A stylus maze was constructed such that all possible paths seemed to lead without obstruction from the starting point to the finish, whereas in reality all paths but one were blocked by invisible stops.In the first part of the experiment one group of Ss was permitted one min for a visual inspection of the maze before beginning trials with the maze screened from view; four other groups were allowed to see the maze during trials one, two, three and five, respectively; while group six learned the maze entirely without the co-operation of sight. The data from the last group were used as norms of comparison. In subsequent parts of the experiment variations in the kinds of visual guidance were introduced. Comparisons were also made with the results of two Ss totally blind from birth.Visual control was more effective in eliminating errors than in reducing the number of trials. Preliminary visual inspection was detrimental when measured in terms of trials, but beneficial in reducing the number of errors. Trials with visual guidance were effective in reducing the amount of error. The introspective results showed that visual memory operated advantageously during trials where the maze was not seen. The earlier the introduction of visual guidance, the greater was its effect upon the error score. Indirect visual cues tended to eliminate error, especially when maps enabled the S to determine the true path. The scores for trials of the two blind Ss exceeded the average range of normal variability, but lay within the maximal range. The error and time values, however, exceeded the maximal range considerably. The results as a whole indicated that visual modes of control were much more effective in the acquisition and performance of acts of skill than hitherto suspected. From Psych Bulletin 19:07:00478. (PsycINFO Database Record (c) 2006 APA, all rights reserved)