Logical fallacies
Logical fallacies are recurring patterns of argument in which the conclusion does not follow from the reasons given, even though the argument sounds as though it should. A fallacy is a fault in the link between premises and conclusion, not a fault in the facts. An argument can be built entirely from true statements and still be fallacious, and a fallacious argument can still land on a true conclusion by luck.
What a logical fallacy is, and what it is not
A logical fallacy is a defect in inference: the move from what has been granted to what is being claimed. Three things are commonly mistaken for fallacies and are not. A false premise is not a fallacy, it is an error of fact. A rude remark is not a fallacy, it is bad manners. A conclusion you dislike is not a fallacy, it is a disagreement. The question that identifies a fallacy is narrower than any of those: if every reason given here were true, would the conclusion have to be true?
Take a concrete case. A city commissions a bridge inspection, and someone objects: "The inspection was written by the same engineering firm that built the bridge, so the bridge is unsafe." The conflict of interest is real and worth raising. But the conclusion drawn does not follow from it. A tainted source is a reason to check the report, not a reason to believe the opposite of what the report says. This is the shape of the fallacy called ad hominem in its circumstantial form, and the repair is not to dismiss the objection but to send a second inspector.
Formal and informal fallacies: the two families
Fallacies divide into two families, formal and informal, and the division decides how you detect them.
A formal fallacy is invalid in its structure alone. You can spot it without knowing anything about the subject, because the pattern fails whatever you substitute into it. The commonest is affirming the consequent: "If the pump failed, the tank is empty. The tank is empty. Therefore the pump failed." The tank could be empty because a valve was left open. The conclusion overreaches the structure, which is why the whole family is often gathered under the Latin label non sequitur, meaning it does not follow.
An informal fallacy has a structure that would be fine if the content behaved, and the failure is in the content, the context or the relevance of what is offered. Informal fallacies include ad hominem, the straw man, the red herring, the false dilemma, the slippery slope and hasty generalization. You cannot detect them by inspecting the form, because the form is often unobjectionable. You detect them by asking what the premises are actually about, and whether that has anything to do with the conclusion.
How many logical fallacies are there
There is no fixed number of logical fallacies, and any source that gives you an exact total has chosen a level of detail rather than discovered a fact. The earliest surviving catalog, Aristotle's Sophistical Refutations, sorts 13 fallacies into 2 groups: 6 that depend on language, such as equivocation and ambiguity, and 7 that do not. Modern introductory logic textbooks typically teach between 20 and 30. Large online compendia list well over a hundred, mostly by naming narrower sub-cases of the same underlying error.
The count rises because fallacies nest. Ad hominem alone can be sliced into abusive, circumstantial, tu quoque and guilt by association, and each of those can be listed separately. For practical purposes a working set of a dozen names covers most of what you will meet in a week of ordinary reading. Learn those properly and the rest read as variations.
A chart of the common logical fallacies, with definitions and examples
The chart below gives 13 common fallacies, the move each one makes, and a one line example. Each example is invented and deliberately dull, because the point is the shape of the reasoning, not the subject matter.
| Fallacy | The move it makes | Example in one line |
|---|---|---|
| Ad hominem | Answers the arguer instead of the argument | "She only says the road is unsafe because she lives on it." |
| Straw man | Refutes a distorted version of the claim | "You want a speed limit, so you want nobody to drive." |
| Red herring | Changes the subject to something easier | "Forget the leaking roof, look how much we saved on the paint." |
| False dilemma | Offers two options where more exist | "Either we replace the whole roof or we let the building rot." |
| Slippery slope | Asserts a chain of consequences without support | "Allow one late submission and no one will hand anything in." |
| Hasty generalization | Draws a broad rule from too few cases | "Two of these batteries were flat, so the brand is junk." |
| Circular reasoning | Assumes in the premises what it sets out to prove | "The manual is reliable because it says it is reliable." |
| Appeal to authority | Cites a source with no standing on the question | "A famous architect endorsed this soil test kit." |
| Appeal to emotion | Substitutes feeling for evidence | "Think how disappointed the team will be if you audit this." |
| Bandwagon | Treats popularity as proof | "Forty thousand shops use this till system, so it works." |
| Equivocation | Shifts the meaning of a word mid argument | "The test was positive, and positive things are good." |
| False cause | Reads sequence or correlation as causation | "Sales rose after we repainted, so the paint sold the shoes." |
| Non sequitur | Draws a conclusion the premises do not reach | "The warehouse is clean, so the stock count is accurate." |
Logical fallacies in advertising and media
Advertising and media use fallacies more often than argument does, because both are working under a severe shortage of space and are not obliged to prove anything. Four appear constantly.
- Bandwagon. "The country's best selling kettle." Sales volume measures distribution and price, not whether the kettle boils water well.
- Borrowed authority. A person in a white coat holding a laundry detergent. The costume signals expertise that the speaker has not been shown to have, which is the fallacy of appeal to authority in its purest commercial form.
- False dilemma. "Either this stain remover, or throw the shirt away." A third option, washing it twice, is quietly removed from the frame.
- Equivocation on a test word. "Laboratory tested" and "laboratory proven" differ by one word and by everything. Tested says a test happened. Proven claims a result.
In news media the recurring pattern is different: a headline reports a correlation in a study as though the study had established a cause. The study may be careful and the headline still wrong, because the fallacy is introduced in the compression, not in the research. Reading a claim in media well means asking which sentence in the source actually supports the sentence in the headline.
Naming a fallacy is not the same as refuting the argument
Naming a fallacy proves one thing only: that a particular argument fails to support its conclusion. It does not show the conclusion is false. Treating it that way is itself a recognized error, the argument from fallacy, sometimes called the fallacy fallacy. If someone defends a sound conclusion with a bad argument, the conclusion survives and the argument does not.
There is a second failure worth guarding against, which is why this site treats fallacy names as tools rather than trophies. A person who has learned twelve fallacy labels tends to find them fastest in claims they already disliked, and slowest in claims they already accepted. That asymmetry is confirmation bias operating on the toolkit itself. The label is a prompt to look harder, not a verdict, and the honest test is whether you apply it as readily to an argument you were pleased to hear.
Three tests to run on the next argument you meet
Run these three tests in order on any argument, in a report, an advertisement or a conversation. Each takes under a minute and none requires you to know the subject.
- Separate the claim from the reasons. Write the conclusion on one line and everything offered in support underneath it. Most fallacies become visible the moment the two are on different lines, because the support turns out to be about something else.
- Grant every premise and check the gap. Ask: if all of this were true, would the conclusion have to be true, or merely be more likely? If the honest answer is "not necessarily", identify what would have to be added, and notice whether anyone has offered it.
- Reverse the sympathy. Ask: if the conclusion had gone the other way, would I have accepted this same reasoning? If you would not, the argument is not what convinced you.
These tests will not tell you whether a claim is true. They tell you whether the argument in front of you is a reason to believe it, which is a smaller question and a far more answerable one.