Since the 1970s, various private and governmental agencies have conducted studies to assess potential unconventional gas resources, particularly those resources contained in tight sands, fractured shales, and coal beds. The US Geological Survey (USGS) has assessed the amount of unconventional gas resources in North America, and its estimates are used by other government agencies as the basis for their resource estimates. While the USGS employs a probabilistic methodology, it is apparent from the resulting narrow ranges that the methodology underestimates the uncertainty of these undiscovered, untested, potential resources, which in turn limits the reliability and usefulness of the assessments. The objective of this research is to develop an improved methodology to assess potential unconventional gas resources that better accounts for the uncertainty in these resources. This study investigates the causes of the narrow ranges generated by the USGS analyticprobabilistic methodology used to prepare the 1995 national oil and gas assessment and the 2000 NOGA series, and presents an improved methodology to assess potential unconventional gas resources. The new model improves upon the USGS method by using a stochastic approach, which includes correlation between the input variables and Monte Carlo simulation, representing a more versatile and robust methodology than the USGS analytic-probabilistic methodology. The improved methodology is applied to the assessment of potential unconventional gas resources in the Uinta-Piceance province of Utah and Colorado, and compared to results of the evaluation performed by the USGS in 2002. Comparison of the results validates the means and standard deviations produced by the USGS methodology, but shows that the probability distributions generated are rather different and, that the USGS distributions are not skewed to right, as expected for a natural resource. This study indicates that the unrealistic shape and width of the resulting USGS probability distributions are not caused by the analytic equations or lack of correlation between input parameters, but rather the use of narrow triangular probability distributions as input variables. Adoption of the improved methodology, along with a careful examination and revision of input probability distributions, will allow a more realistic assessment of the uncertainty surrounding potential unconventional gas resources.