112 PSYCHOLOGICAL METHODS OF TESTING INTELLIGENCE 



course, be developed here. 1 We merely point out 

 here that a correlation = 1.00 means that there is 

 complete correspondence between the two rank-or- 

 ders (or groups). A correlation = means com- 

 plete absence of correspondence. The size of the 

 decimal fraction between and 1, then, shows the 

 degree of correspondence. The probable error 

 (P. E.) is a measure of reliability: only when the 

 correlation amounts to at least three times the size 

 of its probable error is a real significance to be 

 ascribed to it. In details the methods used by differ- 

 ent investigators for calculating the coefficient of cor- 

 relation show considerable differences. The appen- 

 dix contains an example of the simplest method of 

 calculating rank-correlation as I have used it. 



All the different rank-orders that have been named 

 can be brought into correlation with one another : of 

 these possible correlations we shall have to deal for 

 our purposes with : correlation of test with test, cor- 

 relation of estimates with school performance, cor- 

 relation of tests with estimates. 



Correlations of tests with tests have been worked 

 out in particular by Spearman (77, 79, 80). If the 

 pupils of a class have been tested by means of sev- 

 eral different tests and the resulting rank-orders 

 show mutual high correlation, this is, in Spearman's 

 opinion, a sign that the capacity operative in the 

 tests depends upon a common factor (Spearman 

 uses the expression " general intelligence ' 9 or "gen- 

 eral ability"). This brings it about that Pupil A 



X A general exposition of these methods will be found in Betz 

 (70) and Stern (1, chs. 19 and 20). [Also in the translator's 

 Manual of Mental and Physical Tests, Ch. 3.] 



