A barrage of information is now emerging about one of the thorniest and most significant problems in biology: the nature of the genetic events that trigger neoplasia. While the notion that cancer reflects somatic mutations is several decades old, only in the last few years could the relevant 'oncogenes' be identified and put to molecular dissection. The concept of oncogenes first solidified when the ability of tumourigenic viruses to induce experimental tumours in animals proved to reside in only one or two viral genes. Over the last five years, increasing evidence has accumulated that spontaneous tumours, including those of man, also reflect the action of only a few genes but, remarkably, these constitute part of the normal genome: the 'cellular oncogenes' [see Land et af. (1983) for a short review and Bishop (1983) for a longer one]. A cellular oncogene represents, in current dogma, a gene that can promote neoplasia if expressed inappropriately; that is, at the wrong level, in the wrong cell lineage, at the wrong stage of differentiation, or in altered form. Thus somatic mutations, or more rarely germline alterations, that change the regulation or structure of an oncogene may be conducive to the cancerous state, presumably because the gene product alters the normal growth control of the cell. The number of genes with oncogenic potential is not known, but 30-100 would be a plausible estimate.