been invented, borrowed or embroidered.

The preface to the main volume contains the sentence everybody quotes. Ranke wrote that history had been assigned the office of judging the past and instructing the present for the benefit of future ages, that his book aspired to no such office, and that it wished only to show, in his phrase, "wie es eigentlich gewesen".

Six German words, translated into English for a century as "how it actually was", and used ever since as a slogan for a doctrine Ranke did not hold. Getting this right is worth a lesson, because the misreading still shapes what people think objectivity in history means.

What the sentence was doing

Read it as a refusal rather than a programme. The first half rejects two jobs then thought proper to history: passing moral judgement on the dead, and mining the past for lessons to improve the present. Ranke is declining both. The claim of the second half is modest by comparison, and its force is comparative: this book will report what the evidence supports, not what the reader would like to hear.

Then there is the word eigentlich. It can mean actually, in the sense of really happened. It can also mean essentially, in the sense of what the thing was at bottom. Ranke was a religious man who believed that every age stands in a direct relation to God, and who saw the historian's task as discerning the particular character of an epoch rather than assembling a chronicle of verified incidents. On that reading, the sentence promises insight into what an age essentially was, which is nearly the opposite of a demand for bare facts.

Key idea: Ranke's famous phrase is a rejection of moralising and of history as a source of lessons, and an assertion that each age must be understood in its own terms. It is not a claim that historians can escape interpretation, and he never made one.

The method underneath the slogan

What Ranke did contribute, and what has genuinely lasted, is a discipline of source criticism. It runs in four steps, and you already met three of them in this course's first lesson.

1. Prefer the source closest to the event and to the decision. Ranke wanted the report written by the participant for internal use, not the polished narrative composed afterwards for readers.
2. Establish the filiation of texts. When two accounts agree, ask whether that is independent confirmation or one copying the other. Much of the Guicciardini critique consists of tracing where a supposedly independent witness is repeating an earlier chronicle.
3. Test the writer's position. What did this author know, when, and from whom, and what did he stand to gain?
4. Cite so the reader can check. Ranke's footnotes are not decoration. They put the evidence where a hostile reader can reach it, which is what converts an assertion into a claim.

His archive of choice was diplomatic. In Vienna, Venice, Rome and Florence he worked through the relazioni, the reports Venetian ambassadors were required to deliver on returning from a posting. These are extraordinary documents: shrewd, detailed, written by professionals for a government that would act on them, and largely unread by historians before him. From them came his studies of the Ottoman and Spanish empires and his history of the popes.

The seminar, and how a method became a profession

Ranke went to Berlin in 1825 and began holding historical exercises for a small group of students. They were given manuscripts and made to do the work: date the document, establish its transmission, compare it with the printed account, defend the reading in front of the room.

That format is now so ordinary that its invention is invisible. It produced a generation of trained researchers rather than gifted amateurs, and it made history a discipline with an apprenticeship and a standard of proof. When American universities built graduate programmes in the 1870s and 1880s, they imported the seminar directly. Herbert Baxter Adams ran one at Johns Hopkins that trained a large share of the first professional generation in the United States.

The transplant carried a mistranslation with it. American and British admirers took Ranke to be the founder of a scientific history that would establish facts and refrain from interpretation, and they built professional norms on that reading. Peter Novick's study of American historical writing devotes close attention to what he treats as a cult built on a misunderstanding: the idealist Ranke of Berlin was recast as an empiricist by people who mostly had not read his theoretical writing and could not read his German.

What the choice of archive cost

Every method carries the shape of its evidence, and Ranke's evidence was produced by states.

Two consequences follow, and both drive the rest of this module. First, if the documents are diplomatic, foreign policy will look like the motor of history, which is why the doctrine of the primacy of foreign policy is so at home in this tradition. Second, if the state is the record-keeper, the state becomes the natural unit of analysis, and national history looks like the obvious container for the past.

The next two lessons are both revolts against that. Thompson goes after the assumption that the people who left the documents are the people who made the history. The Annales group goes after the assumption that the event, and the year, and the nation are the right scales at all.

What survives, and what does not

It is easy to treat Ranke as a period costume. That is a mistake; a large part of what you will actually do at a desk comes from him.

Still standing: the requirement to work from the record rather than from earlier narratives; the insistence on tracing where an account came from before treating agreement as corroboration; the citation that lets a reader check; the idea that a period must be understood in its own categories rather than scored against ours; and the seminar itself.

Not standing: the belief that the state archive is the archive; the assumption that political and diplomatic history is the trunk from which other histories branch; and the vision, which Ranke held sincerely, of a providential order visible through the sequence of states. And of course the slogan, which is quoted almost exclusively by people arguing for a position its author did not hold.

Common misconceptions

• Ranke believed historians could report facts without interpreting them. He believed each age had a character that the historian must discern, and he wrote about the hand of God in history. The bare-facts Ranke is a nineteenth-century Anglo-American construction.
• Wie es eigentlich gewesen means only how it actually was. Eigentlich also carries essentially, and Ranke's other writing makes that reading at least as plausible. The translation choice is itself a historiographical argument.
• Ranke invented the use of archives. Scholars had used documents for centuries, and antiquarians and the Maurist Benedictines had developed diplomatic criticism long before. What he built was a training system that made archival criticism the entry requirement of a profession.
• Objectivity in history was born with Ranke and died with postmodernism. The objectivity question has been contested continuously since, and the lesson on Novick and Hayden White later in this course takes up the argument in its modern form.

Looking back

A young man published a book with a critique of Guicciardini attached to it, then spent his life in the Venetian dispatches and taught students to date manuscripts in a room in Berlin. What he asked for was a history built from the papers people wrote while acting, checked for filiation, cited so a reader can verify, and judged by the standards of its own age rather than ours. What he was later said to have asked for, a history of bare facts without interpretation, is a mistranslation that hardened into a doctrine.

Bottom line: take the source criticism and the seminar, which are his, and leave the slogan, which is not what he meant. Then notice that his archive decided his subject, and ask what your archive is deciding for you.

Sources

1. Ranke, L. von (1887). History of the Latin and Teutonic Nations, 1494 to 1514 (P. A. Ashworth, Trans.). George Bell & Sons. archive.org
2. Novick, P. (1988). That Noble Dream: The Objectivity Question and the American Historical Profession. Cambridge University Press. archive.org
3. Iggers, G. G. (1962). The image of Ranke in American and German historical thought. History and Theory, 2(1), 17-40.
4. Ranke, L. von (1973). The Theory and Practice of History (G. G. Iggers & K. von Moltke, Eds.). Bobbs-Merrill.
5. Grafton, A. (1997). The Footnote: A Curious History. Harvard University Press.

A crowd that hands over the money

In the autumn of 1766 crowds moved through markets across the west of England. They stopped grain wagons, opened sacks, and sold the contents at a price they announced themselves, well below what the dealer was asking. Then, in a great many recorded cases, they handed the dealer the proceeds and went home.

That last detail wrecks the standard explanation. If hunger produces riot the way a stimulus produces a reflex, the grain is stolen, not sold, and nobody stops to count out coins for a man they have just overpowered. Something else is going on, and the something else is a claim about what a price ought to be and who has the right to set it.

E. P. Thompson called the thing he was arguing against the spasmodic view of popular history: the assumption that ordinary people appear in the record only as a reaction, that their politics is their stomach, and that the historian's job is to correlate riots with the price of bread. Against it he set what he named the moral economy of the crowd, a set of shared and quite specific expectations about marketing, forestalling, and the obligations of dealers, expectations with statutory ancestry that magistrates themselves half accepted.

What the crowd knew about the law

Look at what the rioters were doing and the coherence is unmistakable. They targeted middlemen and exporters rather than farmers selling directly. They enforced the customary requirement that grain be sold in open market before being sold by sample. They invoked, sometimes by name, the old statutes against forestalling and engrossing, which had been repealed but which magistrates still occasionally revived under pressure. They set a price rather than seizing goods, and they were often orderly enough that participants could be named in depositions afterwards without much dispute about what had happened.

So the riot is not a failure of order. It is the enforcement of a competing order, by people who believed the authorities had abandoned an obligation, and it is legible only if you grant that the participants had a theory. Granting that is the whole method.

The point: when a group's behaviour looks irrational, the first hypothesis to test is that you have not yet found their reasons. Look for what they did not do as well as what they did; the restraint is where the theory shows.

Class as a relationship, not a box

Thompson published The Making of the English Working Class in 1963, covering roughly 1780 to 1832. Its preface contains two sentences that reoriented a generation.

The first announces the purpose: he is seeking to rescue the poor stockinger, the Luddite cropper, the obsolete hand-loom weaver and the rest from what he calls "the enormous condescension of posterity". The second states the theory: the working class "was present at its own making".

