The Case for Repeated Measures in Ergonomic Experiments: Another Look

Abstract

Repeated measures designs have been suggested as being inappropriate in most ergonomic research because of the presence of unwanted order effects which bias the results. Poulton has discussed a number of studies in several areas in which such order effects have rendered the interpretation of the results unreliable. The present study investigated order effects by combining in a single experiment a between-subjects design and a within-subjects design. Eight subjects were exposed to four combinations of two stressors, noise and ice-pressor, in a character recognition task based on reaction time measures. The experimental layout included order as an independent variable and found a significant main effect for treatments (F = 10.89, p < .005). More importantly treatments were found to interact with order (F = 3.26, p < .05). Because of the economy of repeated measures designs in terms of subject number, it is suggested that rather than categorically abandoning their use, research should be undertaken to determine: (1) if order effects are always present; (2) with what variables and combinations of variables they occur; (3) the consistency of such effects in terms of magnitude and direction; and (4) the practical as well as the statistical significance of particular order effects. In addition, the issues of individual differences and intra-subject consistency are discussed in relation to repeated measures designs. Finally, it is suggested that the discrimination of those situations where repeated measures designs can be used effectively is a more practical and fruitful activity than the wholesale elimination of repeated measures from erogomics research.