Darwin believed that sexual selection accounts for the evolution of exaggerated male ornaments, such as the sword-like caudal fin extensions of male fishes of the genus Xiphophorus, that appear detrimental to survival. Swordtails continue to feature prominently in empirical work and theories of sexual selection; the pre-existing bias hypothesis has been offered as an explanation for the evolution of swords in these fishes. Based upon a largely morphological phylogeny, this hypothesis suggests that female preference to mate with sworded males arose in ancestrally swordless species, thus pre-dating the origin of the sword itself and directly driving its evolution. Here we present a molecular phylogeny (based on mitochondrial and nuclear DNA sequences) of Xiphophorus which differs from the traditional one: it indicates that the sword originated and was lost repeatedly. Our phylogeny suggests that the ancestor of the genus is more likely to have possessed a sword than not, thus questioning the applicability of the pre-existing bias hypothesis as an explanation for the evolution of this sexually selected trait.