Both sentences are attacks on a way of doing Marxist history that Thompson found intolerable. In that way of working, a class is a position in the relations of production, occupied by whoever the economy places there, and consciousness either matches that position or is false. Thompson refuses the whole apparatus. Class, he insists, is a relationship and not a thing. It happens when people, out of shared experience, come to feel and articulate an identity of interests among themselves and against others whose interests differ. It has no existence apart from the people doing that, and therefore it has a history rather than a definition.

Notice what this commits you to in an archive. If class is made rather than allocated, you cannot establish it by counting occupations in a census. You have to find people articulating it: in club rules, hymn books, oaths, trial testimony, subscription lists, letters to radical newspapers, the minutes of a corresponding society. Thompson's book is built from exactly those materials, together with Home Office files, informers' reports and prosecution briefs, which are hostile sources read against the grain in the sense of the earlier module.

Where this stands in the Marxist tradition

Thompson is arguing inside a tradition, not against it. Marx's own formulation in 1852 is that men make their own history, but not under circumstances of their own choosing, and Thompson is pulling hard on the first half of that sentence. His polemic against structuralism, published in 1978, attacks a version of Marxism in which structures do the acting and human beings are the places where structures meet.

The historiographical school this produced is usually called history from below, and it has a simple procedural core: take the beliefs of the people you study seriously enough to reconstruct them, and treat evidence of their agency as evidence rather than as background. Its practitioners after Thompson include historians of slavery, peasant revolt, gender and colonial resistance, and the subaltern studies group is a direct descendant, as a later lesson shows.

Two worked examples of the method

Thompson's essays are better teaching material than the book, because in an essay you can watch the whole operation.

The second is worth dwelling on. Thompson noticed that before industrial employment, work was task-oriented: you worked until the job was done, hard at some times and hardly at all at others, and Monday was frequently taken off. Factory employment is time-oriented: you sell hours. He then found the evidence of the transition in unlikely places, including fines for lateness, the spread of cheap watches, sermons against idleness and the disappearance of Saint Monday. A change in the experience of time is not an abstraction once you have located the objects that carried it.

The objections that stuck

The book is nearly sixty years old and it has taken damage. The important criticisms are not that Thompson was political, which he never concealed, but that his source base determined his findings in ways he did not fully acknowledge.

• The working class is largely male. Women appear as supporters. Barbara Taylor's work on Owenite socialism and later studies of gender in the artisan trades show that a differently chosen source base yields a different class formation, one in which the exclusion of women from trades was internal to the making.
• The articulate stand in for everyone. Corresponding societies, radical newspapers and trial testimony over-represent literate artisans. The agricultural labourer and the domestic servant left far less, and their absence from the book is a property of the archive as much as of the period.
• The unity is overstated. Regional and trade differences were large; a Lancashire spinner and a London compositor did not obviously share an identity of interests, and treating them as one formation involves an argument that the book sometimes assumes.
• The narrative has a destination. If class is made, it can also be unmade or made differently, but a book called The Making, ending in 1832, invites reading as a story with an arrival point.

Common misconceptions

• History from below means writing only about poor people. It means treating the beliefs and agency of the ruled as evidence rather than as noise. You still have to read the Home Office files, and Thompson read a great many of them.
• Thompson denied that economic structures matter. He denied that they determine consciousness. The book is full of wages, prices, machinery and the reorganisation of trades; what he refuses is the inference from position to belief.
• The moral economy argument says food riots were not about hunger. Hunger is the occasion. The argument is about form: what the crowd did, whom it spared, and what it thought it was entitled to enforce.
• Class consciousness in Thompson is the same as class identity today. His concept is historical and relational: it exists when particular people articulate shared interests against others, in particular conditions, and it can dissolve.

What you now know

A crowd that sets a price and hands over the takings is not a hunger reflex, and noticing that is a method rather than an anecdote. Thompson's claim is that class is not a slot people occupy but a relationship they make, which means the historian must find it being articulated in the sources rather than deduce it from occupations. That commitment produces a distinctive archive: clubs, oaths, hymns, trials, informers' reports. It also produces distinctive blind spots, because the people who leave articulate records are not a cross-section, and the class he made visible was largely male and largely skilled.

Worth holding on to: when your subjects look irrational, look harder for their reasons, and when they look unified, ask which parts of your evidence supplied the unity.

Sources

1. Thompson, E. P. (1963). The Making of the English Working Class. Victor Gollancz. archive.org
2. Thompson, E. P. (1978). The Poverty of Theory and Other Essays. Merlin Press. archive.org
3. Thompson, E. P. (1971). The moral economy of the English crowd in the eighteenth century. Past and Present, 50, 76-136.
4. Thompson, E. P. (1967). Time, work-discipline, and industrial capitalism. Past and Present, 38, 56-97.
5. Taylor, B. (1983). Eve and the New Jerusalem: Socialism and Feminism in the Nineteenth Century. Virago.

A thousand-page book written in a prison camp

Fernand Braudel was captured in 1940 and spent five years in German officers' camps at Mainz and Lubeck. He had no library. He wrote the draft of a book about the Mediterranean in the age of Philip II in school exercise books, largely from memory and from notes taken before the war, and posted the notebooks in instalments to Lucien Febvre in France. The book appeared in 1949, ran to well over a thousand pages, and rearranged what a work of history was allowed to look like.

Its first two hundred pages contain almost no people. They are about mountains, plains, coastlines, the distance a ship could cover, where olives will grow and where they will not. Philip II, the nominal subject, does not die until the final page.

What the founders were against

The journal that gave the group its name was founded at Strasbourg in 1929 by Marc Bloch and Lucien Febvre. What the Annales opposed was the kind of history the previous lesson traced to the state archive: a narrative of treaties, ministers and battles, organised by year, written from diplomatic papers.

Their objections were three. Such history takes the actor's own account of what mattered; it mistakes the surface for the whole; and it cannot ask most of the interesting questions, because those questions are about things that change too slowly to appear in a chronicle. In place of narrative they wanted problem history, in which the historian poses a question and then goes looking for whatever evidence answers it, from any discipline: geography, economics, sociology, linguistics, climate data, field patterns visible from the air.

Two of their books show the programme working. Bloch's study of the royal touch examined the belief, held in France and England for centuries, that a king could cure scrofula by touching the sufferer. He was not asking whether it worked. He was asking what kind of monarchy is sustained by a population that believes it does, which converts a curiosity into evidence about the structure of political authority. Febvre's book on Rabelais asked whether a sixteenth-century writer could have been an atheist, and answered that the mental equipment of the period, its available vocabulary and concepts, did not yet make the position formulable. That is Trouillot's unthinkable history arriving from a different direction.

The three durations, side by side

Braudel's contribution was to insist that these are not competing histories but different speeds, and that a serious account needs all three at once.

Read the table across a single case and the argument becomes concrete. Lepanto in 1571 is an event: a fleet action, a date, casualties. At the level of conjoncture it is one moment in the long contraction of Mediterranean galley warfare and the shift of European wealth and attention towards the Atlantic. At the level of the longue duree, the fact that governs everything is that a galley cannot carry enough water to stay at sea for long, which ties Mediterranean naval power to coastlines and seasons in a way no admiral could change.

In short: the three durations are not three opinions about what caused something. They are three questions: what was possible, what was likely, and what happened. Confusing them is the commonest error in causal argument, and the lesson on causation later in this course returns to it.

The price of the long view

Braudel's own metaphor gives the objection away. Events, he wrote, are surface disturbances, crests of foam carried on the strong backs of the tides. It is a wonderful image and it makes human beings into weather.

Three costs follow. Agency thins out: if structures do the work, then decisions, arguments and revolts are decoration, which is precisely the complaint Thompson made against structuralism in the previous lesson. Politics vanishes, and with it the study of power as something exercised by identifiable people. And the method has an appetite for data it can serialise, so it gravitates towards prices, baptisms and tonnages, and away from anything that survives only as one document.

Bloch himself never took the position that far. His unfinished book on the historian's craft, written while he was in the Resistance and left incomplete when the Gestapo shot him in June 1944, is a defence of history as a human discipline in which the historian's own questions are the starting point. It remains the best short statement of the case.

How the tradition answered itself

The third generation, working from the 1960s, turned back towards people without abandoning the long view. The turn ran through the history of collective attitudes and then to the very small scale.

Emmanuel Le Roy Ladurie published Montaillou in 1975. Its source is a single inquisition register compiled by Jacques Fournier, bishop of Pamiers and later Pope Benedict XII, who between 1318 and 1325 interrogated the inhabitants of a Pyrenean village suspected of Cathar sympathies with exceptional thoroughness. Out of it Le Roy Ladurie reconstructed households, sheep migrations, sexual conduct, attitudes to childhood and death, and the politics of a settlement of perhaps two hundred and fifty people. The book sold in hundreds of thousands.

