Nature 543, 239–242 (2017); doi:10.1038/nature21383

In this Letter we omitted to cite a paper1 that also used recently developed thermodynamic models2 to predict the melting process in Archaean metabasaltic rocks. Importantly, the average enriched Archaean tholeiite used by ref. 1 as a proposed source rock3 for tonalite–trondhjemite–granodiorite rocks has a magnesium number (Mg#) of 57, significantly higher than the average value for the CF-2 basalts (with Mg# of 35)4. This difference has profound implications for the results of these studies. We regret not citing ref. 1 to emphasize the clear distinction between their findings and those of our study. The original Letter has not been corrected.