[Test strategies in evaluation of quantitative psychological hypotheses: the example of total learning time in...

In most textbooks on statistics and experimental design, the analysis of trend hypotheses is incomplete or at least not satisfactory from an hypothesis testing point of view. Quantitative psychological hypotheses usually predict that a particular trend component will be significant, while all other trends are expected to be absent. Only a conjunction of these two results, however, eventually confirms the psychological hypothesis. This fact is not addressed in popular textbooks, and so this article deals with some statistical testing strategies that can be used to examine functional psychological hypotheses via trend tests. Under some circumstances, however, the strategies discussed may fail when several psychological hypotheses are examined simultaneously. This failure can be avoided by adding a further test to the strategies which allows a comparison of predicted and actual correlations. The different strategies for testing trend hypotheses are then applied to the simultaneous examination of two simple quantitative psychological hypotheses which address the role of total presentation time and its pacing in text learning ("total-time hypothesis"). Although Bredenkamp (1975) has convincingly argued on a theoretical basis that the form most often encountered in the literature must be false, this variant has been discussed most often in various fields of psychology. Therefore, two experiments were planned and performed in which this well-known variant of the hypothesis is compared with a modification claiming a less steep linear function than the classical variant. The article shows how the necessary tests can be planned to prevent the cumulation of error probabilities from being too large. The experiments in which N = 180 and N = 216 students have to learn and recall until they have mastered a short prose passage, demonstrated convincingly that the classical form of the total-time hypothesis does not hold, whereas the modification can be regarded as confirmed.