DISTRIBUTION AND ELIMINATION OF ERRORS IN MAZE 151 



cul de sacs. All other factors being equal, we should expect 

 that the relative number of entrances into the various cul de 

 sacs will be roughly proportionate to their nearness to the point 

 of entrance. This influence of returns is further indicated by 

 two features of the data: a. In mazes I-d and I-e there is a 

 negative correlation between number of errors and nearness to 

 the entrance, the greater number of errors being made in the 

 final cul de sacs (table I). The opposite relation, however, 

 obtains for mazes I-a, I-b, and I-c, in which the greater number 

 of entrances were made in the initial cul de sacs. This differ- 

 ence in the distribution of the errors among the cul de sacs is 

 due entirely to differences of the sensory character of the mazes, 

 for exactly the same maze pattern was used throughout and the 

 objective environment of the mazes was identical in all cases. 

 Mazes I-d and I-e differed from the others in that an olfactory 

 trail was laid, either in the true pathway or in the cul de sacs. 

 This trail produced several characteristic peculiarities of beha- 

 vior, one of which was a noticeable diminution of the number 

 of returns. Miss Vincent kept no separate record of the number 

 of returns, but she noted this feature of the behavior and fre- 

 quently discussed its significance with the writer at the time. 

 The degree of returning also accounts for the different distribu- 

 tion of errors in mazes Il-a and Il-b (table III). Again the 

 same pattern was used and the maze was located in the same 

 objective environment in both experiments. The only difference 

 consisted in the presence and absence of sides to the runways. 

 Maze Il-b was without sides and their absence caused the ani- 

 mals to follow one edge of the platform with their paws or 

 vibrissae much in the same manner as did the olfactory trails. 

 Following an edge caused a noticeable diminution in the number 

 of returns and thus accounts in part for the difference of error 

 distribution in the two mazes, b. The returning tendency is 

 quickly eliminated in the course of maze mastery, and we should 

 thus expect that the relative frequency of entering the initial 

 cul de sacs will be gradually diminished during learning for 

 those mazes in which returns are an effective factor in error 

 distribution. Such a diminution obtains for mazes I-a, I-b, I-c, 

 III and IV (table III). In all five cases there is a high positive 

 correlation between number of errors and nearness to the en- 

 trance, but this correlation keeps decreasing with the number 



