Statistical procedures and molecular genetic techniques have attained a fine degree of resolution. Their ability to find disease genes has revolutionized medicine and raised hopes for breakthroughs in psychiatry. However, such breakthroughs may require an equally discriminating nosology. A psychiatric genetic nosology seeks to classify patients into categories that correspond to distinct genetic entities by addressing the problem of diagnostic accuracy: the degree to which a diagnosis correctly classifies people with and without a putative genetic illness. We review methods that deal with misclassification in genetic studies. These are clinical and epidemiological approaches that deal directly with how to define the observable manifestation of a putative genotype. We discuss two groups of methods: those that use known phenotypes and those that design new phenotypes.