River density and landscape roughness are universal determinants of linguistic diversity

Proc. R. Soc. B 281, 20133029 (7 June 2014; Published online 16 April 2014) (doi:10.1098/rspb.2013.3029)

In this paper [1], there were a few minor graphical errors and omissions regarding figure 1. We have supplied the correct figure 1 with an extended caption rectifying the mistakes. The entire text of the article including results, discussion and conclusions remains unchanged as we performed the calculations on the correct raw dataset.

Figure 1.

Illustration of the current heterogeneity of LD. The 100 largest landmasses are depicted in this map, which is a coarse-grained representation of actual linguistic domains in squared grid cells of side length 222 km (see the electronic supplementary material, figure S3). The sinusoidal projection is area-preserving, maintaining, in particular, the circumference of the Earth at each latitude (see the electronic supplementary material, figure S2). In this illustration, we have interpolated the raw density map to emphasize the continental trends in LD variability. Note that LD within cities is not annotated in the Ethnologue database (see Material and methods).

The corrections are: (i) it is not mentioned that we applied an interpolation to the raw data in figure 1 for the sake of visual clarity, (ii) the previous figure 1 erroneously indicated the presence of languages in large areas of the Canadian Arctic and Greenland that are known to be uninhabited, (iii) the numbering on the colour bar was not correct and (iv) it should have been mentioned that the original data source (the Ethnologue) does not annotate which languages are found in cities.