STIMULI DURING DELAYED REACTION IN DOGS 277 



compartment. Only three of the light trials on No. 2 obtained 

 correct reactions. This seemed to indicate that the cues to 

 No. 2 were almost lost. In the "Two Light" experiments 

 (discussed in the next section) No. 1 and No. 3 compartments 

 had been discriminated after a delay period of much longer 

 duration, but here the cues to each of the three compartments 

 did not seem to be localized well enough to be differentiated 

 after they had had the chance to become indistinct during the 

 delay period of forty-five seconds. The cues to No. 2 had so 

 lost their individuality during this long delay that the dog 

 could not differentiate them from those of No. 1 and No. 3 in 

 even a "chance" number of attempts. As a desire was felt to 

 increase the number of discriminated objects, and not to increase 

 the delay period on the three compartments, no more trials 

 were given on this type of experiment. 



"Two Light Experiments." 



By November 19th, 1913, the experimenters saw T that the 

 dog had not received proper training to do well at the three 

 light problems, and that success there would mean spending 

 much longer periods of training on each phase than was at 

 their command, so it was decided that a thorough training on 

 the two lights would gain time and also give a test for com- 

 parison with the results of Hunter (13) on the length of delay 

 possible to obtain with two lights. Another cause for the adop- 

 tion of the two light method was the fact that the trials so far, 

 had shown that the cues to No. 2 compartment were very weak 

 and could easily be dropped out. The method used in the 

 "Two Light" trials were slightly different from those in the 

 "Three Light" types. From the beginning, the stimulus was 

 retained at each bowl after a mistake until the dog made the 

 correct response. He was also kept at each type until each 

 series showed five successive correct reactions, before the diffi- 

 culty of the problem was increased. These methods, it was 

 hoped, would bring the association between stimulus and re- 

 action to the correct compartment, much more strongly into 

 mind. Previous training had failed to strongly set up this 

 association. After the two light association was thoroughly 



