Post Hoc Fallacy: Examples and Why "After" Doesn't Mean "Because"

Rangi manages the shop floor at a small bike shop in Wellington. In early September, the owner, Bettina, paid a designer to rebuild the shop's website. By the end of October, sales were up by almost a third compared with September, and Bettina was delighted. "Best money we ever spent. The new site did that."

Rangi wants to agree. He liked the designer, and the site does look good. But he also remembers what happens every year in New Zealand when September turns into October: the days get longer, the rain eases, and people pull their bikes out of the garage. He opens last year's sales report. October was up on September last year too, with the old website.

The new site came first and the sales came after. That is all Bettina actually knows. Treating "after" as "because" is the post hoc fallacy, and it is one of the most natural mistakes a mind can make. This guide covers what it is, why it feels so convincing, how it differs from "correlation is not causation," and five questions that show whether one thing really caused another.

What is the post hoc fallacy?

The post hoc fallacy is concluding that one event caused another simply because it happened first. The full Latin name, post hoc ergo propter hoc, means "after this, therefore because of this."

The pattern has two steps:

1. A happened, then B happened.
2. Therefore A caused B.

The first step is a fact. The second is a guess dressed as a conclusion. A cause has to come before its effect, but coming first is nowhere near enough to prove it.

The Internet Encyclopedia of Philosophy files post hoc under a family it calls false cause, along with three cousins: cum hoc (assuming two things that occur together are cause and effect), reversing causation, and the regression fallacy. We'll meet all of them below. Post hoc is also entry number 15 in our guide to logical fallacies, with correlation vs causation right after it. This article goes deeper, because the hard part is not the definition. It is telling a real cause from a coincidence of timing.

Post hoc fallacy examples

You've heard most of these, and you've probably said a few.

Every conclusion in the table might be true. Post hoc doesn't mean the cause is wrong. It means the timing alone hasn't shown it's right.

Why "it happened after" feels like proof

If the mistake is so simple, why is everyone so prone to it?

We look for causes by habit. Noticing that one thing follows another is how we learn that dark clouds mean rain, so we trust the instinct even when it misfires.

We remember the hits. When the lucky jersey "works," the win feels meaningful. When it doesn't, you shrug and forget. This is confirmation bias doing quiet work in the background: the cases that fit the story get counted, the rest don't.

We usually don't see what would have happened anyway. You took the zinc, so you never find out how your cold would have gone without it. The comparison you'd need is invisible, and the story you have fills the gap.

The flight instructor's mistake

The most famous version of this comes from Daniel Kahneman. In Thinking, Fast and Slow, he describes a flight instructor in the Israeli Air Force who disagreed with him about praise. When he praised a cadet for an excellent maneuver, the instructor said, the next attempt was usually worse. When he shouted at a cadet for a bad one, the next attempt was usually better. So, he concluded, shouting works and praise doesn't.

Kahneman's point was that the instructor had observed the pattern correctly and explained it wrongly. An unusually good maneuver is partly luck, so the next one tends to be closer to average, whatever the instructor says. An unusually bad one tends to be followed by a better one for the same reason. This is regression to the mean, and it makes any reaction to an extreme result look effective. The shouting came before the improvement, but the improvement was coming anyway.

The same trap catches anyone who acts at a low point, like a club that replaces its coach after a terrible season. Things often get better next, and the change gets the credit.

Post hoc vs correlation vs causation

These ideas overlap, and people often use them as if they were the same. Here's how they fit together.

The reverse causation example is adapted from the Internet Encyclopedia of Philosophy. The coincidence row has a famous illustration: Tyler Vigen, who collected odd pairings while finishing law school, found that the number of films Nicolas Cage appeared in each year tracked US swimming pool drownings for about a decade. Nobody thinks one caused the other. With enough data, some pairs line up by pure chance.

The short version: post hoc is about timing, correlation is about co-occurrence, and both are about mistaking a pattern for a cause. In both cases, the fix is the same question: what else could explain it?

When "after" really is evidence

Here's the part many explanations skip. Timing and correlation are evidence. The Internet Encyclopedia of Philosophy makes this point directly: the fallacy is the hasty leap to a causal conclusion, not the use of the pattern at all. If you rejected every "B followed A," you could never learn anything from experience.

In 1965, the English statistician Austin Bradford Hill gave a talk, published in the Proceedings of the Royal Society of Medicine as "The Environment and Disease: Association or Causation?", that epidemiologists still discuss today. He listed nine viewpoints for judging whether an association is likely to be causal, including:

• Strength: a big effect is harder to explain away than a tiny one.
• Consistency: the pattern shows up across different people, places and studies.
• Temporality: the cause comes before the effect. This is the one thing post hoc gets right.
• Biological gradient: more of the cause brings more of the effect.
• Plausibility: there's a believable mechanism.
• Experiment: changing the cause on purpose changes the effect.

Hill was careful to say that none of these is required, and none proves causation alone. They are ways to weigh the evidence, not a checklist that settles it.

