Probability That a Fossil Absent from a Sample Is Also Absent from the Paleolandscape

Abstract

A procedure to calculate probability limits for the inference that the absence of a taxon from a woodrat (Neotoma) midden implies absence from the paleolandscape uses paired samples of modern vegetation communities and paired samples of paleocommunities. Assumptions are: (1) each member of a sample pair is an independent measure of the same vegetation assemblage; (2) behavioral patterns of woodrats are the same as each midden in a paired sample is constructed; and (3) the probability of fossilization is zero when a taxon is absent from the vegetation. The procedure provides a logical test of data consistency: the upper probability limit of making a false inference should be greater than the lower limit. Averaged over 140 plant taxa, the upper and lower probability limits for a false inference were 11 and 7%, respectively. More than 70% of taxa passed the logical test, indicating a reliable procedure. For many taxa that failed the logical test, four potential explanations account for this failure, two of which can be solved by simply increasing sample sizes. Using analogous assumptions, the procedures are applicable to other types of stratigraphic sampling such as macrofossils from sediment cores or fossils from biostratigraphic units.