36 



Indiana University Studies 



ability to read connected discourse. Just what that relation is 

 is best shown by the coefficients of correlation for ability in the 

 two tests. These are shown for the several grades in Table 14. 



TABLE 14. — Coefficient of Correlation for the Vocab- 

 ulary AND Understanding of Sen- 

 tence Tests 



Grade 



8 



7 



6 



5 



4 



3 



Coefficient of correlation 



.22 



. 14 



. 15 



.34 



.56 



.65 



According to these figures, the score in one test is a fair measure 

 of ability in the other for the grades 3 and 4, but less so for each of 

 the succeeding upper grades. In grades 6, 7, and 8 the relation is 

 neghgible, and the figure of .34 for grade 5 is hardly great enough 

 to justify the estimate of reading ability on the basis of one test 

 alone. A coefficient of .65 (P. E. .1) in grade 3 (perfect correla- 

 tion would be 1.00) warrants us in regarding the score in one test 

 as a valid measure of ability in both. The ease with which the 

 vocabulary test can be given would make it a very important 

 test if it also tested the more complicated reading functions, 

 and the evidence of this very considerable amount of indirect 

 measurements is significant. It may be possible to construct 

 a scale for the measurement of vocabulary that will measure 

 many of the more complicated abilities adequately. If we could 

 construct a vocabulary scale that would correlate to a very high 

 degree, say to .85 to .95 with an Understanding of Sentences scale, 

 it would be sufficient to apply the vocabulary scale and get at 

 one stroke all that we now get thru the two scales. The conse- 

 quent reduction in the labor of measurement would be very great. 



The construction of two scales with this high degree of cor- 

 relation is not a simple task. Neither is it certain that we should 

 not want additional scales for more technical measures, but there 

 is good reason to believe that we could save enormously on the time 

 now consumed in the more complicated scales. 



In order to test out the validity of the scale, the scores of more 

 than 1,000 children were computed for each word in the scale. 

 The results are shown in Table 15. Clearly these words do not 

 have the same values for these Indiana children as they did for 

 the children originally tested by Thorndike.^ Thus, among 238 



7. Thorndike, E. L. "Meastirement of AbiUty in Reading." Teachers College 

 Record, XV. No. 4, p. 25. 



