The Methodology

We learned the order by getting it wrong. The first period we wrote up, we assembled from the top: we knew what we wanted to say and went looking for evidence. What came out was too clean. "The machine decided who survived" is a sentence that sounds correct and explains nothing. Rebuilt from the bottom, the same period said something narrower and much stranger, and that version survived.

Every conclusion we like goes through five checks before it stays.

Are we smuggling in intent? The test is whether the supposed orchestrators also lost money. If they did, it was a relay and a matter of timing, not a plan.

Is this actually about art, or is it a generic financial bubble wearing an art costume? The art-specific part has to be separated from the part that would look the same in any asset class.

Is the striking word the right word? A vivid term that is slightly wrong will be the first thing an opponent takes apart.

Are we ignoring the other side? Somebody usually did get paid, some of it usually was real, and leaving that out costs more than including it.

And is the top line still about appointment? Arguments drift toward inequality and finance, which are more familiar and easier to knock down.

The evidence list and the list of comparison cases are fixed before the conclusion is written. Anything that turns up after we know what we want to prove doesn't get to join.

Each period gets read four times.

The art. What was made, by whom, what was good. This layer is here because the alternative is a book about machines with no art in it, and that book would be false. Talent is a factor in the formula, not a zero.

The appointers. Who decided, what they were paying with, and why they had the standing to decide. This is the axis of the project, and it is where the formula and the comparison cases live.

The context. What else was happening and why it happened then. This layer does more work than it looks like it does. Once you can see that a given break came out of a new buying class, a change in materials, and a new technology taking a job away from painters, the conspiracy reading collapses without anyone arguing against it. Context is the vaccine.

The counter-evidence. Where the pattern fails, where our figures are contested, where the machine ran over its own operators. This is not a disclaimer at the bottom of the page. A writer who volunteers the inconvenient cases earns the convenient ones.

Two things follow.

Period boundaries are drawn where the appointer mutates, not where a style changes. The test is one question: did the way a thing becomes art change? That is why our borders don't line up with the ones in a survey textbook.

And the four layers hold at three levels of zoom. A period can be a paragraph, a long essay, or a chapter about a single object. The detail changes and the position doesn't. If the argument doesn't survive twenty thousand characters, four hundred pages won't save it.

Every fact is verified once, with a source and a date, and it lives in one file. Essays, cards, posts, and chapters are drawn from that file. Nobody re-checks and nobody re-rounds.

One figure has one source, and it is identical everywhere it appears. A painting that sold for two guilders thirty cents in 1881, sourced to the museum's provenance record, is two guilders thirty cents in a two-line post and in a thirty-page chapter. The only thing that changes is how much sits around it.

When sources disagree, the disagreement goes into the base as a disagreement. One well-known sale is reported at over eight million dollars in one place and at twelve in another. So we publish the range, or we publish nothing until it is closed. What we don't do is let the larger number drift into the short formats while the smaller one stays in the book.

Anything unresolved is flagged and phrased carefully. Works were "assessed as having lost value," not "are worthless."

No source, no claim. On this subject one unchecked figure costs more than a year of good writing.

Money is a late special case. It becomes the fuel of appointment around the nineteenth century, and even now it isn't the fuel in every case.

Before that, appointment ran on other things. Ritual authority and organized labor. Political capital. The institutional prestige of a state academy. In the postwar decades, ideology and covert funding on one side of a border and direct coercive power on the other.

So we ask one question of every period: what resource paid for the appointment, and where did it flow? That question shows the wiring — who made it, who certified it, and who took the money, which are often three different parties.

It is not a question about who profited unfairly. We are describing a machine, and we find it a good one.

It does not measure talent and it does not rank artists. There is no score at the end and no number to quote.

It cannot tell you who was erased completely. Anyone the machine never touched left no documents and is invisible to us by construction. The missing denominator is real, and we say so before anyone says it to us.

It cannot prove sufficiency. We can show a machine ran, name who ran it, and date it. Nobody gets to rerun the century without it.

And it does not settle whether the work is good. That question is not ours, and answering it would break the only thing this method has going for it.