Notice the family resemblance to Ginzburg's miller: the same kind of source, produced by the same kind of coercion, read for the same kind of thing. Italian practitioners such as Giovanni Levi gave the approach a name, microhistory, and a rationale: reducing the scale of observation reveals mechanisms that aggregate data smooths away. A price series tells you that grain cost more in 1590. One village's records tell you who then borrowed from whom, on what terms, and what that did to the standing of a family, which is the process the series is a summary of.

Le Roy Ladurie also wrote, in the early 1970s, that the historian of tomorrow would be a programmer or would be nothing. He was half right, in the way that predictions usually are, and the digital lesson earlier in this course is the audit of the other half.

Common misconceptions

• The Annales rejected narrative in favour of statistics. They rejected the primacy of the political event. The output includes Bloch on the royal touch and Le Roy Ladurie on a village, neither of which is quantitative.
• The longue duree means nothing changes. It means some things change on a scale that a chronology of reigns cannot register. Climate, soil exhaustion and shipping technology all change; they change over centuries.
• Microhistory abandoned the Annales project. It was largely carried out by people trained inside it, and it kept the commitment to structures while changing the scale at which they are observed.
• Braudel wrote his book from documents gathered during the war. He was a prisoner without a library. The archival research came before 1940, and the drafting was done from memory and notes, which is worth remembering when the book's confidence intimidates you.

Summing up

A journal founded in 1929 argued that the history of events takes the actors' word for what mattered, and proposed instead a history built from questions. Braudel turned that into a structure: the almost motionless history of the environment, the slow rhythms of prices and empires, and the froth of events, each with its own sources and its own kind of cause. The cost was human agency, and the tradition corrected itself by shrinking the scale until individual people came back into focus in a Pyrenean village and a Friulian mill.

What matters here: before you explain anything, decide which duration your question belongs to, and do not answer a question about constraint with evidence about contingency, or the reverse.

Sources

1. Braudel, F. (1972). The Mediterranean and the Mediterranean World in the Age of Philip II (S. Reynolds, Trans.). Collins. archive.org
2. Le Roy Ladurie, E. (1978). Montaillou: The Promised Land of Error (B. Bray, Trans.). George Braziller. archive.org
3. Bloch, M. (1953). The Historian's Craft (P. Putnam, Trans.). Alfred A. Knopf.
4. Febvre, L. (1982). The Problem of Unbelief in the Sixteenth Century: The Religion of Rabelais (B. Gottlieb, Trans.). Harvard University Press.
5. Braudel, F. (1958). Histoire et sciences sociales: la longue duree. Annales. Economies, Societes, Civilisations, 13(4), 725-753.

A joke you do not get

Some time in the late 1730s, in a printing shop in the rue Saint-Severin in Paris, the apprentices decided to kill the cats. They collected the strays of the neighbourhood and the mistress's favourite, held a mock trial with a confessor and an executioner, and hanged them. Afterwards they re-enacted the whole thing for weeks, and the account we have, written by a former apprentice named Nicolas Contat, records it as the funniest episode of his working life.

You do not find it funny. Robert Darnton, who put the story at the centre of a book in 1984, argued that the failure is the opening. Where a joke does not travel, you have located the edge of a system of meaning, and the way in is to reconstruct what cats, apprentices, masters and mock trials meant to the people laughing.

Do that, and the massacre stops being random cruelty. Cats carried associations with witchcraft, with sexuality and with the mistress of a household. Apprentices slept in freezing rooms, ate scraps while the master's cats ate well, and were losing any prospect of ever becoming masters themselves. A mock trial mimics the legal ceremony of a society saturated with legal ceremony. The killing is an attack on the master's household that could not be prosecuted as such, and its comedy is the comedy of an insult delivered under cover.

Thick description

The method has a name and a source. Clifford Geertz borrowed thick description from the philosopher Gilbert Ryle and made it the programme of interpretive anthropology in 1973.

Ryle's example: two boys rapidly contract the right eyelid. One has a twitch. The other is winking at a friend. A camera records identical movements. The difference between them is not in the muscle but in a public code, and a third boy might be parodying the second, and a fourth rehearsing the parody. Thin description reports the movement. Thick description reports what the movement is doing inside a system that the participants share and that the observer must learn.

Geertz's famous demonstration is the Balinese cockfight. He treats it not as a sport but as a text the Balinese have composed about themselves: about status, about kinship rivalry, about what may be expressed only sideways. He borrows Bentham's phrase deep play for wagering at stakes so high that entering is irrational by any calculation of utility, and argues that the irrationality is the point, because what is really staked is standing rather than money.

For historians the transfer was electric, and the consequence is a shift in what counts as evidence. If culture is a system of public meanings, then a ritual, a riot, a funeral, a shop-floor prank and a cat massacre are all readable. The archive stops being a place to find out what happened and becomes a place to find out what things meant.

Why this matters: thick description converts your failure to understand into a research question. When a source produces bafflement, that reaction is data about the distance between two systems of meaning, and the distance is what you are trying to measure.

How far past the evidence may you go?

The approach has an obvious hazard, and it produced the sharpest methodological quarrel of the 1980s.

Natalie Zemon Davis published The Return of Martin Guerre in 1983, on the sixteenth-century case of a peasant impostor accepted for years as a woman's returned husband. The trial record is rich but silent on the question everyone wants answered: did Bertrande de Rols know? Davis argued that she probably did, and reconstructed her calculations from what a woman in her position could and could not do.

Robert Finlay replied in the American Historical Review in 1988 that Davis had written fiction with footnotes: that the record shows Bertrande protesting her deception, and that Davis had preferred a more interesting woman to the documented one. Davis answered in the same issue that the sources contain a great deal about what was possible, that a historian who reports only explicit statements will reproduce the categories of the court, and that inference is legitimate when it is signalled as inference.

Both positions are defensible and the disagreement has never been resolved, which is why it is worth carrying with you. The test that follows from it is procedural: mark the inference. A reader must be able to see, sentence by sentence, where the record stops and your reconstruction begins.

The linguistic turn: emplotment

Hayden White pushed the argument one level further in Metahistory in 1973, and his target was not the sources but the historian's own prose.

White's claim, stated as he would state it: a chronicle is a list; a history is a story; and turning the first into the second requires choices that the evidence does not make for you. You must decide where the account begins and ends, what is foreground and what is background, and what shape the whole has. He called that emplotment, and argued that historians draw on a small stock of forms, romance, tragedy, comedy and satire, along with characteristic modes of argument and political implications. Two historians with identical facts can produce a tragedy and a comedy, and no additional fact settles which is correct, because the choice is made at the level of form.

This is not the claim that historians invent evidence. White accepted that statements about the past can be true or false. His claim is that the narrative shape is underdetermined by them.

Peter Novick arrived from another direction in 1988 with a history of the American profession's commitment to objectivity, treating it as a founding myth: professionally necessary, philosophically confused, and by the 1980s no longer commanding assent. He ends by quoting the line from Judges about there being no king in Israel, every man doing what was right in his own eyes.

The reply, and a courtroom that had to decide

Richard J. Evans published In Defence of History in 1997. His answer concedes a great deal: historians select, narrate, and bring questions of their own, and no account is neutral. What he denies is that this makes all accounts equal. Evidence constrains. Some reconstructions are ruled out by documents, and a historian who says the trains ran to Auschwitz is not making the same kind of move as one who says they did not.

Then the argument acquired a test with legal consequences. In 1996 David Irving sued Deborah Lipstadt and Penguin Books for libel over her description of him as a Holocaust denier. Under English law the defendants had to prove the truth of what had been written. That turned the case into something no seminar can be: a proceeding in which a court had to decide whether a historian had misrepresented evidence, using the evidence.

Evans spent two years on an expert report that worked through Irving's footnotes: mistranslations, sources cited for claims they do not contain, a document's meaning reversed by omission, statistics altered. In April 2000 Mr Justice Gray's judgment found that Irving had persistently and deliberately misrepresented and manipulated historical evidence, and that he was an active Holocaust denier, antisemitic and racist.

The trial did not refute White, who never claimed that evidence is irrelevant. What it demonstrated is that the practices historians already use, checking a citation against its source, are sufficient to establish that a particular account is false, and that the choice of emplotment does not reach far enough to protect an argument built on doctored footnotes.

What would settle the dispute

Sort the claims and the fight shrinks to something manageable.

