STUDIES or CliiEEBKAL FUNCTION IN LEARNING. VII i 



operation. The reliability of the method for quantitative work 

 must therefore be considered. 



The discrimination habit provides only two criteria of 

 learning, the number of trials necessary to reach some stand- 

 ard of achievement and the number of errors made during 

 this number of trials. Two methods of estimating the relia- 

 bility of a measure have been employed. Hunter and Heron 

 ('23) have computed the correlation in the scores of indi- 

 vidual animals at different stages during training on the 

 assumption that individual differences in learning ability 

 will constantly influence the scores at different stages of 

 learning. This method is inapplicable in the present case, 

 since the discrimination method gives data only at the com- 

 pletion of learning. A second method employed by these 

 writers compares the learning records of the same animals 

 on different problems and determines the reliability of a 

 method by the consistency of its results with those obtained 

 by other methods. This is laborious and not altogether satis- 

 factory, since we have as yet no certainty that individual 

 differences in learning ability are constant for different types 

 of problems. A third method of estimating reliability, fully 

 as valuable as these, may be applied here. If individual 

 variations in learning, as determined by a learning test, are 

 found to correlate highly with any other variable which itself 

 is not a function of the learning test, this correlation is evi- 

 dence for the reliability of the test. A negative result is, 

 of course, meaningless by this criterion, but a constant posi- 

 tive correlation, if statistically valid, is convincing evidence 

 of the reliability of the method. Several quantitative studies 

 have been made with the discrimination apparatus. That of 

 Dodson ('17) is the most extensive. It is internally consis- 

 tent and agrees in general with the results of other writers 

 (Yerkes, Hogue and Stocking, and Cole),^ so that it justifies 

 the use of the measure, although it gives no indication of its 

 fineness. 



' For references see Dodson, '17. 



