Statistical power of psychological research: what have we gained in 20 years?

Power was calculated for 6,155 statistical tests in 221 journal articles published in the 1982 volumes of the Journal of Abnormal Psychology, Journal of Consulting and Clinical Psychology, and Journal of Personality and Social Psychology. Power to detect small, medium, and large effects was .17, .57, and .83, respectively. 20 years after Cohen (1962) conducted the first power survey, the power of psychological research is still low. The implications of these results concerning the proliferation of Type I errors in the published literature, the failure of replication studies, and the interpretation of null (negative) results are emphasized. An example is given of the use of power analysis to help interpret null results by setting probable upper bounds on the magnitudes of effects. Limitations of statistical power analyses, suggestions for future research, sources of computational information, and recommendations for improving power are discussed.