A confounding variable is an extraneous factor that varies along with the independent variable and that could itself produce the observed effect on the dependent variable. When a confound is present, the experimenter cannot be sure what caused the result.

Suppose a study compares a new teaching method with an old one and finds better test scores in the new-method group. If the new-method group also happened to contain more high-achieving students, prior ability is a confound. The improvement might be due to the method, to the students, or to both. Good experimental design prevents confounds through random assignment, holding conditions constant, or measuring and statistically controlling the potential confounds.

Confounds are especially treacherous in correlational and quasi-experimental research, where random assignment is impossible. Age, socioeconomic status, and motivation frequently operate as confounds in developmental and educational studies. Identifying them requires both methodological vigilance and substantive knowledge of the domain.

A finding is only as clean as the design that produced it. Uncontrolled confounds leave the causal interpretation ambiguous.