Canon formation
A Socratic walk-through of canon formation — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do a handful of works get taught for centuries while equally admired contemporaries disappear entirely?
Open a nineteenth-century review journal and you find writers praised in the same breath, by the same critics, in the same terms. One of them is now on every syllabus. The others are not merely less read; they are unobtainable, unedited, untranslated, absent. The comfortable explanation is that time sorted them and quality won. But notice what that explanation quietly requires — that survival is a measurement of merit — and notice that we have no way to check it, because the works that would supply the comparison are precisely the ones we cannot read. What would we expect to see if survival were not a measurement?
Reasoning it through
REASONING #We can actually answer that, because someone built the experiment. Salganik, Dodds and Watts recruited 14,341 people to a music download site with 48 unknown songs, and split them into separate parallel "worlds". In one world people saw only the songs; in the others they also saw how many times each song had been downloaded by the people before them in that world. Same songs, same starting conditions, independent runs. If quality determined outcome, every world should have converged on roughly the same ranking. They did not. Inequality rose and, more damningly, so did unpredictability: a song that came first in one world came near-bottom in another. Quality was not irrelevant — the best songs rarely did terribly and the worst rarely triumphed — but inside that band, the winner was whoever got an early lead.
So consider what an early lead does over long spans. A work that gets taught acquires a scholarly edition, a translation, an introduction, a set of exam questions, a body of criticism to argue with. Every one of those makes it cheaper to teach next year. Its unlucky contemporary has none, so a lecturer who wanted to teach it would have to build all of that first, alone, for a work no student has heard of. The competition is not between the two books. It is between one book with infrastructure and one without, and the gap widens by itself. That is cumulative advantage, and it needs nobody to believe the canonical work is better — only that it is easier.
Underneath the teaching runs something more physical. Reprinting requires a publisher expecting sales; sales require the book being known; and the copyright status matters enormously, since a long-dead author's work can be reprinted by anyone at no licence cost while a neglected contemporary's estate may be untraceable. Libraries discard what is unborrowed. Anthologies are compiled largely from previous anthologies. The record thins in a way that is correlated with attention, not with worth.
Does that make the loop permanent? Evidently not, and the exceptions are instructive. Bach's choral music was a specialist memory until Mendelssohn staged the St Matthew Passion in Berlin in 1829. Melville died with Moby-Dick long out of favour, and its standing was made in the 1920s, seventy years after publication. Zora Neale Hurston was out of print and buried in an unmarked grave when Alice Walker's 1975 essay began the reversal. Look at what those cases share: not a new discovery of the text, which was always available to anyone who wanted it, but a champion with the institutional standing to make others look. Revival is not the loop running backwards. It is an external push large enough to restart it.
Which leaves the honest difficulty. Everything above explains how a canon could be arbitrary; none of it proves this one is. The works we retained may genuinely be the better ones, and the drift may be riding on real differences it then exaggerates. We cannot settle it from inside, and we should be wary of both parties: those who defend the canon are largely people whose expertise is invested in it, and those attacking it are largely people whose careers advance by widening it. Both are describing something true, and both have a stake in the answer.
The analogy
THE ANALOGY #Think of a bus route laid down early, along one of two equally suitable streets. Houses, shops and jobs accumulate along the served street, which raises ridership, which justifies more buses. Fifty years on, the served street is obviously the right one for a bus — and it is right, now, because of the route, not before it. The other street is not worse. It is unaccumulated.
Bus routes have a planner who can redraw the map and a passenger count to check against, whereas no one plans a canon and there is no independent measure of literary merit to hold it up to — so the analogy gives us a correction mechanism the real case lacks.
Clarifying the model
THE MODEL #The claim is not that canonical works are bad, nor that merit does nothing. It is that merit sets loose bounds and history picks inside them, after which the pick becomes self-confirming: taught works generate the apparatus that makes them teachable, and the apparatus is then read as evidence of the merit that justified teaching them.
Two refinements are worth keeping. First, the mechanism is about infrastructure rather than opinion — editions, translations, syllabi, criticism — which is why a change in taste alone rarely dislodges anything, and why cheap reproduction and digitisation genuinely alter the odds. Second, the loop's silence is not evidence of judgement having been made. A work can vanish without ever being weighed, and the absence of criticism about it is often mistaken for a verdict against it, which is the same error as reading an empty archive as an empty past.
A picture of it
THE PICTURE #How to readStart at Published, top left, and notice there are two ways out and they behave differently. The upper path contains a genuine cycle — being taught produces the editions and criticism that keep it anthologised, which keeps it taught — so once a work is inside, staying is close to free. The lower path has no cycle of its own; neglect only deepens. The single arrow climbing back from Unobtainable is the revival case, and it is drawn thin on purpose: it needs an outside champion with standing, since nothing inside the lower states can generate it.
What became clearer
WHAT CLEARED #Survival is not a measurement — it is the output of a self-financing loop, in which being taught makes a work cheaper to teach next year. The parallel-worlds experiment shows what such a loop does: quality sets loose bounds, and early luck decides the ranking inside them. The vanished contemporaries were mostly never weighed and rejected; they were never weighed. And since escaping the lower path takes an outside champion rather than a rereading, absence from a canon tells you far less about a work than its presence tells you about the machinery.
Where to go next
ONWARD #- What digitisation and public-domain reprinting have actually done to the reprint threshold.
- How disciplines that periodically rebuild their syllabus from scratch differ from those that inherit one.
Key terms
TERMS #| Term | What it means |
|---|---|
| Cumulative advantage | the tendency for an early lead to generate the conditions for a larger lead, independently of merit. |
| Canon | the set of works an institution treats as required, and therefore keeps in print, in translation and under commentary. |
| Path dependence | the property of an outcome being determined by the order in which events happened rather than by their intrinsic weights. |
| Deaccession | a library's removal of an item from its permanent collection, typically driven by borrowing statistics. |
Every term the collection defines is gathered in the glossary.