The several thousand-fold range in genome size among animals has remained a subject of active research and debate for more than half a century, but no satisfactory explanation has yet been provided. Many one-dimensional models have been postulated, but so far none has been successful in accounting for observed patterns in genome size diversity. The recent model based on differences in effective population size appeared to gain empirical support with a study of genome size and inferred effective population size in fishes, but there were several questionable aspects of the analysis. First, it was based on an assumption that microsatellite heterozygosity indicates long-term effective population size, whereas in actuality these markers evolve quickly and are sensitive to demographic events. Second, it included both ancient polyploids and non-polyploids, the former of which did not gain their current genome sizes through the accumulation of slightly deleterious mutations as required in the model. Third, the analysis neglected the tremendous influence that Pleistocene glaciation bottlenecks had on heterozygosities in freshwater (and far less so, marine) fishes. In sum, it is apparent that genomes reached their current sizes in most fishes long before contemporary microsatellite heterozygosities were shaped, and that ancient polyploidy rather than the accumulation of mildly deleterious transposon insertions in small populations is the dominant factor that has influenced the large end of the range of genome sizes among fishes.