The g Factor: What General Intelligence Is and How It Relates to IQ

What the g factor means

Give a large group of people a dozen different mental tests and the scores correlate: every test goes with every other, some strongly, some weakly, but almost never negatively. Psychologists call this the positive manifold, and one of the papers cited below calls it the most replicated result in intelligence research. Factor analysis, the statistical method Spearman helped develop, condenses that web of correlations into one common dimension. That dimension is g.

So g is defined by the tests, not by a theory of the mind. Saying a person is “high in g” means they tend to score well across many kinds of mental task. It does not say why.

Spearman's discovery and the two-factor theory

Spearman's 1904 paper in the American Journal of Psychology compared schoolchildren's marks in different subjects with their results on simple sensory tests. His conclusion, in his own words, was that all branches of intellectual activity have in common one fundamental function (or group of functions), while the remaining specific elements of each activity seem wholly different from those of the others. How much the general part mattered varied by subject — from 15:1 to 1:4 across the ten areas he examined.

That split — one general factor (g) plus a specific factor (s) for each task — became known as his two-factor theory. Spearman drew a practical conclusion straight away: different examinations could be treated as independent estimates of the same common function. Modern full-scale IQ scores rest on the same logic.

The g factor in modern models of intelligence

Louis Thurstone argued in the 1930s for several separate primary abilities rather than one general one. Later hierarchical models combined the two views. John Carroll's 1993 Human Cognitive Abilities reanalysed more than 460 data sets from the factor-analytic literature and proposed a hierarchical, three-stratum model: many narrow abilities at the bottom, a handful of broad abilities in the middle — including fluid and crystallized intelligence, spatial ability and processing speed — and a general factor at the top.

Is g just an artefact of which tests happen to be used? Two studies led by Wendy Johnson compared the general factors from several different test batteries, and their titles give the answer: Just one g (three batteries, 2004) and Still just 1 g (five batteries, 2008).

Is the g factor the same as IQ?

Not quite. g is a factor estimated from the correlations among many tests; IQ is a score on one test. The two are linked because a full-scale IQ combines many subtests, and averaging across varied tasks cancels out much of what is specific to each one — leaving mostly what they share. That is why a full battery tracks g better than any single subtest, and why a short test built on one kind of task (only matrices, say) estimates it less precisely. How IQ tests are built follows from this.

Each subtest's relationship to g is called its g loading. A highly g-loaded task is a good single indicator of general ability; a task with a low loading mostly measures something narrower. Test designers use loadings to choose which tasks to include and how to weight them.

What the g factor predicts

g is studied so heavily because it predicts outcomes outside the testing room. The best-known summary, Frank Schmidt and John Hunter's 1998 review of 85 years of personnel research, compared 19 selection methods and found that the three best combinations for predicting job performance all paired general mental ability with something else — a work sample, an integrity test or a structured interview (mean validities of 0.63 to 0.65).

Those figures have since been revised down. Paul Sackett and colleagues argued in 2022 that the corrections used in earlier meta-analyses systematically overstated validity; their re-estimates lowered most selection methods by 0.10 to 0.20 and put structured interviews, not ability tests, at the top. g still predicts work performance — just less strongly than the older numbers suggested.

Where the g factor comes from

A 2010 review in Nature Reviews Neuroscience by Ian Deary, Lars Penke and Wendy Johnson summarised the biology: twin and family studies show additive genetic contributions to cognitive ability, especially general intelligence; no individual gene had yet been reliably linked to it; and brain-imaging work points to differences in pathways connecting parietal and frontal regions. Genes are not the whole story — see is intelligence genetic? for the heritability figures and what they do and do not mean.

Can the g factor be increased?

Education reliably raises IQ scores. A 2018 meta-analysis by Stuart Ritchie and Elliot Tucker-Drob — 142 effect sizes, over 600,000 people — found about 1 to 5 IQ points per extra year of schooling, lasting across the lifespan. But whether that is a rise in g is a separate question. A 2015 study by Ritchie, Timothy Bates and Ian Deary followed 1,091 people from an intelligence test at age 11 to ten cognitive tests at age 70; the link between education and better scores ran through specific skills, not through g.

That distinction — scores can rise without the general factor rising — also runs through the Flynn effect and the evidence on how to increase IQ, including why practising working memory tasks mostly improves those tasks.

Criticisms: what the g factor might really be

Nobody seriously disputes the positive manifold. The argument is about its cause, and two alternatives to “one general ability” are prominent:

• Mutualism. Han van der Maas and colleagues at the University of Amsterdam showed in 2006 that a positive manifold can emerge purely from separate cognitive processes helping each other grow during development — a model in which a single g plays no role, yet which reproduces the hierarchical structure of test scores.
• Process overlap. Kristóf Kovács and Andrew Conway proposed in 2016 that tests tap general executive processes (the kind studied in working memory research) together with more specific ones, and that the executive processes are shared across so many tests that a general factor appears — without g being a single ability.

A broader criticism is that one number leaves out abilities people value — the argument behind the theories of multiple intelligences. Both things can be true: g summarises a real regularity in test scores, and it is not a complete description of a mind.

Trust and scope notes

This page explains published research in general terms. Our IQ test measures reasoning; it does not report a separate g score, and like any single test its result is an estimate, not a measurement of g itself.

IQ Revealed is not affiliated with, endorsed by or connected to the researchers, universities, publishers or journals named on this page. Their names are used only to identify the work being described.