Optimal use of phenotype information is important in complex disease gene mapping. We describe a method, sequential addition, for the analysis of case-control data by taking into account of a quantitative trait that is measured in cases but not in controls. The method also provides an estimate of the best phenotype definition for future studies. We demonstrate proof of principle, using an example of incorporation of age-at-onset data into a study of a small sample for association between APOE and late-onset Alzheimer's disease. The sequential addition method finds evidence of association when conventional case-control methods fail. We also illustrate the use of the method for taking account of a dimensional measure of psychosis in a study of the schizophrenia susceptibility gene, dysbindin, in bipolar disorder.