Fallacies

This is Unit Eight of Logic: Fallacies. A fallacy has to satisfy three conditions at once. It has to recur — one bad argument is a mistake, and a repeating pattern of them earns a name. It has to persuade. And it has to fail to support what it concludes. Patterns that fail without persuading nobody bothers to catalogue. This unit is organised around something that gets left out of most fallacy lists, and it matters more than the names do. Nearly every fallacy here has a legitimate twin — a closely related move that is perfectly good reasoning. The entries call these near-misses, and learning them is what separates analysis from shouting the name of a fallacy. That is also why the last procedure has five steps rather than four. The fifth asks whether the context makes the move legitimate after all, and without it this whole unit becomes a way of dismissing arguments rather than assessing them. By the end you will be able to do five things. One: say what makes a pattern a fallacy, and why calling an argument fallacious settles nothing about whether its conclusion is true. Two: tell the person-directed fallacies apart, and say when someone's credibility genuinely is the evidence. Three: name the legitimate twin, not only the fallacy. Four: tell the two part-and-whole errors apart, and say which properties really do transfer. Five: run the identification method through all five of its steps.

Two entries of definition, and the second one draws the line around what this unit contains. The important point in the first is about what a fallacy verdict actually claims — it is a judgment about whether the support works, and it says nothing at all about whether the conclusion happens to be true.

The three conditions.

Logic 376 What Makes a Fallacy

It has to recur, it has to work on people, and it has to fail to support the conclusion. One bad argument is only an error — a repeating persuasive error is what earns a name and a place in a catalogue.

Which catalogue this is.

Logic 377 Formal vs. Informal Fallacies

Some patterns fail on shape alone, whatever they are about — those were numbered back in Units 3 and 4. The ones here need you to look at what is being said and in what setting.

Four fallacies that answer the arguer rather than the argument, and this family has the most valuable near-miss in the unit — because there is a real situation where a person's credibility is exactly the evidence in play, and telling that apart from these four is a genuine skill.

The blunt one.

Logic 378 Ad Hominem, Abusive

A proposal comes from someone convicted of fraud, so reject the proposal — but its arithmetic has no idea who wrote it. The twin: where someone offers testimony rather than reasoning, their reliability is the evidence.

The interested one.

Logic 379 Ad Hominem, Circumstantial

The contractor recommends the expensive option, which he would, and it may still be right. Having a stake earns closer inspection of the claims — it never earns skipping them.

The hypocrisy one.

Logic 380 Tu Quoque

Being lectured on smoking by a former smoker leaves the health facts exactly where they were. Hypocrisy is real evidence about the person and never about the claim.

And judging by where it came from.

Logic 403 Genetic Fallacy

An idea first raised by a discredited government is not thereby wrong; ideas outlive their origins. The twin: for testimony, and for claims you cannot check yourself, origin is legitimately part of the assessment.

Two moves that put pressure where a reason belongs. Both have twins, and both twins turn on the same question: is the thing being appealed to actually part of what is at issue?

The first.

Logic 381 Appeal to Force

Approve it or the jobs leave. That tells you what happens next and nothing about whether approval is warranted. The twin: where the consequences are the very thing under discussion, stating them is the argument.

The second.

Logic 382 Appeal to Pity

Needing the grade to keep a scholarship is affecting and says nothing about what the essay earned. The twin: in sentencing or hardship decisions, circumstances are precisely what is being weighed.

Three appeals to a source outside the argument. Each names a real kind of evidence and then leans on it past where it holds — and the third one has the distinction of running equally well in reverse, which is the clearest sign nothing is being established.

The first source.

Logic 383 Appeal to the People

Fifty million users cannot be wrong — history disagrees cheerfully. The twin worth keeping: informed judges arriving at the same view independently is real evidence, and a sales figure is not that.

The second.

Logic 384 Appeal to Unqualified Authority

A celebrated physicist endorsing a diet. Standing does not travel between subjects. The twin is important: citing a qualified expert within their field is not a fallacy — it is how limited people know most of what they know.

The third.

Logic 385 Appeal to Ignorance

Nothing has disproved the lake monster, so it exists — and the same shape argues the opposite equally well. The twin: a competent search that should have found something and did not is genuine evidence.

Three errors about the relationship between rules and cases. One builds a rule from too little. One takes a rule somewhere its purpose never reached. And the third builds a rule out of the exception itself — the same road as the second, travelled the other way.

Too little, too confidently.

Logic 386 Hasty Generalization

Two late contractors do not establish that none finish on time. Note where the fault sits: a small sample legitimately supports a tentative conclusion, so what fails is the confidence rather than the evidence.

A rule taken too far.

Logic 387 Accident

Cutting people is assault, therefore surgeons are criminals. Rules do bind in ordinary cases — this needs a case that the rule's purpose never contemplated, which is what surgery is.

And a rule built from an exception.

Logic 388 Converse Accident

Morphine is permitted for the dying, so opioids should be freely sold. The previous error pushed a rule down onto a case it never covered; this one manufactures a rule out of the case that was always special.

Three failures of the inductive reasoning Unit Seven built, and all three have legitimate twins that are not merely acceptable but essential — the courts run on the third one.

Sequence read as cause.

Logic 389 False Cause

Sales rose after the rebrand, so the rebrand did it — along with the season, the promotion, and the economy. Coming first is genuinely required for causing, and nowhere near sufficient for it.

Links asserted rather than shown.

