DISTRIBUTION AND ELIMINATION OF ERRORS IN MAZE 153 



during each run the greater number of errors will be made in 

 the final cul de sacs in these three mazes. It is also possible 

 that the same principle, a shift of motives, will account for the 

 greater frequency of entrance into the initial cul de sacs in 

 the normal mazes; in this case the shift will be from fear and 

 curiosity over to the food odor. The odor of food as the rat 

 reaches the final sections should be sufficiently directive as to 

 minimize the number of errors relative to those made when the 

 animal's acts are controlled primarily by curiosity and caution. 



3. The distribution of errors is influenced by certain peculiar- 

 ities of the cul de sacs other than their positional relation to the 

 point of entrance. Since the influence of this factor may vary 

 according to the stage of mastery, a separate treatment is neces- 

 sary for its effects upon the initial and final distribution of 

 errors. 



a. Initial distribution. — From the data of table III it is obvious 

 that the initial errors (first five trials) are never distributed in 

 the exact order of the spatial arrangement of the cul de sacs. A 

 few blind alleys are generally responsible for the deviations, and 

 these are listed in table IV. The first columns give the various 

 mazes with the number of cul de sacs belonging to each. Each 

 cul de sac is numbered in order from the point of entrance. In 

 the column headed ' plus ' are listed those alleys in which 

 the number of entrances exceeds that to be expected on the 

 basis of a perfect correlation. In the " minus " column are those 

 alleys in which the number of entrances is less than the normal. 

 The first group of mazes are those in which a positive correla- 

 tion obtains between the distribution of errors in the first five 

 trials and the proximity of the cul de sacs to the point of entrance. 

 In the last group of mazes the initial errors are distributed 

 proportionate to the spatial contiguity of the cul de sacs to 

 the food box. The five mazes, I-a, I-b, I-c, I-d, and I-e, are 

 identical in pattern; in the first three the sensory conditions 

 were such that the greater number of errors were made in the 

 initial cul de sacs; in the last two, the first cul de sacs were the 

 least attractive. In spite of this radical difference of error dis- 

 tribution in the two groups, the same cul de sacs are responsible 

 for the deviations from a perfect correlation in every case. The 

 first cul de sac is much less alluring in the initial trials than its 

 position would justify. On the other hand No. 6 was invari- 



