A new approach is developed for the estimation of insect population densities which eliminates the necessity of counting every individual per sample. Although the method was developed for populations of corn rootworm eggs, it seems as applicable to other organisms featuring contagious distributions. Estimation by the proposed method is based upon the proportion (p) of samples containing at least t insects each, where t, termed threshold density, can be any positive integer specified. For example, if t is 1, counting is entirely unnecessary since the only information required is the proportion of samples in which the organism is present. The prediction equation which links mean density to the variables ° and t must be estimated beforehand from a series of samples extending over a wide range of population densities. Expressing the model in terms of logarithms premits the estimation of its parameters by linear least squares. The study indicates that there is an optimal threshold level associated with every population density and specified allowable error. Estimation near these optima is apparently more efficient than conventional estimation involving complete sample enumeration. By enlisting a two—stage sampling scheme, one should consistently be able to approach the optimum threshold and thereby produce estimates comparable in efficiency to direct counts but free from much of the labor.