Knowing the past

Sources come in two kinds. A primary source was made at the time by someone who took part or was there: a soldier’s letter, a tax roll, a treaty, a coin struck that year. A secondary source was made later by someone studying the primary ones, as a modern book on the Crusades draws on chronicles written some eight hundred years ago. The primary source is closer to the event, and its information has passed through fewer hands. The secondary source has the advantage of seeing many primary sources at once.

The label does not belong to the object. A schoolbook printed in 1890 is a secondary source on ancient Rome, and a poor one. Ask instead what children were taught in 1890, and the same book is a primary source. Which kind a source is depends on the question you are asking.

Most people who ever lived left no writing. They still left evidence. The objects people made and used, such as pots, tools, coins, jewellery and the foundations of houses, are called artifacts, and they can speak where documents are silent. A coin sealed under a floor shows the floor was laid after the coin was struck. Worn tools and mended pots show how ordinary people lived, which no king’s chronicle troubled to record.

The other source is memory. Before writing, and long after it, peoples kept their past in spoken form: genealogies, poems, sayings and stories passed from mouth to mouth. This is oral tradition. It can carry real memory across centuries, but it has a weakness. Each retelling can add, drop or polish, and after enough tellers nobody can say what the first one said, unless someone has kept track of who told whom.

A source is a claim about the past made by somebody, and every somebody stands somewhere. So a historian questions each source before trusting it: who made this, when, and why? The habit is called source criticism, and criticism here means testing, the way a jeweller tests gold.

The first thing to ask is where the source came from and whose hands it has passed through. That history of origin and custody is its provenance, and a charter that surfaces from nowhere deserves suspicion. The second is the maker’s loyalties. A chronicle written at court to please a king will dwell on his victories and hurry past his defeats. That slant is bias. A biased source is still useful. The historian reads it knowing which way it leans, and learns something from what it tries to hide.

One source, however good, can be mistaken or lying. So historians look for a second source that reports the same thing without having copied the first. Backing a claim with independent sources is called corroboration, and it is strongest when the sources had no wish to agree.

Take the year 701 BC, when the Assyrian king Sennacherib marched on Jerusalem. His own inscription boasts that he took forty-six towns of Judah and shut its king up in Jerusalem ‘like a bird in a cage’. It does not say he took the city, and Assyrian kings did not leave victories out. The Hebrew account, written by his enemies, says Jerusalem survived. Two records from opposite sides agree on that one point, so historians treat it as settled, whatever else in either account they doubt.

Oral reports have a weakness: after a few retellings nobody knows who first said what. Early Muslim scholars met the problem head on. Whoever passed on a report about the Prophet or the first Muslims was expected to name the person he heard it from, who had named the one before him, and so on back to an eyewitness. That list of names is the report’s isnad, an Arabic word meaning support, and it travelled with the report as its credentials.

The isnad made a spoken report something that could be checked. A scholar could now ask of each name in the chain who this person was and whether he could be trusted. Historians took up the habit too. Al-Ṭabarī, who died in 923, filled his great history with reports that each come with their chain, so that readers can weigh the tellers for themselves.

A chain is only as good as its links. The scholars of hadith, the reports of what the Prophet said and did, studied the narrators one by one. Biographical dictionaries recorded when each one lived, where he travelled, who taught him, how exact his memory was and whether he was honest. A chain with a liar, a careless man or a gap in it was graded weak or rejected.

Dates did much of the work. If one narrator died before the next was old enough to learn from him, the chain was broken. Al-Bukhārī, who died in 870, is known for the strictest standard. That two narrators had lived at the same time was not enough: he wanted it shown that they had actually met. The chain he called the soundest of all runs from the jurist Mālik to Nāfiʿ, and from Nāfiʿ to his teacher, the Companion Ibn ʿUmar.

Checking the narrators, wrote the North African historian Ibn Khaldūn in 1377, is the proper test for religious reports. Most lay down what Muslims are to do, and that rests on the honesty and exactness of those who handed them on. Reports of events, of armies and kings, face another test first. Before asking who reported it, ask whether it is possible: whether it could have happened.

