Nature 529, 80–83 (2016); doi:10.1038/nature16447

It has come to our attention that in this Letter, there were some errors in the categorization of some of the modern datasets (R. Telford et al., personal communication). These errors affect 19 datasets and are highlighted in the Supplementary Information to this Corrigendum. Eight datasets were labelled as unknown and left out of the analyses, but should have been included. Two of these were mainland datasets, and six were island. One dataset labelled as island should have been mainland. Two datasets that were labelled as mainland should have been island. One dataset was largely a duplicate of another because of publication in multiple years and should have been excluded. One was a matrix transposition of another and should have been excluded. We have updated the ages of seven datasets (Supplementary Information). Three of these are island datasets that were not included in the original analyses. Two were the duplicated (and now excluded) mainland datasets. Four datasets had incorrect numbers of species or sites. We have rerun our analyses using the corrected dataset and did not find any appreciable differences in our results (see Table 1).

Table 1 Reanalysis using corrected dataset