16.4 — How to Read History Without Being Fooled

Five accounts survive of the sermon Pope Urban II preached at Clermont in 1095, launching the First Crusade.

All five were written after the crusade succeeded. Each attributes to him the motives that its success had made attractive (Chapter 7.5).

We do not have what he said. We have what it became useful to believe he had said.

That is the ordinary condition of historical evidence, and this chapter is the practical method for handling it — applied equally to a medieval chronicle and to a news report this morning.

Start with the source

Four questions, in order, of any claim about the past or the present.

Who produced this? Not their name — their position. A court chronicler, a colonial administrator, a defeated general writing a memoir, a company's press office, an opposition party.

When? Contemporary, or written afterwards knowing the outcome? A source written after the result frequently reads the result backwards into the causes.

For whom? A royal inscription is addressed to posterity and to rivals. A private letter is addressed to one person. A diplomatic cable is addressed to a superior who has views.

And what did they gain? Not as an accusation — as a description of the pressure they were under.

The categories, with examples from this volume.

Official records of achievement inflate. Mughal and Sultanate chroniclers counted temples destroyed and infidels killed because those were the achievements their patrons wanted recorded (Chapter 6.9). Assyrian kings advertised atrocity as policy (Chapter 4.7).

Records of routine administration are far more reliable, because nobody was performing. Ea-nasir's copper complaint (Chapter 4.1). The Reichsbahn's invoices (Chapter 11.9). The slave trade voyage registers (Chapter 8.6). Temple inscriptions listing employees and salaries (Chapter 6.7).

Physical evidence has no motive. Archaeology, isotopes, ice cores, tree rings, genetics. It has interpretation problems and it does not lie about itself.

And an admission against interest is the strongest textual evidence available. Ashoka recording the death toll of his own conquest (Chapter 6.4). Aurangzeb's farmans protecting temples, which cut against his own chroniclers' framing (Chapter 6.12). Dyer's testimony that he fired to produce a moral effect (Chapter 6.21).

The specific moves to watch

Eight, each with an example.

1. The impossible number. Chroniclers report armies of a million and casualties exceeding the population of the region. The check is arithmetic: how many people could that area feed, how much water does an army of that size need daily, how far can it march. Chapter 6.9 applied this to the Persian chronicles and Chapter 7.6 to the Mongol figures.

2. Absence read as evidence. "The earliest known X" almost always means the earliest found (Chapter 1.1). Fossilisation and preservation are extraordinarily selective, and drowned coastlines contain a great deal we will never see (Chapter 3.6).

3. The single cause. Every collapse in Chapter 16.2 has been attributed to one thing by someone. A monocausal explanation of a large event is a signal that the account has a purpose.

4. The name as the argument. Mutiny, revolt, rebellion or first war of independence (Chapter 6.18). Riot or uprising. Terrorist or freedom fighter. The facts are frequently agreed and the name carries the verdict.

5. Reading the present backwards. Nations, borders and identities are treated as though they always existed. Chapter 10.7 showed nations being manufactured; Chapter 6.19 showed the census hardening caste categories. "India has always..." and "the Chinese have always..." are almost always false.

6. The golden age. Chapter 6.6 tested one. The questions are: golden for whom, compared with what, and who is telling you. The Gupta period produced Aryabhata and hardened untouchability, and both belong in the same paragraph.

7. Survivorship. We study the empires that left records, the buildings that stood, and the texts that were copied. Chapter 7.8 showed African states omitted partly because they built in mud brick.

8. The quotation that was never said. Marie Antoinette's cake, Voltaire on defending your right to say it, Churchill's various aphorisms, Twain's rhyming history at the top of Chapter 16.1. A quotation that perfectly captures a modern point is usually a modern invention.

Real uncertainty and manufactured uncertainty

The distinction Chapter 14.2 turned on, stated generally.

A real uncertainty has a stated range, a method for narrowing it, and researchers actively arguing about it. Climate sensitivity is 2.5 to 4 °C and people are working on it. The scale of Sultanate-era temple destruction is unknown and historians dispute the method.

