Correlation vs. Causation: How to Read a Health Headline

"Coffee linked to longer life." "People who skip breakfast at higher risk of heart disease." Headlines like these are everywhere, and they're easy to read as instructions: drink coffee, eat breakfast. But most of them rest on a single slippery word — linked — that means far less than it sounds. Here's how to read them without being misled.

What "linked to" really means

When a study finds that two things tend to happen together — coffee drinkers living longer, say — that's a correlation. Correlation only says the two move in step. It does not say one causes the other. That's causation, and it's a much higher bar. Almost every "linked to" headline is describing a correlation while quietly hoping you'll hear causation.

Three reasons two things move together without one causing the other

1. A hidden third factor. Maybe coffee drinkers also tend to have office jobs, regular healthcare, and higher incomes — and those drive the longer life. Researchers call this a "confounder," and it's the usual culprit.
2. Reverse cause. The arrow may point the other way. "Sick people exercise less" doesn't mean exercise prevents that illness — being sick is what reduced the exercise.
3. Pure coincidence. Search hard enough through enough data and you'll find things that line up by chance, with no connection at all.

The kind of study matters

Most health headlines come from observational studies — researchers watch what people already do and look for patterns. These are good at spotting correlations but can't prove cause, because the groups differ in a hundred uncontrolled ways. The gold standard is a randomized controlled trial, where people are randomly assigned to do one thing or another, which evens out those hidden factors. If a headline is built on observation alone, treat "causes" claims with caution. (More on telling the two apart in our guide to reading a study.)

Red flags in the headline

• Cause-and-effect verbs on thin evidence: "boosts," "triggers," "prevents," "causes" — when the study underneath only found a link.
• A single study presented as the final word. One study is a data point, not a consensus.
• A big scary risk with no baseline. "Doubles your risk" is meaningless if the risk went from 1 in a million to 2 in a million.
• No mention of who was studied or how many. A small or narrow group rarely applies to everyone.

Questions to ask before you believe it

• Was this an experiment (randomized) or just observation?
• What else might explain the link — income, age, lifestyle, existing health?
• How big is the effect in real terms, not percentages?
• Is this one study, or does it match a larger body of evidence?

The bottom line

"Linked to" means they happened together, not one caused the other. Before you change what you eat or drink because of a headline, check whether the study could actually rule out the alternatives — or whether it just found two things standing next to each other.