TRAINING IN WHITE RATS UPON VARIOUS SERIES OF MAZES 61 



The fourth group learned the A maze, which was more 

 difficult than the D maze, in fewer trials than were needed 

 in the D maze, its saving in the A maze being 74 per cent 

 and in the D maze 64. The fifth group failed to learn 

 any of the intervening mazes in fewer trials than the re- 

 learning of the first maze called for, the nearest approach 

 to the five trials in the relearning being the seven trials 

 in C; but the C maze was much easier than the E. The 

 fifth group was the only one that furnished any evidence 

 in favor of the retention of the habit through a series 

 of four mazes. With the other four groups, the weight 

 of the evidence inclines to the other side. 



When we consider error-records, we find that here also 

 the weight of the evidence is against the retention of a 

 habit through a series of this number of mazes. The third 

 and fourth groups both learned mazes of greater difficulty 

 than the original maze with fewer average total errors 

 per rat than they made in the relearning of the original 

 maze. The first group made the low average total of 

 1.31 errors per rat in relearning its first maze; but it made 

 no errors in learning E. The second group made an average 

 3.97 per rat while relearning B, but it learned A with an 

 average of 1.22 and E with an average of .31. It saved 

 94 per cent in errors as compared with its former learning 

 of B, but it made 69 per cent fewer errors in A than in B, 

 although A was but 29 per cent easier. It learned E with 

 92 per cent fewer errors than it made in A, although E 

 was but 39 per cent easier. Coming to the fifth group, 

 it will be found that its error-record in E was far better 

 than its error-record in any of the other mazes of the 

 series; and with regard to errors, as well as to trials and 

 time, this group is the only one of the five that furnishes 

 evidence of the retention of the beneficial effect of the 

 original learning through four intervening learnings. 



Looking at the matter from the point of view of the time- 

 criterion, we find that the third and fourth groups learned 

 one or more mazes of greater difficulty than the first of 

 the series with a smaller time-average than the relearning 

 of the first necessitated; and that the second group learned 

 A, which was 7.8 per cent easier than B, in 37 per cent 

 less time than was needed in B. With the first and fifth 