A manufactured uncertainty argues that because something is not certain, nothing is known. It does not propose a test. It does not narrow over time. And it is frequently funded.

The tobacco industry's internal strategy documents state it explicitly: doubt is the product, because doubt is sufficient to delay. The same approach was applied to lead in petrol, to acid rain, to ozone and to climate (Chapter 14.2), in several cases by the same consultants.

The test: what specific observation would settle this, has anyone made it, and what did they find? If the answer is that no evidence could settle it, you are not looking at a scientific dispute.

Statistics

Five checks that catch most misleading numbers.

Relative against absolute. "Doubles your risk" means nothing without the base rate. A rise from one in a million to two in a million is a doubling.

The denominator. Chapter 12.10's Qatar death figure covers all deaths of migrant workers from all causes over a decade and is quoted as construction deaths. Always ask: out of what, over what period, defined how?

Correlation and cause. The standard warning, and the useful follow-up is to ask what would produce the correlation without the cause — a common third factor, reverse causation, or selection.

Survivorship in the data. Studies of successful companies, successful people and surviving buildings.

And the base rate. The most-repeated example is the medical test: a test with 99 percent accuracy for a disease affecting one person in ten thousand produces, in a population of a million, about 100 true positives and about 10,000 false ones. A positive result means roughly a 1 percent chance of having it. Most people, including many professionals, get this wrong, and the same structure applies to security screening, fraud detection and profiling.

The biases that make this necessary

Volume VI covers them properly. The four that most affect reading about the world.

Confirmation bias. We seek and remember what fits. The correction is to look specifically for the strongest version of the opposing case rather than the weakest.

Availability. We judge frequency by how easily examples come to mind, and vivid, recent and reported events dominate. This is why people fear the things that make news and not the things that kill them.

Motivated reasoning. Where a conclusion is inconvenient, the standard of evidence rises silently. The self-check is: what would change my mind, and can I state it in advance?

And the fundamental attribution error. We explain others' behaviour by character and our own by circumstance. Applied to history, it produces accounts in which events happen because people were brave or corrupt or decadent, rather than because of incentives and constraints. Chapter 6.15 was written specifically against this: Mir Jafar was not unusually wicked, he was a noble in a system where switching sides was a normal route to advancement.

The working checklist

Applied to any claim about the world, historical or current.

What is the specific claim, stated precisely enough that it could be wrong?

What evidence supports it, and is that evidence physical, administrative, or narrative?

Who produced the evidence and what pressure were they under?

What is the strongest counter-argument, stated by someone who holds it rather than by its opponents?

Is the disagreement about facts, about interpretation, or about values? These require completely different responses and conflating them is the commonest failure in public argument.

And what would change my mind?

Where this is hardest

Two cases from this volume, named because they are the ones where the method matters most.

Contested national history. Chapters 6.2, 6.12 and 6.21 were written to a stated standard: the documents on both sides, the archaeology, the statistics, and an explicit separation of what is established from what is asserted. The method does not produce a comfortable answer and it produces a defensible one.

And current events, where the method is hardest because the evidence is incomplete and the emotional stake is highest. The specific discipline is to notice when the intensity of your conviction exceeds the quality of your evidence — which is the most useful sentence in this chapter.

Where this shows up in your life

Every day, in what you are told.

And the point is not to believe nothing. Corrosive scepticism is as useless as credulity and is frequently its twin — the person who believes nothing official believes anything unofficial.

The point is calibration: believing things in proportion to the evidence, holding the confidence the evidence supports and no more, and being willing to say that you do not know.

This volume has tried to model that. Where the evidence is strong it has said so. Where historians disagree it has named the disagreement. Where a number is unknowable it has given the range and said why. A reader who takes nothing else from it should take the habit.

What the next page covers

Chapter 16.5 closes the volume. One thread, from a cloud of dust orbiting a young star to the news on your phone this morning, with every link in it drawn from a chapter you have read — and then the question of what any of it is for.