First the link to this week’s complete list as HTML and as PDF.
Beckes et al. offer a sensible hypothesis and a reasonable method, so at first sight there is little reason to doubt their result. Looking closely they rely on modest correlations arrived at from a limited number of data points, n=22. As they themselves rightly conclude the r=0.59 and r=0.43 in their figure 3b are essentially one and the same. So their whole result hangs on their figure 2b with a r=0.6 versus r=-0.06. As they do not provide raw data even in their supplement I used Datathief to pick them from their diagram and was able to reproduce figure 2b left. Following the Feynman criterion that a result must not rest on a few points at the extremes of the range alone, I then eliminated three of the 22 data points and ran the regression again. This left only r=0.26 against the unedited r=-0.06. Is this difference significant considering the small number of data points? I’m not sure but then I did not succeed in taking it away either.
With the choice of peaks and averages, different latitudes, and different times of year there are numerous Milankovich cycles to choose from for driving climate. At least the common choice of peak northern summer insolation has been shown to yield good predictions over a very long time span and for many individual incidents. Peaks depend on the yearly amplitude and have very little to do with the annual average as chosen for WAIS’ figure 3. They also restrict themselves to a few thousand years near the end of the LGM. This looks a lot like cherry picking to me. The least they ought to have done, but failed to do, is demonstrate that the changes in average insolation as per their figure S10 are a meaningful driver over the whole 50 ka. Comparing to figure 1 they might be. If so we’re probably witnessing the rapid onset of a strong cooling now.
Neanderthals in the Middle Paleolithic did not make bone tools. This plain and simple rule, imbibed by generations of undergraduates, seems not to hold after all, if Soressi’s result will be corroborated by other finds.