16 STELLA BURNHAM VINCENT 



by a more gradual decline the significance of which has been 

 fully discussed by others. Aside from these two main features, 

 the great variability of the curves prevented any generalization 

 from them, save the mere fact of variability. The graphic 

 method as used in other sciences has usually been applied to 

 masses of statistics of which the curve was the general concrete 

 exponent and later investigators seizing upon Thorn dike's 

 method have grouped their results and shown them in general 

 curves plotted from many animals. 



To my mind it is manifestly unfair to call this curve a learning 

 ctirve, What the curve really shows is physiological and environ- 

 mental conditions which affect the whole group from day to day. 

 But often the curve is plotted from data obtained from groups 

 of animals which have been used at widely separated intervals 

 so that it does not have even this daily significance. 



A problem is said to be learned when it can be performed 

 relatively free from error. A coordination never becomes 

 entirely free from error and the degree of excellence must be 

 based upon a standard artificially set, dependent upon the con- 

 ditions and the complexity of the problem. Because of their 

 great variability one animal may learn a problem in ten trials 

 which may take another forty. Suppose these animals are given 

 fifty trials and the numerical results either of time or errors are 

 plotted in a curve. Something is obtained then which is very 

 unlike either of the original records. The automatisms of the 

 one come in to modify the errors of the other, the errors of the 

 one to disturb the automatisms of the other. Taken individually 

 up to the place where the problem is learned the curves if plotted 

 separately for each animal are much alike except that the irregu- 

 larities in the one due to the process of learning are distributed 

 over a greater length of time. A curve has been plotted here 

 which seems to be a truer representation of the learning process. 

 It does not exhibit daily conditions as does the other nor does it 

 show the actual length of time taken which may easily be shown 

 in other ways, but it is believed it does demonstrate that under 

 similar experimental conditions practically the same course is 

 taken by animals in learning a problem whether it takes them 

 ten or fifty trials to accomplish the task. 



In this experimental work an animal's trials were discontinued 



