Changes in sample size may confound interpretation of faunal change in the fossil record. The record of early Eocene mammals from the Wasatchian Land-Mammal Age in the Clark’s Fork Basin, Wyoming, shows a high correlation between the square root of sample size and species richness. This correlation suggests that Schankler’s (1980) Biohorizon A, a faunal turnover composed mainly of disappearances, is largely an artifact of sampling fluctuation. Within the interval of Biohorizon A, sample size drops from record high to record low values. Monte Carlo simulation of the drop in sample size across Biohorizon A demonstrates the role of sampling variation in producing artifacts of faunal change. The distribution of missing species resulting from the simulations provides a reliable estimate of the number of species that are likely to be missing at a specified sample size, even if they were present in the original population. Results of the simulations indicate that most of the 16 disappearances observed in the 200-m interval above Biohorizon A can be explained by low sample size alone. For each species that disappeared in the actual record, the frequency of absence in the simulations is a basis for comparing the likelihoods of two hypotheses: (1) that the species was present but not represented by fossils, and (2) that the species was absent Evaluation of the likelihood ratio for these hypotheses indicates that Arctodontomys wilsoni-A. nuptus, Phenacodus vortmani, and Homogalax n. sp.-H. semihians are the lineages most likely to have disappeared over the interval of low sample size. Reconsideration of the biostratigraphic correlation between the central Bighorn Basin and the Clark’s Fork Basin based on the first appearances of Homogalax protapirinus and Tetonius matthewi/steini supports a higher stratigraphic position for the interval corresponding to Biohorizon A in the Gark’s Fork Basin than its original placement.