That the sporadic and inherited forms of a particular cancer could both result from mutations in the same gene was first proposed by Knudson1. He further proposed that these mutations act recess-ively at the cellular level, and that both copies of the gene must be lost for the cancer to develop2. In sporadic cases both events occur somatically whereas in dominant familial cases susceptibility is inherited through a germline mutation and the cancer develops after a somatic change in the homologous allele. This model has since been substantiated in the case of retinoblastoma3, Wilms' tumour4–7, acoustic neuroma8 and several other tumours9,10, in which loss of heterozygosity was shown in tumour material compared to normal tissue from the same patient. The dominantly inherited disorder, familial adenomatous polyposis (FAP, also called familial polyposis coli), which gives rise to multiple adenomatous polyps in the colon that have a relatively high probability of progressing to a malignant adenocarcinoma11, provides a basis for studying recessive genes in the far more common colorectal carcinomas using this approach. Following a clue as to the location of the FAP gene given by a case report of an individual with an interstitial deletion of chromosome 5q, who had FAP and multiple developmental abnormalities12, we have examined sporadic colorectal adenocarcinomas for loss of alleles on chromosome 5. Using a highly polymorphic 'minisatellite' probe13 which maps to chromosome 5q we have shown that at least 20% of this highly heterogeneous set of tumours lose one of the alleles present in matched normal tissue. This parallels the assignment of the FAP gene to chromosome 5 (see accompanying paper14) and suggests that becoming recessive for this gene may be a critical step in the progression of a relatively high proportion of colorectal cancers.