ABSTRACT
A replicated ABBA design is a well-known psychological design in which the average stage number is constant for two treatments, A and B. For example, with four stages, the A condition in a ABBA sequence has
(l ; 4) = 2.5 as an average stage number; the B condition has (2; 3) = 2.5 as well. The original purpose of such designs was to give an equal average amount of practice effect at each treatment when performance under it was tested. If one could assume that practice effects were linear and carryover effects did not distort treatment mean differences, then the difference between average performance under A and average performance under B would be the difference between effects of Treatments A and B, uncontaminated by differences in practice effects.
