Corrections and Clarifications

See allHide authors and affiliations

Science  20 Dec 2013:
Vol. 342, Issue 6165, pp. 1446
DOI: 10.1126/science.342.6165.1446-a

Reports: “Anthropogenic seismicity rates and operational parameters at the Salton Sea geothermal field” by E. E. Brodsky and L. J. Lajoie (2 August, p. 543, published online 11 July 2013). There are two typographical errors in Table 1: The reported phase lag in the time interval of 1982–1991 associated with injection should be 0 instead of 6. Also, the correlation between injection and seismicity in the 1991–2006 time window should be 0.25 instead of 0.26. The HTML and PDF versions online have been corrected.

Reports: “Mapping the origins and expansion of the Indo-European language family” by R. Bouckaert et al. (24 August 2012, p. 957). The authors are grateful to William Chang and Andrew Garrett for informing them that there was a problem with the data matrix they used. The error occurred when 13 languages were removed from the original 116-language data matrix (http://ielex.mpi.nl) because they were colonial varieties or doculects, for which the authors had a better source. Removing these languages produced 283 “empty” columns of zeros (out of 6279), which the authors neglected to omit. Columns full of zero entries can potentially bias rate estimates from model-based phylogenetic inference. In addition, this revealed an error in the ascertainment bias correction for all-zero columns in the BEAST code [A. J. Drummond, A. Rambaut, BMC Evol. Biol. 7, 214 (2007)]. The authors have therefore rerun the analyses with corrected data and BEAST code. The covarion model is now the best-fitting model of cognate evolution [C. Tuffley, M. A. Steel, Math. Biosci. 147, 63 (1998); D. Penny et al., J. Mol. Evol. 53, 711 (2001)]. Under this model, the basic inference about the geographic origins of Indo-European remains unchanged (revised Table 1 shown below); however, the tree topology differs slightly (revised Fig. 2 shown below) and date estimates are younger, although still showing a better fit with the Anatolian hypothesis than with the Pontic steppe hypothesis (median = 7579 years BP; 95% HPD interval = 5972 to 9351 years BP). The date ranges under the different models of cognate evolution, including the previously best-fitting model (the stochastic-Dollo), are shown in a newly added fig. S13. Revised supplementary material with revised versions of all affected tables and figures, as well as updated xml code, has been posted online. Two points in the main text also need correcting. First, the analysis, in which the authors constrain the tree topology to fit with an alternative pattern of diversification, still shows strong support for an Anatolian origin, but the Bayes factors are slightly different (BFSteppe I = 174.02, BFSteppe II = 145.35). Second, in the revised analysis, the five major Indo-European subfamilies—Celtic, Germanic, Italic, Balto-Slavic, and Indo-Iranian—all emerged as distinct lineages between 4000 and 7000 years ago, not between 4000 and 6000 years ago as previously stated.

Navigate This Article