On Severe Tests of Trend Hypotheses in Psychology

Abstract

The distinction between scientific and statistical hypotheses is resumed. It is stressed that psychologists are usually interested in scientific rather than statistical hypotheses, and that statistical tests are, for the most part, performed in order to arrive at a decision about the psychological research question or scientific hypothesis. By means of two examples it is argued that various statistical hypotheses can generally be derived from a scientific hypothesis under study. The problem which of these statistical hypotheses should be chosen and tested is discussed showing that the choice should be based on considerations concerning the validity of the empirical study, the error probabilities connected with the evaluation of the scientific hypothesis, and the subsequent principle: Always test those statistical hypotheses that lead to the severest scrutiny of the scientific hypothesis.