SummaryWhen subjects act as their own controls in drug experiments there is a risk of asymmetrical transfer between treatment conditions, with the result that treatment effects are determined largely by order of administration. In such cases the effect of a drug may be reduced or exaggerated, and prior treatment with a drug may affect placebo conditions. Asymmetrical transfer is probably much more common than is generally realized. Suggested causes include inadequate wash-out between treatments, state-dependent learning, adoption of learning strategies, and practice effects. The problem is illustrated by the re-analysis of a paper on the effects of benzhexol on memory, published recently in this journal (Potamianos and Kellet, 1982). The advantages of separate-group experimental design are reconsidered.
Cue use in state-dependent learning