Tumor cell populations displaying metastatic properties often have higher gene dosage than their less malignant progenitor tumors, as shown by increased ploidy levels, chromosome duplication and gene amplification. The acquisition by tumor cells of high chromosome numbers may be due to endoreduplication or somatic hybridization either between tumor cells or between tumor and host cells. All such mechanisms increase genetic variability and instability in tumor cells since they trigger a polyploidization-segregation cycle. Among the wide variety of segregants which may emerge from high-ploidy cells, variants with increased malignancy can be positively selected in vivo. Evidence for in vivo fusion of tumor and normal host cells has been reported in different tumor systems. However the attainment by tumor-host hybrids of a higher degree of malignancy has only been observed following substantial chromosome segregation. The involvement of a cell of bone marrow origin as preferential host partner in the fusion process has been proved both by studies on tumor-host hybrids in bone marrow radiation chimeras and in vitro hybridization experiments between non-metastatic tumors and normal lymphoreticular cells which have led to the establishment of metastatic variants. Several different segregational mechanisms may bring about homozygosity or hemizygosity of recessive alleles in tumor-host hybrids, leading to their expression. The marked chromosome dynamics of tumor-host hybrids are also responsible for extensive chromosome rearrangements. At the molecular level these may represent mechanisms causing altered oncogene activity. The activation of new oncogenes by transposition or amplification as well as the amplification of previously activated oncogenes are the mechanisms most likely to be responsible for transition from low to high malignancy, occurring through ploidy changes, such as those produced by somatic mating.