Results of Chamberlain and Meyer are combined to extend Meyer's location-scale condition to portfolio choice models where the distribution of returns is elliptically symmetric. This extension implies that mean-variance choice is consisteht with expected utility maximizing choice for such models. All expected utility maximizing portfolios lie on the mean-variance efficiency frontier which can be generated with quadratic risk programming. A test for elliptical symmetry is also described. This test enables one to determine whether a given set of portfolio data satisfies the conditions which make mean-variance choice consistent with expected utility maximization.