1. Taphonomic Uncertainties

In the case of CLAMP several papers have examined this, recognizing that the loss of information during the fossilization process means, by definition, that it is not measureable. However the maximum uncertainty of single or multiple character loss can be measured in extremis by seeing how much predictions are altered by loss of complete character states across all taxa, even if this means that some of the resulting estimates are highly theoretical because complete character loss would render the material unscoreable. Because CLAMP is a multivariate technique that incorporates information from across 31 character states it has proven to be remarkably robust to taphonomic losses. Empirical tests of CLAMP’s resilience to taphonomic issues were conducted by Wolfe (1993), but a statistical approach has been used by Spicer et al. (2005) and Spicer et al. (2011). Leaf Margin Analysis(LMA) is also prone to taphonomic uncertainties but is potentially more susceptible than CLAMP because of its univariate nature.

This 'spidergram' shows the effect of character loss on the MAT estimate using the Physg3br data set. The scale in °C is shown on the vertical line running from the centre of the plot to the '12 o'clock' position. The complete dataset yields an MAT of 9°C. The experiment was carried out on the modern Bolshoi Canyon sample and more details can be found in Spicer et al. (2005).