The SAT was designed as a college admissions test, not an IQ test. Empirical work over the past two decades has shown that the two are closely related — close enough that, with adjustment, an SAT score can serve as a respectable estimate of a person’s general cognitive ability. The most-cited result in this literature is Frey and Detterman’s 2004 finding of a corrected correlation of .82 between SAT scores and a g-factor derived from a battery of cognitive tests. That number deserves to be unpacked, qualified, and read against the broader evidence on what the SAT measures.
The Frey and Detterman (2004) findings
The 2004 paper in Psychological Science ran two complementary studies. The first analyzed data from 917 participants in the National Longitudinal Survey of Youth 1979 (NLSY79), who had taken both the SAT and the Armed Services Vocational Aptitude Battery (ASVAB). From the ASVAB the authors derived a general cognitive ability factor (g) and correlated it with SAT scores. Corrected for nonlinearity in the score scaling, the correlation was .82 — the principal number from the paper.
The second study used a sample of undergraduates who had taken both the revised SAT and Raven’s Advanced Progressive Matrices (APM), one of the most widely used non-verbal measures of fluid reasoning. The correlation, corrected for the restricted range of an undergraduate sample, was .483.
The authors concluded that the SAT functions, to a substantial degree, as a measure of general cognitive ability, and they proposed conversion equations allowing researchers to estimate IQ from SAT scores. This procedure has since been used in research contexts where direct IQ testing is impractical — for example, when historical SAT records are available but no IQ data exists.
How to read the .82 number
Two qualifications are essential for interpreting the headline .82 correlation:
- It is corrected, not raw. The .82 figure comes from a correction for nonlinearity that adjusts for the SAT’s score-distribution shape. The uncorrected correlation is lower. Whether the correction is appropriate is a methodological judgment; most subsequent work has accepted it.
- It depends on the g extraction. The g factor in study 1 was derived from the ASVAB. Different test batteries produce different operational definitions of g. The .82 estimate is specific to the SAT–ASVAB-g relationship in NLSY79.
The .483 correlation with Raven’s APM in study 2 is more conservative and arguably more transferable to other populations because the sample is undergraduates, the cognitive measure is a well-known fluid-reasoning test, and the correction is for restricted range only — not for nonlinearity.
The full picture, then, is that the SAT is heavily g-loaded, with the strength of the relationship varying with sample, correction method, and the specific g operationalization. .82 is the upper end of plausible estimates, .48 the more conservative end. Both establish that the SAT is, in substantial part, an intelligence measure dressed in admissions-test clothing.
The non-g component: additional variance the SAT measures
Calling the SAT “a measure of g” is not the same as saying it measures only g. Subsequent work by Coyle and colleagues has shown that the SAT carries information beyond general cognitive ability:
- Coyle and Pillow (2008) in Intelligence showed that, even after statistically removing g, SAT and ACT scores retain meaningful predictive validity for college GPA. The SAT therefore predicts academic performance both through its g component and independently of it.
- Coyle, Purcell, Snyder, and Kochunov (2013) further demonstrated that the non-g residuals of SAT and ACT subtests predict specific abilities — verbal, mathematical, spatial — that are domain-relevant for college coursework.
This is the more sophisticated picture of the SAT: a heavily g-loaded test that also captures verbal and quantitative aptitude beyond g. The .82 correlation indicates what the test mostly is; the Coyle results indicate what the test also is.
The cognitive ability layer matters for outcomes
Why it matters that the SAT is so g-loaded is that g is one of the most consistently predictive variables in social science. Roth, Becker, Romeyke, Schäfer, Domnick, and Spinath’s 2015 meta-analysis in Intelligence examined the relationship between intelligence and school grades across hundreds of samples, finding correlations in the range commonly reported for cognitive ability and academic achievement (population-corrected estimates around the .50s). Cognitive ability also predicts adult outcomes including occupational attainment, income, and aspects of health.
What this means for SAT interpretation: the strong link between SAT and g is not a curiosity but a load-bearing feature. To the extent the SAT taps g, it inherits g’s broader predictive validity for outcomes far beyond the freshman-year GPA the test was designed to predict.
Durability of the SAT–performance link over time
A common concern is that admissions tests measure abilities relevant only to the first year or two of college, after which other factors take over. Dahlke, Kostal, Sackett, and Kuncel’s 2018 longitudinal analysis in Journal of Applied Psychology examined this directly. By tracking both test scores and college performance over time, they tested whether the predictive validity of cognitive tests degrades through the college years. The general finding is that the validity is durable — the SAT does not become uninformative for sophomore, junior, or senior performance. The cognitive-ability signal carries through.
Practical implications for researchers and parents
A few practical points emerge:
- SAT scores can be used to estimate IQ, with caveats. Frey and Detterman’s conversion equations are the standard approach. Estimates derived this way are reasonable for group-level comparisons and as control variables; for individual diagnostic claims they are not a substitute for a direct cognitive assessment.
