After two hundred and fifty trials on problem 2 had been given Julius, it seemed desirable to introduce a radical change in method in order to stimulate him to maximal effort.  It was therefore decided to force him to make a round trip through the apparatus in connection with each choice, and to let this forced labor serve, in the place of confinement, as punishment for mistakes.  This new method yielded peculiar and characteristic results.  They differ from those previously obtained largely because of the orang utan’s remarkably strong tendency to reenter the box through which he had just passed.  This occurred so persistently, as may be seen in table 9 (June 17, second series, June 18, etc.), that a further modification of method was introduced in that after the same wrong box had been entered five times in succession, the experimenter on the next choice of the box confined the animal for a stated interval, say sixty seconds, in it, and then allowed it to escape by way of the exit door and choose repeatedly until it finally located the right box.  Were it not for this particular feature of the method, the number of choices recorded after June 17 would unquestionably be very much greater than the table indicates.

The new method proved a severe test of the orang utan’s patience and perseverance, for he had to work much harder than formerly for his reward, and often became much fatigued before completing the regular series of ten trials.  Early in the use of this method, he developed the habit of rolling around from exit door to starting point by a series of somersaults.  When especially discouraged he would often bump his head against the floor so hard that I could hear the dull thud.  As has been noted, I found it desirable to vary the procedure repeatedly.  It proved especially interesting to give one series per day with the round trip as punishment and another series with confinement as punishment.

Day after day, as the experiment progressed, slight or great fluctuations of the ratios of right to wrong choices appeared, but without consistent improvement.  There was, to be sure, as the last column of table 9 shows, a radical improvement during the first six hundred and fifty trials, for the number of right choices per series increased from 0 to 8.  But, as the observations were continued from day to day, it became more and more evident that the animal was merely passing from tendency to tendency—­method to method—­mixing tendencies, and occasionally developing new ones, without approach to the solution of the problem.  This fact would have led me to discontinue the work much earlier than I actually did had it not been for the peculiarity of the results obtained with problem 1.  It seemed not improbable that at any time Julius might succeed in perfectly solving this problem over night precisely as he had solved the first problem.