We discuss two classes of bias that arise in case–control studies, selection bias and information bias. A third source of bias, confounding, is not considered in this article. Selection bias arises either when cases in the study sample are not representative of cases arising from the source population (“study base”) or when controls are not representative of corresponding noncases in the study base. Sources of selection bias include bias in identifying cases or noncases in the study base, nonrepresentative sampling techniques, and selective refusal to participate (nonresponse bias). Use of nonrepresentative controls, such as hospital-based controls, can lead to bias if the exposure distribution differs from that of controls in the study base. Information bias arises from inaccurate reporting of exposures and other covariates by cases or controls, and can result in serious distortion of estimated odds ratios when the errors in reporting by cases and controls are differential (recall bias).