This paper explores the determinants of the tests on the coefficients of the systematic risks on Taiwan's stock market. The Fama-Frenchfs (1992) two-stage regression analysis is used. The empirical evidence reveals that, ceteris paribus, the conclusions will not be altered no matter (a) the portfolio returns are value- or equally-weighted, (b) the market returns take the account of cash dividends, and (c) the GreTai stocks are included in the sample. However, the estimates of betas are largely affected by the way in which the portfolios are formed in the first stage, so that the results of the second-stage regressions will be affected.