64 



the effect on efficiency if rata which had learned the problem 

 were caiiaed to continue their runs in the maze for a long period 

 i.e., would continued i^ractise cause a marked increase in effi- 

 ciency evinced by a lowering of the absolute time rocord, or had 

 the highest possible level already been reached in the last six 

 trials of the learning process? 



To test the matter six tats of the sixty-five day group 

 were kept at work for more than one hundred sixty trials after 

 learning was complete, the average time for each six trials was 

 computed and appears in Table XI as twenty-seven tests. Taken 

 individually the results show that in every case a lower record 

 than the absolute time record was made, but in no case maintain- 

 ed. If the group average be noted, the absolute time is never 

 quite reached, the curve (Fig. 8) starting a little above it 

 and continually rising. In other words, final efficiency de- 

 creases rather than increases when practice is continued, ^n 

 interesting point is that errors will be made even after the 

 problem is learned. Of the six rats used, three made errors in 

 the first test of six trials after the problem was learned, one 

 in the second test, and one not until the fifth test. Lrrors 

 increased as the work was continued. During the last half of 

 the one hundred sixty additional trials twice as many errors 

 were made as in the first half. 



A closer examination of the table shows: Ist , that the 

 best record in each case was made during the first fourteen 

 testa; 2nd, that the last test was better than the first in 

 only one case (rat 15]; 3rd, that rats fourteen and fifteen 



