What Survives, and Why
A canon is the set of works a culture keeps reading, teaching and arguing about. The word is borrowed from theology, where it named the books admitted to scripture, and it carries that inheritance awkwardly: a canon looks like a list of the best works, and is in fact a list of the works that stayed in circulation.
The distinction matters because the two are not the same, and the gap between them is enormous. Athenian tragedy is the clearest case. Something on the order of a thousand tragedies were performed in fifth-century Athens. Thirty-two survive complete, around three per cent, by three authors out of dozens. We call Aeschylus, Sophocles and Euripides the great tragedians, and they may well be; but we are also describing the shape of a survival accident, since a play's chances depended on being copied, and being copied depended on being taught.
Tradition is the mechanism, not the result. T. S. Eliot's argument in "Tradition and the Individual Talent" is that tradition is not inherited but obtained with great labour, and that a genuinely new work rearranges the whole existing order rather than being appended to it. On this account the canon is not a queue.
What makes a classic? The proposals that survive scrutiny share one feature, they are all about rereading rather than reading:
| Proposal | The test |
|---|---|
| Sainte-Beuve | Enriches the mind, adds to its store |
| Eliot | Maturity of mind, manners and language; comprehensiveness |
| Calvino | A book that has never finished saying what it has to say |
| Coetzee | Whatever survives the criticism of successive generations |
| Kermode | Sustains readings its author could not have intended |
Calvino's is the most useful for study, because it is a claim you can test: a classic yields something on the second reading that the first could not have given, and a merely good book does not.
The canon also churns. Works enter and leave. Donne was minor for two centuries and central after 1920; Longfellow was world-famous and is now barely read. A canon of a hundred works that admits five and drops three per generation is not a fixed list but a slow current, and after ten generations it holds a hundred and twenty works of which many are new.
Common pitfall: treating the canon as a verdict on quality, or as a conspiracy that has nothing to do with quality. Neither survives contact with the evidence. Survival is filtered by institutions, languages, printing and teaching, and the filter is not random, because works that reward rereading are the ones people bother to copy.