For diseases with a complex mode of inheritance, such as schizophrenia, traditional linkage analysis assuming Mendelian inheritance of a single gene may be expected to have relatively low power, but to occasionally detect (in fortuitous pedigree samples) either a rare gene which by itself causes the disorder, or a common contributing gene which displays Mendelian cosegregation with the disease only in those rare pedigrees in which the other requisite environmental or genetic factors are present in nearly all family members. Since special populations and/or sampling luck are required to produce samples enriched enough for these cases to reveal the linkage, valid linkage reports may not be as quickly replicated as with Mendelian diseases. Risch [ 19901 correctly stresses that the early linkage findings for schizophrenia [Sherrington et al., 19881 should not be overinterpreted; neither should we overinterpret the early replication failures, which at this point merely confirm that the disease is not a simple Mendelian one. If valid, the 5q linkage may well be confirmed by using more sensitive analysis methods which test simultaneous linkage of multiple disease genes to multiple markers, or by physical approaches such as cloning the 5q translocations [Bassett et al., 19881 which motivated the linkage studies, before it is replicated by the traditional methods.
The more important theoretical issue arising from the contradictory reports is what rate of false positive LOD scores we should expect in such analyses. If there are inherent biases in the LODs, replication is no assurance of validity. Further methodological work is needed in this area and I offer here only a few observations:
- The theoretical justification for Morton's LOD 3 criterion [Morton, 1955; Rao et al., 19781 is based on an estimated prior probability of linkage and therefore neces-