Abstract
The effect of misclassification of patients by interview or questionnaire methods is analyzed statistically. It is shown that so long as misclassification is unbiased, reported associations between disease and other factors are always smaller than the true associations.