Common misconceptions

• Hayden White said history is fiction and facts do not matter. He said narrative form is chosen rather than found, and he accepted that statements about the past have truth values. The stronger claim is usually attributed to him by people arguing against it.
• The Irving trial proved the postmodernists wrong. It proved that a particular body of work misrepresented its sources, established by ordinary source criticism. That is a large thing and a narrow one.
• Thick description means adding more detail. Thin and thick are not degrees of detail. Thin reports the act; thick reports what the act does within a code that participants share.
• Davis was refuted by Finlay. Neither side conceded and both articles remain on reading lists. The transferable lesson is procedural: signal where inference begins.

Recap

Apprentices killed cats and thought it hilarious, and the way into that is to treat incomprehension as a measurement. Geertz supplied the vocabulary: an act has meaning inside a public code, and describing the act without the code explains nothing. Turning that on historians themselves, White argued that the shape of a historical narrative is chosen rather than discovered, and Novick showed how much professional identity had been resting on a philosophically shaky idea of objectivity. Evans conceded the selection and denied the relativism, and a London courtroom then established, footnote by footnote, that one historian's account was dishonest.

The core of it: the evidence does not choose your plot, and your plot cannot survive evidence that contradicts it. Both halves are true, and a historian who forgets either one writes badly in a predictable direction.

Sources

1. Geertz, C. (1973). The Interpretation of Cultures: Selected Essays. Basic Books. archive.org
2. White, H. (1973). Metahistory: The Historical Imagination in Nineteenth-Century Europe. Johns Hopkins University Press. archive.org
3. Darnton, R. (1984). The Great Cat Massacre and Other Episodes in French Cultural History. Basic Books. archive.org
4. Finlay, R. (1988). The refashioning of Martin Guerre. American Historical Review, 93(3), 553-571.
5. Evans, R. J. (2001). Lying About Hitler: History, Holocaust, and the David Irving Trial. Basic Books.

An army that brought its scholars

When Napoleon sailed for Egypt in 1798 his expedition carried, alongside the troops, a commission of more than a hundred and fifty scholars, engineers, draughtsmen and naturalists. Their survey was published over the following decades in a series of enormous volumes: measurements, plans, botanical plates, inscriptions, an account of a country rendered legible.

The invasion failed militarily within three years. The survey did not. It became a model for how a European power would come to know the places it governed, and the knowledge outlived the occupation by a century and a half.

That relation, between the scholarship and the government, is what Edward Said set out to describe in 1978, and the argument that followed reorganised how a generation thought about the position from which history gets written.

1978: Orientalism

Orientalism makes three claims, and they are worth separating because critics have accepted some and rejected others.

First, that a body of European writing about the Middle East and Asia constitutes a discourse: a set of assumptions, images and authorities that reproduce themselves, in which each new work cites earlier works rather than starting from observation. Second, that this discourse works by a founding distinction between an Orient and an Occident, assigning to the first a fixed essence, typically despotic, sensual, unchanging and incapable of self-government. Third, that this was entangled with imperial administration rather than parallel to it: philologists trained officials, officials commissioned philologists, and the claim to know a people supported the claim to rule them.

Said drew on Michel Foucault for discourse and on Antonio Gramsci for hegemony, and worked through Silvestre de Sacy, Ernest Renan, Edward William Lane and the statesmen who quoted them.

The objections came quickly and some have force. Bernard Lewis argued that Said misread particular scholars and ignored the philological achievements the tradition genuinely produced. Aijaz Ahmad, writing from the left, argued that Said had replaced class analysis with a story about representation and had made the West as monolithic as the Orient it invented. Sadik Jalal al-Azm pointed out that the same essentialising move can be run in reverse by nationalists in the region. A recurrent technical criticism is that German Orientalist scholarship, arguably the largest body of it, belonged to states with no empire in those regions, which weakens a tight link between knowledge and rule.

Remember: the durable part of Said is not the verdict on any individual scholar. It is the demand that you ask who is licensed to describe whom, and what institutional arrangement makes that description authoritative.

1982: the subaltern turn

In India, Ranajit Guha launched the Subaltern Studies series in 1982, borrowing the word subaltern from Gramsci. The target was double. Colonialist historiography credited British administration with whatever modernity India possessed. Nationalist historiography credited the Congress elite. Both, Guha argued, treated peasants as material to be mobilised by somebody else.

Against this the group asserted an autonomous domain of subaltern politics. Guha's 1983 book reads the record of a century of peasant insurgency for its internal logic and finds one: the choice of targets, the inversion of symbols of authority, the way rumour spread, the territorial limits of a rising. Its evidence is almost entirely official, which is why the essay on the prose of counter-insurgency, met earlier in this course, had to be written first.

1988: the objection from inside

Gayatri Chakravorty Spivak published Can the Subaltern Speak? as a friend of the project and a critic of its central hope.

Her argument runs like this. The subaltern is not a person waiting behind the archive to be released. The category is constituted by the discourses that record it, so the recovered voice arrives already shaped by the recording. Her case is the British prohibition of widow immolation, where the colonial justification took the form she summarises as white men saving brown women from brown men, while the indigenous defence claimed the women wished to die. Between two authorised accounts, no third position is available. She closes with a young woman who hanged herself in Calcutta in 1926, having waited for her period so the death could not be read as an illicit pregnancy, and whose family read it that way anyway. Her conclusion, that the subaltern cannot speak, is not a claim that such people were silent. It is a claim that speech requires a listening position, and that the institutions did not supply one.

She also offered the phrase strategic essentialism for the tactic of using a simplified collective identity for political purposes while knowing it is a simplification, and later distanced herself from it because it was taken as a licence rather than as a temporary expedient.

2000: two books that answer the same question differently

Two influential books published in the same year show the fork in the road.

Dipesh Chakrabarty argued in Provincializing Europe that European political thought is at once indispensable and inadequate for writing the histories of the rest of the world: indispensable because the categories of citizen, state, rights and modernity are unavoidable, inadequate because they carry a schema in which other societies are always arriving late. He calls that schema the waiting room of history, the not yet that colonial administrators used to defer self-government. His most demanding point is about writing: a discipline committed to secular explanation must translate a peasant's gods into social causes, because the gods cannot be actors in an academic history. The translation is not neutral, and there is no version of the discipline in which it is unnecessary.

Kenneth Pomeranz answered the Eurocentrism problem with economics instead. The Great Divergence compares the most advanced regions of eighteenth-century China and Europe rather than comparing continents or nations, and finds them broadly similar as late as about 1750 in life expectancy, market development, consumption and agricultural productivity. If the two were comparable then, no long-run European cultural superiority can explain what followed. His candidates are contingent and material: coal that happened to lie near English industry and water transport, and the land-saving resources of the New World.

The method matters more than the conclusion. Pomeranz calls it reciprocal comparison: instead of taking Europe as the norm and asking why China failed to follow, you ask each question in both directions, treating each case as the deviation from the other. It is a cheap technique and it exposes an enormous amount of hidden Eurocentrism in ordinary comparative questions.

The case against global history

By the 2010s global history was the most fashionable subfield in the discipline, and the criticisms had accumulated.

The replies are serious too. Richard Drayton and David Motadel answered Adelman by pointing out that global history has never been a single political project, that connected histories were written by scholars from the global south long before the fashion, and that the alternative, methodological nationalism, has a far worse record.

Common misconceptions

• Said argued that Western scholarship on Asia is worthless. He argued that it forms a self-referential discourse entangled with rule. Several of his critics accept that and still defend particular philological achievements.
• Spivak was attacking Subaltern Studies from outside. She wrote as a participant, and her essay is a limit case for the project rather than a dismissal of it.
• Provincializing Europe calls for abandoning European categories. Chakrabarty explicitly denies that. His claim is that they are indispensable and inadequate at once, which is a harder position to occupy than rejection.
• Global history means writing about everywhere. Pomeranz compares two regions, closely. Scale is not the method; the method is refusing to let one case supply the standard for the other.

Where this leaves us

A survey of Egypt outlived the occupation that produced it, and Said used that pattern to argue that knowledge about the colonised was not innocent of the arrangements that made it possible. Guha's group tried to write the history of people the archive records only as targets. Spivak showed that recovery cannot be as simple as reading harder, because the category itself is produced by the recording. Chakrabarty made the problem one of translation rather than of evidence. Pomeranz answered with a technique: compare in both directions and refuse to let one case be the norm. The critics of global history, in turn, ask who is doing the comparing and in whose language.

Bottom line: ask of any project, including your own, what position it is written from, what unit it treats as natural, and which comparison it has quietly made the standard. Then run the comparison the other way and see what changes.

