Testing extra-binomial variations for proportion responses in a generalized linear regression is considered. With only the first two moments specified in the model, a unified quasi-score test that can be applied to more general alternatives is proposed. The test statistic is based on the binomial deviances and its null distribution can be approximated by a weighted sum of chi-square distributions. Simulation study shows that the quasi-score test is as powerful as the ordinary score test in some situations. Two data examples are illustrated.