140 PSYCHOLOGICAL METHODS OF TESTING INTELLIGENCE 



mean that we must combine together tests that cor- 

 relate less with one another than each one of them 

 correlates with estimated intelligence, and that com- 

 bination whose amalgamated rank-order shows the 

 highest and most constant correlation with estimated 

 intelligence is the system of tests that we seek. Nat- 

 urally, we shall not limit ourselves to two tests in 

 making our system, but shall combine a larger num- 

 ber into one compensation-system. 



This was the idea that incited Hylla to the investi- 

 gations previously mentioned which are still in prog- 

 ress. The idea of compensation as a principle in as- 

 sembling tests has already arisen simultaneously 

 both in England and in France. 



Thus, from the tests that had afforded the highest 

 correlations with estimated intelligence in his in- 

 vestigations, Burt worked out an amalgamated rank- 

 order whose correlation with estimated intelligence 

 considerably exceeded all the single correlations 

 (Table XVIII). In the elementary school the single 

 correlations ranged between 0.52 and 0.76, that of 

 their combination amounted to 0.85; in the higher 

 schools the single correlations ranged between 0.54 

 and 0.84, while the correlation for the combined tests 

 rose to 0.91. From this Burt draws the conclusion 

 (pp. 158-9) : "By means, then, of some half-dozen 

 tests, we are able independently to arrange a group 

 of boys in an order of intelligence, which shall be de- 

 cidedly more accurate than the order given by 

 scholastic examination, and probably more accurate 

 than the order given by the master, based on personal 

 intercourse during two or three years, and formu- 

 lated with unusual labor, conscientiousness and 



