How to Read a Book You Disagree With

Forty pages into a book you think is wrong, the pencil starts to press harder. The marginalia change tone. Early on they are neutral — claim, see ch. 3, a small checkmark. Then the author says something you find not merely mistaken but galling, and the next note is a single furious NO, underlined twice, the graphite almost tearing the paper. I have a shelf of books defaced like this. For years I thought the angry notes were the mark of a reader fully awake. I have come to suspect they are the opposite.

What the NO records is not an argument. It is a flinch. I have stopped reading the sentence and started defending myself from it, and the defense is so quick and so total that no actual thinking occurs. The book becomes a sparring partner I am beating in my head, and the warmth I feel afterward is the warmth of having won a fight no one else attended.

Reading to win

There are two postures you can take toward a page you dislike, and from the inside they feel nearly identical.

The first is reading to win. You scan for the weakest sentence — the overreach, the unguarded claim, the statistic that smells off — and you pounce. You read the way a prosecutor reads a deposition, hunting for the line that, isolated and held up to the light, makes the witness look ridiculous. This is enormously satisfying and almost completely useless. It teaches you nothing you did not already believe, because you have used the whole book as raw material for a conclusion you brought with you. You have read ten thousand words and learned the contents of your own mind.

The second posture is reading to understand, and it is harder because it asks a small act of self-betrayal. You have to grant the author their strongest case before you are allowed to dismantle it. You have to find the version of the argument that an intelligent, decent person could hold without being a fool or a villain — and assume, until proven otherwise, that this is the version they meant.

This is what people now call steelmanning, an ugly word for an old virtue. The steelman is the inverse of the strawman: instead of attacking the flimsiest reading, you build the sturdiest one and attack that. If it falls, you have actually learned something. If it doesn't, you have learned something better.

You have not understood a position until you can state it so well that the person who holds it would thank you for the phrasing.

The test is brutally simple, and almost nobody passes it on the first try. Could you write the opposing case as a love letter? Not a parody, not a grudging summary with the contempt leaking through the seams — a version its believers would read and feel seen by. If you can't, you don't disagree with the book. You disagree with a cartoon you have drawn of the book, and cartoons are very easy to defeat.

Charitable but not credulous

The obvious objection is that this is a recipe for getting talked into anything. Read charitably enough and you will end up nodding along to nonsense, having steelmanned yourself into a corner. But charity and credulity are different instruments, and the confusion between them is where most people abandon the project.

Credulity lowers your standards. Charity raises the author's. The credulous reader believes the claim because it was made confidently; the charitable reader assumes the claim was made for a reason, finds the best version of that reason, and then holds it to the highest standard in the room. You are not being kinder to the argument. You are being more demanding of it. A strawman is a coward's standard — you beat it precisely because it cannot fight back.

So the order of operations matters:

• First, find the strongest form of the claim, the one you'd be embarrassed to have misrepresented.
• Then, and only then, bring everything you have against it.
• If it survives, change your mind a little. If it dies, you have killed something real, not a scarecrow.

Done this way, disagreement stops being a wall and becomes a door. The books I have actually been changed by are, without exception, books I started out wanting to argue with. The ones I agreed with from page one I have mostly forgotten; agreement is frictionless, and frictionless things leave no mark. It is the grit that does the polishing.¹

This is also, quietly, why marginalia matter — not as a record of your verdicts but as a record of your resistance, so you can go back and audit it. The NO in the margin is worth keeping precisely so that later, calmer, you can ask the only useful question: was that a thought, or was that a flinch? Most were flinches. The few that survive a second reading are the ones to trust.

The point of reading something you disagree with was never to confirm that you were right. It was to find out, under controlled conditions, whether you still are. A book you cannot be changed by is not a book you are reading. It is a mirror you are admiring — and the face in it keeps getting older while you tell it how right it is.