Evaluating the success of astronomical tuning: Pitfalls of using coherence as a criterion for assessing pre‐Pleistocene timescales

The imprint of orbital variations on the geological record of climatic variability is well documented, especially for the Plio‐Pleistocene. There is considerable interest in developing very high resolution timescales through the Cenozoic and into the Mesozoic by tuning geological records to the assumed astronomical forcing. Since the precession signal is so highly amplitude modulated, it is widely believed that high coherence between record and assumed forcing in the precession band is an indication that the timescale is probably correct, because coherence is supposed to provide a measure of the degree of common amplitude modulation. We show that this is misleading; even a sinusoidal variation that has been “tuned” to an insolation record shows highly significant coherence at the 23‐kyr and 19‐kyr precession frequencies. Coherence is a good indication that the tuning has generated a consistent phase relationship, but complex demodulation is a better tool for assessing the relationship between amplitude modulation in the data and in the hypothetical forcing.