Any method for establishing genetic relatedness by identifying similar constructions in different languages must provide a specification of similarity. As long as the identification of regular sound correspondences is our measure of similarity in linguistic form, there can be no avoiding appeal to sound correspondences in genetic arguments. This is true even for the type of evidence termed shared aberrancy. If early comparative historical linguists thought otherwise, as is often suggested, then by modern standards their proofs are invalid. Yet I argue that many such claims regarding early 19th comparativist practice involve a distorted and selective view of the history of linguistics. I also argue that claims regarding Meillet's supposed privileging of shared aberracy over sound correspondence in the early 20th century result from a misinterpretation of his notion fait particulier 'singular fact'.