TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 11 



dissimilar sections was 71.7 per cent as compared with 

 89 in the identical part. The same method applied to 

 maze E discloses that the average number of errors in the 

 dissimilar sections was .71 and the average saving 85 per 

 cent as compared with .11 and 99 in the identical part. 

 It may be remarked that the smaller average savings for 

 the three dissimilar sections were due in the case of both 

 mazes principally to the low records of the fourth section; 

 to which a defender of this method might reply that the 

 low gains in the fourth sections may have resulted from 

 the same cause as the high gains in the identical parts: 

 namely, the presence of the identical part, which exerted 

 more of a beneficial effect in the learning of other sections 

 than it did in the learning of the fourth. But we have 

 seen reason to believe that the smaller gain in the fourth 

 section of the B maze was due to a peculiarity of its pattern 

 in connection with the maze-habit acquired in A; and the 

 smaller gain in the fourth section of E as compared with 

 D may be explicable according to the same principle. 



The Dependence of Transfer upon the Relative Difficulty 



of the Two Mazes 



The query is a logical one, whether as between mazes 

 of different degrees of difficulty it matters from the point 

 of view of savings if the more difficult or the less difficult 

 is learned first. 



Two of the transfers in this experiment were from the 

 harder to the easier maze, judging by the records of the 

 control-groups; and four were from the easier to the harder. 

 In the case of one of the transfers from the harder to the 

 easier, the saving in errors was less than in any of the 

 transfers from an easier to a harder maze, the saving in 

 trials was less than in two of these transfers, and the saving 

 in time was also less than in two of these transfers, the 

 computations being on both the basis of the average total 

 number of errors and the average total time per rat and 

 the basis of the average errors and the average time per 

 trial. In the case of the other transfer from a harder 

 to an easier maze, the savings in trials and in errors were 

 greater than in any of the transfers from an easier to a 



