VALIDATION OF MEASURING INSTRUMENTS 189 
Opinions differ on interpreting the significance of the size 
of the correlation coefficient. Some authorities contend for 
a correlation of .60, while others claim that even if the cor- 
relation is as low as .23 it indicates an important relation- 
ship between the two variables. Rugg (157, p. 256) de- 
scribes the significance of different amounts of correlation 
in the following terms: 
less than .I5 or .20.....negligible or indifferent 
.I5 or .20 to .35 or .40.....present but low 
.35 or .40 to .50 or .60.....markedly present 
above .60 or .70.....high 
One cannot generalize about the size of the correlation 
coefficient in this way. The coefficient must be interpreted 
in each instance in terms of the total situation, the nature of 
the group, the thing measured, the reliability of each vari- 
able, and the size of the probable error of the coefficient. 
The following are some of the factors which account for 
low correlation between variables: 
Actual absence of relation; 
Restriction in the range of either variable; 
Failure to keep a third variable constant; 
Inaccuracy of the original measures; 
Loaded sampling. 
In practical use an analysis of the distribution of cases 
in the scatter diagram may reveal far more significant in- 
formation than the correlation coefficient. If there is any 
correlation at all, the scatter diagram should be analyzed in 
order to set off critical scores and critical sections. 
We have given consideration to the various methods 
whereby the extent of the general relationships between 
measurements of abilities and the criteria of vocational 
accomplishment are established. In the next chapter we 
shall deal with the methods of determining the probability 
of vocational success implied by definite test scores.