General intelligence (g) represents variance common to mental tests, which largely explains the predictive power of tests for work and school criteria. In contrast, non-g factors represent unique variance of tests (after removing g) and generally have negligible predictive power. An exception is the non-g residuals of the SAT and ACT, two college admissions tests. The non-g residuals of the SAT and ACT are unrelated to g but still predict school and work criteria (grades and jobs). Whereas prior studies have examined the non-g residuals of individual tests (SAT and ACT), the current study is the first to systematically examine the predictive power of non-g residuals of group factors. Group factors are based on multiple tests and measure specific abilities (e.g., math or verbal). In general, group factors should yield more accurate estimates of non-g effects than individual tests (e.g., SAT and ACT), which are loaded with unique test-specific variance. In the current study, non-g residuals of group factors predicted math/STEM criteria and verbal/humanities criteria (test scores and college majors). Predictions were based on investment theories, which assume that non-g factors reflect investment in specific abilities. Such theories predict that investment in a specific ability (math) boosts similar abilities but retards competing abilities (verbal), yielding negative relations between competing abilities (math and verbal).
Data were obtained from the 1997 National Longitudinal Survey of Youth, a representative sample of US youth (N = 1950). The 12 tests of the Armed Services Vocational Aptitude Battery (ASVAB) estimated two academic abilities (math and verbal) and two non-academic abilities (speed and shop skills). All abilities were based on three or more tests (e.g., math ability was based on numerical reasoning, word problems, and equation solving). The ASVAB abilities were residualized after removing g (based on all tests) and were correlated with four criteria: college majors in STEM and humanities; jobs in STEM and humanities; SAT/ACT test scores; and SAT/ACT ability tilt scores. Tilt scores were based on the difference between math and verbal scores, which yields math tilt (math>verbal) and verbal tilt (verbal>math). STEM included engineering, computers, and physical sciences; humanities included English, history, and foreign languages. Effects are reported as standardized coefficients (betas). Significant effects are reported at p < .05.
Consistent with investment theories, the non-g residuals of the ASVAB academic abilities (math and verbal) yielded a domain-specific pattern. Math residuals correlated positively with math/STEM criteria (SAT/ACT math scores, math tilt, STEM majors and jobs) and negatively with verbal/humanities criteria (SAT/ACT verbal scores, verbal tilt, humanities majors and jobs). Verbal residuals showed the opposite pattern. The effects of the math and verbal residuals were significant and strong for all criteria (|M|=.51, range=-.99 to .81). In contrast, the effects of the non-academic residuals (speed and shop) were nonsignificant and weak (|M|=.07, range=-.24 to .40), demonstrating divergent validity.
The results are the first to demonstrate the predictive power of non-g residuals of group factors (based on multiple tests) for diverse criteria. The findings support investment theories. Such theories assume that non-g residuals reflect investment in specific abilities (e.g., math/STEM), boosting similar abilities and inhibiting competing abilities (e.g., verbal/humanities). Future research should examine whether non-g effects increase over time (with continued investment), and whether such effects are influenced by other factors (e.g., scholastic interests, course choices, ability level).
