132 



HELEN B. HUBBERT AND K. S. LASHLEY 



In the individual records we can not be certain either that a 

 single correct turn made by an animal in the maze proves that 

 the animal has learned the turn, or that a single error proves 

 the absence of the habit for the particular turn involved. The 

 former may be due to chance, the latter to a distracting stim- 

 ulus whose threshold of reaction is lower than that of the maze- 

 habit. In a question of this sort the most dependable results 

 are therefore to be obtained from averages, which rule out, to 

 a certain extent at least, the chance successes and errors. It 

 may be that while individual records do not show an exact pro- 

 gressive elimination of errors, the larger averages do reveal this 



Figure 1. — Ground plan of circular maze, showing the position of the errors studied. 



as one of a number of factors influencing the course of learning. 

 There is some doubt as to the comparability of the different 

 alleys in the circular maze and no great significance can be 

 attached to the progression except in the case of alleys b, c, d 

 and e, which are strictly comparable. 



Vincent ('15, IV) has given data which indicates a more pro- 

 nounced progressive elimination than appears in the work of 

 Hubbert. In the form in which this is presented, however, 

 Vincent is not justified in applying the data to the problem. She 

 makes no statement as to what constituted a trial in her experi- 

 ments and in the two ' ' typical records ' of trials which she 

 reports in detail ('15, I) we find that the rat, after reaching the 

 food, was allowed to return and re-explore the maze. In the 

 first trial of her rat " No. 1 " the animal was allowed to leave 



