Some observations on methods of estimating the proportions of morphologically similar pollen types in fossil samples

For pollen types whose morphological characters overlap markedly, it is often helpful to be able to estimate the proportions of each type present in a sample of pollen grains. Several numerical methods of estimation have been proposed in recent years. This paper criticizes some procedures that are based on the identification of each pollen grain, rather than the direct estimation of proportions. A class of maximum likelihood models is described and illustrated by the analysis of eight samples of Picea pollen from late Quaternary profiles in eastern North America. The maximum likelihood approach provides an indication of the precision of the estimates, a point considered to be important.