146 K. S. LASHLEY 



taken as the more trustworthy index to the amount of practice. 

 Since the total distance traversed in learning is greatly 

 influenced by the number of trials, irrespective of whether or 

 not errors are made in all the trials, the total distance traversed 

 during learning is not an accurate index to the number of errors 

 made. The total distance in excess of the direct path through 

 the maze is, on the contrary, closely correlated with the num- 

 ber of errors made, and may perhaps be looked upon as the 

 most accurate measure of error available since we can in no 

 other way evaluate the distances to which the animals penetrate 

 the blind alleys before returning to the direct path. 



The relative values of the number of trials and the number 

 of errors made as indices of the effort consumed in learning can- 

 not be determined at present. It is not necessary that a blind 

 alley be entered in order that the alternative entering of the 

 correct path be fixed in the habit. On the other hand it is 

 possible that the failure to obtain food in the blind alley is an 

 important factor in the fixation of the movements required for 

 running over the true pathway. The data on excess distance 

 are given therefore as contributory evidence, the value of which 

 cannot be determined at present. Records were made, also, of 

 the time taken on each trial and of the total distance traversed, 

 thus giving an index of the relative activity of the different 

 groups as determined by their rate of running, and also of the 

 approximate number of errors made by the different rats. 



The rdle of chance in determining the results. In every experi- 

 ment where the course of learning is studied in groups of sub- 

 jects trained under diverse conditions some control of chance 

 variations in the subject and in the conditions of training is 

 necessary. In very few of the recorded experiments of this 

 type, whether performed on man or animals, have enough sub- 

 jects been used to make the differences found significantly 

 greater than their probable errors, computed by the usual 

 formulae. Consequently, the results are suggestive rather than 

 conclusive. 



The significance of the results based upon few subjects may 

 be increased by a careful control of possible variable factors 



