--- title: "Post Hoc Fallacy" description: "The post hoc fallacy concludes that because one event followed another, the first caused the second. Sequence is treated as proof of causation. Since something must always precede any event, mere order is extremely weak evidence — coincidence, reverse causation, and common causes all produce the..." last_updated: "2026-08-02" --- # Post Hoc Fallacy *Also known as: post hoc ergo propter hoc* The post hoc fallacy concludes that because one event followed another, the first caused the second. Sequence is treated as proof of causation. Since something must always precede any event, mere order is extremely weak evidence — coincidence, reverse causation, and common causes all produce the same pattern. Assuming that because one event followed another, the first caused the second. Sequence is mistaken for causation. *Example: You wore your lucky socks and your team won, so you decide the socks did it.* ## What it looks like **In a conversation about health** > I took the supplement and my cold cleared up in three days, so the supplement cured it. *Why it fails:* Colds resolve on their own in roughly that time. Recovery would have followed the supplement whether or not it did anything, so the sequence carries no information. **In a business review** > We redesigned the homepage in March and sales rose in April, so the redesign drove the increase. *Why it fails:* April may carry seasonal demand, a campaign, or a competitor's stumble. Without a comparison group that did not see the redesign, the rise has many candidate causes. **In a discussion about policy** > Unemployment fell after the law passed, so the law created those jobs. *Why it fails:* Economies move for many reasons at once, and the pre-existing trend may already have been downward. Attributing the change to one intervention requires ruling the others out. ## Practice ### Question 1 A city installs new streetlights in a neighbourhood and burglaries drop 20% over the next year. Which conclusion is best supported? A. The streetlights caused the drop, since the drop followed installation B. The streetlights may have contributed, but the drop is also consistent with a citywide trend, regression to the mean after a bad year, or other changes in the area C. The streetlights had no effect, because correlation never implies causation D. Burglaries would have dropped 20% regardless **Answer: B.** Option B holds the causal claim open while naming the specific alternatives that need ruling out — a comparable neighbourhood without new lights would settle much of it. Option A is the post hoc error. Option C overcorrects into a slogan: correlation is genuine evidence for causation, just not sufficient on its own. Option D asserts a counterfactual with no more support than option A. ### Question 2 Which of these is a post hoc fallacy? A. "I flipped the switch and the light came on, so the switch controls the light" — confirmed by repeating it ten times with the same result B. "The team started winning after we changed the locker room music, so the music is why we are winning." C. "The medication reduced blood pressure in a randomized trial against placebo." D. "The glass broke when the ball hit it, so the ball broke the glass." **Answer: B.** Option B has nothing but sequence, and plenty of competing explanations — roster changes, easier opponents, regression after a bad run. The others each have more: option A adds repeatability under control, option C adds a randomized comparison group, option D has a known physical mechanism and immediate contact. Sequence plus a plausible mechanism plus ruled-out alternatives is causal evidence; sequence alone is not. ### Question 3 What are the three main alternatives to a direct causal link when B follows A? A. Coincidence, reverse causation (B caused A), and a common cause driving both B. Bad data, bad reasoning, and bad faith C. Correlation, causation, and confusion D. Sampling error, measurement error, and rounding error **Answer: A.** These three cover most cases and are worth memorizing as a checklist. Coincidence: with enough events, some sequences occur by chance. Reverse causation: sales training follows rising sales rather than causing it. Common cause: ice cream sales and drownings both rise with summer heat, and neither causes the other. ### Question 4 Why is regression to the mean a common source of post hoc errors in medicine? A. Because doctors are poorly trained in statistics B. Because people seek treatment when symptoms are at their worst, and extreme states tend to move toward average anyway — so improvement follows treatment even when the treatment did nothing C. Because medications take time to work D. Because patients tend to misremember when they started treatment **Answer: B.** The timing of treatment-seeking creates the illusion. You go to the doctor on your worst day, and the natural course of most conditions is to drift back toward baseline — so improvement follows almost any intervention, including none. This is a large part of why placebo-controlled trials exist rather than simply tracking whether treated patients improve. ### Question 5 What is the relationship between post hoc and correlation-versus-causation? A. They are unrelated concepts B. Post hoc is a specific case: inferring causation from temporal sequence, which is one particular kind of correlation C. Correlation-versus-causation applies only to numbers; post hoc only to events D. Post hoc is the correct version of correlation-versus-causation reasoning **Answer: B.** Post hoc is the temporal special case of the broader problem. Any correlation might reflect coincidence, reverse causation, or a common cause; post hoc reasoning is what happens when the correlation in question is "B came after A" and that ordering alone is treated as settling the direction of causation. ## Frequently asked ### Does post hoc mean sequence is irrelevant to causation? No. Temporal order is a genuine necessary condition — a cause cannot follow its effect — so sequence does real work in ruling things out. The fallacy is treating it as sufficient. It narrows the candidates without identifying which one is responsible. ### How do I establish causation rather than just sequence? The strongest tool is a comparison group that differs only in the suspected cause, ideally assigned randomly. Short of that, look for a plausible mechanism, a dose-response relationship, replication across settings, and explicit consideration of confounders and reverse causation. ### Is post hoc the same as the correlation-causation error? It is a subtype. The general error treats any statistical association as causal; post hoc specifically treats temporal succession as causal. Every post hoc claim is a correlation-causation error, but not every correlation-causation error involves a sequence in time. ## Related - [Correlation vs. Causation](https://www.criticalthinkingexercise.org/pages/glossary/correlation-vs-causation.md) - [Slippery Slope](https://www.criticalthinkingexercise.org/pages/glossary/slippery-slope.md) - [Base Rate](https://www.criticalthinkingexercise.org/pages/glossary/base-rate.md) - [Correlation vs Causation](https://www.criticalthinkingexercise.org/pages/exercises/correlation-vs-causation.md) - [Scientific Method Basics](https://www.criticalthinkingexercise.org/pages/exercises/scientific-method-basics.md) - [Lesson: traps-and-pitfalls/logical-fallacies-presumption-ambiguity-causal](https://www.criticalthinkingexercise.org/pages/learn/traps-and-pitfalls/logical-fallacies-presumption-ambiguity-causal.md) ## Sitemap See the full [sitemap](https://www.criticalthinkingexercise.org/sitemap.md) for all pages.