Using a sample of U.S. firms over five decades, we find robust evidence of temporal variations in R&D excess returns. Initially low but still significant from the 1970s to 1980s, R&D excess returns gradually increased and peaked in the early 2000s. After 2005, R&D excess returns declined dramatically and were rarely significant. These findings are inconsistent with the implicit assumption in the extant literature that R&D excess returns are persistent across time. We focus on a 22-year period (1983 to 2004) during which R&D excess returns were consistently significant and find that the temporal variations in excess returns were more in line with risk than mispricing explanations. We also find no evidence that the existence of and the temporal variations in R&D excess returns are explained by other return anomalies. Overall, we provide insights into the existence and temporal variations in R&D excess returns and caution against assuming their consistent predictability over time.