This paper examines the strengths and limitations of the proband-sibling design. As controls, siblings have two major advantages over biologically unrelated subjects. First, they provide better matching and, second, the sibling design is free from regression effects. Only discordant identical co-twins are stronger controls. However, discordant identical twins are not as available for psychiatric research as siblings; the design is sensitive to environmental differences only; and the conclusions reached have less external validity than those reached through proband-sibling comparisons. The limitations of the sibling design include a skewing towards larger sibships and an inability to detect between-family effects. These strengths and limitations are explored and their implications for psychiatric research discussed. Two technical problems associated with this design are examined. First, a statistical technique is presented which allows it to be determined whether the proband is different from his siblings without regard to sibship size or the form of the dependent variable. Second, it is shown how the role of geographic mobility can be clarified in follow-back studies using the proband-sibling design.