Logical fallacies: how to spot an argument that only pretends to be one
Seven questions help you check facts. But there is a second half: even true facts can be arranged so that the conclusion is wrong. These arrangements have names — logical fallacies — and they repeat so often that once you learn a few, you start seeing them everywhere: in comments, in advertising, in politicians' speeches, sometimes in your own head.
Here are the ones you will meet most often when reading news and numbers, each with an example. The list is neither academic nor complete — a fuller one in Lithuanian is published by the Lithuanian Skeptics' Society.
When the person is attacked, not the argument
Ad hominem
— the author is attacked, not the claim
Instead of answering the argument, the person who made it is attacked: "what would they know", "look who's talking". It sounds like an answer, but it never touches the claim — a fact does not become false because someone you dislike said it. Example: "That statistic is worthless — officials published it." A number is either right or wrong; you check it, you don't decide by sympathy.
Straw man
— what gets refuted is not what was said
The opponent's claim is replaced with a weaker caricature — and that caricature is triumphantly defeated instead of the real argument. Example: "You want lower speed limits in town? Why not ban cars altogether!" Nobody proposed banning cars — but that is easier to answer.
No true Scotsman
— the definition shifts so the exception doesn't count
When a counterexample breaks the claim, the claim is not withdrawn but rewritten: "a real X would never do that". Example: "Specialists in our field don't make mistakes." — "Here's one who did." — "Then he wasn't a real specialist."
When numbers are bent toward a conclusion
Hasty generalization
— from a few cases to everyone
A conclusion about everyone is drawn from a handful of cases: three acquaintances got sick — "everyone is sick now"; two neighbours moved abroad — "the whole country is leaving". How many cases you saw and how many exist are always two different numbers. That is why our pages print the absolute number next to the percentage.
Cherry picking
— only the convenient data is shown
Out of all the data, only the part that supports the desired conclusion is shown. The most common form is choosing the year: "the indicator has grown since 2022" may be true, but if 2021 was a record low, the comparison tells a different story than the whole curve. That is why our tables show several years in a row, not two points.
Texas sharpshooter
— the target is painted where the shots landed
First you shoot at the fence, then you paint the target around the holes. Data always contains random coincidences; the mistake is declaring them a law. Search long enough for the months when something "usually" happens and you will find something — randomness has patterns too.
Correlation is not causation
— moving together does not mean one causes the other
Two indicators move together and it looks like one causes the other. But both may depend on a third thing, or coincide by chance. In summer both ice-cream sales and drownings rise — ice cream doesn't drown anyone; both rise with the heat.
Post hoc
— after does not mean because
If B happened after A, it does not follow that A caused B. Prices rose after a decision — maybe because of it, maybe because of what was happening in the whole world at the time. Sequence is a hint, not proof.
The single cause
— one cause for a complex thing
Emigration, prices, birth rates — such things never have one cause. Whoever says "it's all because of X" is usually also selling the X solution. Complex phenomena have a list of causes, and the weights on that list keep shifting.
When the appeal is to something other than facts
Appeal to authority
— true because a famous person said so
Expert opinion matters — but it is an argument only when the expert speaks about their own field and can show the data. "The professor said so" is not proof; professors argue with each other too. Ask not who said it, but how they know.
Appeal to emotion
— fear or pity instead of proof
Fear, anger or pity are switched on in place of evidence. The emotion may be entirely justified — but it proves nothing. If a text frightens first and shows numbers later or never, that is a sign to pause.
Appeal to popularity
— everyone thinks so, therefore it's true
"Everyone knows that…" — but truth is not decided by vote. Majorities can be wrong and have been, many times. Popularity measures how many people believe something, not whether it is true.
Appeal to nature
— natural, therefore good
"Natural" does not mean "good" or "safe": a death cap mushroom is natural, a helmet is not. In advertising, the word "natural" works on emotion, not fact.
When the frame itself is wrong
False dilemma
— two options offered where more exist
Only two choices are offered — "you're either with us or against us" — when reality holds more. If a question sounds like either–or, first check whether there truly is no third, fourth, fifth way. Usually there is.
Argument from ignorance
— not disproven, therefore true
"Nobody has proven it doesn't work — so it works." Absence of proof is not proof: a claim needs support, not a wait for someone to refute it.
Slippery slope
— one step and supposedly catastrophe
"Allow this and in a year we'll have that, and in two everything collapses." A chain of "inevitablys" sounds logical, but every link is a separate claim that needs its own proof. The longer the chain without it, the less it is worth.
A fallacy does not automatically make an argument a lie — a true conclusion can be defended badly. But when a text stands on nothing but such constructions, that says a lot. And when you spot a fallacy, don't rush to triumph in the comments — first check you are not committing it yourself. That is the hardest part.