The exaggeration of petroleum hydrocarbon by ground water monitoring wells is a well known problem that introduces significant errors in the estimation of recoverable hydrocarbon. Farr (1990) and Lenhard (1990) demonstrated that the volume of hydrocarbon in the formation is a function of the observed (exaggerated) thickness of hydrocarbon in the well and the saturation/capillary pressure characteristics of the soil. These characteristic curves for soils are commonly determined in the laboratory using a pressure plate apparatus, but Mishra (1989) suggest a method of determining the characteristic curve using sediment grain-size analysis. The purpose of this investigation was to evaluate the influence of sediment variability on estimation of hydrocarbon volumes. Forty-five samples of a relatively homogeneous sandy deposit were collected within a small area (less than 1,100 square meters). Characteristic drying curves were determined for 10 of these samples using a pressure plate apparatus and grain-size analysis was completed for all 45 samples. Graphical plots relating estimated hydrocarbon volume versus observed hydrocarbon thickness were constructed for each of the samples using the approaches of Lenhard (1990). The results using only the characteristic curves determined from the pressure plate show significant variability in the prediction of hydrocarbon volumes. For example, a formational hydrocarbon volume of 3 centimeters cubed per square centimeter (cm3/cm2) could produce anywhere between 45 and 200 centimeters of measurable hydrocarbon within a ground water monitoring well. This suggests that use of an "average" soil sample to characterize hydrocarbon exaggeration within even a very small site can lead to substantial errors in volume estimation. It also suggests that maps of apparent hydrocarbon thickness can be extremely misleading, causing hydrologists to place remediation wells in areas of greatly exaggerated thicknesses produced by fine-grained materials, rather than in areas of greatest hydrocarbon volume. Comparison between characteristic curves calculated from grain-size analysis using the approach of Mishra (1989) to those measured using the pressure plate apparatus shows poor correlation that can markedly influence the estimation of formational hydrocarbon volumes. The typical error introduced by using characteristic curves developed from grain-size analysis is similar to the total variation in saturation/capillary pressure relations developed from the pressure plate samples for the entire site. This work suggests that characteristic curves should be generated from a pressure plate apparatus for each well that encounters free hydrocarbon product, and that apparent hydrocarbon thickness should be corrected prior to contouring.