Two approaches of statistical testing are critically reviewed. A new approach, which is a hybrid of the two, is proposed. The new approach requires the researcher to think about the two types of potential inferential errors and an explicit alternative hypothesis of interest. (VM)