As opposed to the constant-effect procedure, the”constant-treatment”method would allow each subject a fixed number of practice trails in which to learn as many of the correct responses as he could. His score would then be the number of nonsense syllables mastered in that many trials.
The chief weakness of the constant-treatment procedure is our ignorance of the characteristics of the units in which the dependent variable is measured.
What does it mean to say that a subject has learned eight of twelve words in a list when using one practice procedure, but only six of twelve when using another procedure ?
At most, we can conclude that the first method is the more efficient, but we cannot carry out such manipulations of the data as are involved when we attempt to say how much better is one method than the other.
By way of contrast, the dependent variables in the constant-effect procedure usually are measured in units whose characteristics are well known, and the experimenter does have some confidence that his manipulations of the data (adding, dividing, and so forth) are valid ones. Of course, many problems in psychology are still at the stage of development where it is necessary to discover whether Treatment A produces any effect at all, or whether A has a greater or lesser effect than Treatment procedure will continue to find applications.