For many years, the uncontested opinion in the field has been that there is not an empirical linguistic difference, but rather a societal one, between dialects and languages. Recently, however, it has been reported that distances between linguistic varieties are typically bimodal, suggesting that a classification into language vs. dialect might not be exclusively a social construct. Here, I review these papers and attempt to replicate them with a larger and more suitable dataset from Romance languages. Results fail to reveal any empirically salient boundary between languages and dialects. A look at other language families supports this conclusion. Simulations suggest that the distribution of linguistic distances in a family is primarily driven by its phylogenetic structure, with no additional mechanisms needed beyond vertical and horizontal diversification.