In this paper, we conduct a comprehensive study of tests of mean-variance spanning. We provide both a comparison and a geometrical interpretation of three asymptotic tests (likelihood ratio, Wald, and Lagrange multiplier) of mean-variance spanning under the regression based framework of Huberman and Kandel (1987). For the case of normality, we provide the exact distributions and a comprehensive power analysis of the three tests. For the general case, we provide the GMM version of the spanning tests and evaluate their performance using simulation. In addition, we compare the performance of the spanning tests under the regression approach with those under the stochastic discount factor approach. Our results suggest that the two approaches have similar properties when returns are normally distributed but the regression approach is superior to the stochastic discount factor approach when returns follow a multivariate Student-t distribution.