120 



CARNEGIE INSTITUTION OF WASHINGTON. 



"The left-hand first choices (Hght lines in figure 1) will be considered. In 

 the right-hand problem the tests and controls started with about equal left- 

 end choices, and reduced this number at about the same rate. In the first 

 20 trials of the left-hand problem, when the number of right-hand end- 

 compartments chosen first was suddenly reduced, the number of left-hand 

 end- compartments chosen first was suddenly increased. But after this original 

 increase, instead of showing still further improvement, the following set of 

 20 trials actually lowered the proportion of left-hand first choices; that is, 

 the continued training seems to have the opposite from the expected effect. 

 The tests (broken line) made fewer left-hand choices than the contrDJs, but 

 the same tendencj^ to go against the direction of the training is obvious in 

 both sets of rats. 



>ER CENT 



50 _ 



CORRECT DOOR^ 

 AT RIGHT 



CORRECT DOOR 

 AT LEFT 



PERCENT 



r 



5tf 



CHOices 



BEFORE 

 TRAINING 



3-4 5-6 7-e 9H0 11-12 13-14 15-16 17-18 19-20 21-22 23-24 1-2 3-4 

 DAYS IN TRAINING RETENTION 



Fig. 1. — Sho-n-ing the percentage of right-hand and left-hand end-compartments that were 

 chosen first in each successive group of twenty trials per rat. The heavy lines show right- 

 hand end-compartments chosen first, the light lines show left-hand compartments chosen first; 

 solid lines are the controls, broken lines the tests. After the twelfth day of training the cor- 

 rect door was made the one on the left end, instead of the one on the right end. 



"The retention trials show the tests making about as many right-hand and 

 left-hand choices as in their preliminary trials, although the number of right- 

 hand choices was larger in spite of the 120 trials immediately preceding, when 

 food was given only after a left-hand compartment was chosen. The controls, 

 on the other hand, show very plainly the retention of the tendency to go to the 

 left end-com-pai-tment more frequently than the right end-compartment. 



"When the cui-ves like the above were drawn for each rat separately, 5 

 controls and 8 tests were found that did not seem to learn at all. If these 

 rats are omitted from the summaries, the curves present the same general 

 situations, but an exception appears in that the tests in the left-hand problem 

 make fewer right-hand choices than the controls, instead of more, and there is 

 less difference in this problem between the numbers of left doors chosen by the 

 tests and controls. In other words, the elimination of these rats reduces 



