Most studies of the demand for meat in Australia have used some measure of total income or expenditure, but two recent studies have assumed weak separability of a meat group and used expenditure on the meat group instead. These specification differences are of interest to the extent that they affect the economic interpretation, goodness-of-fit, elasticity estimates, predictive performance or hypothesis tests in empirical demand equations. In this paper, non-nested hypothesis testing procedures are used to test the alternative specifications of the income variable and the hypothesis of separability. The results favour the use of the expenditure variable implied by separability but are mixed concerning whether separability holds.