THE WHITE RAT AND THE MAZE PROBLEM 149 



into a cul de sac and down went his nose to the trail which he 

 followed for its entire course, to the end of the alley. He moved 

 along by jerks, as described before, and when he reached the 

 end, he turned and in the same irregular, slow, halting way 

 returned to the entrance of the alley. Between the cul de sacs, 

 he ran; but when in them, slow movements were the rule. As 

 a result more time was spent in a single cul de sac than had been 

 the case in any of the other experiments. Still, from the first, 

 these excursions from the true path were lessened in number 

 as compared with the normal maze. The blind alleys seemed 

 to be marked for the animal in some way. He began to go less 

 and less deeply into them and finally, as he was running more 

 and more confidently in the true path, I have seen him, time 

 and again, actually thrown back on his haunches if chance 

 running flung him into the entrance of a cul de sac. Or, he 

 might be running quickly, swerve into an entrance, and there 

 would be seen an instant decisive turning the minute he struck 

 the trail. It looked like a real discrimination. Surprisingly 

 enough, however, after the problem was learned, and the animal 

 was making 90% correct trials, these errors began to reappear 

 and it took almost as long to get rid of them the second time as 

 it did the first. The meaning of this will be discussed later. 

 There were many more returns in this experiment than there 

 were in the one where the trail was laid in the true path — five 

 times as many in the first trial, It was a long time before the 

 rats learned to pass the food-box without lingering. The numer- 

 ical results for accuracy confirmed the conclusions drawn from 

 the observed behavior. 



2. Comparative Accuracy 



Under the conditions of this experiment, the accuracy was 

 decidedly greater than in the normal maze in the first fifteen 

 trials. If we now make a comparison with the other olfactory 

 experiment, we find that more errors were made in the first 

 nine trials than were made by the animals which followed the 

 trail in the true path but that the next six trials were more perfect. 

 From the fifteenth trial on, the accuracy was far less than in 

 Experiment 1, or in the normal maze, and it was only toward 

 the end of the experiment, that it again approached their standard. 

 The curve (Fig. 2) shows this variation exactly. The total 



