The Toulmin card method

Stephen Toulmin's layout of argument is taught in every rhetoric course and used in almost none of the writing that follows, and the reason is practical. The layout is a diagram — claim, grounds, warrant, backing, qualifier, rebuttal, with arrows — and prose is not a diagram. A writer who has learned that every claim needs a warrant still writes paragraphs, and paragraphs hide warrants beautifully. The diagram sits in the notes from the course, and the argument in the thesis has the same unexamined joints it would have had anyway.

Cards are a diagram you can write on. Each of Toulmin's six elements is a kind of card, each arrow is a link, and the layout stops being something you draw once and becomes the shape of the deck. The result is that the element prose hides best — the warrant — becomes a card that is either there or not, and a missing one is visible on the table.

The six elements, as cards

The claim. What you are asserting. One sentence, with a verb, that could be disputed. A claim card, as in every guide before this one.

The grounds. The evidence — the data, the quotation, the finding, the observation — on which the claim rests. Evidence cards linked to the claim with supports. These are the cards undergraduates learn to put under a claim, and at this level they are the same cards.

The warrant. The principle that licenses the move from the grounds to the claim: why does this evidence support this claim? If the grounds are the sample showed effect X and the claim is X is a general property, the warrant is the sample is representative of the population — and stated like that, it is obviously a thing that might be false. The warrant is a card of its own, linked supports to the claim with the link's note reading warrant (so it is told apart from the grounds), and it is the card this guide exists to make you write.

The backing. The support for the warrant itself: why should anyone accept the principle? The sample was drawn by the method the field regards as representative (S12). Evidence cards linked to the warrant, exactly as grounds are linked to the claim. A warrant with no backing is a claim with nothing under it, one level down.

The qualifier. How strongly the claim is held: probably, in laboratory settings, for adults, unless… Not a card but a field on the claim card, because a qualifier is a property of the claim and not a thing in its own right. A claim with no qualifier is either certain or unexamined, and it is almost never certain.

The rebuttal. The conditions under which the claim would not hold, or the objection that would defeat it. A counterpoint card linked with contradicts — the card student guide 5 is about — with the difference that a Toulmin rebuttal is often aimed at the warrant rather than at the claim: the sample is not representative, because… Where a rebuttal points is information, and the link's target says.

So: two card types you already have (claim, evidence), one new one (warrant), one field (qualifier), and the links you already use — supports (grounds to claim; backing to warrant; the warrant itself to the claim it licenses, with the link's note saying warrant) and contradicts (the rebuttal, aimed at the claim or at the warrant). That is the whole apparatus.

The method

Step 1: Claims with grounds, as usual. Build the row: claims with evidence under each (student guides 1 and 2). At the end of this step you have what every card method produces, and it is not yet a Toulmin argument.

Step 2: For each claim, write the warrant. Take one claim and its evidence. Ask: why does this evidence support this claim? Write the answer as a sentence on a warrant card and link it. Do this for every claim in the row.

Three things happen, and they are the method's payoff:

• Some warrants are obvious and boring. A direct measurement of X is evidence of X. Write it anyway; a warrant that is obvious costs one card and a warrant that is not obvious looks identical until it is written down.
• Some warrants are surprising. You write the sample is representative of the population and realize you do not believe it. The claim has just been narrowed, or the qualifier has just been found, and it happened on a card instead of in a viva.
• Some warrants cannot be written. You stare at the evidence and the claim and cannot say why one supports the other. This is the joint prose was hiding. Either there is a warrant you have not found — go and look — or the evidence does not support the claim, and the claim needs different grounds or has to go.

Step 3: Back the warrants that need it. A warrant that a reader in your field would grant without argument needs no backing card. A warrant that is contestable — and in most graduate work the contestable warrants are the whole argument, because the grounds are rarely in dispute and the inference from them always is — needs its evidence, linked to it exactly as grounds are linked to the claim. Now there are evidence cards at two levels, and the table shows which level is thin.

Step 4: Qualify. Go down the row and give every claim its qualifier. For this population. Under these conditions. Probably. Unless the alternative explanation in card X holds. A claim that resists qualification is a claim you have not examined; a claim whose qualifier turns out to be in the one setting where it was measured is a claim that has just been given its true size.

Step 5: Rebut, and say where the rebuttal points. For each claim, write the strongest objection (student guide 5) — and link it to the element it attacks. A rebuttal of the grounds (the data are wrong) is a different thing from a rebuttal of the warrant (the data are right and do not show what you say), and the answer to each is different: the first needs better evidence, the second needs a better principle or a narrower claim. The link tells you which.

What the table shows

Laid out, a Toulmin deck has a shape: claims in a row; under each, grounds; beside each, a warrant with its own backing beneath; the rebuttals pointing at grounds or warrants. Read it as a reviewer would:

• A claim with grounds and no warrant — the commonest gap, and invisible in prose.
• A warrant with no backing that a reader would contest — the second commonest, and the one examiners find.
• A rebuttal aimed at a warrant, with the answer aimed at the grounds — an objection answered in the wrong place, which reads in prose as a paragraph that changes the subject.
• A qualifier that does not match the grounds — X holds generally over evidence from one setting.

Each of these is a finding you can make yourself, on the table, before anyone else does.

Writing from a Toulmin deck

The temptation is to write the diagram: My claim is X. The grounds are Y. The warrant is Z. Do not. Nobody reads that. Write the paragraph — the claim as its first sentence, the grounds as its body — and let the warrant surface where a reader would ask why: as the sentence that follows the evidence and precedes the claim's restatement, or as the qualifier that limits the claim to what the warrant licenses. The backing goes into the paragraph only when the warrant is contested; otherwise it is a citation. The rebuttal gets its own paragraph, beside the claim, answered at the level it was made.

The deck's job was to make sure every one of those sentences exists and points at the right thing. Once it does, the prose can be as continuous as you like. Toulmin's layout is not a way to write; it is a way to check what you have written against what an argument needs, and cards are the only form in which the check is cheap.

Doing this in software

Software that works from cards will usually let you define a card type of your own — a Warrant, with its fields and what it owes (a supports link out to a claim, say) — so that a warrant linked to nothing is a structural fact the software can point out when you ask, exactly as it points out a claim with no grounds. The link types are the ones you already have: the warrant supports the claim, with the link's note saying so; the backing supports the warrant; the rebuttal contradicts whichever it attacks. The qualifier is a field on the claim. What the software cannot do is write the warrant, judge whether one is contestable, say whether the backing is enough — or, yet, notice a claim that has grounds and no warrant: that check is the table's, and yours. The argument is yours.