Sources

1. Said, E. W. (1978). Orientalism. Pantheon Books. archive.org
2. Chakrabarty, D. (2000). Provincializing Europe: Postcolonial Thought and Historical Difference. Princeton University Press. archive.org
3. Pomeranz, K. (2000). The Great Divergence: China, Europe, and the Making of the Modern World Economy. Princeton University Press. archive.org
4. Adelman, J. (2017). What is global history now? Aeon. aeon.co
5. Spivak, G. C. (1988). Can the subaltern speak? In C. Nelson & L. Grossberg (Eds.), Marxism and the Interpretation of Culture (pp. 271-313). University of Illinois Press.

A decree with no women in the room

On 30 October 1793 the French National Convention closed all women's political clubs. The case for the decree was made by Jean-Pierre-Andre Amar on behalf of the Committee of General Security, and he did not argue from disorder or from a particular incident. He argued from nature: that women's constitution unfitted them for public deliberation, that their proper sphere was the household, and that the moral order of the Republic depended on the distinction being kept.

Now here is the question that organises this lesson. You can write the history of that decree as an episode in the history of women, and you should. But notice that the argument Amar used was also doing something else. It was defining what a citizen was, and what public reason was, and what the Republic was, by saying what those things were not. There were no women in the Committee of General Security. The category was at work anyway.

What was wrong with the way it was being done

Joan Wallach Scott published an article in the American Historical Review in December 1986 that named the problem precisely. By then women's history was two decades old and had recovered enormous amounts of material. Her complaint was about what that recovery had and had not changed.

Two failure modes. The first she called the descriptive use of gender: the word functioning as a polite synonym for women, so that a book with gender in the title is a book about women, and the field remains a specialism that other historians can decline to read. The second is the additive model: the master narrative stays intact and a chapter is inserted. The chapter on women in the French Revolution sits beside the chapters on the Terror and the war, and the account of the Terror is unchanged by its presence.

Scott also rejected three theoretical solutions then on offer. Patriarchy theory grounded subordination in a universal male drive to control reproduction, which explains everything and therefore predicts nothing historical. Marxist feminism kept subordinating gender to class and treated it as a by-product of the mode of production. Psychoanalytic accounts, in both the object-relations and the Lacanian versions, generated the same family drama in every century and had trouble with change.

The point: the test of a category of analysis is whether using it changes what you say about things that are not obviously about it. If gender only tells you about women, it is a subject and not a category.

The definition, in two halves

Scott's definition has two clauses joined by an and, and most misreadings come from taking only the first.

1. Gender is a constitutive element of social relationships based on perceived differences between the sexes. Perceived is the load-bearing word. The claim is not about bodies but about what a society makes of them.
2. Gender is a primary way of signifying relationships of power. This is the half that does the work outside the household. Differences between men and women supply a ready vocabulary for expressing hierarchy of every kind, and rulers, revolutionaries, unions and empires have all used it.

The first clause is then unpacked into four elements which she insists must be examined together.

Scott's point about the third row is aimed at a habit. Historians who accepted that gender mattered tended to confine it to family and reproduction. She insists that a labour market, a franchise and a bureaucracy are gendered institutions, and that treating them as neutral is itself the finding.

Studying gender without studying women

This is the article's most useful and most contested move. If gender signifies power, then a body of evidence with no women in it can be analysed as gendered.

Consider the possibilities that opens. Political legitimacy described through the family: the king as father, the nation as mother, the rebel as an unnatural son. Diplomatic and military language in which states are firm or supine, virile or effeminate, and in which the humiliation of defeat is figured as emasculation. Trade unionism organised around the family wage, which is a claim about masculinity as much as about pay, and which structured labour markets for a century. Colonial administration that described governed populations as childlike or effeminate and rule as a manly responsibility, which is where this lesson meets the previous one.

Scott's own later book on French feminists shows the method in a harder case: the very language of universal rights, framed around an abstract individual, turned out to have a sex, so that women claiming rights had to argue both that sexual difference was irrelevant and that they spoke as women. She called that a paradox that could not be argued away, only inhabited.

The objections

Taking the article seriously means knowing what it cost.

• Experience and agency thin out. If gender is a system of signification, the woman who worked, organised or resisted can become an effect of discourse. Scott pressed this further in 1991 by arguing that experience is not a foundation but something itself produced, and social historians objected that this dissolves the ground on which political claims are made.
• Method. Historians trained in social-science techniques argued that deconstructing texts had displaced counting, sampling and demographic reconstruction, and that the two should not be traded against each other.
• Gender alone is not enough. Evelyn Brooks Higginbotham argued in 1992 that race functions as a metalanguage that shapes what gender means, so that an analysis running on gender by itself will describe some women's history as though it were all women's history. That is the same argument the intersectionality literature developed from the late 1980s.
• The binary. Scott's framework is built on a two-term opposition. Work on trans and non-binary histories, and on societies with more than two recognised categories, has pressed on whether the framework can accommodate what it was not designed for. Scott returned to the question herself in 2010 and concluded that the category stays useful only while it remains a question rather than an answer.
• The word did the work of the politics. Scott noted at the time that gender sounded neutral and academic in a way that feminist did not, and that departments found it easier to adopt. That advantage is also a way of defanging the argument.

Common misconceptions

• Gender history is women's history under a new name. The article exists to break that identification. Masculinity, and institutions with no women in them, are inside the field.
• Using gender as a category means claiming it explains everything. It means treating it as one of the primary ways power is expressed, alongside class, race and others, and testing what changes when you look.
• Scott denies biological difference. The definition turns on perceived differences. What is at issue is the meanings a society builds on them, which vary and have a history.
• The additive approach is simply wrong. Recovery work was necessary and produced the evidence the analysis needs. The objection is to stopping there and leaving the main narrative untouched.

What to carry forward

A committee with no women on it justified closing women's clubs by defining what a citizen was, which is what it means to say that gender signifies power. Scott's article separates a subject from a category: adding women to the record is necessary and insufficient, while asking how perceived sexual difference organises meaning changes accounts that never mentioned women at all. Her four elements keep the analysis honest by forcing symbols, doctrines, institutions and identities to be examined together rather than one at a time.

So what?: take any body of evidence you are working on that appears to have nothing to do with gender, and look for the metaphors of family, body and virility in which its arguments about authority are conducted. If they are there, you have a finding that the additive approach could never have produced.

Sources

1. Scott, J. W. (1988). Gender and the Politics of History. Columbia University Press. archive.org
2. Scott, J. W. (1996). Only Paradoxes to Offer: French Feminists and the Rights of Man. Harvard University Press. archive.org
3. Scott, J. W. (1986). Gender: A useful category of historical analysis. American Historical Review, 91(5), 1053-1075.
4. Scott, J. W. (1991). The evidence of experience. Critical Inquiry, 17(4), 773-797.
5. Higginbotham, E. B. (1992). African-American women's history and the metalanguage of race. Signs, 17(2), 251-274.

A list of demands written five weeks into a war

On 9 September 1914, with German armies still advancing in France, Kurt Riezler drafted a memorandum for Chancellor Bethmann Hollweg setting out what Germany should obtain when it won. The September Programme proposed a French indemnity heavy enough to prevent rearmament, a customs union under German direction, buffer arrangements in Belgium, and territory in the east.

For forty years this document was known and not much used. Then, in 1961, a Hamburg professor put it near the centre of a book, and the resulting argument tells you more about causal reasoning in history than any methods text.

What Fischer claimed, and why the roof fell in

Fritz Fischer argued that Germany bore the principal responsibility for the outbreak of war in 1914; that its war aims were not improvised after the fighting began but expressed ambitions already present in Wilhelmine politics; and that there was a continuity of expansionist purpose running from the empire into the Third Reich. He had an evidential advantage: he had worked in archives in the east that West German historians had largely left alone.

The reaction was not scholarly disagreement so much as a public emergency. Gerhard Ritter led the counter-attack, and when Fischer was invited to lecture in the United States in 1964, the Foreign Office withdrew its support for the trip. The heat came from what the claim was taken to imply. Since 1919 the consensus had been that Europe slid into war collectively, and after 1945 that view supported another: that National Socialism was a rupture rather than a culmination. Fischer's continuity thesis threatened both at once.

Where the argument stands now is instructive. Substantial German responsibility for the July crisis is widely accepted. The strong continuity thesis, running a straight line from 1914 to 1939, is not; and the associated idea of a German special path was itself dismantled in the 1980s by historians who showed that the supposedly normal British and French developments were nothing of the kind. Christopher Clark's account in 2012 swung the emphasis back towards a multipolar crisis in which several governments took risks, and was criticised in turn for letting Berlin off.

Why this matters: in a dispute of this kind, separate the empirical question from the implication people fear. Fischer's evidence about war aims can be right while his continuity thesis is wrong, and the reception mixed the two together for a decade.

The kinds of cause, and the mistake that follows from mixing them

Most causal arguments in history go wrong before any evidence is examined, because the claim has not been specified.

