In this paper we suggest using a modified version of the time reversibility (TR) test of Chen, Chou and Kuan (2000) as a complementary diagnostic test for time series models. The modified CCK test is easy to compute and requires weaker moment conditions than existing tests. Our simulations demonstrate that this test is powerful against asymmetry in volatility but the BDS test is not. In the empirical study of US stock index returns, we first find that these returns are all time irreversible. Applying the GARCH and EGARCH models to these returns, we also find that the Q and BDS tests always accept the null hypothesis and cannot distinguish between these models. By contrast, the modified CCK test accepts only the EGARCH model with an order that can accommodate the underlying asymmetry in volatility. Our results suggest that the detected time irreversibility in these return series may be attributed to volatility asymmetry and that such asymmetry may be captured using a proper EGARCH model. Copyright © 2002 John Wiley & Sons, Ltd.