Criterion-related validity examines whether test scores predict or relate to relevant external outcomes (criteria). Predictive validity refers to forecasting future performance; concurrent validity relates to current outcomes. Correlation coefficients, regression coefficients, and utility analysis quantify these relationships.
You've studied the reliability-validity relationship and know that validity comes in multiple forms, each answering a different question about what a test measures. You've also worked with linear regression, which lets you quantify the relationship between a predictor and an outcome. Criterion-related validity brings these concepts together in the most practically grounded form of validity evidence: does this test actually predict something that matters in the world?
The question criterion validity asks is concrete. If you have a cognitive ability test for job applicants, does it predict job performance? If you have an anxiety measure, does it predict who responds to treatment? Criterion-related validity is quantified as the correlation (or regression relationship) between test scores and a separate, meaningful outcome measure — the criterion. A test with high criterion validity is genuinely useful; one with low criterion validity, however theoretically motivated, gives you little practical traction.
Two forms are distinguished by timing. Predictive validity tests whether scores forecast future outcomes: administer the test now, wait, then measure the criterion outcome months or years later. The classic example is SAT scores predicting college GPA — a forward-in-time relationship. Concurrent validity measures the relationship between test scores and a criterion collected at the same time, such as a depression scale correlated with current clinician diagnosis. Concurrent validity is faster and cheaper to establish; predictive validity is usually more important, because the practical value of a test in selection or screening contexts is its ability to forecast, not just correlate with current standing.
Your regression background applies directly here. The validity coefficient — the correlation r between test and criterion — tells you the direction and strength of the relationship. But r² (the coefficient of determination) tells you the proportion of criterion variance accounted for, which is the more interpretable effect-size metric. A validity coefficient of 0.40 sounds substantial but accounts for only 16% of criterion variance. Utility analysis then asks a practical question: even a modest validity coefficient may justify using a test if the stakes are high, selection is competitive, or errors are costly. The economic value of a selection instrument depends jointly on the validity coefficient, the base rate of success in the population, and the selection ratio — how many positions there are relative to applicants.