CLAMP (Climate Leaf Analysis Multivariate Program) is a powerful paleoclimate proxy with the ability to yield quantitative data on past temperatures, precipitation, growing season length, and humidity, as well as enthalpy (a property of a parcel of air that is useful in studies of paleoaltimetry). Commonly quoted uncertainties in CLAMP predictions relate to the statistical uncertainty inherent in the combined quality of the modern calibration data sets and the relationship of foliar architecture to the various climate parameters. This minimum uncertainty assumes that the fossil assemblage represents faithfully the foliar physiognomy of the source vegetation. Taphonomic processes degrade this physiognomic fidelity. Differential selection for size, shape, and species composition during transport and post-depositional processes biases the physiognomic profile of the fossil assemblage. The sensitivity of CLAMP precision to taphonomic filtering was assessed empirically using a modern data set from the Crimean Peninsula as a proxy fossil site. Elimination of leaf margin, apex, base, size, and shape character-state categories, singly and in combination, changed the predictive capability of CLAMP. Loss of margin characters had the greatest effect, particularly on temperature-related variables (mean annual, warm-month mean and cold-month mean temperatures, length of the growing season, and enthalpy). Taphonomic selection against large leaf sizes had little effect even on moisture-related estimates (precipitation during the growing season, mean monthly growing season precipitation, precipitation during the three wettest and driest months, relative humidity). Loss of taphonomically sensitive characters (apex, base, or shape) also had little effect on CLAMP predictions. Copyright © 2005, SEPM (Society for Sedimentary Geology).