Find the step nobody wrote down.

This page's argument

Here's what this page argues, laid out the way it teaches. Six parts, each labeled:

The rest of the page is the support. The next section is the data. The four-day week example tests the warrant, and Toulmin's own example is the backing.

The step nobody writes down

Here's an argument the way most of them get made, in a team meeting or a memo:

It has data and a claim. It sounds reasonable. But four of the six parts are blank, and the one that matters most is the warrant: the step that gets you from the data to the claim. Nobody said it out loud, so nobody checked it.

Stephen Toulmin called this missing piece the warrant. In David Hitchcock's summary of Toulmin's book, the data answer "What do you have to go on?" and the warrant answers "How do you get there?" Toulmin described warrants as "general, hypothetical statements, which can act as bridges" between the two.

Lay it out in six parts

Toulmin's layout has six parts, in a fixed arrangement:

• Data, also called grounds: the facts you start from.
• Claim: what you want the listener to accept.
• Warrant: the general rule that links the data to the claim.
• Backing: what supports the warrant itself.
• Qualifier: how sure the claim is. "Probably," "presumably," "certainly."
• Rebuttal: the conditions under which the claim would fail.

Here's the same argument with every part filled in:

Writing the warrant down changes what you can claim. "What worked for the pilot group will work for the rest of the team" is only true if the pilot group is like the rest of the team. The backing for that is thin: the 12 volunteered. People who wanted a four-day week were the ones who got one. So the rebuttal is real: the claim fails if the gain came from who volunteered, or from the novelty of any change, which may fade. That's why the qualifier is only "presumably."

With the rebuttal on the page, the claim you can defend is smaller. Not "switch all 60," but "test it properly first," which is the box under the figure. A random pilot is exactly what would tell you whether the rebuttal holds.

The data didn't change. The numbers are the same. Only the step became visible.

Toulmin's own example

Toulmin's book, The Uses of Argument (1958), uses one example throughout its chapter on the layout of arguments. Here it is in the same layout, as Bart Verheij and David Hitchcock report it:

The warrant says "generally," not "always," so the claim can only follow presumably. The rebuttal names the exceptions. Every part of the argument now says how much it can carry.

This isn't the Pyramid Principle, which orders a set of reasons for a busy reader. And it isn't an explorable explanation, which lets the reader change the numbers. The Toulmin model opens up one inference, and shows what it rests on and when it fails.

Watch it laid out

The clip starts with the four-day week argument as it was first written. Watch the box that appears under the step, and what happens to the claim once the qualifier goes in.

Check every argument

Every argument on this page is checked when the page is built. Each laid-out argument must fill all six parts, and the worked example's numbers must agree with each other:

Argue this way

1. Write the claim. One sentence: what do you want the listener to accept?
2. Write the data. The facts you're starting from, with numbers and sources.
3. Write the warrant. Finish the sentence "This data supports this claim because..." in general terms. This is the step most arguments leave out.
4. Back the warrant. What makes the warrant true here? If the backing is thin, say so.
5. Write the rebuttal. Under what conditions would the claim be wrong? Name them.
6. Set the qualifier. Given the backing and the rebuttal, how sure can you be? Change the claim if you have to.
7. Read it to someone who disagrees. Ask which part they'd attack first. It's usually the warrant.

Ask your agent

Lay out the argument below using Stephen Toulmin's model (The Uses of Argument, 1958). Give each of the six parts its own labeled line:
1. Claim: what the argument wants the reader to accept.
2. Data: the facts it starts from, as stated in the text.
3. Warrant: the general rule that links the data to the claim. If the text doesn't state one, write the warrant it must be assuming, and mark it "UNSTATED".
4. Backing: what supports the warrant. If the text gives none, write "NONE GIVEN".
5. Qualifier: how sure the claim can be, given the backing.
6. Rebuttal: the conditions under which the claim would fail.
Then name the weakest link. If the warrant was unstated, say so first: it's usually the weakest link because nobody checked it. Finally, rewrite the claim so it matches the qualifier.
Don't add facts that aren't in the text.

Argument:
[paste here]

What good output looks like: a warrant you could argue with, marked UNSTATED if the author never wrote it, and a rewritten claim that's no stronger than its backing.