DISTRIBUTION AND ELIMINATION OF ERRORS IN MAZE 149 



for representative stages. The first horizontal column states 

 the correlation values for the various mazes between the order 

 of elimination of the alleys and the relative number of entrances 

 made by the group as a whole for the first trial. The second 

 column gives similar data in relation to the total number of 

 errors made during the second and third trials. The last three 

 columns state the results in reference to the total number of 

 errors made in the first five runs, the second five runs, and 



4 



from the eleventh trial until the maze was mastered. 



TABLE II 



Correlations Between Quickness of Elimination and Number of 

 Errors at Different Stages of Learning 



Trials I-a I-b I-c I-d I-e Il-a Il-b III IV 



1 -.571 -.821 -.535 -.714 -.892 -.321 -.642 -.654 -.133 



2-3 -.535 -.643 -.892 -.857 -.643 -.107 -.321 -.300 -.166 



4-5 -.642 -.571 -.821 -.535 -.500 -.642 -.964 -.764 -.350 



1-5 -.714 -.892 -.750 -.857 -.857 -.143 -.642 -.758 -.116 



6-10 -.857 -.857 -.678 -.750 -.857 -.857 -.571 -.518 -.583 



11 -until learned.. -.500 -.785 -.857 -.535 -.321 -.928 -.000 -.973 -.866 



With one exception, a uniform negative correlation was ob- 

 tained, the values ranging from —.107 to — .964. The excep- 

 tion refers to the final stage in the mastery of maze I I-b. This 

 record hardly invalidates the uniformity of the results as this 

 maze was mastered very easily and but few errors were made 

 after the tenth run. In considering the validity of this correla- 

 tion, the uniformity for all mazes and for the various stages of 

 learning must be emphasized. A single correlation value may 

 well be due to chance in view of the paucity of data for which 

 the correlation was calculated. But chance is pretty well elim- 

 inated when the computation is repeated 54 times and consistent 

 results are secured. We may then safely conclude that some 

 degree of negative correlation obtains for all stages of learning 

 between quickness of elimination and the tendency to enter the 

 cul de sacs, that in a general way those blind alleys which for 

 some reason offer little inducement to entrance are easily elim- 

 inated, while cul de sacs which present the most enticement to 

 exploration are the hardest to master. This correlation is a 

 priori logical, for it is reasonable to expect that the strongest 

 tendencies will be the hardest to overcome. Our results thus 

 establish one of the factors underlying the order of error elimi- 



