History: Methods and Tools | Quomodo Scimus : IBTOK

areas of knowledge - history

Methods and tools in History

How is history made?

THE PROVOCATION

History is in part, invention.

Natalie Zemon Davis is one of the most respected historians of her generation, a specialist in sixteenth-century France. The Return of Martin Guerre is serious scholarship, published by Harvard University Press in 1983. And yet she opens her account of a peasant imposture case with the admission that what follows is "in part my invention."

In 1560, a man appeared in a French village claiming to be Martin Guerre, a peasant who had disappeared eight years earlier. He lived with Guerre's wife for three years, managed his property, fathered a child. Then the real Martin Guerre returned. The impostor was tried and hanged. What Davis wanted was not just an account of what happened, but an explanation of why - what motivated the wife, how the village community responded.

The problem is that the evidence does not say what motivated the wife or the community. The trial records for the Parlement of Toulouse before 1600 are almost entirely lost. Davis worked from two published accounts written by men close to the case, alongside notarial contracts and church records from the region. When she could not find her individual man or woman in the archives, she drew on broader sources from the period and place "to discover the world they would have seen and the reactions they might have had." Davis insists the invention is "held tightly in check by the voices of the past." But what sits between the evidence and the account is a reconstruction of probable motive and plausible context, something that draws on more than the surviving documents. At some point it draws on imagination.

Davis was also a historical consultant on the 1982 French film Le Retour de Martin Guerre, and this is where the pressure becomes visible. The historical complexities she had spent months in the archives reconstructing were smoothed into a more coherent, emotionally satisfying story. She found this troubling enough to write the book. But the film's departures from the historical record followed the same pull towards narrative that shapes all historical reconstruction, only further along the same axis.

Then came Somersby (1993), an American remake transposing the story to the post-Civil War South, with Richard Gere. The sixteenth-century French peasant village became Tennessee in 1867. The same imposture, the same question about identity, but now entirely detached from any archival trace. Whatever "held tightly in check" had meant for Davis, it no longer applied.

The chain from event to book to film to remake maps something the page that follows examines directly. Every step away from the sources involves more imagination and less constraint. The question is where legitimate reconstruction ends - and whether that line can ever be drawn clearly.

Big idea 1 - The historian reads what was usually never meant to be read.

The problem of historical sources

History’s first epistemological weakness is that most of the past is unknowable. Historians cannot observe past events directly; they rely on surviving traces, and most traces have disappeared. What remains is often accidental, fragmentary, unrepresentative, or produced by the powerful. Sources are not neutral windows onto the past: diaries, minutes, reports and archives already contain interpretation, memory, bias and self-justification. Historians therefore reconstruct the past from limited evidence that may not answer their questions.

Carlo Ginzburg (1939-2026), who died whilst I was writing this page, spent his career asking how we recover the lives of people who left almost no trace. His essay "Clues: Roots of an Evidential Paradigm" argued that the historian works exactly as a detective does - backwards from fragments to a reconstruction that can never be directly verified. Neither the detective nor the historian has access to the event itself, only to what it left behind. What makes historical evidence distinctive is that almost none of it was produced to serve as testimony. A tax record, a will, a grain price, a church register were written for other purposes entirely.

Historian Marc Bloch called this category of source "witnesses in spite of themselves": traces that reveal things their authors never intended to disclose. A medieval lord's account book, kept to record rents and dues, tells us more about peasant life than the lord meant to share. The most productive historical evidence is often not the official record of what people wanted posterity to know, but the incidental trace of how they actually lived.

This is the first and most fundamental constraint on historical knowledge. The past does not preserve itself. What survives is what someone chose to keep and what happened to be durable enough to outlast everything else. The fires, floods, purges and simple neglect that consumed the rest left no record of what they took. Herbert Butterfield described what remains as a "heap of broken fragments," and the metaphor captures both the incompleteness and the randomness of the archive.

The constraint falls unevenly. Elites write, keep records, commission documents. The poor, the illiterate, the colonised and the enslaved mostly do not. Writing a social history of medieval peasants, or of women before the modern period, or of those who appear in the archive only as defendants, requires reading sources produced by people with entirely different interests and purposes. Ginzburg's own masterpiece, The Cheese and the Worms, reconstructs the cosmology of a sixteenth-century miller named Menocchio entirely from his Inquisition trial records - documents produced by the Church to condemn him, read by the historian to hear him. That gap between what a source was for and what it can be made to yield is where social history lives.

The problem does not disappear when sources do survive. E.H. Carr made this concrete with the Stresemann archive. When a historian cites the correspondence of the German foreign minister Gustav Stresemann, they are reading letters Stresemann chose to preserve, edited after his death by an archivist with political purposes of his own. The abundance of elite documentation creates its own distortion: the sources speak, but they speak for those who produced them. The evidence arrives already shaped, and what it chooses not to say is as significant as what it does.

