As previously, the detailed results are presented in tabular form (table 4). From this table it appears that, whereas the expected ratio of right to wrong first choices for this problem is 1 to 2.5, the actual ratio for Sobke's first series was 1 to .67. This surprisingly good showing is unquestionably due to his marked tendency to choose the end box of a group; and this tendency, in turn, may in part be the result of the preliminary training, for during that only one box was open each time. But, if the preliminary training were responsible for Sobke's tendency, it should be noted that it had very different effect upon Skirrl, and, as will be seen later, upon Julius. The results for the ten different settings of the doors for problem 1 as they appear in table 4 are of interest for a number of reasons. In the first place, the setting 1. 2. 3 appearing twice,--at the beginning of the series and again at the end--yielded markedly different results in the two positions. For whereas no mistakes were made in the case of setting 1, there were fifty per cent of incorrect first choices for setting 10. Again, satisfactory explanation is impossible. It is conceivable that fatigue or approaching satiety may have had something to do with the failures at the end of the series, but as a rule, as is indicated by settings 1, 2, and 6, if correct choices were made at the beginning, they continued throughout the day's work. In this problem, Sobke's improvement was steady and fairly rapid, and in the eighth series, trials 71 to 80, only correct first choices appear. Consequently, seventy trials were required for the solution of the problem. This number is in marked contrast with Skirrl's one hundred and thirty-two trials. Immediately following the first perfect series, Sobke was given two series of control tests on April 28. Conditions were unfavorable, since the day was stormy and the rain pattering on the sheet-iron roof made a great din. Nevertheless, he worked steadily and well up to the sixth trial, which was preceded by a slight delay because of the necessity of refilling some of the food boxes. After this interruption, wrong choices occurred in trial 6. And again after trial 9, there was brief interruption, followed by wrong choices in trial 10. The ratio of right to wrong choices for this first control series was therefore 1 to .25. TABLE 4 Results for Sobke, _P. rhesus_, in Problem 1