The use of herbarium specimens as vouchers to support ethnobotanical surveys is well established. However, herbaria may be underutilized resources for ethnobotanical research that depends on the analysis of large datasets compiled across multiple sites. Here, we compare two medicinal use datasets, one sourced from published papers and the other from online herbaria to determine whether herbarium and published data are comparable and to what extent herbarium specimens add new data and fill gaps in our knowledge of geographical extent of plant use. Using Brazilian legumes as a case study, we compiled 1400 use reports from 105 publications and 15 Brazilian herbaria. Of the 319 species in 107 genera with cited medicinal uses, 165 (51%) were recorded only in the literature and 55 (17%) only on herbarium labels. Mode of application, plant part used, or therapeutic use was less often documented by herbarium specimen labels (17% with information) than publications (70%). However, medicinal use of 21 of the 128 species known from only one report in the literature was substantiated from independently collected herbarium specimens, and 58 new therapeutic applications, 25 new plant parts, and 16 new modes of application were added for species known from the literature. Thus, when literature reports are few or information-poor, herbarium data can both validate and augment these reports. Herbarium data can also provide insights into the history and geographical extent of use that are not captured in publications.Electronic supplementary materialThe online version of this article (doi:10.1007/s12231-017-9367-1) contains supplementary material, which is available to authorized users.