Continuity failure modes, and which ones zero drafting catches

The eight modes

• **1. Fact collision.** Two established facts that cannot both be true. Maren is nine; Maren is twenty.
• **2. Arithmetic drift.** Ages, dates, durations, and distances that do not reconcile once written out as a sequence. The most common failure in long-running work, and the easiest to introduce, because each number is correct in isolation.
• **3. Knowledge-state error.** Who knows what, and when. A character acts on information they could not have. The most common continuity failure in produced television, and the least visible in an outline.
• **4. Spatial impossibility.** Someone is somewhere they could not have reached, or a room's geography changes between scenes.
• **5. Rule violation.** Your own world rules broken — the magic system, the technology, the cost the story established for using it.
• **6. Object drift.** Props and artifacts that move, change, or reappear after being destroyed. Endemic in visual media.
• **7. Voice and characterization drift.** A character gradually stops sounding or behaving like themselves across a long work. No single scene is wrong.
• **8. Accidental retcon.** Canon changed without anyone deciding to change it — usually by improvising an answer in one scene that contradicts an answer improvised in another.

What zero drafting catches well

**Modes 1 through 6** — everything a scene is forced to instantiate.

The mechanism is the same in each case. A scene written at full resolution cannot stay vague. It has to place people somewhere specific, at a specific time, knowing specific things, holding specific objects, under rules that either permit the action or do not. Every one of those is a query against the record, and a query is what surfaces a collision.

Arithmetic drift and knowledge-state errors are where it is most dramatically better than the alternatives, and for the same underlying reason: both are invisible in summary. An outline can say "Eli confronts Maren" without ever needing to know how old either of them is or what Eli has worked out. A scene cannot write itself without deciding both. The worked example in zero drafting a pilot is an arithmetic drift that had passed every previous check because the facts had only ever been examined in isolation.

Object drift and spatial impossibility are caught well too, though in visual media they need the dedicated space pass described in zero drafting for animation pipelines, because much of the relevant canon was drawn rather than written.

What it catches partially

**Mode 8, accidental retcon**, is a mixed case and worth understanding precisely.

Zero drafting is excellent at *preventing* accidental retcons and mediocre at *finding* ones that already happened. It prevents them because the entire discipline is refusing to answer questions mid-draft: you log the question, you resolve it into the record afterwards, and the answer exists in one authoritative place. Improvising an answer inside a scene — which is how accidental retcons are born — is exactly what the method forbids.

Finding an existing one is harder, because you will only catch it if the scene you happen to be drafting touches both contradictory answers. If the improvised answer lives in episode two and the original in episode six, the pass will only surface it if the scene needs both. That is luck, not method.

What it will not catch

**Mode 7, voice and characterization drift**, is genuinely outside its range, and no amount of skill at the method changes that.

Drift is not a property of any scene. It is a property of a trajectory. A character who has slowly become someone else over eighteen episodes is behaving perfectly consistently in the scene you are drafting — the scene is not wrong. There is nothing for the record to contradict, because the character is contradicting an earlier version of themselves that no document ever captured as a rule.

Catching it requires reading at length and comparing across time, which is a different operation: a read-through, an editor, or someone who knew the character in episode one coming back fresh. If your problem is that a character has stopped sounding like themselves, zero drafting is the wrong instrument and running it will produce a clean pass that tells you nothing.

The second blind spot is subtler: **anything your record is confidently wrong about**. A zero draft tests the scene against the record. If the record itself contains a settled error, the scene will agree with it and the pass will come back clean. This is not a small caveat — it is the reason the record has to be maintained deliberately rather than accreted.

Covering the rest

No single instrument covers all eight, so the practical answer is a small stack where each part does what the others cannot.

• **An authoritative record** handles the base condition. One place the facts live, that people check rather than remember. Everything else assumes this exists; nothing works without it.
• **Zero drafting** surfaces what the record has not decided and what a scene cannot support — modes 1–6, and prevention for 8.
• **A systematic scan across the whole world** catches collisions no scene happened to touch. This is where a tool beats a person, because it can hold every pair of facts at once and a person provably cannot. CanonBoard's Conflict Detection does this against your canvas, on Core and above.
• **A read at length by a fresh reader** is the only thing that catches mode 7.

The reason to keep the taxonomy in mind is that it tells you which instrument to reach for. A team that runs zero drafts diligently and still ships a character who stopped sounding like themselves has not failed at the method — it has used a scene-resolution instrument on a trajectory-level problem. Why that class of error is so persistent in the first place is the subject of the logic problem.

Frequently asked questions

What is the most common continuity error in produced television?

Knowledge-state errors — a character acting on information they could not have had at that point. They are the most common because they are invisible in an outline: summary does not track information as it moves between characters, so nothing flags it until a scene forces the question.

Does zero drafting catch every continuity error?

No. It catches fact collisions, arithmetic drift, knowledge-state errors, spatial impossibilities, rule violations, and object drift — anything a scene is forced to instantiate. It does not catch voice and characterization drift, which is a property of a trajectory rather than any single scene.

What catches voice and characterization drift?

Reading at length and comparing across time — a full read-through, an editor, or someone who knew the character early coming back to the later material fresh. No scene-level instrument can detect it, because the drafted scene is not itself wrong.