Specific Abilities vs g: What New Reviews Actually Weigh

A modern IQ test does not hand back one number. A test like the WISC-V reports a full-scale score plus several index scores — verbal comprehension, visual-spatial reasoning, working memory and others — each built from its own subtests. The question this raises is not new, but it has not settled: once you know the full-scale score, do those specific-ability scores tell you anything useful on top of it, or are they mostly the same signal measured five different ways? Recent reviews land in genuinely different places, and the disagreement traces less to new data than to which statistical method each one uses to ask the question.

What does "beyond g" actually mean here?

The g factor is the general ability that different cognitive tasks all correlate with — it is why someone who does well on a vocabulary test also tends to do reasonably well on a spatial-reasoning task, even though the two look unrelated. Every subtest score is partly g and partly something specific to that subtest. "Incremental validity" is the statistical question of whether the specific, non-g part of a subtest or index score adds real predictive power for an outcome like grades or job performance, once g itself is already accounted for.

What does the newest review find?

A 2025 review in the Journal of Psychoeducational Assessment, by Ryan J. McGill, Gary L. Canivez and Stefan C. Dombrowski, examined what is called the successive-level approach to test interpretation — comparing what full-scale scores, index scores and individual subtest scores each contribute when predicting school achievement. Their conclusion leans toward g: specific cognitive abilities did contribute some unique variance in achievement, but that contribution was described as limited, and frequently overshadowed by the global ability composite. In practice, that means the single full-scale score does most of the predictive work for school achievement specifically, with the more granular index and subtest scores adding comparatively little on top of it.

So why do other reviews reach the opposite conclusion?

A 2017 analysis in the Journal of Intelligence, by Harrison Kell and Jonas Lang, makes the disagreement itself the finding. Using the traditional incremental-validity method on the data they reviewed, specific abilities added only about 4.4 percent of variance in job performance beyond general mental ability — a result that reads as "g dominates," consistent with the newer review above. But re-analyzing comparable data with a different statistical technique, relative-importance analysis, the picture reversed: verbal comprehension outweighed g itself in predicting performance, and number and reasoning ability came close to matching it, depending on the job being studied. Kell and Lang's point is not that one answer is right and the other wrong — it is that the two methods are built to answer subtly different questions, and a study that only runs one of them reports only one side of a genuinely two-sided finding.

• Traditional incremental-validity analysis: specific abilities add roughly 4.4 percent of variance beyond g (one meta-analytic estimate)
• Relative-importance analysis, same underlying data: verbal comprehension can outweigh g itself, depending on the job
• Conclusion depends heavily on which method a given study runs, not only on which dataset it uses

Does context change the answer too?

A separate 2018 analysis in the Journal of Intelligence, by Thomas Coyle, looked at non-g components of specific math and verbal ability scores predicting college GPA, choice of college major, and eventual occupation. Coyle reports non-g effects on the order of beta approximately 0.30 for predicting college GPA specifically — a size too large to dismiss as noise — and argues these findings run against what he calls the "primacy of g" view. Read alongside the two reviews above, a pattern emerges: specific abilities appear to matter more for predicting a particular life outcome, such as a college major or a job type, and matter less for predicting a single, general outcome like overall school achievement, which is closer to what g was built to predict in the first place.

Does g matter less at higher ability levels?

One more wrinkle adds to why "it depends" is the honest answer. Spearman's law of diminishing returns describes a pattern where g's statistical pull across different tasks weakens among higher-scoring groups, meaning specific abilities may carry relatively more independent weight exactly among the people for whom a single full-scale number matters most, such as candidates for gifted programs or highly specialized roles. That pattern does not show up in every dataset it has been tested on, but where it holds, it is a structural reason — distinct from the choice of statistical method — for specific scores to matter more in some populations than in others.

How does this show up on an actual test report?

In practice, on a test like the WISC-V, this is why a profile of index scores is reported alongside the full-scale number rather than instead of it — the index scores exist because specific abilities plainly vary within a person, even when the debate over exactly how much predictive weight to give them stays unresolved. The broader taxonomy behind those index scores, CHC theory, is built explicitly around this layered structure: one general factor at the top, several broad abilities beneath it, and narrower specific skills beneath those.

What is the practical takeaway from a single test report?

Treat the full-scale score as the most stable, best-replicated number on the page, and treat a specific index or subtest score as informative but noisier — useful for noticing a real strength or weakness, less useful for fine-grained prediction of a specific outcome on its own. The review evidence does not support ignoring index scores, and it does not support treating a low score on one index as equivalent in weight to the full-scale number.

For the arithmetic behind converting a raw index or subtest score onto a familiar 100-point scale, see the IQ bell curve guide, and take the IQ test to see your own full-scale result.

Common questions