Wichmann (2005) discusses the power-law distribution n=ar−b as a description of the relationship between the number of languages n in a language family, and the rank r of that family in a list ordered by decreasing n. Two datasets are used by Wichmann, one from Ethnologue (Grimes 2000), which lists 130 language families, and one from Ruhlen (1987), listing 21 families. We have reanalysed these data and find that the method of fitting a power-law used in the paper is not optimal because it does not allow for a sensible maximum value for the family size n.