dramatically altered performance in comparison with reproduction data, showing significantly better perceptual ability than would have been suspected from the reproduction data alone.
Garner, Hake, and Eriksen ( 1956), in discussing how response processes can affect the nature of response rather than the perceptual system, note the importance of the set of responses E provides S. In one example, they note how the number of response alternatives afforded S may affect the amount of error in discrimination. That such an issue has direct relevance to research with retardates can be noted in the study of Ohwaki ( 1953). Ohwaki was interested in examining how retarded Ss discriminate weight when size of the stimuli is held constant. He first had them compare 3-, 15-, and 30-gram boxes. He noted that all of his Ss could successfully discriminate between the 3- and 30-gram boxes. He then required them to arrange, in order of weight, three, four, and five such boxes, all differing in weight by 30 grams, a difference all were capable of discriminating when only two boxes were involved. No S with an MA under 4 could perform any of these tasks. With the higher-MA Ss, the task was harder the more boxes involved, and the higher the MA, the better the performance with any one set of boxes. Even in his highest-MA group (6.0 to 11.0) 2 out of 12 failed with three boxes, 3 out of 12 with four boxes, and 7 out of 12 with five boxes. It is clear that these failures did not reflect the inability to discriminate a difference of 30 grams, but rather other response characteristics which were elicited by the way the task was constructed.
The issue that probably demands the greatest vigilance, offers the most subtle confounding potential, and yet is the most difficult to control and manipulate is that of S's set to the experiment as a whole and the tasks given in particular. If a crucial element in perceptual research is the contingent relationship between stimulus and response, it behooves E to maximize the probability that S understands what he is to do and is attending and responding to the stimulus as defined by E. S enters with certain expectations regarding success and failure, he establishes a relationship with E, and he wonders how he is doing and what will happen to him as a consequence of what he does. Initial work has suggested that retarded Ss do have different expectancies regarding success and failure and that this affects their response to success and failure in the experimental situation ( Cromwell, 1959). The findings of Stevenson and Zigler ( 1957) and Keller ( 1958) indicate that both BI and NBI retarded Ss are sensitive to the investigator as a stimulus to which they respond and from whom they extract cues as to how they "should" perform. In all likelihood, it is the rare study that includes in its printed form the various exceptions to the formal instructions necessary in actually preparing S. Perhaps we all owe it to our colleagues to draw as complete a picture as possible of what was actually said to S, and of those difficulties encountered in working with him.
The studies completed to date reveal a number of varied and interesting findings and raise even more challenging questions. However, the paucity