44 



E. L. THORNDIKE. 



pened to be pushing against the door. She repeated this and 

 formed the association as shown in the curve on page 34. She 

 had had experience only of escaping by pulling a loop of string. 



Out of 6 cats who were put in the box whose door opened 

 by a button, not one failed, in the course of its impulsive activity, 

 to push the button around. Sometimes it was clawed one side 

 from below ; sometimes vigorous pressure on the top turned it 

 around ; sometimes it was pushed up by the nose. No cat who was 

 given repeated trials failed to form a perfect association between 

 the sight of the interior of that box and the proper movements. 

 Some of these cats had been in other boxes where pulling a 

 loop of string liberated them, 3 and 4 had had considerable ex- 

 perience with the boxes and probably had acquired a general 

 tendency to claw at loose objects. 10, 11 and 12 had never 

 been in any box before. The curves are on page 19. 



Of two dogs, one, when placed in a similar but larger box, 

 succeeded in hitting the button in such a way as to let the door 

 open, and formed a permanent association, as shown by the 

 curves on page 33. No one who had seen the behavior of these 

 animals when trying to escape could doubt that their actions 

 were directed by instinctive impulses, not by rational observa- 

 tion. It is then absolutely sure that a dog or cat can open a 

 door closed by a thumb-latch or button, merely by the acci- 

 dental success of its natural impulses. If all cats, when hun- 

 gry and in a small box, will accidentally push the button that 

 holds the door, an occasional cat in a large room may very 

 well do the same. If three cats out of eight will accidentally 

 press down a thumb-piece and push open a small door, three 

 cats out of a thousand may very well open doors or gates in 

 the same way. 



But besides thus depriving of their value the facts which 

 these theorizers offer as evidence, we may, by a careful exami- 

 nation of the method of formation of these associations as it is 

 shown in the time-curves, gain positive evidence that no power of 

 inference was present in the subjects of the experiments. Surely 

 if I and 6 had possessed any power of inference they would not 

 have failed to get out after having done so several times. Yet 

 they did (see p. 43). If they had once even, much less if they 



