Correlation and causation
Correlation means two variables move together. Causation means changing one changes the other. Observational studies can establish the first reliably and the second only rarely, because a third factor, reverse causation or how the sample was selected can produce the same pattern.
Two things moving together does not mean one caused the other. Confounding, reverse causation and selection explain most wellbeing headlines.
Key takeaways
- Confounding: an unmeasured third factor drives both variables.
- Reverse causation: the outcome may be causing the exposure.
- Selection effects: who ends up in the sample can create the pattern.
- Adjustment helps but never removes unmeasured confounding.
- Causal language in an observational study is a warning sign, not a finding.
Confounding
People who meditate regularly also tend to sleep more, drink less and have more discretionary time. If meditators report better mood, any of those could be responsible. Statistical adjustment can account for measured confounders, but it cannot account for the ones nobody recorded - and in lifestyle research the unmeasured ones are usually the interesting ones.
Reverse causation
Poor sleep is associated with depression. Poor sleep can contribute to depression, and depression reliably disrupts sleep. Cross-sectional data - everything measured at one moment - cannot separate the directions. Longitudinal designs help, but only if the exposure was clearly measured before the outcome appeared.
Selection and survivorship
Who volunteers, who stays enrolled and who is still available years later all shape results. If people who found an intervention unhelpful dropped out, the remaining sample will look as if it worked.
What strengthens a causal reading
Observational evidence becomes more persuasive when several conditions line up.
- A plausible biological mechanism
- A dose-response relationship
- Consistency across different populations and methods
- Correct temporal order, with exposure clearly preceding outcome
- Agreement with any experimental evidence that exists
Reading the language
Careful observational research says 'associated with', 'linked to' or 'predicted'. When a press release turns that into 'reduces', 'boosts' or 'prevents', the claim has outrun the design. This applies to positive wellbeing findings exactly as much as to alarming ones.
Frequently asked questions
Can an observational study ever show causation?
Occasionally, when the effect is very large, the mechanism is clear, the timing is unambiguous and no plausible confounder could explain it. Methods such as Mendelian randomisation can also strengthen causal inference. These are exceptions rather than the rule.
Why are so many wellbeing claims correlational?
Because randomising people to years of exercise, loneliness or a particular diet is usually impossible. Observational research is often the only ethical option, which makes reading it carefully more important, not less.
Does adjusting for confounders solve the problem?
Only partly. Adjustment handles the variables that were measured, and measurement is often imprecise. Residual confounding remains the standard explanation for modest observational associations.
Related on Beiing
- Understand the evidence - All five research-literacy explainers.
- Research methodology - How Beiing sources and reviews claims.
- Neuroplasticity Evidence Explorer - Evidence strength and context, rated in one place.
Sources
- MedlinePlus - Evaluating Health Information - U.S. National Library of Medicine
- Centers for Disease Control and Prevention (CDC) - Mental Health - CDC
- PubMed - National Library of Medicine - NIH
- Cochrane Handbook for Systematic Reviews of Interventions - assessing certainty of evidence - Cochrane
- Ioannidis (2005), Why most published research findings are false - PLOS Medicine / PubMed
Written by Beiing Editorial Team · Editorially reviewed by Research Desk · Published · Last reviewed . Editorial and research review by Beiing writers and researchers. This is not medical or clinical review, and no clinician signs off on our content.
Educational content only - not medical advice. In a crisis, call or text 988 (US) or your local emergency number. See crisis resources.