Conventional genetic theories have failed to explain why cancer (1) is not found in newborns and thus not heritable; (2) develops only years to decades after 'initiation' by carcinogens; (3) is caused by non-mutagenic carcinogens; (4) is chromosomally and phenotypically 'unstable'; (5) carries cancer-specific aneuploidies; (6) evolves polygenic phenotypes; (7) nonselective phenotypes such as multidrug resistance, metastasis or affinity for non-native sites and 'immortality' that is not necessary for tumorigenesis; (8) contains no carcinogenic mutations. We propose instead that cancer is a chromosomal disease: Accordingly, carcinogens initiate chromosomal evolutions via unspecific aneuploidies. By unbalancing thousands of genes aneuploidy corrupts teams of proteins that segregate, synthesize and repair chromosomes. Aneuploidy is thus a steady source of karyotypic-phenotypic variations from which, in classical Darwinian terms, selection of cancer-specific aneuploidies encourages the evolution and subsequent malignant 'progressions' of cancer cells. The rates of these variations are proportional to the degrees of aneuploidy, and can exceed conventional mutation by 4-7 orders of magnitude. This makes cancer cells new cell 'species' with distinct, but unstable karyotypes, rather than mutant cells. The cancer-specific aneuploidies generate complex, malignant phenotypes, through the abnormal dosages of the thousands of genes, just as trisomy 21 generates Down syndrome. Thus cancer is a chromosomal rather than a genetic disease. The chromosomal theory explains (1) nonheritability of cancer, because aneuploidy is not heritable; (2) long 'neoplastic latencies' by the low probability of evolving competitive new species; (3) nonselective phenotypes via genes hitchhiking on selective chromosomes, and (4) 'immortality', because chromosomal variations neutralize negative mutations and adapt to inhibitory conditions much faster than conventional mutation. Based on this article a similar one, entitled 'The chromosomal basis of cancer', has since been published by us in Cellular Oncology 2005;27:293-318.