How the Austronesian tree was reconstructed to compute evolutionary distinctiveness more accurately, using Germanic languages as an example. A tree (a) of Germanic languages with (here invented) branch lengths can be used to compute evolutionary distinctiveness (ED), but missing languages (Dutch and Swedish) will bias this score. Language classifications into families and subfamilies by the Ethnologue (simplified for illustration) can partially compensate for this bias. It can be used to infer a tree (b) with no meaningful branch lengths. Those languages or groups of languages missing from tree (a) are imported from tree (b) to form a reconstructed tree (c). ED, as calculated from tree (c), is usually more accurate than when calculated from tree (a); see main text. This method does not allow computing ED of languages missing from tree (a). In this analysis, we used the Austronesian equivalent of tree (c). Details in the electronic supplementary material.