Asymmetric transfer in within-subjects experiments on stress interactions

Abstract

A comparison is made between the results of three similar within-subjects experiments on mild heat paired respectively with the arouser low-frequency noise, and with the depressants 1 mg of l-hyoscine hydrobromide, and the loss of one full night of sleep. The comparison shows that the level of performance in the control condition of a within-subjects experiment may depend upon the nature of the second stress. Bias in the results of the control condition may produce a reliable main effect, or change the sign of a main effect from positive to negative. The direction of transfer between the conditions of a within-subjects experiment may be positive or negative, depending upon the order of presentation of the conditions. The sign of the interaction between two stresses may reverse between the first and second halves of the experiment. The reliable main effects may be quite different from those of a comparable separate-groups design. Thus the transfer bias should always be measured in a within-subjects design. This can be done using separate groups.