Correlation vs. Causation: Why “Linked To” Isn’t “Causes”

Georden Jones, Founder, The Peer Review · Last updated: July 28, 2026

---

The Short Answer

When a headline says a product is “linked to” a disease, it is almost always describing a correlation — two things that tend to appear together — not proof that one causes the other. Those are different claims, and the gap between them is where most health scares live. Ice cream sales and drowning deaths rise together, but ice cream does not cause drowning (summer heat drives both). A correlation can happen because A causes B, because B causes A, because a hidden third factor causes both, or by pure chance. Establishing real causation takes far more than a single “link” — which is exactly why “study links X to Y” should make you curious, not alarmed. This page explains how to tell a real causal finding from a coincidence dressed up as one.

---

Table of Contents

• What “Correlation” Really Is
• Four Reasons Two Things Correlate
• The Confounder: The Hidden Third Factor
• How Scientists Build a Case for Cause
• A Real Example: The Paraben Scare
• How to Spot the Difference in a Headline
• FAQ
• The Bottom Line
• References

---

What “Correlation” Really Is

A correlation means two things move together: when one goes up, the other tends to go up (or down). That is useful information — correlations are how scientists spot patterns worth investigating, and many real causal discoveries started as a correlation [1]. The problem is not correlation itself; it is treating a correlation as if it already proved cause.

“Linked to,” “associated with,” and “tied to” are the language of correlation, and they are honest words — the trouble is that a headline can pair them with a scary outcome and let your brain fill in “causes.” A study showing that people who use a certain product have slightly higher rates of a condition has found an association. Whether the product caused the condition is a separate, much harder question that one study almost never settles.

---

Four Reasons Two Things Correlate

When A and B reliably appear together, there are four possible explanations — and only one of them is “A causes B” [1]:

1. A causes B. The intuitive one, and sometimes true — but it has to be shown, not assumed.
2. B causes A (reverse causation). The arrow points the other way. For example, if “people with anxiety use more of a product,” it may be that distress drives the buying, not the product driving distress.
3. A third factor causes both (confounding). Something hidden drives A and B at the same time, creating a link between them with no direct connection. This is the big one, covered next.
4. Coincidence. With enough data, some things line up by pure chance. Comb through thousands of variables and you will find striking correlations that mean nothing.

The headline almost always implies explanation #1. Good reading means holding the other three in mind until the evidence rules them out.

---

The Confounder: The Hidden Third Factor

A confounder is a lurking third variable that influences both things you are comparing, making them look connected when they are not directly [1]. It is the single most common reason a correlation is not causation.

The classic illustration: ice cream sales correlate with drowning deaths. Ban ice cream and drownings would not fall, because neither causes the other — hot weather (the confounder) drives both ice cream buying and swimming. In health research, confounders are everywhere and subtler: people who use a particular “wellness” product may also exercise more, eat differently, earn more, or see doctors more often, and any of those could explain a health difference that gets blamed on the product. Strong studies try to adjust for known confounders, but they can never rule out every unknown one — which is why even a good observational study (see the Evidence Hierarchy) shows association more confidently than cause.

---

How Scientists Build a Case for Cause

Because a single correlation cannot prove cause, scientists look for a pattern of evidence pointing the same way. The most-used framework is a set of considerations proposed by epidemiologist Austin Bradford Hill in 1965, still referenced today [2]. In plain terms, causation becomes more believable when:

• The association is strong and shows up consistently across different studies and populations.
• The cause comes before the effect in time (not the reverse).
• More exposure produces more effect (a dose-response relationship — the Hazard vs. Risk idea).
• There is a plausible biological mechanism for how it could work.
• Removing the cause reduces the effect, ideally shown in a controlled experiment.

No single one of these is proof, and none is strictly required — they are a way of weighing whether a correlation has earned the word “causes.” When a claim satisfies few of them and rests on one observational study, “linked to” is as far as the evidence goes. When many line up across strong studies, cause becomes reasonable to accept.

---

A Real Example: The Paraben Scare

The classic case in personal care is the paraben-and-breast-cancer worry, which we cover in depth in Parabens. It traces largely to a 2004 study that detected parabens in samples of breast-tumour tissue. That detection is a correlation — parabens were present in the tissue — and it spread as if it proved parabens cause breast cancer.

But detection is not causation. The study had no healthy-tissue comparison group, so it could not even show parabens were more common in tumours than elsewhere, and finding a widespread chemical in a sample does not show it caused the disease. Twenty years of follow-up has produced mixed, inconclusive results rather than confirmation. The paraben story is a textbook example of a correlation — a real, measurable one — being read as a cause it never established. (The mechanistic research since is more specific and worth knowing, which is why we rate parabens “Mixed Signals” rather than dismissing them — but that is a separate point from the original leap.)

---

How to Spot the Difference in a Headline

A few reflexes catch most correlation-as-causation errors on sight:

1. Watch the verbs. “Linked to,” “associated with,” “tied to” mean correlation. “Causes,” “leads to,” “triggers” claim cause — and usually claim more than one study can support.
2. Ask “what else could explain this?” If you can think of a plausible confounder in ten seconds (people who do X also tend to do Y), the study probably had to wrestle with it too.
3. Check the study type. An observational study can show association but struggles to prove cause; that is baked into the Evidence Hierarchy.
4. Look for the word “detected.” Finding a chemical somewhere is a correlation at best — it says nothing about harm on its own.

None of this means dismissing correlations — they are how real discoveries begin. It means not letting a “link” masquerade as a proven cause, which is one of the most common moves in fear-based health content (Fear vs. Evidence).

---

FAQ

---

The Bottom Line

“Linked to” is not “causes,” and the distance between them is where most health scares operate. Two things can move together because one causes the other, because the arrow runs the other way, because a hidden third factor drives both, or by chance — and only careful, converging evidence can tell you which. A single correlation, especially from one observational study or a “detected in tissue” finding, is a reason to ask questions, not to panic. Watch the verbs, hunt for the confounder, and remember that real causation has to be built, not assumed.

Stay curious, stay critical.
Georden

---

References

1. Health Knowledge (UK public health textbook). “Causation in epidemiology: association and causation.”
2. “Modernizing the Bradford Hill criteria for assessing causal relationships in observational data.” PubMed.
3. “Minireview: Parabens Exposure and Breast Cancer.” International Journal of Environmental Research and Public Health (NCBI).

---

Georden Jones is the founder of The Peer Review. Read the full story.