The frequency at which age data heap at round ages can be used to infer people’s ability to count. Földvári, Van Leeuwen and Van Leeuwen-Li (FVV) contend that gender specific trends in numeracy derived from age heaping in census data are unreliable because women’s ages are adapted to those of their male household heads. This paper reassesses this finding by comparing two independently constructed age data sources for the case of rural Ireland in the nineteenth century: prison registers and corresponding census districts, where the former has the unique advantage of being self-reported by newly incarcerated male and female prisoners. We find that women are substantially less numerate than a comparison based solely on census data would suggest. We conclude that the FVV bias is a concern for the age heaping literature and recommend that female numeracy estimates made for societies where the census is the only available source be used with caution.