Bertrande de Rols appears in the historical record as a wife and a defendant. Her own testimony at the Toulouse trial is gone with the records that were lost. What survives about her comes from men who judged her case and wrote about it for their own purposes. Without Davis's willingness to infer - from what the period and place reveal about honour, marriage, Protestant conversion and village life in sixteenth-century France - Bertrande's voice would be as silent as the vast majority of women who lived and died in that world and left nothing behind. The imagination the provocation admits is not a failure of rigour. It is the only thing that keeps certain lives in history at all.

Big Idea 2 - The historian knows more than the participants and less than they think

The problem with historians

History’s second epistemological weakness lies in the historian’s method. Historians do not simply recover the past; they select, sort and interpret its surviving traces. Unlike those who lived through events, they benefit from hindsight, archives, earlier scholarship and emotional distance. This can make their understanding stronger, but it also means history is constructed. As Croce argued, all history is contemporary history: each generation asks new questions, notices different people, and rewrites the past from its own perspective.

The previous lesson on perpective examined how the historian's own position shapes what they see and what they miss. The problem this section examines is different: it concerns the limits that apply to the method itself, independent of who is using it.

As I describe in the film accompanying this section above, imagine the staff who clean a railway station after the morning rush. They arrive when the crowd is gone. They cannot know exactly what happened on the platform: who was there, what was said. But they can read what was left behind: discarded tickets, coffee cups, a newspaper left on a bench. They have access to the whole space and are not distracted by the event itself. What they cannot recover is what the participants had directly: the experience.

The historian's situation is structurally similar. This historian Herbert Butterfield saw this as an advantage as much as a limitation. The historian can understand past actors better than those actors understood themselves, with hindsight and access to archives that were sealed at the time. Margaret MacMillan makes the same point: participants are too close to events to see their shape. The historian writing about the origins of the First World War in 1970 knows things the statesmen of 1914 could not: how it ended, and at what cost.

But the method has a structural limit that hindsight does not remove. The historian cannot run a controlled experiment. They cannot stop the Archduke's car from turning into Franz Josef Street on 28 June 1914 to see what would have happened differently. Every causal claim the historian makes is untestable in the way that Karl Popper required of scientific claims. This is a structural feature of the discipline. The core lesson 5 examines what Popper's falsification principle demands; history cannot meet that demand, and the discipline's honesty requires acknowledging it.

Collingwood went further. He argued that all history is the history of thought: the historian re-thinks the thoughts of past actors in their own mind rather than reconstructing external events. To understand why Caesar crossed the Rubicon, the historian must think through what Caesar was thinking: the risks he calculated, what would happen if he failed. In core lesson 3 philosopher Thomas Nagel described the aspiration to pure objectivity as "the view from nowhere": knowledge achieved from a standpoint outside any particular perspective. History cannot reach it. The historian's own mind is always inside the method.

This is where concepts from outside the discipline become useful. Leon Festinger's theory of cognitive dissonance, developed in psychology in the 1950s, describes the mind's tendency to resolve contradictions by quietly adjusting one belief until the tension disappears rather than confronting them directly. The philosophers Hugo Mercier and Dan Sperber, whose work features in Lesson 6, take this further: reason, in their account, evolved not to help individuals find truth but to justify positions to others. Even conscious self-examination is likely to produce rationalisation rather than honest self-knowledge. If Collingwood is right that the historian re-thinks past thought, what they re-enact may be only the rationalised version.

Bertrande de Rols is the test case. The question of whether she knew that the man she was living with was not Martin Guerre but the impostor Arnaud du Tilh may have no recoverable answer, less because the evidence is thin than because the mental state itself may never have been clear. She may have told herself she was not sure. Cognitive dissonance allows people to hold incompatible beliefs by keeping one of them just out of reach. The historian who claims to know her motivation is claiming access to something that may not have been fully available even to Bertrande herself.

Big idea 3 - The historian produces literature, not a record

The problem of history as literature

History’s third epistemological weakness lies in narrative and truth. Historians do not simply list accurate facts; they turn fragments into meaningful stories. Carr argued that facts are like fish: selected, prepared and arranged. But history cannot rely on correspondence theory, because the past never existed as a narrative against which an account can be checked. Events happened, but their structure, meaning and significance are created retrospectively. History is therefore reconstruction: the plausible plastered over the forgotten.

In January 1905, Russian imperial troops fired on a peaceful procession of workers approaching the Winter Palace in St Petersburg. Orlando Figes opens his account of the event in A People's Tragedy (1996) like this: "Snow had fallen in the night and St Petersburg awoke to an eerie silence on that Sunday morning, 9 January 1905." One of the marchers, Ivan Vasilev, left a note for his wife before he left her sleeping: "If I fail to return and am killed, Niusha, do not cry." He never returned. This is history at its most compelling. It is also history making a series of choices about what to include, in what order, and in what register, choices that are doing as much work as the evidence.

