EMPLOYMENT PSYCHOLOGY 
age, or failure—there should be an appropriate proportion 
of subjects in each group. 
If the criterion is a continuous variable, such as quantity 
of output or average school mark, a representative dis- 
tribution of workers throughout the range of this criterion 
should be selected. If all the men engaged in the occupa- 
tion were to be judged by the criterion which has been 
fixed upon, the resulting distribution would resemble the 
normal probability distribution; that is, a great many 
workers would cluster about the average of the group, 
while fewer and fewer cases would be found as the lower 
and higher extremes of ability were approached. The 
problem of the investigator is to discover useful methods 
of measuring those abilities of the workers which are related 
to the criterion, so that those who stand high in the mea- 
surement of the ability will as a general rule also stand high 
in the criterion, the measure of actual accomplishment on 
the job. It is important then to obtain test records on 
workers known to represent various degrees of vocational 
accomplishment expressed in terms of the criterion which 
has been adopted. 
If only successful, competent workers are available at the 
time, the alternatives are to test also a wide sampling of 
the population not engaged in that occupation, to see 
whether significant differences can be found between these 
highly selected workers and the general run of the work- 
ing population; or to accumulate over a long period of time 
test records on new employees as they are hired. 
The investigator should bear in mind when choosing his 
subjects, that a wide range of vocational ability will tend 
to increase the correlations with the tests, and that a large 
number of cases will tend to increase the reliability of these 
correlations. It is impossible to make a generalization re- 
garding the number of cases necessary. Investigators have 
generally hesitated to publish results where less than 30 
cases are used. But it is to be borne in mind that the num- 
ber of cases required depends on the reliability of both 
Jo)