The one-idea test: when to split a card

Every method that uses cards has one rule in common, and it is the rule people break first: a card holds one thing.

It sounds like a rule about neatness. It is not. It is the rule that makes everything else work — sorting, ordering, finding what is missing, trying a different order — because all of those depend on being able to pick up a card and move it without moving anything else. A card that holds two things cannot be moved without moving both. It is, for every purpose the cards exist for, two cards stuck together.

This guide is about how to tell, and what to do.

The test

Try to give the card a title. One line, across the top, that says what the card is about.

If you can, the card holds one thing.

If you cannot — if the title needs an and, or a semicolon, or you find yourself writing notes on chapter 3 — the card holds more than one thing, and the and is where the cut goes.

That is the whole test. A title is a claim about what a card contains, and a card that resists a title is telling you something.

Why a card with two ideas costs more than two cards

Suppose a card says: The canal cost three times its estimate, and the merchants who backed it knew.

Two things: the overrun, and the merchants' knowledge. They may belong together in the paper. They do not belong together on the card, because:

• They may need different evidence. The overrun is in the treasury figures (S2). The merchants' knowledge is in a letter (S7). Under one card you have to pile both, and you can no longer see which claim is thinly supported. (The overrun has four cards; the merchants' knowledge has one, and it is the whole argument.)
• They may go in different places. The overrun might open the paper; the merchants' knowledge might be the last thing you reveal. One card can be in one place.
• One may survive and the other not. You may find the merchants did not know. The card now has a true half and a false half, and every pile it sits in is contaminated.
• A reader — or a check — cannot tell which half a link points at. S7 supports this card: supports what? The overrun or the knowledge?

Two cards fix all four. The cost is a second card. There is no other cost.

The three ways cards get fat

Notes on a source. The first card most people write is notes on Smith chapter 3, with six things on it. This is not a card; it is a page from a notebook in a card's shape. Cut it into six. Give each one Smith's number and the page. Guide 2 is about doing this for all your sources at once.

A claim and its reason. The vote was rigged because the count was held in private. The claim (rigged) and the reason (private count) are two cards, linked: the private count is evidence for the rigging. Kept as one, the paper can never argue that the count was private but the vote was fair, or ask whether a private count proves anything — both of which are questions a good paper asks.

A scene that keeps going. In fiction, a card that starts at breakfast and ends with the phone call at noon is two scenes: the one where she decides, and the one where she is told. Each has its own turn. The cut goes at the turn.

When not to split

The test cuts the other way too. A card whose title is one line and true does not need splitting because it is long. A quotation of two paragraphs is one card if it is one quotation. A claim with three sentences of explanation is one card if it makes one claim. Length is not the measure; number of things is.

And do not split so far that a card holds nothing: the canal is not a card, it is a topic. A card has to say something — a claim makes a claim, a fact states a fact, a quotation quotes. A title you could put on twenty different cards is not a title yet.

How to cut

Take the fat card. Write the two (or six) titles first, on fresh cards, before copying any text. If you cannot write the titles, you have not found the seam; read the card again for the and.

Then move the text: each sentence to the card whose title it belongs under. Sentences that belong under neither are usually the connective tissue between the two ideas — and this matters because — and they belong in the paper, not on a card. Write them on the back of the card that comes first, as a note to yourself about the bridge.

Keep the links. If S7 supported the fat card, decide which of its children S7 actually supports, and move the link there. Usually it is one of them, and that is the whole point.

Renumber nothing. Card numbers are names, not positions.

What splitting does to the paper

The paper gets more paragraphs and shorter ones, at first. That is the cards showing you that the fat card was two paragraphs pretending to be one. Later, some of the split cards will merge back — not into one card, but into one paragraph, in the draft, with a sentence between them. That is fine. The cards' job was to let you see the two things and decide their order. Once the order is decided, the prose can be as continuous as you like.

What the paper does not get is the sentence that argues two things at once and proves neither. Those sentences come from fat cards, and they are the ones a reader — or a marker — stops on.

Doing this in software

Software that works from cards usually gives a card a title line and a body, which makes the test automatic: a title that will not fit on its line is the and showing. Some software also lets a card be split at a point in its text, with the links deciding which half they follow, and will notice when a card's body has grown past the length a single idea usually needs — not to stop you, but to ask. Nothing about the test changes. The card still has to be able to say what it is about in one line, and you still decide where the cut goes.