A recent study by Goyal and Santa-Clara [J. Finance 58 (2003) 975] finds a significantly positive relationship between average stock returns and pre-determined average return volatility measures, while finding no relationship between the average return and its own volatility. The result is interpreted as evidence that idiosyncratic risk matters in asset pricing. We re-examine the issue in extended sample periods and find the proclaimed positive relationship is not substantiated. Our analysis indicates that the above-mentioned positive relationship is mainly driven by the data in the 1990s. The trading strategy suggested by Goyal and Santa-Clara to exploit the return predictability by pre-determined volatility does not yield sustained economic gains.