Does exceptional preservation distort our view of disparity in the fossil record?

Feb 2019 Proceedings of the Royal Society B

Flannery Sutherland, J.T., Moon, B.C., Stubbs, T.L. & Benton, M.J. 2019 Does exceptional preservation distort our view of disparity in the fossil record? . Proceedings of the Royal Society B 286: 20190091 doi:10.1098/rspb.2019.0091

What started out as a summer project led by Tom Stubbs started growing into an exploration of methods for calculating diversity. This was done by Joe Flannery Sutherland from Summer 2017.

Initially, the plan was to look at the effect of Lagerstätten on disparity calculations in ichthyosaurs. We use discrete-character disparity calculations to measure the disparity of ichthyosaurs through time and then remove taxa that are from sites of exceptional preservation – predominately those with particularly complete specimens, but also those that preserve lots of material. The argument is that taxa from these exceptional sites may selectively increase the disparity, both by presenting more material and better-preserved material.

Ichthyosaur disparity with (black) and without (red) taxa from Lagerstätten included. Lagerstätten do increase the observed disparity, however, the patterns for both series are broadly the same. From Flannery Sutherland et al. (2019).

The positive news is that the trends we found in ichthyosaur evolution were consistent whether all taxa were included or not, even though removing taxa from Lagerstätten did reduce the disparity in some bins.

One of the things that Joe noticed doing this is that the disparity and morphospace region occupied decreased as taxa became less complete.

Morphospace occupation collapses towards the centroid with decreasing completeness of the input data. From Flannery Sutherland et al. (2019)

This was also explored by Lehmann et al. (2019), who went into the underlying methods and found similar trends with their analysis. This particularly reflects the used of Generalised Euclidean Distances that fill missing values with averages across a range of taxa, and so creates an artificial averaging effect as taxa are less complete and more values are replaced.


Lehmann, O.E.R., Ezcurra, M.D., Butler, R.J. and Lloyd, G.T. 2019. Biases with the generalized Euclidean distance measure in disparity analyses with high levels of missing data. Palaeontology 62 (5): 837–849. doi:10.1111/pala.12430