next up previous contents
Next: References Up: MPR-online 1996Vol.1, No.4 Previous: Recommendations Summarized

Some Final Remarks

All other things being equal predictions of a, say, linear trend have a greater empirical content (as understood by Popper, 1980 [53]) than predictions of a trend that is 'only' monotonic, because linearity implies not only a particular rank order for parameter values but also precisely specifiable distances among them, whereas (strict) monotonicity only refers to the rank order without specifying any distances. From this perspective, the preference for quantitative trends is compatible with Popper's demand to focus on the theories and hypotheses with the greatest empirical content possible (Popper, 1980 [53]). Whether this demand is adequate for the domains of psychology has already been questioned elsewhere (Hager & Westermann, 1986 [27]). And Popper himself has pointed out, too, that conceptualizing theories which are more and more precise and more likely to be falsified empirically are not conducive to scientific progress (Popper, 1981, p. 244 [54]).

The repeated reference to Popper as a philosopher whose methodology is continuing a matter of debate should not be interpreted as meaning that the testing strategies discussed in the present paper serve the better falsification of substantive statements. My focus is on the examination of psychological hypotheses and whether the empirical data fit them or not. If a hypothesis is valid or 'true' the probability that it is 'confirmed' should be high, and if it is not valid or 'false' the probability that it is 'disconfirmed' should be high. The lower the validity of the study, in the sense used by Cook and Campbell (1979) [], and the poorer the correspondence between predictions derived from the hypothesis and the statistical methods and tests applied, i.e., the lower the hypothesis validity (cf. Wampold et al., 1991 [63]), the lower the probabilities of correct decisions concerning the psychological hypothesis will be, all other things being equal. Furthermore, the probabilities of correct decisions will be lowered if the derivation of (psychological and statistical) predictions do not take the empirical content of the hypothesis into full account, that is, if the predictions and statistical partial hypotheses are not derived appropriately and exhaustively. Since adequate explanations and descriptions of psychological phenomena can only be achieved through hypotheses and theories which have passed valid empirical tests successfully, there is no good reason to continually try to falsify these hypotheses, as strict falsificationists would demand. The better and more general rule demands to plan and execute experiments in a way that gives hypotheses a good chance to be 'confirmed' if they are 'true' and that leads to a high probability of 'disconfirming' them if they are 'false'. Overall, it can be said that correct decisions are more likely the higher the validity of the experiment (see also Westermann, 1988 [64]). Considerations like these also seem valid in the realm of 'applied' psychology: More is gained if one knows that an intervention program is effective than if one knows that it is not. Since intervention research, as one possible example, can also be designed as examining psychological hypotheses referring to effectiveness (see Hager, 1995 [24]), there is no great difference between testing hypothesis about phenomena in basic psychology and testing hypotheses in 'applied' or technological psychology, though hypotheses serve different aims in both realms and their theoretical background may be quite different. In both instances, however, predictions can and should be derived from them which refer to the same statistical constructs and which can be submitted to the same statistical testing strategies and tests. If the psychological hypotheses are precise, the same holds for the predictions, and if they are imprecise, also the predictions are less precise.


next up previous contents
Next: References Up: MPR-online 1996Vol.1, No.4 Previous: Recommendations Summarized

Methods of Psychological Research 1996, Vol.1, No.4
© 1997 Pabst Science Publishers