THE CRITERION OF LEARNING IN EXPERIMENTS 



WITH THE MAZE 



K. S. LASHLEY 



The Department of Psychology of the Johns Hopkins University 



In comparative studies of the rate of learning in which ani- 

 mals are trained in the maze the selection of a proper criterion 

 by which to judge the progress of habit-formation in different 

 groups of animals offers a rather difficult problem. There can 

 be little doubt that the ability to thread the maze without error 

 is the final test of learning, but whether a single trial without 

 error, three successive trials as used by Hubbert, or a still larger 

 number of errorless runs should be required before the habit 

 is considered as established has so far been determined largely 

 by the convenience of the experimenter. The question is chiefly 

 one of economy of the experimenter's time, but not wholly so, 

 for, although all animals may become automatic in running the 

 maze after long training, an occasional error still appears and 

 no method of evaluating these has been devised. 



In some tests dealing with the effects of drugs upon the rate 

 of learning I have recently trained 94 rats in the Watson cir- 

 cular maze, obtaining data which makes possible a limited com- 

 parison of such criteria of learning. 



The animals were all given five trials per day in the maze 

 with food at the end of each trial. At the beginning of the 

 experiments, as an arbitrary standard of ' perfect learning," 

 a single record of three successive errorless trials on the same 

 day was selected. After this degree of proficiency is once at- 

 tained the animals make very few errors, so that this standard 

 actually represents very nearly the limit of training, but it was 

 chosen simply because it could be attained after about ten 

 days' training. 



To test the reliability of this standard in estimations of the 

 difference beween groups of animals its results have been com- 

 pared with those of another standard, that of the number of 