Logic 390 Slippery Slope

A chain of consequences with every link stated and none supported. The twin: where each link comes with a documented mechanism, the argument is legitimate and common in policy work.

Similarities that do not bear.

Logic 391 Weak Analogy

Employees compared to family, where the respects they differ in are exactly the ones the conclusion needs. The twin is indispensable — case law is reasoning from resemblance — and Unit 7's criteria are what settle which is which.

Four ways of quietly helping yourself to the thing in dispute. The last is the most dangerous in practice, because every premise in it is true — which is why it survives the checks people actually run.

The circle.

Logic 392 Begging the Question

Guilty because he did it — a single claim stated twice. Wide circles hide well: the repeat may arrive paraphrased and several steps away from where it started.

The question you cannot answer.

Logic 393 Complex Question

Asked whether you have stopped falsifying timesheets, both answers enrol you. The remedy is refusing the frame rather than picking an answer — say the assumption is false.

The two options.

Logic 394 False Dichotomy

Cut the programme or go bankrupt, and bankruptcy is unacceptable, so — while the third and fourth options go unlisted. Some pairs really are exhaustive; the fault is presenting one that is not as though it were.

And the one with true premises.

Logic 395 Suppressed Evidence

Every stated fact about the car checks out, and the two floods went unmentioned. Nothing false was said, which is what makes it dangerous — the question for any advocate is what is missing, not what is wrong.

Unit Two named two ways a term goes slack. Here they are doing damage inside arguments, and the remedy for both is mechanical rather than clever: fix one reading and read the argument again.

In the word.

Logic 396 Equivocation

A word changes sense partway through, so the premises connect in wording only. Unit Two's warning about a meaning shifting mid-argument, now realised — and the fix is to pin one meaning and reread.

In the grammar.

Logic 397 Amphiboly

A sentence whose structure permits two readings, with the conclusion drawn from whichever suits. The sibling of the previous entry: the trouble lives in the sentence's construction rather than in any word.

Two errors that mirror each other, and one question settles both: does this particular property survive the trip between a thing and its pieces? Some do, and the fallacy is assuming they all do.

Upward.

Logic 398 Composition

Every player outstanding, therefore an outstanding side — ask anyone who has managed a room full of stars. Some properties really do travel up: every part having weight does make the whole weigh something.

And downward.

Logic 399 Division

The mirror image. The same test decides it: bulk properties come apart happily, while averages and totals do not — a department's overall figure says nothing about any individual in it.

Three ways of ending up somewhere else. They are often lumped together and they are genuinely distinct: one distorts the position, one abandons the topic, and the third does neither — its premises are working perfectly, for a conclusion nobody drew.

Distort, then refute.

Logic 400 Straw Man

Someone asks to review the budget and is answered as though they opposed all spending. This is the principle of charity from Unit One, violated deliberately — and the tell is that the refuted view belongs to no one.

Abandon, then declare.

Logic 401 Red Herring

Asked about the audit, the director speaks about staff dedication, and moves on as though answered. Nothing was distorted — the subject was exchanged, with a victory lap attached.

And the honest-looking one.

Logic 402 Missing the Point

Break-ins have tripled, therefore triple the sentences — but the evidence supports doing something, and picks out nothing in particular. The premises are doing real work; the conclusion is not the one they earned.

Three moves aimed at making a position safe rather than making it right. The middle one is the purest: it saves a generalisation by redefining the category, and in doing so converts a claim about the world into a claim that cannot be wrong and says nothing.

The exemption.

Logic 404 Special Pleading

One standard for everyone else and another for oneself, with no reason offered for the difference. The twin matters: a relevant difference does justify different treatment, and naming it is what turns this into an argument.

The redefinition.

Logic 405 No True Scotsman

A counterexample arrives, and the category quietly narrows to exclude it. Repeat that and the claim can no longer be false — which also means it has stopped telling you anything.

And the vocabulary itself.

Logic 406 Loaded Language

One memo's streamlining is another's gutting, describing the same reorganisation. This is raw material for the whole family rather than a pattern of its own, and Unit Two's opening move is the remedy.

Five steps, and the last one is why this unit does not turn into a reflex. Having named a gap, you go back and ask whether the setting makes the move legitimate after all — because most of these names describe a good move used in the wrong place.

Watch the method run.

Logic 407 A Method for Fallacy Identification

State the conclusion, state what is offered for it, ask whether that support bears on it, name the gap — and then check whether the context licenses the move anyway. Skipping the fifth step is what turns analysis into fallacy-spotting.

The capstone. One editorial, twice.

Logic 408 Capstone: Fallacy Hunt

Read an argument against a cycle lane straight through, then go back and tag each failure by name — and acquit the one move that only resembles a fallacy. The acquittal counts as much as the convictions; an analysis that convicts every sentence has stopped analysing.

Five things you can now do. You can say what makes a pattern a fallacy — recurrent, persuasive, and failing to support — and you know the verdict is about the support, so a fallacious argument's conclusion may well be true. You can tell the person-directed fallacies apart by what they attack — character, interests, hypocrisy, origin — and you know when a person's reliability genuinely is the evidence, which is when they are giving testimony rather than reasons. You can name the legitimate twin rather than only the fallacy: a qualified expert in their field, a search that should have found something, a slope with documented mechanisms, an analogy the law itself runs on. You can tell the two part-and-whole errors apart, and you know the test is whether that particular property survives the trip — bulk properties do, averages and totals do not. And you can run the method to its fifth step, which asks whether the context licenses the move — without it this unit becomes a way of dismissing arguments instead of assessing them.