Validity, fairness, and group differences - Tough questions for selection testing

Selection tests, such as those used for college admissions, present multipl e dilemmas for psychometricians, who grapple with intractable problems in m easurement, and the lay public, whose lives are affected by test scores the y often do not understand or trust. Criterion studies utilize convenient cr iteria that have Little meaningful significance, such as grades, and the va lidity coefficients are necessarily low because of range restrictions and t he low reliability of grades. Both supporters and detractors of college adm issions tests are correct in their assessment: College admissions tests can account for only a small percentage of the variance in success in school a nd in Life, but even a small reduction of variance substantially reduces un certainty and improves admissions decisions. Admissions testing can and sho uld be improved. Numerous suggestions for improving admissions testing are presented, including ways to reduce group differences without sacrificing c onstruct validity. The effect of the suggested changes on predictive validi ty is an empirical question and a question about fairness and values.