TRANSFER OF RESPONSE IN THE WHITE RAT 5 



much complexity of objective factors and a lack of accurate 

 analysis and definition of the relation of the two problems 

 in terms of similarities and differences. 



It seems to the writer, therefore, that if the experiments 

 upon this problem are to have any finality and secure for 

 us the solution of the problem they will have to begin with 

 such conditions as will either keep the response as a whole 

 constant and vary the stimulus or situation one factor at 

 a time, or keep the stimulus or situation as a whole con- 

 stant and vary the response one factor at a time. From 

 some such simple beginning we could then pass, as experi- 

 ence and technique indicate, to tests where more than one 

 factor of the stimulus or situation was varied while the 

 response remained constant, or to tests where more than 

 one factor of the response was varied while the stimulus 

 or situation remained constant, or to tests where one or 

 more factors in the stimulus or situation and one or more 

 factors in the response were varied while all others in the 

 stimulus and in the response remained constant. Most 

 experiments which have been performed would presumably 

 fall under the fourth class mentioned, although no careful 

 analysis, especially of the response side, has been made in 

 any of them, and in many no careful analysis of the stimu- 

 lus or situation. We have already cited Bogardus and 

 Henke's tests. For further example, how much, and what 

 change is made in the stimulus and how much and what 

 change is required in the response in passing from one of 

 the boxes used by Yoakum to another, or from one of 

 Hunter's mazes to another? Shall we, following Thorn- 

 dike's principle of identical elements, decide upon the 

 amount of likeness or difference of the two problems by 

 the amount of transfer shown when one is learned after 

 the others? The true course would require first such an 

 analysis and control of factors as is suggested above and 

 a long series of experiments in which control was present, 

 in order to furnish the only sound basis for Thorndike's 

 principle, or any other. 



It appears to the writer, moreover, that after all, the 

 problem back of the transfer of learning finds one of the 

 best points of attack, as suggested above, in the form in 



