EMPLOYMENT PSYCHOLOGY 
If the new methods prove to be more efficient and are 
recommended for adoption, he installs them as a part of the 
employment procedure. He also assures himself from time 
to time that they are being used properly, and checks their 
predictive accuracy frequently in order to adjust them if 
necessary to changes in industrial demands or type of 
applicant. 
Not every research in vocational selection will require that 
all of these 11 steps be included in the program. The 
investigator will, however, be more apt to avoid pitfalls if 
he examines his outline of proposed procedure critically with 
each of these steps in mind, and assures himself that pro- 
vision has been made for every stage of investigation essen- 
tial to the sound progress of his particular study. 
We are primarily concerned with the scientific measure- 
ment of aptitudes for an occupation. The method outlined 
may also be used -in developing differentiating tests for 
placement. It can be used in distinguishing between several 
occupations which differ in the degree of difficulty of their 
work. The method may be employed, for example, in 
validating tests for placing office help at either typewriting, 
stenography, or secretarial work; or at filing or comptometer 
work. 
The examinations of which this manual treats in greatest 
detail are psychological tests. The accuracy with which a 
test measures an ability to succeed in a job varies with the 
extent to which it calls out the same or similar types of 
adjustments as the job. If there is a close relationship 
between test scores and success at the job, disturbing vari- 
ables being held constant, the assumption may be made that 
the relationship is caused by the fact that the two call out 
the same ability or abilities. 
A causal connection between two variables is merely the 
observed concomitant variation of the two. A causal rela- 
tionship between test performance and success at the job 
may be inferred if the correlation between them is high. 
Even this modicum of theory may be omitted. If the corre- 
Q