- The SAT is not “just” a coachability test. The strong g loading means a substantial portion of SAT performance reflects underlying cognitive ability that is not easily moved by short-term preparation. Test prep has effects, but they are typically modest in magnitude.
- SAT performance is meaningful for outcomes beyond admissions. Because g predicts a broad set of life outcomes, an SAT score carries information about more than freshman GPA. This is part of why test-optional debates have been contentious: removing the SAT removes a fairly informative cognitive signal from the admissions process.
- The SAT is not a complete intelligence test. It samples primarily verbal and mathematical reasoning. It does not assess working memory in detail, processing speed, or spatial cognition independent of math. Treating it as a comprehensive cognitive battery would overstate what it covers.
What this evidence does not establish
Several boundaries are important:
- The conversion is not exact. Frey and Detterman’s equations produce point estimates of IQ from SAT scores with non-trivial standard errors. Treating a converted IQ as a precise individual measurement overstates its precision.
- Group differences in SAT performance are not direct evidence about group differences in g. Test scores reflect cognitive ability, exposure, education, and the conditions of test-taking. Inferring causes of group differences from group means is a much harder task than the simple SAT–g correlation suggests.
- The SAT continues to evolve. The redesigned SAT, the digital SAT, and the move toward test-optional admissions have all changed the testing landscape. The .82 correlation derives from earlier SAT versions and may shift modestly as the test changes.
- Predictive validity differs by use case. The SAT predicts college GPA, but its predictive validity for graduate school admission, career performance, or life outcomes follows different patterns and requires separate evidence.
Frequently Asked Questions
Is the SAT really an IQ test?
Substantially, yes. Frey and Detterman’s 2004 study found a corrected correlation of .82 between SAT scores and a general cognitive ability factor in a nationally representative sample. Subsequent work has confirmed the SAT is heavily g-loaded while also containing some non-g information.
Can I convert my SAT score to an IQ score?
Frey and Detterman published equations for doing this. The converted IQ is a reasonable estimate but carries non-trivial measurement error. For research and group-level comparisons it is useful; for individual diagnostic decisions it is not a substitute for a direct cognitive assessment.
Why does the SAT correlate so strongly with IQ?
Because the SAT samples verbal reasoning, mathematical reasoning, and reading comprehension — all heavily g-loaded abilities. When a test taps multiple g-loaded subdomains, it inevitably becomes a strong overall measure of g.
Does test prep change SAT scores enough to break the IQ link?
Test prep produces measurable but modest improvements in SAT scores. The effects are typically smaller than the standard error of measurement of an IQ test, so prep does not meaningfully invalidate the SAT–g correlation at the group level.
Is the SAT a good measure of intelligence for everyone?
The SAT–g correlation is robust on average but the relationship can be weaker for individuals with test anxiety, language barriers, or specific learning differences. For these individuals, an SAT score may underestimate cognitive ability that a direct, untimed assessment could detect.
What about the ACT?
The ACT shows comparable g loadings to the SAT. Coyle and colleagues’ work on non-g residuals has covered both tests. For purposes of cognitive-ability estimation, the two are largely interchangeable.
Does this mean colleges should weight the SAT heavily?
That is a policy question rather than a psychometric one. The empirical evidence is that the SAT is informative about g and predicts academic outcomes through that loading. Whether to use that information, and how heavily, depends on values and goals colleges hold beyond pure prediction.
References
- Frey, M. C., & Detterman, D. K. (2004). Scholastic Assessment or g? The Relationship Between the Scholastic Assessment Test and General Cognitive Ability. Psychological Science, 15(6), 373–378. https://doi.org/10.1111/j.0956-7976.2004.00687.x
- Coyle, T. R., & Pillow, D. R. (2008). SAT and ACT predict college GPA after removing g. Intelligence, 36(6), 719–729. https://doi.org/10.1016/j.intell.2008.05.001
- Coyle, T. R., Purcell, J. M., Snyder, A. C., & Kochunov, P. (2013). Non-g residuals of the SAT and ACT predict specific abilities. Intelligence, 41(2), 114–120. https://doi.org/10.1016/j.intell.2012.12.001
- Roth, B., Becker, N., Romeyke, S., Schäfer, S., Domnick, F., & Spinath, F. M. (2015). Intelligence and school grades: A meta-analysis. Intelligence, 53, 118–137. https://doi.org/10.1016/j.intell.2015.09.002
- Dahlke, J. A., Kostal, J. W., Sackett, P. R., & Kuncel, N. R. (2018). Changing abilities vs. changing tasks: Examining validity degradation with test scores and college performance criteria both assessed longitudinally. Journal of Applied Psychology, 103(9), 980–1000. https://doi.org/10.1037/apl0000316