Logical Fallacies: Where an Argument Actually Breaks

A fallacy is a break in the wiring, not a wrong conclusion

An argument has two separate parts: the premises (the reasons offered) and the structure (how those reasons are supposed to lead to the conclusion). A logical fallacy is a fault in the structure — the reasons don't actually support the conclusion, even if they're all true, even if the conclusion happens to be correct.

That last point matters and trips people up: a conclusion can be right and the argument for it still be broken. "The sky is blue, therefore you should trust me" reaches a true-ish first clause and a non-sequitur. Spotting a fallacy doesn't prove the conclusion false — it proves this particular argument doesn't establish it. You've found where the wire is cut, not decided what's on the other end.

Why learn to name them? Two reasons. Defensively, a named fallacy is one you can no longer be steered by — the move stops working once you see it. Offensively — and more importantly — knowing them makes your own arguments harder to break, because you stop making these mistakes yourself. The goal isn't to win comment sections by shouting "straw man". It's to see structure.

The ten, and where each one breaks

1. Ad hominem — attacking the person, not the argument

Rejecting a claim by attacking who made it. Where it breaks: the truth of a claim doesn't depend on the character of the person saying it — a hypocrite can still be right. (Note the honest exception: when someone's credibility is the actual issue — a witness, an expert's conflict of interest — questioning them isn't a fallacy.)

2. Straw man — beating up a weaker version

Replacing someone's actual position with a distorted, easier-to-attack version, then knocking that down. Where it breaks: you refuted a claim nobody made. The antidote is its opposite, the steelman — restate their argument in its strongest form first, then respond to that.

3. False dilemma — "either this or that" when there's a third door

Presenting two options as the only two when more exist. "Either we do X or we've given up." Where it breaks: the two options weren't exhaustive; the whole force of the argument came from hiding the alternatives.

4. Slippery slope — the ungrounded chain reaction

Claiming one small step must lead to an extreme end, with no reason given for why each link follows. Where it breaks: each step in the chain needs its own justification; "A leads to Z" is asserted, not shown. (Sometimes a slope is real — then you show the mechanism for each step, and it's an argument, not a fallacy.)

5. Appeal to authority — "an expert said so", misused

Treating someone's say-so as proof. Where it breaks: authority is a reason to take a claim seriously, not to treat it as settled — and it collapses entirely when the "authority" is famous rather than qualified on this topic, or has a stake. A celebrity's view on medicine carries the weight of a stranger's.

6. Appeal to emotion — feeling instead of reason

Substituting fear, pity, outrage or flattery for evidence. Where it breaks: how a claim makes you feel is not evidence about whether it's true. The emotion may be appropriate and the argument still empty.

7. Hasty generalisation — the sample of one

Drawing a broad rule from too few, or unrepresentative, cases. "My grandfather smoked and lived to 95, so it's fine." Where it breaks: an anecdote isn't a dataset; one vivid case can't overturn a rate (this is availability bias, DT-07, wearing an argument's clothes).

8. Circular reasoning — the conclusion hiding in the premise

Assuming what you're trying to prove. "It's the best because nothing's better." Where it breaks: the premise already contains the conclusion, so no independent reason was ever given — the argument goes in a circle and never touches the ground.

9. Equivocation — one word, two meanings

Sliding between two senses of the same word mid-argument. "Evolution is just a theory, and a theory is just a guess" — "theory" means something different in science than in everyday speech. Where it breaks: the argument only works because the word quietly changed meaning between premises.

10. Correlation as causation — the two lines again

Treating "A and B move together" as "A causes B". Where it breaks: covered fully in DT-03 — correlation has four possible explanations and the argument silently assumed one. It's listed here because in the wild it usually arrives as an argument, not a chart.

The tu quoque and the whataboutism trap

Two more you'll meet daily, both cousins of ad hominem:

• Tu quoque ("you too"): dismissing a point because the person doesn't live up to it. "You eat meat, you can't talk about the environment." Their hypocrisy, real or not, has no bearing on whether the claim is true.
• Whataboutism: deflecting a criticism by pointing at someone else's fault. "What about their side?" It changes the subject; it doesn't answer the point.

Both work by moving your attention off the argument and onto a person. Naming them puts it back.

The steelman: the move that makes you sharper

The habit that beats every fallacy on this list — including your own — is the opposite of the straw man. Before you respond to any argument, restate it in the strongest, most charitable form you can — ideally so well the other person says "yes, that's it, better than I put it." Then engage.

Two things happen. You stop wasting energy on distortions nobody's defending, and you find out whether your counter survives the best version of their case or only the weak one. If your objection only beats the weak version, you don't actually have an objection yet. Steelmanning isn't politeness — it's how you avoid fooling yourself.

What this means for you

• A fallacy breaks the structure, not necessarily the conclusion. Finding one means "this argument fails", not "you're wrong".
• Attack the argument, not the arguer — and notice when others do the reverse to you (ad hominem, tu quoque, whataboutism).
• Steelman before you strike. If you can't state their best case, you're not ready to answer it.
• "Just a theory", "either/or", "an expert said" — these are structural tells you can spot mid sentence.
• Watch your own. The real payoff is arguments of yours that no longer break.

Try it: dissect three real arguments (30 min)

1. Find three real arguments — a comment thread, an opinion piece, an advert, a debate clip, a persuasive post.
2. For each, locate where the structure breaks and write:

ARGUMENT (in one line): ........................................
PREMISES OFFERED: ..............................................
THE FALLACY: (which one, and where exactly it breaks) .........
IS THE CONCLUSION STILL POSSIBLY TRUE? (yes/no — why) .........

3. Take one argument you disagree with and steelman it — write its strongest honest version in 3–4 sentences, then your actual response to that.

✅ Finish check: three arguments with the fallacy named and the break located, an honest note on whether each conclusion could still be true, and one genuine steelman with your response to the strong version.

Summary card

• Fallacy = broken structure. A right conclusion can have a broken argument; naming the fallacy refutes the argument, not the claim.
• 1. Ad hominem — attack the point, not the person. 2. Straw man — answer the real position.
• 3. False dilemma — find the third door. 4. Slippery slope — show each link or drop it.
• 5. Appeal to authority — qualified on this, or just famous? 6. Appeal to emotion — feeling ≠ evidence.
• 7. Hasty generalisation — anecdote ≠ data. 8. Circular — conclusion hidden in the premise.
• 9. Equivocation — one word, two meanings. 10. Correlation→causation — see DT-03.
• Steelman first. The strongest counter-move on the list, and the one that improves your thinking.

Sources

1. Weston, A. — A Rulebook for Arguments, 1986
2. Walton, D. — Informal Logic, 1989
3. Copi, I. & Cohen, C. — Introduction to Logic, 1953
4. Schopenhauer, A. — The Art of Being Right, 1831

Next lesson: DT-04 — Sampling, Survey Design and Why Most Surveys Are Junk (L2) Related: DT-07 Ten Cognitive Biases · DT-03 Correlation Is Not Causation · PS-09 The Mechanics of Persuasion · DT-02 How Charts Lie Path: Hard to Fool — extended