Genetic epidemiology has provided consistent evidence over many years that schizophrenia has a genetic component, and that this genetic component is complex, polygenic, and involves epistatic interaction between loci. Molecular genetics studies have, however, so far failed to identify any DNA variant that can be demonstrated to contribute to either liability to schizophrenia or to any identifiable part of the underlying pathology. Replication studies of positive findings have been difficult to interpret for a variety of reasons. First, few have reproduced the initial findings, which may be due either to random variation between two samples in the genetic inputs involved, or to a lack of power to replicate an effect at a given alpha level. Where positive data have been found in replication studies, the positioning of the locus has been unreliable, leading no closer to positional cloning of genes involved. However, an assessment of all the linkage studies performed over the past ten years does suggest a number of regions where positive results are found numerous times. These include regions on chromosomes 1, 2, 4, 5, 6, 7, 8, 9, 10, 13, 15, 18, 22 and the X. All of these data are critically reviewed and their locations compared. Reasons for the difficulty in obtaining consistent results and possible strategies for overcoming them are discussed.