This paper considers the issue of testing for symmetry of the business cycle. It is demonstrated that findings of symmetry should be interpreted with caution since tests tend to have low power to detect asymmetries when applied to data that have been filtered to isolate their stationary business-cycle component. This implies that asymmetries are likely to be detected in practice only when they are particularly prominent. An empirical application examines the properties of the cyclical component of real GDP for the G7 countries.