Relations among general intelligence (g), aptitude tests, and GPA: Linear effects dominate

a r t i c l e i n f o This research examined linear and nonlinear (quadratic) relations among general intelligence (g), aptitude tests (SAT, ACT, PSAT), and college GPAs. Test scores and GPAs were obtained from the National Longitudinal Survey of Youth (N = 1950) and the College Board Validity Study (N = 160670). Regressions estimated linear and quadrat-ic relations among g, based on the Armed Services Vocational Aptitude Battery, composite and subtest scores of aptitude tests, and college GPAs. Linear effects explained almost all the variance in relations among variables. In contrast, quadratic effects explained trivial additional variance among variables (less than 1%, on average). The results do not support theories of intelligence (threshold theories or Spearman's Law of Diminishing Returns), which predict that test scores lose predictive power with increases in ability level or at a certain threshold. Suppose a large group of students takes an academic aptitude test (SAT or ACT). Further suppose that the ability level of the group ranges from very low (IQ b 70) to very high (IQ N 130). Should relations between the aptitude test and academic outcomes (e.g., college GPAs) decrease at higher levels of ability or beyond a certain ability threshold (e.g., IQ N 120)? The purpose of the current study was to address this question and its implications for intelligence. This study examined linear and nonlinear relations among general intelligence (g), college grade point average (GPA), and aptitude tests. The aptitude tests included the SAT (formerly, Scholastic Aptitude Test), ACT (formerly, American College Test), Preliminary SAT (PSAT), and Armed Services Vocational Aptitude Battery (ASVAB). These tests are used for selection in college admissions and the armed services. In addition, the tests are strongly related to g (e.g., β ≈ .78, Coyle & Pillow, 2008, Fig. 2), which reflects variance common to mental tests. A test's g loading (correlation with g) predicts outcomes in everyday life: Tests with strong g loadings strongly predict work and school performance , and tests with weak g loadings weakly predict such outcomes (Jensen, 1998, pp. 274–294). Predictions were guided in part by threshold theories of cognitive only up to a point. Once someone has an IQ of somewhere around 120, having additional IQ points doesn't seem to translate into any measureable real-world advantage. " The important point for the current study is that threshold theories predict that test scores lose (linear) predictive power …