Do We Really Need More Replications?

Abstract

A plea is often made that psychologists should perform more replications and that replications should be granted more favorable consideration in publication decisions. The present article argues that, despite these assertions, replications are performed and published. Many psychologists apparently do not recognize them as such because they expect replications to be close copies of prior research designs. However, replications are often quite different from previous research; what is important is that they may replicate theoretical hypotheses, not research designs. In addition, the point is made that statistically significant results are more important than nonsignificant results and that many nonsignificant results are published, also contrary to popular wisdom. Thus, even if replications were in short supply, we would not be in great danger of falsely rejecting too many null hypotheses. The conclusion is reached that further complaints about the need for more replications should be based on better data than now are available and that we would better spend our time improving other aspects of the research process than in needless debate about and production of replications qua replications rather than tests of hypotheses.