124 VINNIE C. HICKS AND H. A. CARR 



to cut down the values of each trial, it necessarily increases the 

 number of trials. It is probably true that, other things being 

 equal, the human can learn the maze in a fewer number of 

 trials than the rats, but that this direct connection of intelli- 

 gence with the number of trials is overpowered by their indirect 

 relation through the factor of " total values." This inverse 

 relation between the number of trials and total values is, of 

 course, not absolute; no one maintains that the adults would 

 have learned the maze in one trial provided they had bunched 

 all their fifty-eight errors in the initial attempt. The inverse 

 relation is not absolute because the element of " distribution " 

 is an effective factor. In all probability the fact that the rats 

 stand midway between the two human groups as to " rate of 

 elimination " is explicable in some similar manner. The " initial 

 drop " may be directly related to the rational processes as is 

 assumed by Thorndike and Hobhouse. Yet, on the other hand, 

 because the rational processes function to prevent any cut de 

 sac from being repeatedly explored, they necessarily prevent 

 the rapidity of this initial drop. As a consequence, this feature 

 of the curves is related to intelligent ability in two antago- 

 nistic ways. 



VII. CONCLUSIONS 



For the maze problem, there are only two features of the 

 numerical representatives of the learning process which can 

 serve by themselves as direct indices of intelligent capacity. 

 These are the surplus values of time and error, and the rapidity 

 with which errors are eliminated in relation to distance. 



These criteria are valid only for certain problems, and their 

 accuracy and delicacy as indices are limited to relatively gross 

 comparisons. Whether they serve best in estimations of rational 

 status, general intelligent capacity, or of the special ability 

 for motor acquisition is largely a matter of opinion. 



Any accurate index of intelligent capacity must consist of 

 some formula which shall give due weight to all factors : it must 

 represent the " distribution " as well as the amount of effort. 

 Such a formula must necessarily vary according to the problem. 



The complex interrelations of the various features of any 

 numerical representation of a learning process, the organic com- 



