On the basis of two studies, one published in this issue (4 ) and the other unpublished, Starker expresses doubt as to the validity of our beliefs regarding the part attention plays in the propensity for experiencing autokinesis, and the relationship of autokinesis to some aspects of personality organization. In his unpublished study it appears that he used questionnaire data as a measure of S's personality qualities. Since such data are highly subjective (people tend not to see themselves as others see them), it does not surprise me that he could not duplicate our findings. There are difficulties with the present study which require comment. There appears to be a misunderstanding of our theoretical model and there are some methodological weaknesses. Starker believes he has gotten his Ss to withdraw attention from the light and direct it inwardly. If S is really thinking about something else-unrelated to the light-it is probable that autokinesis might cease or be reduced, not increased as he predicted (his Group A ) . Paying closer attention to the light, thus taking attention away from the spatial frame of reference should increase autokinesis, not decrease it as Starker predicted (his Group C) . He has omitted any reference to the frame of references in his design. Asking Ss continuously to generate phantasies about the light would be a better test of our theoretical model (3 ) than having them think about something unrelated to the light. The test conditions for Group C are a good test of our model provided S does not also concentrate (unknowingly) on the test room (frame of reference) as he concentrates on the test light. Combining Condition C with a condition wherein S is asked to produce phantasies about the light would best test our model. Starker correctly notes the possible difficulty inherent in his condition of internal rehearsal. His idea that distracting auditory stimuli should reduce autokinesis is consistent with our theoretical model. He must be certain that Ss pay attention to the distracting stimuli throughout the test period. We have observed that high movers may remain oblivious to very loud noises outside the test room apparently because of the degree to which they are absorbed by test experience. I suggest using moderate movers in the experimental and control groups. For very high movers the autokinetic effect is powerful, and such persons also have a remarkable ability to ignore distractions. Very low movers are in reality no movers most of the test period and unless the distracting stimulus is continuous, the distracting event may not coincide with the experience of autokinesis. Starker's use of a difference score as the dependent variable (change), makes it difficult to guess what the results would have been had a more appropriate score (or analysis) been used ( 2 ) . Regression effects make a difference score questionable as a measure of change and are particularly likely to occur in groups selected for their extremity as in this experiment ( 1 ). In addition, when Ss cannot be randomly assigned to the (High-Low AK)-(Experimental Condition) treatment combinations, a split-plot analysis is more appropriate than a factorial design.