The g factor — "general intelligence" — is the single most replicated finding in cognitive psychology. When people take multiple different mental ability tests, their scores correlate strongly across tasks that look very different on the surface: vocabulary, math, pattern recognition, working memory, processing speed. The statistical pattern shared across all of them is what psychologists call g. This guide explains what g is, how it was discovered, what it predicts in the real world, what it doesn't predict, and the honest case for and against the concept.
What Is the g Factor?
The g factor (or general intelligence factor) is the statistical observation that performance on virtually any mental ability test positively correlates with performance on any other mental ability test. Someone who scores well on vocabulary also tends to score well on math reasoning, pattern recognition, and digit-span memory — and the converse. The shared variance underlying all these correlations is what gets called g.
The key word is statistical. The g factor is not a substance, not a location in the brain, and not a single ability that has been directly measured. It's the latent variable that emerges when you do factor analysis on a battery of cognitive tests. The reliability of the pattern across populations, ages, cultures, and decades is what gives the concept weight — it's the single most replicated finding in psychometrics.
Two equivalent ways psychologists describe it:
- "Positive manifold": the fact that all mental ability tests correlate positively with each other
- "General factor of intelligence": the latent variable that explains the positive manifold
Who Discovered g and How
The g factor was first described by the British psychologist Charles Spearman in 1904. Spearman noticed that schoolchildren's scores on apparently unrelated subjects (Latin grades, math scores, sensory discrimination tasks) were all positively correlated. He proposed a two-factor model: each test measures a general factor (g) plus a specific factor (s) unique to that test.
Spearman's discovery has been replicated thousands of times since. Modern factor models are more sophisticated — they typically describe g as the apex of a hierarchy, with broader abilities (verbal reasoning, fluid reasoning, working memory, processing speed) underneath, and specific test scores at the bottom — but Spearman's basic observation has held up for 120 years.
What g Predicts in the Real World
This is where the topic gets serious. The correlations are real but easy to misinterpret. The honest summary:
Where g predicts strongly (r > 0.4):
- Academic performance — especially in cognitively demanding subjects (math, physics, formal logic)
- Standardized test scores (SAT, GRE, professional licensing exams)
- Speed of learning new complex skills, especially in technical domains
- Job performance on cognitively complex jobs (engineering, medicine, research, software)
Where g predicts moderately (r ~ 0.2-0.4):
- Lifetime earnings (mediated heavily by education and occupation choice)
- Job performance on jobs of moderate cognitive complexity
- Some health outcomes (literacy → ability to follow medical instructions → adherence)
Where g predicts weakly or not at all (r < 0.2):
- Career satisfaction
- Relationship quality and marital stability
- Job performance on jobs that aren't cognitively complex
- Creativity past a moderate IQ threshold (the "threshold hypothesis")
- Ethical behavior and trustworthiness
- Leadership effectiveness
The pattern: g is a strong predictor of cognitive throughput — how fast you can absorb and apply complex information. It's a weak predictor of nearly everything else humans care about in their daily lives.
How Big the Effect Actually Is
A correlation of 0.4 — typical for the strongest g-outcome relationships — explains about 16% of the variance in the outcome. That's meaningful but not overwhelming. For job performance specifically:
- g predicts about as much variance in performance as conscientiousness (the personality trait covering self-discipline and dependability), which is uncorrelated with g
- Combining g and conscientiousness explains substantially more variance than either alone
- Adding domain knowledge, opportunity, and social skill explains more still
The honest implication for someone trying to predict their own life outcomes: g matters, but it's one of perhaps four to six roughly co-equal factors. People who treat their IQ as the dominant variable end up with worse predictions than people who track all the inputs.
The "Threshold Hypothesis"
An interesting finding that softens the strong-g view: past a certain IQ (typically around 120), additional IQ stops predicting much of anything outside of academic and test-taking outcomes. The "threshold hypothesis" — supported by several studies of high-IQ populations — suggests that for most adult goals, having "enough" g (somewhere in the 110-125 range) matters, but having more doesn't.
