Correction of an error in earlier simulations which show how anisogamy could evolve by selection on individuals (Parker et al., 1971) now indicates that anisogamy can evolve when the range of gamete size is very much smaller than previously thought. These models assumed random fusion of gametes, external fertilization, and that zygote viability is dependent on the volume of provisioning it receives from one or both gametes. The present analysis concerns the success of strategies for selective fusion of gametes arising in a randomly-fusing parental population. On a priori grounds selection is expected to favour assortative fusion in ova but disassortative fusion in sperm; anisogamy can persist only if genes for assortative fusion of ova will not spread, and “perfect” anisogamy where genes for disassortative fusion fixate. Mutant strategies for assortatively-fusing ova may not be successful if such ova must compete with sperm for fusions with the randomly-fusing ova. Particularly at high levels of anisogamy, very few of the mutant ova will be fused by the time all other ova have become zygotes; hence their spread may be checked by the enhanced chances of death before fusion, or by problems associated with selfing if they do manage to fuse. In contrast, disassortatively-fusing sperm generally have an advantage when anisogamy would be favoured under random fusion. Genetic simulations (involving two loci, one with alleles for fusion behaviour and the other with alleles for gamete size) were used to confirm these conclusions. Where there is some degree of asynchrony of spawning, disassortative fusion alleles do even better than with perfect synchrony. Simulations with various sex-limited fusion strategies show that non-limited disassortative fusion, i.e. for both ova and sperm, is likely to be an ESS at high anisogamy against all strategies but the one which plays random fusion in ova, disassortative fusion in sperm. This is the ultimate ESS and it does not disrupt anisogamy, but at high anisogamy it has an extremely small advantage over non-limited disassortative fusion. The reasons for the establishment of non-limited disassortative fusion are probably related to avoiding selfing, and to the cost of maintaining random-fusion in ova (in terms of motility, etc.) outweighing the benefits of becoming obligatorily disassortative (non-motile).