122 PSYCHOLOGICAL METHODS OP TESTING INTELLIGENCE 



Personally, I do not find much to recommend in 

 the method. Even though most of the pupils can be 

 located without difficulty in one of the five groups, 

 there still remains a fairly large number of pupils 

 with whom the teacher remains in doubt as to which 

 of two adjacent groups they may belong to. The 

 final decision is then an arbitrary one, which, if often 

 enough repeated, can destroy the value of the whole 

 distribution. If, for instance, three such doubtful 

 cases are assigned to Group II the proportioning of 

 cases and the resulting correlations are quite differ- 

 ent from what they would be if these cases had been 

 assigned to Group III, where they might just as 

 properly be placed. It follows that so small a num- 

 ber of groups as this is entirely inadequate .for the 

 most important problems of correlation. Many of 

 our tests yield finely graded rank-orders of pupils, 

 and so it is to be desired that the estimated intelli- 

 gence with which these tests are to be related should 

 also take the form of a rank-order. 



The construction of a rank-order of all pupils 

 based on estimations of their intelligence in this way 

 naturally presents many difficulties. Many persons 

 hold it as a matter of course to be quite impossible. 

 But experience has shown that the task can be done. 

 The division into groups that we have just dis- 

 cussed can, indeed, be carried out as a preliminary 

 step ; then the pupils within each of the groups must 

 be further arranged in a scale, so far as practicable. 

 At the limits of the groups, however, care must be 

 taken, for it has been shown many times that shifts 

 will have to be made at these points, as, for instance, 

 a child originally assigned to Group II has to be 