Now recall Braudel's three durations. A claim about constraint, a claim about trend and a claim about contingency are answers to different questions, and a great many historical arguments are two people answering different questions loudly. Alliance systems and the mobilisation timetables are structural; the decisions taken in Vienna and Berlin in July 1914 are contingent; and the pistol in Sarajevo is a trigger. Establishing one of the three does not refute the others.

A more useful vocabulary comes from thinking about a fire: oxygen, dry timber and a match are all causes, and the argument is about which one we could have removed. Historians usually mean by cause the factor whose absence would have changed the outcome and which somebody could have altered. That is a moral and practical selection as well as a logical one, and it should be stated rather than smuggled.

When a causal claim carries a moral payload

In June 1986 Ernst Nolte published an article in a German newspaper arguing that the Bolshevik class murder was, in his phrase, logically and factually prior to the Nazi race murder, and that Auschwitz was in some measure a reaction to what had been done in Russia. He suggested the gas chambers were the technical distinguishing feature of an event otherwise not unique.

Jurgen Habermas replied in July, accusing Nolte and others of an attempt to manage the past for the sake of a usable national identity. The exchange that followed, the Historikerstreit, ran through the West German press for eighteen months.

Look at the structure rather than the heat. Nolte's core move is a causal ordering: A came first and B resembles A, therefore B is derivative. Three separate things are bundled in it. There is an empirical question about whether Nazi leaders in fact modelled anything on Soviet practice, on which the evidence is thin. There is a logical question about whether priority in time establishes causation, which it does not. And there is a rhetorical question about what follows morally from calling something a reaction, which is where the argument was actually being fought.

Andreas Hillgruber supplied a second front, publishing a short book pairing the collapse of the German Reich in the east with the murder of European Jewry, and suggesting the historian should adopt the standpoint of the German army defending the eastern population. Critics pointed out that the choice of standpoint is a historical argument in itself, made without evidence.

The dispute settled nothing formally and it changed a good deal. Comparison of the Nazi and Soviet regimes continued as ordinary scholarship. What was rejected was the inference from comparison to derivation, and the use of a causal ordering as a way of redistributing responsibility.

Counterfactuals: unavoidable, and dangerous

Every causal claim contains a counterfactual. To say that the mobilisation timetables mattered is to say that without them something else would have happened. Max Weber saw this a century ago and argued that judging a cause requires imagining the course of events with the factor removed.

E. H. Carr disagreed, dismissing might-have-beens as a parlour game played by the losing side. Niall Ferguson's collection in 1997 answered that the counterfactuals are already there, unstated, and are better made explicit and disciplined. Richard Evans replied in 2014 that most published counterfactual history is undisciplined, politically motivated, and unfalsifiable.

Three rules make the tool usable. First, the minimal rewrite: change one thing, and change it as late as possible. Second, restrict yourself to alternatives contemporaries actually considered and could have chosen, which keeps you inside the evidence, since the debate about those alternatives is documented. Third, stop early. Consequences compound, and after two or three steps you are writing fiction with no method for adjudicating it.

Applied to July 1914, a disciplined counterfactual asks what follows if Vienna delivers its ultimatum in the days immediately after the assassination rather than three weeks later. That option was discussed at the time, the discussion survives, and the exercise tests a specific claim about whether delay allowed Russian and French positions to harden. An undisciplined counterfactual asks what would have happened if Germany had won, which changes everything at once and answers nothing.

Common misconceptions

• Establishing a long-term cause disproves a short-term one. They answer different questions. Constraint, trend and contingency can all be true of the same outcome.
• Saying an event was contingent means it was accidental. Contingent means it depended on decisions that could have gone otherwise. The decisions still had reasons, and the reasons are the historian's material.
• Counterfactuals are inherently unhistorical. They are implicit in every causal claim. What is unhistorical is a counterfactual that rewrites several variables at once and runs for decades.
• If A preceded B and resembles B, A helps explain B. This is the move at the centre of the Historikerstreit, and it fails on ordinary logical grounds before any evidence about the Nazi leadership is examined.

The takeaway

A memorandum from the fifth week of a war became the centre of a controversy because of what a causal claim was taken to imply about a nation, and a newspaper article twenty-five years later did the same in reverse. Both cases teach the same discipline. Specify what kind of cause you are claiming and in which duration. Separate the empirical question from the implication that makes people angry. Do not accept temporal priority plus resemblance as evidence of derivation. And when you reach for the counterfactual, which you will, change one thing, change it late, keep to what contemporaries actually debated, and stop before the story runs away with you.

In short: a cause is a factor whose removal would have changed the outcome, chosen from among many such factors for reasons you should be able to state out loud.

Sources

1. Gaddis, J. L. (2002). The Landscape of History: How Historians Map the Past. Oxford University Press. archive.org
2. Carr, E. H. (1961). What Is History? Macmillan. archive.org
3. Fischer, F. (1967). Germany's Aims in the First World War. W. W. Norton.
4. Evans, R. J. (2014). Altered Pasts: Counterfactuals in History. Brandeis University Press.
5. Maier, C. S. (1988). The Unmasterable Past: History, Holocaust, and German National Identity. Harvard University Press.

An offer made at dawn

Before first light on 13 July 1942, outside the Polish village of Jozefow, Major Wilhelm Trapp assembled the men of Reserve Police Battalion 101 and told them what they had been sent to do. They were to round up the Jews of the village, take the working-age men away, and shoot the rest. Trapp was visibly distressed. He then said that any of the older men who did not feel up to it could step out.

About a dozen did. The rest spent the day shooting roughly fifteen hundred people at close range in a wood. Over the following months the battalion, some five hundred middle-aged reservists from Hamburg, mostly working and lower-middle class, few of them party activists, shot tens of thousands of people and deported tens of thousands more to Treblinka.

In the 1960s the Hamburg state prosecutor's office interrogated around two hundred surviving members of the battalion. Those files are the evidence. Two historians worked through them and wrote two books with incompatible explanations, four years apart. This lesson is about locating exactly where their reasoning parts.

The two accounts

Christopher Browning published Ordinary Men in 1992. His explanation is situational and cumulative. Conformity to the group, deference to an officer, the wish not to leave comrades with the dirty work, careerism among the ambitious, the availability of alcohol, and progressive brutalisation as the killings continued, all working on men already soaked in wartime racial propaganda. He draws explicitly on the social psychology of obedience and conformity, and his conclusion is that the men were not selected for cruelty, so the explanation must lie mostly in situation.

Daniel Goldhagen published Hitler's Willing Executioners in 1996 and rejected all of that. His claim is that a distinctive form of antisemitism, which he calls eliminationist and traces deep into the German nineteenth century, had made the killing of Jews thinkable and then desirable for ordinary Germans. On this account the men killed because they wanted to. He points to gratuitous cruelty beyond what orders required, to photographs taken as souvenirs, to letters home, to men bringing wives to the killing sites, and above all to Trapp's offer: they could have stepped out and did not.

Where the reasoning diverges

Run the two arguments side by side and the fork is not where the public debate placed it.

The first two rows contain the real problem. Both historians agree the interrogations are unreliable in a specific direction: a man being questioned about mass murder by a prosecutor has every reason to describe himself as reluctant, coerced and peripheral. That is a known bias, and knowing its direction is what makes the source usable.

But look at what each does with it. Goldhagen discounts statements of reluctance as self-serving and accepts statements and traces of enthusiasm as revealing. Browning treats reluctance claims with suspicion but notes that some are corroborated by the behaviour they describe, since men who dropped out were recorded doing so at the time. Neither approach is illegitimate. What would be illegitimate is applying the discount asymmetrically without saying so: taking every self-flattering statement as a lie and every self-incriminating one as truth is a rule that guarantees a particular answer.

The upshot: when a source is biased in a known direction, state the direction and then apply the correction consistently to both kinds of statement. Selective scepticism is the commonest way that honest historians reach predetermined conclusions.

Counting the refusers

The second divergence is a definitional one, and it is worth watching because it recurs everywhere.

How many men refused? The answer depends entirely on what refusal means. Stepping out when Trapp made his offer at Jozefow: about a dozen. Dropping out during that day, by asking to be relieved, drifting away, or shooting deliberately wide: considerably more. Avoiding shooting duty over the following months by seeking transfers, staying at the perimeter or volunteering for other tasks: more again, with estimates commonly running from ten to twenty percent of the battalion. Formally refusing an order in a way that produced a disciplinary record: essentially none.

Each figure is defensible and they support different arguments. A book that wants to show refusal was possible and rare cites the first. A book that wants to show a substantial minority found a way out cites the third. Neither is fabricating. This is why a methodological reader checks the definition before the number, which is the lesson from the whipping ledger in module three arriving in a new setting.

