Nature 495, 495–498 (2013); doi:10.1038/nature12006

In Figure 3c of our Letter, the time axis of the ln(Si/Al) record of ODP Site 658A (orange line) was based on a prior age model that did not take into account an age tie-point at 17.8 kyr ago, derived from oxygen isotope stratigraphy (see Methods). This incorrectly shifted the plotted data between 20 kyr ago and 14.2 kyr ago towards younger ages. The shift is most significant at the onset of the opal peak during Heinrich stadial 1 (HS1), which in the original Fig. 3 appeared to occur at 16.5 kyr ago, whereas the correct age model puts this increase in ln(Si/Al) at 17.8 kyr ago. Figure 2, which shows the full time series, is not affected. The error has no implications for the conclusions of the original Letter. In fact, the earlier onset of the opal peak in the correct age model is even more consistent with our interpretation of the opal peak as an indication of changes in ocean circulation early during deglaciations. The correct age model and data are accessible at Pangaea ( Figure 3 has been corrected in the PDF and HTML of the original paper.