Supernumerary dental elements have been reported in Lynx lynx by several authors. These features have been given different evolutionary interpretations by different commentators. I note here that, since these features are absent in the plesiomorphic sister-groups of L. lynx, they represent a true evolutionary reversal. If they were simply a retention of an evolutionarily older phenotype, we should expect to see them developed in at least one plesiomorphic sister-group. Such development of a previously hidden character can occur if it is genetically linked to features selected for, until it becomes phenotypically expressed, whereupon selection can act on the character itself. Since Dollo's law, which is the theoretical issue behind the present discussion, is not a law, but a rule, and, like all rules based on probabilities, we should expect to find exceptions in the fossil record. Such exceptions are not rare, but few are as spectacular as the present one, in which the redeveloped feature is at least phenotypically identical with one which has been lost in the Felidae since the Miocene.