Great question!  But before we get to that, let's take a little test, shall we?

 

If you only had enough resources to create an assessment having one of these forms of validity, which would you choose?​  Hover your cursor over your choice.   (OK, it's not a realistic scenario, but you can play along anyway...) 

Great question!  But before we get to that, let's take a little test, shall we?

 

If you only had enough resources to create an assessment having one of these forms of validity, which would you choose?​  Hover your cursor over your choice.   (OK, it's not a realistic scenario, but you can play along anyway...) 

Thanks for playing!  

 

If you ended up hovering over more than one answer, you found out that only predictive validity has any correlation at all to on-the-job performance. As far back as 1996, David Foster (then at Novell) presented data that demonstrated that a cognitive test of the NetWare Administrator certification test has correlation coefficients of 0.01 and -0.03 between scores on the test and the subjects' supervisoror's ratings of doing their jobs well, while the performance-based version of the test showed correlation cofficients of 0.62 and 0.71. Essentially, the cognitive-only test results were completely uncorrelated with the supervisor's ratings while the results on the performance-based test were very highly correlated.

​