In sampling from general populations, the skewness and kurtosis statistics are subject to the constraintb2> 1 +b1. A bivariate product-density model for the distribution of √b1andb2is studied, consisting of a JohnsonSuapproximation to the marginal density of √b1, and a gamma density for the conditional distribution ofb2. Equiprobability contours are given for sampling from normal and nonnormal populations. In the normal case, an eight-parameter model is completely specified.