Pollen grains of five species of pine from southwestern United States (Pinus edulis, P. aristata, P. flexilis, P. ponderosa, and P. contorta) differ slightly in the morphology of the pollen wall. Single grains from modern collections of pollen from the five species can be discriminated by a combination of eight qualitative characters and three dimensions. The same criteria were applied subjectively to the pine pollen in 20 stratigraphic samples from Dead Man Lake in the Chuska Mountains of New Mexico. Pinus edulis and P. ponderosa, the only two species of pine that now grow in the Chuska Mountains, account for nearly all the identifiable pine pollen in near-surface samples. In Pleistocene sediments, pollen assigned to P. flexilis is common.
As a check on the subjective identifications, the fossil grains were classified again by a numerical method. The technique estimates the likelihood that a fossil grain could have been drawn from each of the reference populations, and it then assigns the grain to a reference species if it is sufficiently similar to that species and sufficiently dissimilar to all other reference populations. The assumptions made explicit by the numerical method are similar to those implicit in routine pollen analysis. Only half of the fossil grains identified subjectively in the samples from Dead Man Lake were identified by the more stringent numerical method, but the two methods agree well in the proportions identified of the five species. Comparison of the two methods illustrates the difficulty of eliminating bias in making subjective identifications.