VALUES OF CURVES OF LEARNING 139 



On the other hand, Yerkes '' in the following passage is quite 

 as unqualified in his approval of the error method of estimating 

 learning. " Time records are not reported for these and subse- 

 quent labyrinth tests because they proved to be almost valueless 

 as measures of the rapidity of habit formation. At any point of 

 its progress through the labyrinth, the dancer may suddenly 

 stop to wash its face, look about or otherwise examine its sur- 

 roundings ; if a shock be given to hurry it along it may be sur- 

 prised into an error. It is my experience, and this is true of other 

 animals as well as of the dancing mouse, that alongtrip,as measured 

 in time units, does not necessarily indicate the lack of ability 

 to follow the labyrinth path accurately and rapidly. Hence, 

 whenever it is possible (and the experimenter can always plan 

 his tests so that it shall be possible), the number of errors should 

 be given first importance and the time of the tests second place." 



Apropos of our general problem, we wish to urge in the first 

 place that the relative value of any two criteria may depend 

 upon circumstances, and that no dogmatic statements applic- 

 able to all conditions can be made. Consequently, any general 

 conclusions we may reach in this paper must be understood 

 as limited to the white rats and to the problem of the maze. 

 Yerkes in the passage quoted suggests that the value of a cri- 

 terion may depend upon the nature of the animal used in the 

 experiment. However this may be, it is undoubtedly true that 

 the nature of the problem must influence the choice of a cri- 

 terion. Watson in the passage quoted apparently recognizes 

 this fact. The contention is evident from the consideration of a 

 few illustrations. In a maze with no ciil de sacs, the possibility 

 of an error curve (when errors are confined to entrances into 

 the cul de sacs, as is usually the case) is eliminated, and time 

 is the only criterion. It may be contended that such a hypo- 

 thetical maze will present no problem to an animal, and hence 

 that no learning process can exist to be represented. This con- 

 tention is emphatically disproven by an experiment conducted 

 in this laboratory by Misses Hybarger and Cowles with a maze 

 in which all cul de sacs were closed. Two groups of six rats each 

 were used. The average time for the first trial was 6.42 minutes, 

 and this time was gradually decreased for twelve runs. The time 

 results are represented graphically by curve II of fig. 4. The 



» Yerkes, R. M., The Dancing Mouse. New York, 1907, pp. 217-8. 



