CEREBRAL FUNCTION IN LEARNING 93 



The tune required for solving the problem in successive trials 

 during training by the animals in the different groups is sum- 

 marized in table 3. Some of the animals learned the problem 

 quickly and training was discontinued with them soon afterward. 

 If their records are dropped out at the point where they finished 

 practice the end of the learning curve will represent, not the aver- 

 age for the group, but the average of only those animals which 

 learned slowly. To avoid this the time for the perfect records 

 of each animal that finished early is included in the computation 

 of the average for each period of later training with the group. 



Judged by the time alone, the normal annuals are somewhat 

 superior to the operated ones, yet the superiority is by no means 

 significant until after the fiftieth trial. The normal animals were 

 but slightly superior to groups 3 and 5 in the time required for 

 the perfect records (last period of training) and never equaled 

 the record of group 4. Time records alone are probably never 

 an accurate measure of the rate of learning and in this case they 

 are apt to be particularly misleading since some of the paretic 

 animals moved slowly throughout the experiments and, even 

 when they made fewer errors, consumed more time than the 

 normals. 



The tune records, then, give no accurate measure of the relative/ 

 abilities of the groups. They do show roughly, however, that 

 there is no very significant difference in rate of learning between 

 the normal and operated animals. 



The use of the problem box permits of a certain amount of 

 variation in the standard of perfection required. Errors were 

 recorded when the rat deviated from his usual path, but a good - 

 bit of latitude was allowed in the choice of that path. He might 

 go in front of, across the top of, or behind the problem box in 

 passing from platform a to b. So long as a stereotyped path was / 

 followed, learning was considered perfect. This allows some * 

 chance for the personal equation of the experimenter to modify 

 the results. A partial check upon this source of error is provided ^ 

 by experiments upon the rate of formation of a habit of visual 

 discrimination in which a definite invariable standard of error 

 could be adopted. A description of this experiment follows. 