The test that broke the stronger claim

A monocausal cultural explanation carries a heavy commitment: it must predict that people without that culture behave differently.

That prediction is checkable, and it does not hold well. Auxiliary units recruited from Lithuania, Latvia, Ukraine and elsewhere participated in mass shootings, in some places with equal or greater brutality, without a German nineteenth-century intellectual inheritance. Conversely, Germans in other situations, including some of the same men before 1942, did not kill. Most specialists concluded that eliminationist antisemitism was neither necessary nor sufficient, though nearly all agreed that Goldhagen had restored something Browning underplayed: ideology was not merely background noise, and the enthusiasm of many perpetrators is in the record.

Browning's own position moved slightly in response. The later editions carry an afterword conceding more weight to antisemitic conviction while holding that situational mechanisms explain why nearly everyone participated rather than only the convinced.

Why one body of evidence supported two books

Here is the general point, and it is not a comfortable one. The two explanations are pitched at different levels. Browning is describing a mechanism operating on a person in a place. Goldhagen is describing a disposition carried by a population. An interrogation transcript records neither directly. It records what a man said in 1963 about what he did in 1942, and both levels have to be inferred from it.

When evidence underdetermines an explanation, three things follow. Additional evidence of the same kind will not settle it, so more interrogations would not have helped. The decisive test has to come from outside the case, which is why the comparison with non-German auxiliaries did more work than any document in the file. And the choice between the explanations will be driven by what each historian finds it plausible for human beings to do, which is precisely the thing that ought to be argued for rather than assumed.

Common misconceptions

• The dispute was about whether the men were guilty. Both authors describe the same acts and neither exculpates anyone. The disagreement is about the mechanism.
• Browning claims that anyone would have done it. He claims that these particular men were not selected for extremism, so situational pressures must carry much of the explanation. That is a claim about this battalion, not a universal law.
• Goldhagen was simply wrong and the book is worthless. The strong thesis failed the comparative test, and the book put perpetrator agency, cruelty and ideology back at the centre of a literature that had drifted towards bureaucratic explanation.
• Testimony given to prosecutors is worthless because it is self-serving. Bias with a known direction is usable. What is not usable is a bias whose direction you have not established, or a discount applied to only half the statements.

What to remember

Two hundred interrogation files, two books, one set of acts. The divergence begins with how each historian handles a source everyone agrees is self-serving, continues through a definitional question about what counts as refusal, and ends at a level-of-explanation problem that the files themselves cannot settle. What settled the strongest version was evidence from outside the case: other units, without the culture in question, did the same things.

Worth holding on to: when you find yourself discounting the parts of a source that cut against your argument and crediting the parts that support it, stop and write the rule you are applying. If you cannot state it as a rule you would apply to both, you do not have a method, you have a preference.

Sources

1. Browning, C. R. (1992). Ordinary Men: Reserve Police Battalion 101 and the Final Solution in Poland. HarperCollins. archive.org
2. Milgram, S. (1974). Obedience to Authority: An Experimental View. Harper & Row. archive.org
3. Goldhagen, D. J. (1996). Hitler's Willing Executioners: Ordinary Germans and the Holocaust. Alfred A. Knopf.
4. Shandley, R. R. (Ed.). (1998). Unwilling Germans? The Goldhagen Debate. University of Minnesota Press.
5. Browning, C. R. (2000). Nazi Policy, Jewish Workers, German Killers. Cambridge University Press.

A page that is mostly notes

Open Pierre Bayle's Historical and Critical Dictionary, first published in 1697, at almost any page. At the top you will find two or three lines of text in large type. Below them, in smaller type and double columns, runs a page and a half of notes, and the notes have notes of their own. The apparatus is not supporting the text. It is where Bayle argues, corrects his sources, sets rival authorities against one another and makes his jokes.

This matters because of what everybody thinks about footnotes. The story runs that Ranke invented modern citation in Berlin in the 1820s, and that the footnote is the badge of scientific history. Anthony Grafton, tracing the practice in a book published in 1997, found the ancestry to be older and messier: the antiquarian and legal scholarship of the seventeenth century, Jean Mabillon's work on how to tell a genuine charter from a forged one, Bayle's dictionary, and the ironical notes of Edward Gibbon. He also found that Ranke's own notes, examined closely, are less complete than his reputation implies.

The point is not to embarrass anyone. It is that the footnote is a device with a job, and knowing the job tells you how to write one.

What a footnote is for

A historical text makes two claims at once. The first is about the past. The second, carried entirely by the apparatus, is that the first was arrived at by a procedure a reader can repeat. Take away the second and you have assertion backed by the author's standing, which is exactly the authority that early modern scholars were trying to escape when they began putting their evidence at the bottom of the page.

So a footnote is an invitation to be checked. It should let a reader who dislikes your conclusion go to the same shelf, open the same box, turn to the same folio, and see whether it says what you claim. Everything else about citation style follows from that test, and any note that fails it is decoration however correctly it is punctuated.

Key idea: write every citation for the reader who wants to prove you wrong. If it would not help that reader, it is not doing its job.

Building the argument

Before the notes, the argument. A historical argument has six parts, and most problems come from a missing fourth.

The warrant is worth dwelling on. Suppose your claim is that a town's guild controlled entry to a trade, and your evidence is that the guild's records list apprentices. The warrant is the assumption that people not listed were not working in the trade, and that assumption may be false: unregistered work is normal and is invisible to a register. Once stated, the warrant can be tested against other sources, such as prosecutions of unregistered workers. Left unstated, it does its work silently and nobody, including you, notices the hole.

The objection has a rule of its own. Construct it from the best published version, quote it, and answer that. An objection built to be knocked down tells a reader that you have not read the other side, and it is the single fastest way to lose an examiner or a reviewer.

Writing citations that work

Style manuals differ and your discipline will impose one. Underneath the styles are rules that do not vary.

1. Cite what you used. If you took a quotation from a secondary work, cite that work. Reporting the original as though you had read it is misrepresentation, and it propagates errors: a misquotation copied through four books has been checked by nobody.
2. Give a locator precise enough to check. A page number, not a book. A folio or piece number, not a collection. For an archival document, the repository's own reference in the repository's own form, plus enough description that it can be found if the reference system changes.
3. Make the note match the sentence. A citation should support the specific assertion it is attached to. A note listing four works after a paragraph containing five claims tells a reader nothing about which supports which.
4. Quote exactly, and mark every intervention. Ellipses for omissions, square brackets for insertions, and a note when the emphasis is yours. Never silently repair spelling or grammar in a quotation from a period source; the irregularity is often evidence.
5. Cite for the contestable and the borrowed. Anything a competent reader might doubt, and anything that is somebody else's idea, needs a note. Uncontested background does not.
6. Record where the negative result came from. If you searched a series and found nothing, say which series and which years. A documented absence is a finding, and an undocumented one is an excuse.

Two failures are worth naming because they are respectable-looking. The first is the citation that does not support the sentence, which usually happens when a paragraph is rewritten and the notes are not. The second is the padded note, listing works the author has not read, which tends to collapse the moment an examiner picks one and asks what it argues.

The double narrative

Grafton's most useful observation is that a footnoted history tells two stories in parallel. Above the line, what happened. Below it, how the historian came to think so: which archive, whose edition, what was missing, where an earlier scholar went wrong. Readers who work in the field read both, and often read the second first.

That means the apparatus is where you do things the main text cannot. You can register a doubt without breaking the narrative. You can concede that a document is ambiguous while stating why you read it as you do. You can put the counter-evidence where an honest reader will find it. Gibbon put his best insults there, which is a reminder that a note can also carry a voice.

It also means the apparatus is where dishonesty is detectable, which is why the libel trial in the earlier lesson was decided in the notes rather than in the prose.

Defending it

At some point a person will ask you why anyone should believe this. Preparation has three parts.

Know your weakest evidence better than your critic does. There is always one inference doing more work than it should; find it before somebody else does, and be ready to say what would change your mind about it. Distinguish the three honest answers: that is outside what I studied; I do not know and here is what would settle it; and I disagree, and here is why. Confusing the second with the third is the usual failure under pressure. And separate your claim into a core you will defend and a periphery you will trade. A candidate who defends every sentence with equal energy signals that none of them has been examined.

Common misconceptions

• Ranke invented the footnote. The device has a long earlier history in antiquarian, legal and lexicographical scholarship, and Grafton shows Ranke's own notes to be patchier than the legend.
• More citations make an argument stronger. Precision does. One note that lets a reader check the exact claim is worth ten that gesture at a literature.
• Footnotes are for facts, not for argument. From Bayle onwards the notes have carried objections, corrections, qualifications and jokes. They are part of the argument, not a receipt attached to it.
• Conceding a weakness damages your case. An acknowledged limit that you have measured is far stronger than an unacknowledged one your reader finds first.

