26 ROBERT M. YERKES 



he ceased to work. Consequently, obseirvations were inter- 

 rupted on the completion of trial 132, and on May 5, the series 

 was begun with setting 3. On this date, eighteen trials were 

 given in succession, and in only one of them did a mistake occur. 

 Since the -ten trials numbered 133 to 142 were correct, Skirrl 

 was considered to have solved problem 1, and systematic train- 

 ing was discontinued. 



On the following day, as a measure of the extent to which the 

 animal had learned to select the first door at the left no matter 

 what its position or the number of doors in the group presented, 

 a control series was given in which the settings differed from the 

 regular series of settings. These supplementary settings are 

 presented at the bottom of table 1 together with the records of 

 reaction in ten trials. 



Since in only six of these ten control settings was the first 

 choice correct, it is scarcely fair to insist that the animal was 

 reacting on the basis of an ideational solution of the problem. 

 Rather, it would seem that he had learned to react to particular 

 settings. A careful study of all of the data of response, together 

 with notes on the varied behavior of the animal during the 

 experiments, justifies the statement that Skirrl's solution of 

 problem 1 was incomplete and unreliable. It was highly de- 

 pendent upon the particular situation, or even the particular 

 door at the left end of the group, and slightly if at all dependent 

 upon anything comparable to the human idea of first at the 

 left of the group. 



This particular series of observations has been described and 

 discussed in some detail in order to make the chief points of 

 method clear. It will be needless, hereafter, to refer explicitly 

 to many of the characteristics of reaction or to the important 

 points in the construction of tables which have been mentioned. 



A graphic representation of Skirrl's learning process in problem 

 1 is presented in figure 18. The irregularities are most strik- 

 ing, and fairly indicate the erraticness of the animal. The 

 curve is based upon the data in next to the last column of table 

 1 , that is, the column presenting the errors or wrong first choices 

 in each series of trials. 



Unquestionably, the form of such a curve of learning should 

 be considered in connection with the method or methods of 

 selecting tl^e right box employed by the animal during the course 



