Unduplicated counts of female grizzly bears Ursus arctos horribilis with cubs-of-the-year are currently used to estimate minimum population sizes used, in turn, to calculate allowable (assumed to equal sustainable) mortality for grizzly bear populations in the contiguous United States of America. This calculation assumes that unduplicated counts are an unbiased and accurate indicator of population size and that the ratios of minimum population size and known mortality to their respective totals are equal. Neither of these assumptions can be directly tested. However in this paper I use data from the Yellowstone ecosystem, 1977–1990, to evaluate two directly related but alternate hypotheses: (1) annual variation in unduplicated counts is explained by factors extraneous to the number of adult females in the population (i.e. search effort and sightability of females with cubs); and (2) there is >10% risk of allowing unsustainable mortality (actual mortality rate >6%) given a plausible, uniform range of population and mortality ratios. My results are consistent with accepting both of these hypotheses. I therefore concluded that unduplicated counts varied without a known relationship to population size and that, by normal standards, the method currently adopted for management of grizzly bear populations in the contiguous United States was not a conservative basis for calculating maximum allowable mortality. I suggest that using lower mortality rates and conservative bounds of confidence limits for the estimated parameters used in calculations of allowable mortality could substantially reduce the risk of unintentionally allowing excessive mortality.