Abstract
The choice of a one- rather than a two-tailed hypothesis testing strategy can influence research outcomes, but information about the type of test conducted is rarely reported in articles appearing in educational and psychological journals. Because unambiguous standards for using one- and two-tailed tests do not exist, complete reporting of hypothesis testing procedures is essential. In addition, educational researchers need to reevaluate the decision-oriented, “critical experiment” model of science that underlies the use of one-tailed tests. It is the adherence to the arbitrary .05 level of significance as a benchmark for publication decisions, rather than logical or methodological considerations, that largely accounts for the popularity of one-tailed tests. Effect size estimates, accompanied by confidence intervals or exact two-tailed probabilities, are generally more compatible with the growing meta-analytic view of social science as an incremental, cumulative, and shared enterprise.
Get full access to this article
View all access options for this article.