444 JOHN LJNCK ULRICH 



appeared before the latch, an ineffective neck reflex movement 

 was performed. At times increases are due to two or more 

 ineffective neck reflexes produced to raise the latch. When no 

 imperfect responses were noted, as was the case with two rats, 

 the increases and the decreases must result from no decided 

 fluctuations in the rat's reflex mechanisms interacting with 

 other body parts. 



With the occurrence of imperfect responses, it is difficult to 

 say when learning is complete, for it is difficult to obtain many 

 consecutive perfect records. Usually from 5 to 20 are produced. 

 The absence of imperfect responses is not always an indication 

 that learning or the development of interaction of body parts is 

 complete; for the first 15 perfect responses recorded in table 4 

 can not be regarded as a safe criterion that learning is complete. 

 Decreases in reaction time did not take place until the sixteenth 

 trial, and, if training had stopped at this time, there would 

 be no certainty as to what would have been the reaction time 

 after this trial, and whether interaction had been developed. 

 The best criterion indicating that learning is at an end, is, when 

 after a reduction in the reaction time for each trial, no further 

 reduction is possible. Only slight changes in time for each trial 

 appear. 



In this investigation of learning in the latch-box problem, 

 little support is to be had for prevalent theories of learning 

 based upon the working concept of " trial and error" or upon the 

 performance of forced movements or tropisms. The production 

 of forced movements is dependent upon the existence of a bi- 

 lateral symmetry existing in animals, and the majority of rats 

 are asymmetrically developed, showing an unequal antagonistic 

 action of the limbs. As a result forced movements can not be 

 considered in any theory of learning. In respect to other theo- 

 ries of learning, very little support for them can be said to exist. 

 It has previously been stated that upon the observation of the 

 first trial in learning the actual support for the " trial and error" 

 concept of learning with its twofold division of movements into 

 " successful" and "unsuccessful" rests. In the investigation of 

 the first trial in solving the latch-box problem, it is found that 



