Definition · Plain-language
Confounding variable
A confounding variable is an extraneous factor that is related to both the independent and the dependent variable, distorting the apparent relationship between them.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
The third-variable problem
A confounding variable is a hidden influence that sits behind an apparent relationship. To confound, a variable must be associated with the independent variable and also independently affect the dependent variable. The classic example is the link between ice-cream sales and drowning: both rise together, but neither causes the other — hot weather is the confounder driving both. Whenever a confounder is present, you cannot tell how much of the observed effect is due to the independent variable and how much to the confounder, so any causal claim is unsafe. This is why confounding is a central threat to internal validity.
Confounding versus control variables
A confounding variable and a control variable are not the same thing, though they are related. A confounding variable is an extraneous factor that you have failed to account for and that distorts your results. A control variable is an extraneous factor that you deliberately hold constant or otherwise account for so that it cannot distort results. In effect, the goal of good design is to convert potential confounders into controlled variables before they can bias the study. A variable only earns the label "confounder" when it is left uncontrolled and is linked to both the cause and the effect.
How to control confounding
Researchers reduce confounding at the design stage and the analysis stage. Random assignment is the most powerful tool: by allocating participants to conditions by chance, it distributes known and unknown confounders roughly evenly across groups. Restriction limits the study to one level of the confounder (for example, one age band). Matching pairs participants so groups are balanced on the confounder. At the analysis stage, statistical control — such as multivariable regression, stratification or analysis of covariance — adjusts for measured confounders. Each method has limits: only randomisation addresses unmeasured confounders.
Key facts
At a glance
- Definition: an extraneous variable linked to both the IV and the DV
- Also called: confounder, lurking variable, third variable
- Core problem: offers an alternative explanation for the observed effect
- Threatens: internal validity — undermines causal claims
- Contrast: a control variable is deliberately held constant; a confounder is not
- Controlled by: randomisation, restriction, matching, statistical adjustment
Common misconceptions
What people often get wrong
Often heard: A confounding variable is the same as any extraneous variable in a study.
Actually: Only confounders are linked to both the independent and dependent variable. An extraneous variable that affects only the outcome, and is unrelated to the cause, adds noise but does not confound the relationship.
Often heard: Confounding variables and control variables are two names for the same thing.
Actually: They are opposites in practice. A control variable is one you deliberately hold constant or account for; a confounder is an uncontrolled extraneous variable that distorts results. Good design turns potential confounders into controlled variables.
Often heard: Statistical adjustment removes all confounding from a study.
Actually: Statistical control only handles confounders you have measured. Unmeasured or unknown confounders remain, which is why random assignment — the only method that balances unmeasured factors — is so valued.
