BEHAVIOR OF VERTEBRATES 439 



Type D: This involves the error of making more than one 

 separate continuous effort to open a given door during the same 

 trial, but always with an interruption of such repetitions of 

 activity by an interval of effort to open one or more of the 

 other doors. The searching tendency modified by recrudescent 

 motor impulses: regularly increases as we descend the phyletic 

 scale. 



Type E : Includes several different modes of behavior which 

 have a common objective characteristic, viz., automatism. This 

 is a relatively implastic and unadaptive mode of behavior. The 

 tendency towards perseveration of active impulses and inhibitions: 

 increases regularly when both phylogenetic and ontogenetic 

 scales are descended. 



Bogardus and Henke (3) studied in the rat the role played 

 by head and nose contact in learning the maze. The vibrissae 

 were cut off before experimentation began in order to facilitate 

 the observation of the actual head and nose contacts made at 

 the corners of the maze. It was found that the percentage of 

 corners touched is high at first, and then gradually decreases 

 as the maze is learned. There is a high correlation between the 

 number of contacts and the number of errors made, and between 

 the increase and decrease of time and the number of contacts. 

 When the maze is learned, contact at corners is no longer neces- 

 sary. Blind rats (previously trained on another maze) were 

 then required to learn the maze used in the first set of experi- 

 ments. The results obtained were similar to the above, and 

 add weight to the contact theory of acquiring kinaesthetic and 

 organic cues. This general conclusion is further strengthened 

 by experiments in which the pathways of the maze were altered 

 in various ways. 



Hicks (14) gives a timely and exceedingly valuable discussion 

 of the relative values of the different curves of learning. The 

 author holds that the distance curve is a better representative 

 of the progress towards automatic accuracy than the error curve. 

 It, however, is impracticable from the standpoint of recording 

 and manipulating the data. The prevalent practice of omitting 

 all total and partial returns from the error record, and of making 

 no attempt to evaluate varying degrees of error, gives a curve 

 which is not only worthless but false. It is suggested that an 

 error curve which includes returns, and which is constructed on 