A cautionary tale from medicine

From the 1980s on, many observational studies found that women who took hormone therapy after menopause had lower rates of heart disease, and the belief spread that the therapy protected the heart. In 2002, part of the Women's Health Initiative, a large randomized trial in the US, was stopped early. Women taking estrogen plus progestin had higher risks of breast cancer, heart disease, stroke and blood clots, not lower ones.

In a 2008 review, the epidemiologist Jacques Rossouw pointed to one reason the earlier studies misled: women who chose hormone therapy were healthier to start with. Their good health was coming along with the treatment, not from it. That's a confounder, and it fooled careful researchers for years. (Research on hormone therapy has continued since, so this is a lesson about evidence, not medical advice.)

Five questions that test any "it caused it" claim

When someone says A caused B, including you, run these.

1. Would B have happened anyway?

This is the big one. Colds end, seasons change, slumps recover. Ask what the normal course of events looks like without A. Rangi's question was simple: what did October look like last year?

2. What else changed at the same time?

Causes rarely arrive alone. A new website, a new season, a competitor's sale, a local cycling event. List everything that changed around the same moment, then ask why A deserves the credit over the rest.

3. Could it run the other way, or could something cause both?

Maybe B caused A. Maybe a third thing caused both. Healthier women chose hormone therapy. Busy summers bring both ice cream and swimming.

4. Does it happen again, and does more A bring more B?

One sequence is a story. A repeated pattern is evidence. If A is the cause, then more of A should usually bring more of B, and removing A should make B fade. If the "lucky jersey" works, it should work across a whole season, not one memorable game.

5. Is there a comparison?

The best evidence compares a group that got A with a similar group that didn't. That's why randomized trials matter so much in medicine. In everyday life, the comparison might be last year, a similar shop, or simply trying it both ways.

Worked example: Rangi's website

Rangi didn't tell Bettina she was wrong. He brought her last year's numbers and suggested they answer the five questions together.

The last row was the interesting one. Walk-in sales rose by about what they rose last year. Online orders roughly doubled. So the website probably did help, mostly with online orders, and spring did most of the rest.

That's a better answer than "the site did it" or "the site did nothing," and it told Bettina where to spend next.

How to respond to the post hoc fallacy

"That's post hoc!" is true and rarely helps. Questions work better, because they let the other person see the gap for themselves, which is the idea behind Socratic questioning.

A simple pattern: grant the timing, ask about the alternatives, suggest a check.

1. Grant the timing. "Yes, sales went up after the launch."
2. Ask about the alternatives. "What else was going on in October?"
3. Suggest a check. "Could we compare it with last October?"

Here's how that sounds in a few common situations.

If you're writing a reply in a debate or an essay, our guide to writing a rebuttal shows how to answer a causal claim by attacking the link between the evidence and the conclusion, rather than the evidence itself.

How to catch it in your own thinking

You're most likely to make this mistake when the story flatters you, or when you badly want something to work.

• Say "after" instead of "because." "Sales went up after the launch" is accurate. "Because of the launch" is a claim that needs more evidence.
• Be extra careful after extreme results. After the best or worst month, expect the next one to be more ordinary, whatever you do.
• Count the misses. If you believe something works, write down every time you try it, not only the times it seems to. Our guide to hasty generalization explains why a few memorable cases can't carry a rule, and survivorship bias explains why the cases you never see matter just as much.

How to practice this with AI

Post hoc is hard to spot in your own story, because you already believe the explanation. A debate helps because someone is assigned to look for the other explanations. On Debatly, several AI models and characters argue different sides of your question over several rounds, you can step in at any point, and a judge gives a verdict with reasoning.

• Put your cause on trial. Type your causal claim into the debate room as a question, for example "Did our new website cause the rise in sales?" One side argues for the cause, the other looks for alternatives.
• Ask a skeptic for the counterfactual. Have Viktor, the skeptic, ask "what would have happened anyway?" after every causal claim. That one question exposes most post hoc stories within a round or two.
• Pit models against each other. In model vs model, different AI models argue opposite sides, so you're less likely to get one model's favorite explanation presented as the only one.
• Read the judge's reasoning. Judge Alex explains which arguments carried the round. If a point won because of timing alone, run the five questions on it.

When Rangi tried the website question, the side arguing against the website brought up the season in its first reply, and then asked for online and walk-in sales separately. That was the split that settled it.

An honest limit

AI models can list possible confounders, but they can't see your sales data, your weather or your city's events calendar. They can also commit the fallacy themselves, telling a fluent causal story from a simple sequence of events, and many lean toward whatever explanation you seem to prefer (see why AI agrees with you). Use the debate to generate the alternatives, then check them against your own numbers.

Ask what would have happened anyway

The post hoc fallacy survives because sequences are memorable and the alternatives are invisible. The fix is one question: what would have happened without it? If the honest answer is "probably the same thing," the cause hasn't been shown yet.

This week, pick one thing you believe "worked" for you, a habit, a tool, a change at work, and run the five questions on it. If the answer matters, run it as a debate on Debatly and let a skeptic hunt for the other explanations. Give it a try. Plans are on the pricing page.