Nature 445, 429–432 (2007)

In the print version of this Letter, the significance level for hypothesis testing was incorrect for the ‘first-choice’ results (line 19 on page 430) and should be 0.05 instead of 0.01. The online PDF and HTML versions are correct.