350 JOSEPH PETERSON 



quency and recency. There is no situation in which frequency 

 factors operate alone with recency factors equally balanced, but 

 there are situations in which the reverse is true. E.g., suppose a 

 given animal has entered cul de sac 2 (either in forward or in 

 return runs, or both) twelve times, and also that it has passed 2 

 in the forward direction twelve times ; the last of the two possible 

 courses traversed is then the one favored by recency when 2 

 is approached by the animal going in the forwards direction. 

 Frequency factors weigh equally for the two directions. 



The following kinds of scores may be made by an animal at 

 critical points ■ Fr = reactions favoring frequency but contrary 

 to recency expectations; Rf = favoring recency, against fre- 

 quency ; R = favoring recency ; r — against recency ; B = favor- 

 ing both recency and frequency; and b = contrary to both these 

 factors. These symbols are used in the tables (II and III) 

 here given to illustrate the treatment of the choices in the first 

 three trials of each of the seventeen animals whose results are 

 found in a condensed form on later pages. Choices scored 

 " 1 " are uncritical choices. 



In these tables only the first three trials are tabulated. The 

 first trials are the important ones for the matter under considera- 

 tion, in as much as the greatest changes in the behavior occur 

 in the early part of the learning. The reason for this fact is 

 obvious. In the early random stages of the learning any one of 

 several single blunders is likely to throw the animal into con- 

 fusion and bring about returns which lead to numerous other 

 errors. It has been shown that the tendency to return is elimin- 

 ated comparatively early in the process of learning the maze. 11 

 This fact is evidently responsible to a considerable extent for the 

 very marked decrease in errors and time noticeable in the early 

 trials. In the individual tables of the seventeen rats completed 

 to the point of no errors in each record, it is found, as of course 

 would be expected, that frequency-recency choices gradually 

 increase in percentage, reaching at an early stage nearly 100%. 

 Complete learning always gives 100% of such choices. The 

 meaning of this final preponderance of frequency-recency reactions 

 is not that frequency and recency factors have brought about the 

 learning, but, more probably, that some other factors have finally 

 brought about the frequency-recency responses ! 



11 See note 5. 



