x SOME EXPERIMENTAL RESULTS 237 



to take advantage of it. The elephant, for example, who 

 succeeded by a random pull in getting out the bolt at the 

 first trial, failed immediately afterwards. But when, after 

 being twice shown, it succeeded again, the trick was 

 established. Jack seldom failed more than once or twice 

 after learning apparently by perception. But consider the 

 case of the push-back bolt, which Jack learnt principally 

 by trial and error. Here there was a random success at 

 the third trial, followed by no less than twenty-nine 

 failures. A second success later on was also isolated, and 

 it was not till the forty-fourth trial that a third success 

 practically established the trick. I have referred above to 

 the collie, which no less than four times knocked the 

 string down by roaming about, but showed no inclination 

 whatever to pull it down by a deliberate act. In short, 

 these experiments seem to suggest that recent writers have 

 over-estimated the effect of pure accident. In any case 

 it would seem that, whether accidental or not, the success 

 is only effective if attended to. When due to some chance 

 movement disconnected with the efforts of the animal, 1 

 it seemed, as a rule, to leave no impression. 2 



1 As, e.g., the knocking down of the string by the cat with his tail. 



2 So far as they go the "double movement" tricks, if properly arranged, 

 tell in favour of learning by attention to the changes rather than by 

 association of movements. The most instructive of these experiments 

 are those in which an animal first learns an act A, and then learns that B 

 must be done to make A possible. Thus, the otter first learns to pull a 

 lever, then to draw a bolt or raise a catch in order that the pulling of the 

 lever may be effective. If this be explained as regressive association it 

 must be borne in mind that first the association of success with the act A 

 has to be dissolved the animal pulls at the lever and finds it of no use. 

 Then the association has to be reconstituted for A-in-the-circumstances- 

 created-by-the-performance-of-B, and then B has to be learnt. Now, all 

 this is accomplished by the animals in some cases in a very few trials. 



Instances of double movements learnt by Jack were the " loop " and 

 "skewer" and "pull-out bolt." It is noteworthy that where, as in the 

 " push-back bolt," he had great difficulty in learning the preliminary 

 movement, he would every now and then try the other out of its place. 



The cat got the order right in the "loop" experiment, but wrong in the 

 "skewer.'' He also got disheartened with the lever in the " push-back 

 bolt " experiment, and refused to try it. 



The elephant learnt bolt and knob, and gradually got perfect about the 

 order (see table). After learning the catch she was only once wrong as to 

 the order, but in the " push-back bolt " did not properly acquire it. 



The otter learnt some highly complex movements rapidly, but was 

 never " safe " as to the order. Subject to that limitation it may be said 

 to have learnt the bolt and lever in thirteen trials, catch and lever in one. 



