VALUES OF CURVES OF LEARNING 143 



period of time each day. Before the experiment the rats were 

 accustomed to being handled and fed in the food box in th? maze. 



In the time records, the preHminary period (^f nosing about 

 was eliminated. The time was recorded from the moment that 

 the}^ started out from the entrance box until they reached the 

 entrance to the food box. Time consumed in stopping during 

 the course of the run was not eliminated. 



All parts of the maze were divided into small segments of 6 

 inches, and all corners, runways, and segments were labelled 

 according to a logical scheme. This system of labels was out- 

 lined on a large chart which hung in plain sight, and in this 

 manner the experimenter was enabled to jot down an accurate 

 representation of each run in all of its devious details. The 

 error and distance records were computed from these symbolic 

 records. Over one hundred of these runs were reproduced 

 graphically in the manner depicted in figure i. These per- 

 manent graphic records are invaluable for the study of certain 

 qualitative features of the learning process. 



In computing the number of errors, some position must be 

 assumed as to the questions previously raised regarding the 

 definition of an error. We were guidqd by two general propo- 

 sitions : — 



I St. Errors shall include all total and partial returns as well 

 as entrances into the blind alleys. 



2nd. A runway, viz., the distance between two corners, was 

 taken as the unit of error. 



The first proposition is supported by the experiment of Misses 

 Hybarger and Cowles upon a maze with no cul de sacs. Returns 

 constitute the only possible errors. The experiment proves 

 that the elimination of the blinds constitutes only a part of the 

 maze problem. The number of returns made is given in table i, 

 and they are not so numerous in such a case as with a maze with 

 cul de sacs. Unless returns are counted, the error curve neces- 

 sarily does not fully represent the learning process. Observa- 

 tion of the rat's behavior indicates that these returns are an 

 essential and necessary feature of the learning. Apparently, the 

 animals sometimes become confused or lost, and they wander 

 about until by chance they return to a point in the true path 

 that gives them the proper cue. Their behavior strongly sug- 

 gests that they learn the mazq by segments, and after exploring 



