The 20 known transforming onc genes of retroviruses are defined by sequences which are transduced from cellular genes, termed proto-onc genes. Based on these sequences, viral onc genes have been postulated to be transduced cellular cancer genes and proto-onc genes have been postulated to be latent cancer genes that can be activated from within the cell, to cause virus-negative tumours. The hypothesis is popular because it promises direct access to cellular cancer genes. However the existence of latent cancer genes presents a paradox since such genes would be most undesirable for eukaryotes. The hypothesis predicts (i) that viral onc genes and proto-onc genes are isogenic, (ii) that expression of proto-onc genes induces tumours, (iii) that activated proto-onc genes transform diploid cells upon transfection, like viral onc genes, and (iv) it predicts diploid tumours. As yet, none of these predictions is confirmed. Instead: (i) Structural comparisons between viral onc genes, essential retroviral genes, and the proto-onc genes show that all viral onc genes are indeed new genes, rather than transduced cellular cancer genes. They are genetic hybrids put together from truncated viral and truncated proto-onc genes. (ii) Proto-onc genes are frequently expressed in normal cells. (iii) To date, not one activated proto-onc gene has been isolated that transforms diploid cells. (iv) Above all, no diploid tumours with activated proto-onc genes have been found. Moreover the probability of spontaneous transformation in vivo is at least 109 times lower than predicted from the mechanisms thought to activate proto-onc genes. Therefore the hypothesis, that proto-onc genes are latent cellular oncogenes, appears to be an over-interpretation of sequence homology to structural and functional homology with viral onc genes. Here is is proposed that only rare truncations and recombinations, that alter the germline configuration of cellular genes, generate viral and possibly cellular cancer genes. The clonal chromosome abnormalities that are consistently found in tumour cells are microscopic evidence for rearrangements that may generate cancer genes. The clonality indicates that the tumours are initiated with, and possibly by these abnormalities as predicted by Boveri in 1914.