His example was a number. Al-Masʿūdī and other chroniclers wrote that the Israelites, counted in the desert, had six hundred thousand men under arms. The figure is theirs, not the Qur’an’s. Could the land have supported such a host? In battle order it would stretch farther than the eye can see. The Persians, a far greater power, fielded a hundred and twenty thousand at their largest. A miracle, he added, is another matter: the test is for the ordinary course of things.

Why do false reports get written down? Near the start of his Muqaddimah Ibn Khaldūn lists the causes. Writers take sides, and a partisan accepts whatever favours his own party. They trust the people they heard it from. They want to please the powerful, so rulers collect praise. And people love a sensational number. In his own day the armies and fortunes everyone talked of came to a tenth of the talk once the officials were asked. Rumour had made them ten times too large.

The chief cause, he said, lies deeper than all of these: not knowing how human society works. Someone who knows what a harvest can feed and what a treasury can pay stops at an impossible report, the way an accountant stops at a column that does not add up. So he set out to describe those workings, and called the result a new science.

Before a historian can say why something happened, he has to know when. Putting events in their order in time is chronology, and everything else rests on it, because a cause must come before its effect. Get two events the wrong way round and the story of how one led to the other falls apart.

Dates alone make a long, shapeless list, so historians also cut the past into named stretches: the Stone Age, the Abbasid period, the Middle Ages. This is periodization. It is useful, and it is invented. Nobody in the year 900 thought he was living in the middle of anything, and the people of the Stone Age did not know they were waiting for bronze. A period is a label that later generations stick on, and a historian who changes the labels can change how the whole story looks.

To put events in order you need a count of years, and every count starts somewhere. The Muslim count starts with the Hijra, the Prophet’s emigration from Mecca to Medina in the year Christians number 622. The starting point was chosen some years afterwards by the caliph ʿUmar, by tradition after a governor complained that the letters reaching him carried a month but no year, so that he could not tell which came first.

The Islamic year is lunar. Each month begins with the new crescent, and twelve such months come to about 354 days. The lunar year is shorter than the solar year of the seasons by about eleven days. So the months slide slowly through the seasons, and Ramadan falls in summer in some decades and in winter in others. Anyone converting a date from one count to the other has to allow for that drift.

Chronology also catches forgers. A thing that turns up in a time where it could not have been is an anachronism, and forgers produce them without noticing, because they know their own age better than the one they are faking.

In Baghdad in the eleventh century a document was produced, said to date from the Prophet’s time, that freed the Jews of Khaybar from a tax. The scholar al-Khaṭīb al-Baghdādī read its list of witnesses and declared it a forgery. One witness, the Companion Saʿd ibn Muʿādh, had died before Khaybar was taken. Another, Muʿāwiya, embraced Islam only after it. Four centuries later, in 1440, the Italian scholar Lorenzo Valla exposed a famous charter of the Roman emperor Constantine in the same way: its Latin was the Latin of a much later age.

With the order of events fixed, the historian turns to why. The first rule is that great events have many causes. The Roman Empire in the west did not fall for one reason, and scholars have proposed more than two hundred. The work is to weigh them: which causes were deep and slow, which were triggers, which the event could have happened without. An account that names a single cause is usually selling something.

The second rule guards against a tempting mistake. Two things that rise and fall together are said to show a correlation. In sixteenth-century Europe printed books and religious wars both multiplied, and it is easy to say that the first caused the second. They may be linked that way, or the other way round, or both may follow from something else. A correlation is a question to look into. Alone, it proves no cause.

The last source of error is the historian’s own mind. He knows how the story ends, and the people in it did not. Once an outcome is known it looks as if it had to happen, and the warnings look obvious. This is hindsight bias, and it makes the past seem far more predictable than it felt to those living through it. A commander who lost is called a fool, though his decision was a fair one on what he knew that morning.

The second trap is quieter. A scholar who already holds a view notices every source that agrees with it and passes lightly over the rest. This is confirmation bias. The cure is the method this whole reading has described, turned on oneself: go looking for the sources that disagree with you, and ask of your own account what you would ask of anyone else’s.