ESTIMATION AND TESTING OF FINER GRADATIONS 143 



(c) Fairly high correlation, even of the single 

 test, with estimated intelligence. Because tests that 

 of themselves exhibit little or no relation to intelli- 

 gence can not, of course, gain symptomatic signifi- 

 cance for intelligence by combination, however many 

 of them are combined. 



(d) Comprehensiveness of the tests, and that in 

 two directions. First, we should endeavor to bring 

 into action the different functions concerned in intel- 

 ligence (see above, pp. 20 f.). Secondly, we should 

 take care that the numerical records refer not only to 

 quantity, but also to the quality of the performance, 

 e. g., both to the number of units accomplished in a 

 given time and also to the percentage of errors made 

 during the work. 



(e) We should see to it that the estimation of in- 

 telligence be done thoroughly and conscientiously. 



(/) When a considerable number of tests have 

 been carried out upon a group, then combine the re- 

 sults into different amalgamated rank-orders until 

 that combination has been found that yields the 

 strongest correlation with the estimated intelligence. 

 The combination should then be tested out on other 

 groups. 



The construction of an amalgamated rank-order is 

 very easy. The ranks obtained by each subject in the 

 several tests are combined into an average value. 

 These average values themselves do not form the se- 

 ries desired, but must first be revised into ordinal 

 numbers that represent the final rank-order. 



Example : The pupils have been tested in three tests. The best 

 pupil has obtained in the three trials the rank-places 3, 1, 1, the 



