In this paper we study the distributional properties of the earnings surprise and the properties of the earnings-returns relation and their evolution over time. We distinguish between Friday and non-Friday announcements to control for a "Friday effect," reported in numerous studies. Our major findings are as follows. First, we find that during the period 1989-2006 firms tended to report more "bad" news on Friday than during the rest of trading days. Second, we report a temporal shift in the earnings-return relation with stock returns becoming more sensitive to Friday earnings announcements compared to announcements released during the rest of the week. The shift is substantially more pronounced for the negative earnings surprises. Finally, we find that the relative sensitivity of stock returns to Friday versus the non-Friday earnings announcements is related to the quality of the informational disclosure by the firms' management. Overall, our findings suggest that investors have learned about the firms' management strategy to report "bad" news on Fridays. As a result, the benefits from following this strategy have disappeared over time.