The Prosecutors Kept the Record
Joan of Arc has one of the densest records of any medieval life, and her prosecutors produced it. Four properties get collapsed into the phrase well documented, and they trade off against each other rather than moving together.
- You are seeing
-
- A name is described as well covered and the number of accounts is the only figure given
- Coverage count is used as a proxy for how reliably a business is understood
- Several accounts agree and all of them derive from the same disclosure
- A view moves substantially with no new evidence and the move is described as a reassessment
- The most stable narrative in a sector belongs to the company that supplies the most of it
- The mechanism
- How much is known about a subject is four separate measurements — lag to the earliest source, that source's relationship to the subject, how many accounts derive independently, and how far assessment moves without new evidence — which trade off against one another and are reported as one adjective.
- The older apparatus
- Prosopography, which refuses an unqualified well documented, names which of the four axes a comparison is running on, and treats documentary density and reliability as different claims requiring different evidence.
- The false friend
- Genuine independent corroboration, which does raise confidence. It is separated by asking what each account derives from rather than counting accounts, since several sources reprocessing one supplied statement is one route to one assertion in several containers.
- The discriminating test
- Produce the four numbers separately for one subject. Where only the count of sources is available, the axes have been collapsed, and the count is usually the one that moves most easily.
- On your own data
- For a set of covered names, compute the share of published accounts whose distinctive content can be traced to a subject-supplied disclosure, and regress assessment dispersion on that share rather than on the number of accounts.
Joan of Arc’s 1431 trial was recorded daily, and three copies of the transcript survive.
Set beside almost any other fifteenth-century life, that is an extraordinarily dense record. It is also a document produced by her prosecutors, during an adversarial ecclesiastical process whose purpose was to condemn her.
The corrective record exists because of that. The rehabilitation trial of the 1450s collected 115 witnesses. It was run precisely because the first account could not be trusted as a description of her life.
So the densest medieval record on the roster is dense twice over, for two incompatible reasons, and neither density figure is a measurement of reliability.
Four properties, moving in different directions #
Prosopography keeps them apart because they trade off rather than reinforcing each other.
Lag is the gap between the events and the earliest surviving account. Priscus dined at Attila’s court during an embassy in 448 and 449 and wrote it down himself. Tacitus wrote about Boudica roughly forty years after her death and Cassius Dio roughly a hundred and forty. The earliest written attestation of William Tell is around 1474, for an event placed in 1307.
Relationship is who wrote it and why. Ashoka’s edicts are his own policy statements, cut into stone in his own name. Boudica and Spartacus are described entirely by writers from the side that defeated them. Richard III’s skeleton, examined in 2012, sits at a third position again, being evidence with no author at all.
Density counts accounts that derive independently of one another. Boudica is four passages by three authors with no contemporary corroboration. Pytheas of Massalia is thin in a stranger way. His own work is lost, and what remains is roughly thirty-nine fragments quoted inside other writers’ books. No two of them are independent accounts of the same voyage. Strabo, citing Polybius, accused him of making the journey up.
Volatility is how far a standing has moved without new evidence. Richard III was vilified by Tudor chronicles, fixed there by Shakespeare, and partly rehabilitated in the twentieth century well before any skeleton was found. The hostile Roman portrait of Cleopatra outlasted the war that produced it by well over a thousand years.
No ranking falls out of the four #
A subject can have an eyewitness whose account is a single uncorroborated fragment, which is close to Priscus on Attila. A subject can have a dense record produced entirely by one interested process, which is Joan of Arc’s trial before the retrial.
Neither is better documented than the other. They are documented differently, and the phrase that collapses them into one adjective is doing the damage.
The discipline’s own rule is to name the axis before making the comparison.
That sounds like pedantry and it is the entire finding. Four properties that move in different directions cannot be summarised by one word without the word tracking whichever of them is easiest to count.
Coverage is the same adjective #
A name followed by twenty-five analysts is described as well covered, and the count is usually the only figure quoted.
Underneath it are the same four properties. How close to the events the reporting is. Who holds the banking relationship, who has the access, who is on the call list. How many of the accounts contain anything that did not originate in a company-supplied disclosure. How far the consensus view has moved in the absence of new evidence.
Those do not move together either, and the fourth is the one nobody logs.
A sector with high coverage counts and wide dispersion is a different object from one with high counts and none. Reported as coverage, the two are the same number.
The subject is still writing #
Ashoka is the case that inverts the expectation.
His standing has moved less than almost anyone else’s on the roster, across two thousand years and several complete changes in historical fashion. His own edicts survive, and became readable again after Prinsep deciphered them in 1838. Every later commentator has been working from something close to his own words.
The self-authored source stabilised the reputation. In history that is a genuine gain, because Ashoka stopped issuing edicts a long time ago and the corpus is closed.
The transfer stops there.
A company supplying most of the record about itself produces the same stability by the same mechanism. The corpus is not closed. The subject is still authoring, still choosing what to release and when.
So the steadiness measures the authorship rather than the business, and it is the names with the least dispersion that this is most true of.
Boudica has four passages by three authors, none of them contemporary, none of them friendly, and her reputation has been rebuilt more than once since.
Ashoka has his own words in his own name and has barely moved in two millennia.
The steadier record is the one its subject wrote, which is also the one worth checking.
Questions
Does more coverage mean a subject is better understood?
It means one of four things, and coverage count is the weakest of them. Joan of Arc's record is dense — three surviving copies of a daily-recorded trial transcript — and the transcript was produced by prosecutors running a process designed to condemn her. The 115 witnesses of the 1450s rehabilitation exist precisely because the first record could not be trusted as an account of her life. Density bought corroboration there only after a second, differently motivated process was run.
What are the four properties that get collapsed?
Lag between the events and the earliest surviving account. The relationship of that source to its subject. Documentary density, meaning how many accounts exist that derive independently of each other. And volatility, meaning how far assessment has moved without new evidence arriving. They trade off rather than reinforcing. Priscus dined at Attila's court and wrote about it directly, which is a near-zero lag attached to a single uncorroborated fragment. Joan of Arc's 1431 trial is the opposite: a dense record produced entirely by one interested party.
How thin can a well-known subject's record actually be?
Boudica is four passages by three authors with no contemporary corroboration of any kind, all of them Roman, written for a Roman audience about someone Rome had just destroyed. Pytheas of Massalia is thinner in a different way: his own work is entirely lost and survives as roughly thirty-nine fragments quoted by other writers, with no two independent accounts of the same voyage to check against each other. Strabo, citing Polybius, accused him of inventing the journey.
Why is a stable reputation not evidence of a well-understood subject?
Because stability tracks who wrote the record. Ashoka's standing has moved comparatively little across two millennia of changing historical fashion, and the reason is that his own edicts survive and became legible again after Prinsep deciphered them in 1838, so later commentators work from something close to his own words. Richard III, known mainly through hostile Tudor chronicles, has been reassessed twice. Subjects known through adversarial sources swing hardest; subjects who supplied their own record barely move.
What does a lag figure actually buy?
Less time for accretion, and nothing else. Priscus dined at Attila's court during an embassy in 448 and 449 and wrote about it directly, a lag near zero, and his account is still individually disputed on accuracy. At the other end, the earliest written attestation of William Tell dates to roughly 1474 for an event traditionally placed in 1307. Proximity removes one class of error without touching the others, which is why it is one axis rather than the ranking.