FACTORS IN LEARNING BY WHITE RATS 347 



Any differences in such matters, such as are found in all actual 

 mazes, will not affect adversely the explanation as a general 

 principle; they can only disturb the accuracy of our predictions 

 based on it in any given case. 



What, then, can be said of this theory, that pure probability- 

 frequency factors explain the rat's learning in the maze? In 

 the first place, there is an inconsistency in the explanations 

 based on the theory, which on closer examination seems seriously 

 to limit the validity of the theory; and, secondly, rats do not 

 as a matter of fact learn the maze in agreement with the expec- 

 tations of the theory. 



Let us take up the first objection. Assume that a rat on the 

 first tiiai enters cid de sac 1, emerges from it and continues 

 forward in the right direction, that it passes 2 and enters 3. 

 Let us suppose that on coming out of 3 it takes the return direc- 

 tion. Frequency, then, will favor its entering 1 again, if it 

 continue past 2, which would be a matter of probability. Cn 

 emergence from 1 again it should go forward and once more 

 enter 3. Pure frequency (and also recency) effects would there- 

 fore lead us to expect the animal to continue running from 1 

 to 3, 3 to 1, etc., indefinitely. As a matter of fact other con- 

 trol factors develop to throw the animal out of this routine. 

 These may be either changing external circumstances or they 

 may be varying intra-organic factors. These conditions, then, 

 become the directive factors in the learning. When the animal is 

 thus thrown out of the trial conditioned by frequency, prob- 

 ability may again determine the course in the part of the maze 

 beyond the third cut de sac, but here again frequency effects 

 would serve only to fix another circular type of response. Fre- 

 quency does not seem to operate toward the elimination of 

 errors, but only to stereotype action. 



It may be argued that the effect of frequency (and recency) 

 is not strong enough thus to fix wrong acts in the beginning of 

 the learning process. More scope might thus be left for the 

 operation of mere probability law r s to show themselves in the 

 early trials of the animal. This, however, would be tantamount 

 to saying that learning did not take place in the early trials. If 

 frequency is the determiner of the learning, the rate of its opera- 

 tion does not alter the case. Such an assumption, moreover, 

 is contrary to fact; learning effects are marked in the first and 