Putting it together

Bayle's pages, mostly notes, show what the apparatus was for before anyone called it scientific: a place to set authorities against one another where a reader could watch. That is still the job. A historical text asserts something about the past and, through its notes, asserts that the assertion can be checked, which is the only kind of authority available to a discipline that cannot repeat its experiments. Build the argument in six parts, state the warrant that connects evidence to claim, take the objection at its strongest, and cite so that a hostile reader can go to the same folio and disagree with you properly.

What matters here: the footnote is not a courtesy to previous scholars. It is the mechanism by which a claim about the past becomes something other than an opinion held by someone with a job.

Sources

1. Grafton, A. (1997). The Footnote: A Curious History. Harvard University Press. archive.org
2. Fischer, D. H. (1970). Historians' Fallacies: Toward a Logic of Historical Thought. Harper and Row. archive.org
3. Booth, W. C., Colomb, G. G., Williams, J. M., Bizup, J., & FitzGerald, W. T. (2016). The Craft of Research (4th ed.). University of Chicago Press.
4. Turabian, K. L. (2018). A Manual for Writers of Research Papers, Theses, and Dissertations (9th ed.). University of Chicago Press.
5. Bayle, P. (1734-1738). The Dictionary Historical and Critical of Mr Peter Bayle (2nd ed., Vol. 2). D. Midwinter.

A letter in a government file

Among the correspondence that English poor law unions exchanged with London between 1834 and 1900, filed at The National Archives in the series MH 12, there are letters written by paupers. A woman refused relief writes over the guardians' heads to the central board. A man disputes a settlement decision. The clerk's covering note is attached, and so is the reply.

That series is a good place to start this lesson because it is enormous, it is not fully catalogued at item level, and it contains exactly the kind of material this course has been teaching you to want. It is also a trap. A first-year researcher who announces a project on the poor law and the voices of the poor will spend a year discovering how much of it there is.

What follows is one project run through twelve months, with the decisions in the order they have to be taken.

Month one: from a topic to a question

A topic is a region: the poor law, the Terni steelworks, colonial policing. A question has an answer that could turn out to be no. Three tests separate them.

Run the poor law topic through those and you might arrive at: did the introduction of a paid relieving officer in one union between 1870 and 1880 change who was granted outdoor relief, and on what grounds? That question names a place, a decade, an institutional change and an outcome, and it can be answered no.

Remember: a question you cannot fail at is a topic wearing a question mark. If no possible document would embarrass you, keep working on the sentence.

Month two: find the conversation

Nothing you write matters until it is placed against what has already been argued. Work backwards: find the three most recent serious treatments, take their footnotes, and follow them until the same works start recurring. That is the conversation.

Then write a paragraph you will revise for a year, in three sentences. Historians have argued X. They have shown it mainly from evidence of type Y. This leaves open Z, which the sources in front of me can address. If the third sentence will not come, either the gap is not there or you have not read enough to see it. Both are worth knowing in month two rather than month ten.

Month three: survey the sources before you read them

Characterise the series before opening an item, exactly as the lesson on reading along the grain proposed. For each candidate body of records, establish six things: what office created it, its start and end dates, its physical extent, how it is arranged, what finding aids exist, and whether you can actually get to it. The last is not a formality. Access can be blocked by closure periods, by language, by cost of travel, by a catalogue that stops at box level, or by handwriting you cannot yet read.

Two habits pay for themselves here. First, keep a source register from day one, listing every series consulted, the dates covered, and what you found or did not find. The negative entries become the documented absences your argument will need. Second, name the counter-archive: if the union's own minutes are lost for three of your ten years, ask which other body wrote about the same decisions, such as the central board, the local press, a charity or a parish vestry.

Months four to six: the pilot, and the reformulation

Do not begin the main reading. Take a bounded sample first: one year, one bundle, one hundred cases. The pilot answers three questions that nothing else can. Does the source actually contain the variable I care about? How long does one unit take to read and record? What does the material want to be about, which is rarely what I thought?

Then reformulate. Almost every good project changes its question after the pilot, and the change is a finding rather than a failure. You may discover that outdoor relief decisions are recorded only as a grant or refusal with no reasons, which kills the second half of your question, and that the same minutes record at length the arguments among guardians, which is a better question than the one you had.

This is the point at which to fix your note system, because changing it later costs weeks. Whatever the tools, three rules hold. Every note carries the full reference at the moment of writing, never afterwards. Quotation, paraphrase and your own thought are visually distinguished, so that in month eleven you can still tell which is which. And every note that will support a claim records enough context to reconstruct the document's genre and purpose, since the argument may depend on what kind of paper it was.

Months seven to nine: the main reading, and the counting

Now the expensive work, on a question that has survived contact with the material. Two things run in parallel.

The reading proper, with the operations from earlier in this course applied deliberately rather than by instinct: provenance, purpose, audience and survival on the significant documents; along the grain for the series as a whole; against the grain where the record was made by one party about another.

And the counting, if there is anything to count. Before any number goes into a table, run the five questions from the lesson on statistics: who counted and why, what the unit was and whether it changed, where the denominator comes from, who was excluded by definition, and what incentive shaped the total. Write the answers into the notes, not into the eventual footnote only. Half of them will not survive into the thesis, and the half that does will be the half you can defend.

Months ten and eleven: argument, then prose

Build the argument in the six parts from the previous lesson before writing continuous prose: question, claim, evidence, warrant, objection, response. Do it on one page. If it does not fit on one page, you have more than one project, which is common and fixable.

Then write, and observe two disciplines this course has insisted on throughout. Mark inference: a reader must be able to see where the record stops and your reconstruction begins, which is the standard the Davis and Finlay exchange left behind. And write the silences into the text rather than into a preface, saying which series were searched, which years are missing, and what the missing material would have to contain to change the conclusion.

Structure follows from the argument rather than from the chronology of your research. Nobody wants a narrative of your visits to archives. The exception is where the discovery order is itself the argument, which happens more often in this field than people expect and should then be signalled as a choice.

Month twelve: the defence, and the things that go wrong

Prepare by locating your own weakest inference before anyone else does, and by deciding in advance which claims are core and which you will trade.

Four failures account for most of the trouble, and all four are avoidable in the first three months.

• The unreachable source. A collection is closed, uncatalogued, in a language you do not read, or three flights away. Establish access before the question depends on it.
• The question the archive cannot answer. The record notes decisions and not reasons, or names men and not women. The pilot exists to find this in month five.
• Notes that cannot be checked. References added later are references invented later. This ends projects.
• Ethical and legal constraints discovered late. Living subjects, personal data, oral history consent and copyright in unpublished material all have rules, and they bind what you may quote and deposit. Settle them before recording anything.

Common misconceptions

• You should read everything before forming a question. You cannot, and a question is what makes reading selective enough to finish. Form one early and expect to revise it after the pilot.
• Changing the question means the project failed. The reformulation after a pilot is the normal course of research, and a project whose question never changed usually means the sources were never allowed to speak.
• A gap in the literature is a sufficient reason to write. Some gaps exist because the question is not interesting or the evidence does not exist. The third sentence of the historiographical paragraph has to say what the gap costs us.
• Writing comes at the end. Writing is how you find out whether the argument works. Draft the one-page argument in month ten and a chapter as soon as you have enough for one.

Looking back

This course began with a telegram and four questions: who made it, why, for whom, and how it survived. Everything since has been an extension of those questions to larger objects. An archive is made by decisions that can be reconstructed. A number is the output of an administrative process. A school of history is a position with texts behind it and an archive that shaped it. A dispute is usually two people answering different questions until somebody separates them. And a project is the same discipline applied to your own work: state the question so it can fail, find out what exists before committing, test the question against a sample, then read, count, argue, cite, and say plainly what you could not find out.

The point: the finished thing is not the story you tell but the story plus the account of how you came to tell it. Write both, and a reader who disagrees with you will still be able to use what you did.

Sources

1. Tosh, J. (2015). The Pursuit of History: Aims, Methods and New Directions in the Study of Modern History (6th ed.). Routledge. archive.org
2. Booth, W. C., Colomb, G. G., & Williams, J. M. (1995). The Craft of Research. University of Chicago Press. archive.org
3. The National Archives. (n.d.). Research guides. nationalarchives.gov.uk
4. Howell, M., & Prevenier, W. (2001). From Reliable Sources: An Introduction to Historical Methods. Cornell University Press.
5. Bloch, M. (1953). The Historian's Craft (P. Putnam, Trans.). Alfred A. Knopf.