Skip to main content
    Path to Science

    Thinking Dictionary

    Indicators of causation

    What it is

    Indicators of causation are questions to ask before you believe a claim that "X causes Y." A practical way to use them is four questions, a simplified version of the epidemiologist Austin Bradford Hill's 1965 "viewpoints":¹

    1. Timing. Did X come before Y?
    2. Association. Do X and Y reliably go together?
    3. Mechanism. Is there a believable way that X could produce Y?
    4. Alternative explanations. Have other explanations been ruled out?

    Hill's own list was longer.¹ The Correlation vs causation entry explains Hill's ideas in more detail. This entry is the practical version: a short set of questions you can use on any claim.

    Why it matters

    Claims that one thing causes another are everywhere: in adverts, headlines, and conversations. "This pill reduces stress." "Screen time causes poor sleep." These questions help you decide how much weight to give such a claim.

    But they are not a checklist that proves anything. Hill himself said that none of his viewpoints can settle the question on its own, and none is strictly required.¹ Researchers today agree that causal claims can't be strictly proven. The questions raise or lower your confidence. They don't deliver a final verdict.²

    The four questions in more detail

    1. Timing. The cause must come before the effect. This is the only question that is a true must-have.¹˒² If people's sleep got worse before they started using screens late at night, screens can't be the cause.

    2. Association. Do the two things go together reliably, in more than one person and more than one study? A single story ("it worked for me") is weak evidence. But a strong association is not enough on its own either. A strong hidden third factor can produce a strong link.

    3. Mechanism. Is there a believable way X could cause Y? A known mechanism strengthens a claim. But what seems believable depends on what we know today.¹˒³ Sometimes good evidence that something is a cause comes before anyone fully understands how it works. So a missing mechanism is a reason for caution, not a reason to reject a claim outright.

    4. Alternative explanations. This is often the most important question. Common alternatives include:¹

    • Chance. A pattern that appeared by luck.
    • A hidden third factor, called a confounder. See the Confounding variable entry.
    • Reverse causation. Y caused X, not the other way round.
    • Natural recovery or the placebo effect. People often feel better anyway, or because they expect to. See the Placebo effect entry.

    A worked example

    Here is an imaginary advert: "I took StressAway pills and feel so much calmer!"

    • Timing: Yes. The pills came first, then the calm.
    • Association: Weak. It's one person's experience.
    • Mechanism: Unknown. Does the label name an ingredient with tested effects?
    • Alternatives: Plenty. Maybe the stressful week simply ended. Stress goes up and down, and a bad week is often followed by a better one. Or perhaps expecting to feel calmer helped.

    Verdict: not enough to conclude that the pills work.

    In everyday life

    • Health products. "I started this supplement and my joints stopped aching." Run the four questions. Alternatives like natural recovery are often the most likely explanation.
    • Headlines. "Teenagers who use social media are more anxious." Which came first? Could anxious teenagers spend more time online? Could something else cause both?
    • Your own habits. "I slept badly because of that late coffee." Maybe. Did you also go to bed later, or have a stressful day?

    The common trap

    The trap is thinking that if all four questions get a "yes," the claim is proven. It isn't. The questions make a claim more or less believable. They don't prove it.²

    Another trap is the opposite: rejecting a claim because nobody yet knows how it works.

    Some ways out:

    • Treat the questions as a guide to how confident to be, not a pass-or-fail test.
    • Spend the most effort on alternative explanations.
    • Ask what kind of study could rule those alternatives out. Often the answer is a randomized controlled trial.

    Try it

    Take a claim from an advert or a social media post. Score each of the four questions: yes, no, or unknown. If "alternative explanations" comes out as "unknown," ask what study could rule them out. Hint: see the Randomized controlled trial entry.

    Related concepts

    Correlation vs causation · Confounding variable · Randomized controlled trial · Placebo effect · Multiple lines of evidence · Reverse causation

    Sources

    1. Hill, A. B. (1965). The environment and disease: Association or causation? Proceedings of the Royal Society of Medicine, 58(5), 295–300. https://doi.org/10.1177/003591576505800503

    2. Rothman, K. J., & Greenland, S. (2005). Causation and causal inference in epidemiology. American Journal of Public Health, 95(S1), S144–S150. https://doi.org/10.2105/AJPH.2004.059204

    3. Fedak, K. M., Bernal, A., Capshaw, Z. A., & Gross, S. (2015). Applying the Bradford Hill criteria in the 21st century: How data integration has changed causal inference in molecular epidemiology. Emerging Themes in Epidemiology, 12, 14. https://doi.org/10.1186/s12982-015-0037-4