This is why the population of people doing genuinely important creative or leadership work spans a wide IQ range above the threshold. Past 120, what separates Nobel laureates from middle managers isn't more g — it's persistence, domain knowledge, opportunity, and social skill.
The Case Against Strong-g Interpretations
The g factor itself is statistically robust, but the popular interpretations of it have problems worth knowing:
g is a statistical pattern, not a thing. Critics like Howard Gardner (multiple intelligences) and Robert Sternberg (triarchic theory) argue that calling g "intelligence" obscures the heterogeneity inside cognitive ability. Two people with the same g score can have very different cognitive profiles — one stronger in fluid reasoning, the other in crystallised knowledge. The single number flattens real differences.
g doesn't capture all cognitively valuable abilities. Practical intelligence (street smarts), social intelligence (reading people), creativity (generating useful novelty), and emotional regulation all matter for life outcomes and are poorly captured by g-loaded tests.
g may reflect testing artifacts as much as underlying ability. Some critics argue that the positive manifold partly reflects shared task demands (verbal instructions, time pressure, motivation to perform) rather than a unitary cognitive resource.
Cultural fairness is unresolved. Different cultural backgrounds produce different cognitive habits and test-familiarity. Whether g-loaded tests measure ability or test-taking culture remains debated.
The honest centrist position: g is a real statistical phenomenon with real predictive validity, but it's narrower than the popular interpretation of "intelligence" and shouldn't be the dominant variable in self-assessment.
Crystallized vs. Fluid g
A useful subdivision worth knowing. Raymond Cattell distinguished:
- Fluid intelligence (Gf): raw reasoning ability on novel problems. Pattern recognition, abstract logic, working memory. Peaks in the late teens to mid-20s and declines slowly.
- Crystallized intelligence (Gc): accumulated knowledge, vocabulary, reasoning patterns absorbed over time. Rises through middle age and only slowly declines.
Modern IQ tests measure both, weighted differently. The Wechsler scales lean more toward Gc; Raven's Progressive Matrices is one of the purest measures of Gf. The age curve of "general intelligence" you see on most tests is the blended average — Gf falling, Gc rising — which is why total IQ scores look roughly stable through middle age in healthy adults.
How to Measure g (Approximately)
You can't measure pure g — it's a statistical abstraction. What you can do is take a battery of cognitive tests and let g emerge from the shared variance. Common single proxies:
- Raven's Progressive Matrices — purest measure of fluid g, language-independent
- Wechsler Adult Intelligence Scale (WAIS) — full battery, blended Gf+Gc, professionally administered
- Multi-subtest cognitive batteries like the CHC-based Woodcock-Johnson
For a quick directional read on your own cognitive profile across reasoning subscales, our free IQ test takes 20 questions and gives an instant breakdown across numerical, verbal, logical, and pattern-recognition reasoning.
Frequently Asked Questions
Is the g factor scientifically real?
The statistical phenomenon — that mental ability tests positively correlate with each other — is one of the most replicated findings in psychology. Whether the phenomenon should be interpreted as a unitary "general intelligence" is debated. The data are real; the popular interpretation is contested.
What is the difference between g and IQ?
g is the underlying statistical factor. IQ is a specific score on a specific test that's heavily loaded on g. Different IQ tests load on g to different degrees, but all major tests aim to measure something close to it.
Can g change over time?
Adult g is largely stable from the late teens onward, with some predictable shifts (fluid intelligence declines slowly with age, crystallised intelligence rises through middle age). The headline IQ number stays roughly constant because the components offset.
Does high g guarantee success?
No. High g is one of several roughly co-equal predictors of life outcomes, and its predictive power flattens past about IQ 120. Persistence, conscientiousness, social skill, domain knowledge, and opportunity all matter as much or more for most adult goals.
What's a "high g" score?
"High" by population standards usually means at least one standard deviation above the mean — IQ 115+ — covering the top 16% of the population. "Very high" typically means IQ 130+, the top 2%. Past about IQ 150, single-test measurement becomes unreliable.