Every genetic laboratory diagnosis carries some degree of uncertainty, even if only the risk of laboratory error. Increasingly, diagnosis is based on direct tests that tell us whether or not the patient carries a given mutation. In these cases formal risk analysis is scarcely necessary—although it is important to remember that the patient wants to know the risk of disease, not the probability of possessing a certain genotype. Variable expression, nonpenetrance, or late onset may mean that carrying the mutation is not the same thing as having the disease. Unfortunately, there are still many situations where direct mutation tests are not available as part of a routine diagnostic service. Some disease genes have not yet been cloned; for others, although the gene is known, mutations are hard to find (e.g., in neurofibromatosis 1 and adult polycystic kidney disease type 1); sometimes direct tests easily reveal a proportion of all mutations, but the remainder are too diverse to be hunted down in a busy diagnostic laboratory (as happens in cystic fibrosis families). In all these cases family studies require the use of linked markers to track the disease gene through the family. Gene tracking always requires a risk calculation. A further complication arises when the same clinical disease can be caused by mutations at more than one locus, as with polycystic kidney disease or tuberous scleroses.