This is the third epistemological limit the historian faces, and in some ways the deepest. The product of historical inquiry is a text. But the problem is not simply that the past is gone. The probelm is that the past never existed as a text. The workers who marched toward the Winter Palace on 9 January 1905 did not experience a narrative. They experienced cold, noise, fear and confusion. The snow, the silence, the letter: these were not arranged in Figes's order. They were not arranged in any order. There was never a single perspective, a truth that history is failing to recover. Narrative is what the historian brings to events that never had one. The historian's text can be compared to other texts: other accounts, other sources, other interpretations. It cannot be compared to the past, because the past and a text about the past are incommensurable things.

Hayden White's concept of emplotment makes this concrete. In Metahistory (1973) he argued that the historian does not simply select facts but arrange them within a narrative form (Romance, Comedy, Tragedy, Satire) that shapes what the history claims regardless of what the evidence says. Figes emplotted Bloody Sunday as tragedy: snow and silence, hymns and icons, a letter that says farewell. But his own text contains the materials for an entirely different account. He records that Gapon "could not even pronounce the word 'constitutionalism'." He notes that Nicholas II "thought so little of the danger that he even left the capital for his palace at Tsarskoe Selo and another quiet weekend of country walks and games of dominoes" while 150,000 workers marched toward his guards. A historian who foregrounded those details rather than the snow and the letter would write the same events as satire: the gap between what the actors believed about themselves and what was actually happening, exposed without mercy. The same sources, a different history. The emplotment operates within the evidence, in what gets placed at the centre and what gets pushed to the margin. White's point is that this choice is constitutive of what the history claims, not a literary decoration added after the history is written.

Milan Kundera, in Ignorance, describes a character reconstructing a past event from a diary. The surrounding context has been forgotten, and "all he could do was invent them; not to fool anyone but to make the recollection intelligible." The result, Kundera writes, was "only the plausible plastered over the forgotten." Where the evidence runs out, the account continues, because narrative demands continuity. For me, this is best description of what history is and once again it was produced by a novelist.

This is where Zemon Davis's formulation returns. Her laboratory generated "not proofs, but historical possibilities." Again, the plausible plastered over the forgotten. The historian who acknowledges this is describing accurately what historical knowledge is: an argument about the past, constructed from traces and shaped by a narrative form that carries its own epistemological consequences. The Language pages on this site examine what language can and cannot do as a tool for producing knowledge. History is made entirely from language. Whatever those limits are, history inherits them.

Think further: questions and resources

Davis admitted that her account of The Return of Martin Guerre was "in part my invention, but held tightly in check by the voices of the past." If the line between legitimate inference and invention cannot be drawn precisely, can it be drawn at all - and if not, what distinguishes history from historical fiction?
Collingwood argued that to understand past actors, the historian must re-think their thoughts in their own mind. Festinger's research on cognitive dissonance shows that people do not always know their own motivations. If the actor did not fully understand their own mind, what exactly is the historian re-enacting?
Hayden White argued that the narrative form a historian chooses - Tragedy, Satire, Romance - shapes what the history claims regardless of what the evidence says. If two accounts of Bloody Sunday use different emplotments, is there a way to judge which is more historically accurate, or only which is more illuminating?
Mantel argues that "if we want to meet the dead looking alive, we turn to art." If a historical novel gives us knowledge of the past that the archive cannot, does it count as historical knowledge? What would make it more or less reliable than a scholarly account?
Kundera writes that where memory runs out, all we can do is invent, "not to fool anyone but to make the recollection intelligible" - the result being "only the plausible plastered over the forgotten." If every historical account plasters the plausible over the forgotten, is there a way to distinguish better from worse history, or only more from less honest history?

Films
For more see my 10 films for the TOK journey page.

🎬 WATCH — Le Retour de Martin Guerre (1982)

Daniel Vigne

The film that opens this page. Natalie Zemon Davis served as historical consultant and then wrote the book that quarrelled with the film's conclusions. Her dispute was methodological: the film resolved ambiguities the sources cannot resolve. Bertrande de Rols becomes fully legible on screen, her thoughts visible and her choices explained, in ways the archive does not support. The filmmakers needed a coherent character; narrative form demanded one. Davis refused, acknowledging what the evidence would and would not say. The gap between what the film shows and what the sources allow is what Big Idea 1 examines.

🎬 WATCH — Rashomon (1950)

Akira Kurosawa

A samurai is found dead in a forest clearing. Four witnesses give completely different accounts of what happened: the bandit, the wife, the dead man speaking through a medium, a woodcutter who watched from the trees. Each account is internally coherent. Each fits the available evidence. None can be verified against the event itself, because there is no access to the event, only to what people say about it. Kurosawa's film enacts Hayden White's argument: the same events arranged in different narrative forms produce different truths. The film refuses to say which account is correct. That refusal is the argument.