Multi-component RNA viruses have genomes that are segmented into two or more RNA molecules. A viral particle carries only one RNA molecule. Reproduction of a particle requires complementation by particles carrying other segments of the genome. Complementation is achieved when a group of particles co-infects the same host cell and forms a co-infection group. I have previously proposed the hypothesis that multi-component reproduction evolved in RNA viruses as a form of sex. Multi-component viruses may need sex because, like all RNA viruses, they have very high mutation rates. On the other hand, Nee (1987, J. molec. Biol. 25, 277-281.) has proposed the hypothesis that multi-component genomes evolved because smaller RNA molecules are favored by selection on RNAs within a host cell. Nee (1989, J. theor. Biol. 138, 407-412.) also claimed that selection on RNAs alone can account for the evolution of multi-component viruses. He criticized the viral sex hypothesis because, in his view, co-infection groups are not units of selection and are too transient to be engaged in sex. These two hypotheses were further examined through population genetic models. Three evolutionary agents are assumed to operate in the models. Selection on co-infection groups favors retention of the genome on one large RNA molecule because larger RNAs require less complementation. Selection on RNAs favor segmentation of the viral genome into smaller RNAs, which are replicated and encapsidated more rapidly. Mutation pressure also favors smaller molecules because those molecules are smaller targets for deleterious mutations. Analysis of the models shows that (when parameter values argued to be biologically realistic are used) selection on co-infection groups is necessary for the evolutionary persistence of multi-component viruses. Without selection on co-infections groups to oppose mutation pressure and selection on RNAs, a population of multi-component viruses is displaced by a population of parasitic viral RNAs that are replication and encapsidation specialists. These results support arguments that co-infection groups are units of selection in multi-component viruses. Both mutation pressure and selection on RNAs may be responsible for the evolution of genome segmentation in multi-component viruses because there is good evidence documenting the action of both in RNA viruses.(ABSTRACT TRUNCATED AT 400 WORDS)