Theories of Intelligence

In brief

In 1904, Charles Spearman noticed that people who scored well on one kind of mental test tended to score well on nearly all of them, a pattern he called the positive manifold. He explained it with a single underlying factor, g, and psychometrics has argued about that finding ever since. Raymond Cattell split g into a fluid, on-the-spot reasoning ability and a crystallized store of learned knowledge. Howard Gardner and Robert Sternberg went further, arguing that a single number badly compresses what "smart" means, and proposed several independent intelligences instead. Around all of this sits the applied controversy: how valid IQ tests really are, why scores have risen for decades in a pattern called the Flynn effect, and how to interpret group differences in test scores responsibly.

The full treatment

The problem it answers

School grades, reasoning puzzles, vocabulary, and even simple sensory discrimination tasks all correlate with one another more than chance allows. Someone who is faster at mental arithmetic is also, on average, better at reading comprehension and spatial rotation, tasks that share no obvious content. Spearman needed to explain why performance clusters this way rather than scattering into independent skills, and needed a method to test the idea rather than simply assert it.

Spearman's g and the birth of factor analysis

Charles Spearman (1863 to 1945), a British psychologist, invented factor analysis to answer his own question. Analyzing correlations among school children's grades and sensory tests, he proposed a two-factor theory: a general factor g common to every cognitive task, plus a specific factor s unique to each one. Modern IQ scores are built to approximate g, the shared variance running through a battery of different subtests.

Cattell, Horn, and Carroll: splitting g and putting it back together

Raymond Cattell (1905 to 1998), who trained under Spearman, argued in 1963 that g itself splits into two broad abilities. Fluid intelligence (Gf) is the capacity to reason and solve novel problems without relying on prior knowledge; it rises through childhood and tends to decline from early adulthood. Crystallized intelligence (Gc) is accumulated knowledge, vocabulary, and skill drawn from experience and culture; it holds up, and can even grow, well into old age. John Horn extended the model into a fuller Gf-Gc theory, and John Carroll's 1993 survey of hundreds of factor-analytic studies folded it into a three-stratum model: g at the top, broad abilities like Gf and Gc in the middle, and narrow specific skills at the bottom. This synthesis, now called Cattell-Horn-Carroll (CHC) theory, underlies most contemporary IQ batteries, including the Wechsler scales and the Woodcock-Johnson tests.

Gardner and Sternberg: intelligence beyond the test

Howard Gardner (born 1943) argued in Frames of Mind (1983) that a single psychometric g misses most of what makes someone effective in the world. He proposed seven relatively independent intelligences: linguistic, logical-mathematical, spatial, musical, bodily-kinesthetic, interpersonal, and intrapersonal, later adding an eighth, naturalistic, each judged against criteria like isolation by brain damage, a distinct developmental course, and evidence from prodigies and savants. Robert Sternberg (born 1949) took a different route in Beyond IQ (1985), proposing a triarchic theory of analytical intelligence (the componential, test-friendly kind), creative intelligence (handling novelty and generating ideas), and practical intelligence (the tacit, contextual know-how that lets people navigate real environments). Sternberg's Rainbow Project (2006) tested whether adding creative and practical measures to the SAT improved prediction of college performance and found modest gains, though independent replication of the full triarchic structure has been thin.

The Flynn effect

James Flynn (1934 to 2020), a political philosopher based in New Zealand, documented in a series of papers, most fully in 1987, that raw IQ scores had risen substantially across many countries over the twentieth century, forcing test publishers to renorm their tests periodically so the average stayed at 100. Gains ran around three points per decade on average, larger on fluid-reasoning subtests than on vocabulary. Proposed causes include better nutrition, smaller families, more schooling, and what Flynn called a shift toward abstract, "scientific" categorical thinking driven by modern education. The effect is not universal or permanent: Bratsberg and Rogeberg's 2018 study of Norwegian conscription records found scores had begun reversing within families born after the mid-1970s, evidence that whatever drives the effect can also run backward.

Lineage

The tradition descends from Francis Galton's late-nineteenth-century work on hereditary ability and from Alfred Binet's 1905 test, built for French schools to identify children needing extra help. Binet himself warned against treating a single score as a fixed measure of a child's worth, a warning largely ignored when Lewis Terman Americanized the test as the Stanford-Binet and popularized the term IQ. David Wechsler later replaced Terman's ratio-based score with the deviation IQ still used today. Spearman's factor analysis gave the field its central method, Cattell and Horn split g into components, and Carroll's synthesis unified decades of competing models into CHC theory. Gardner and Sternberg reacted against this whole psychometric tradition, drawing instead on neuropsychology and cognitive science.

The strongest case for it

The psychometric tradition's strongest evidence is convergent. IQ scores are highly reliable on retest, and Frank Schmidt and John Hunter's 1998 meta-analysis of eighty-five years of personnel research found general mental ability the single best predictor of job performance across occupations, especially complex ones, ahead of interviews, experience, or education. Twin and adoption studies converge on substantial heritability of g that rises with age into adulthood, though heritability describes variation within a studied population, not a fixed ceiling on any individual. Neuroscience adds a physical anchor: g correlates, modestly but repeatedly, with brain volume, white matter integrity, and processing speed, suggesting the factor tracks something real in the brain rather than a statistical illusion.

The strongest case against it

Stephen Jay Gould, in The Mismeasure of Man (1981, revised 1996), argued that reifying g as a single physical thing is a category error: factor analysis produces a mathematical summary whose apparent "reality" depends on arbitrary choices of rotation and interpretation, and he traced how IQ testing had already been used to justify eugenic immigration policy earlier in the century. Defenders reply that later replications across independent samples, culminating in Carroll's synthesis, strengthened the case for a stable factor structure regardless of past misuse, but the historical warning about overreach stands. Gardner's theory draws its own fire: Perry Klein, in a 1997 critique, argued the proposed intelligences correlate with one another more than the theory predicts for genuinely independent abilities, and that Gardner has resisted building testable instruments, making the theory hard to falsify on its own terms. Linda Gottfredson made a parallel case for the defense of g more broadly, arguing that decades of factor-analytic replication give it far stronger empirical footing than any of its proposed replacements.

The most fraught controversy concerns group differences. Average IQ score gaps between racial groups in the United States have been documented for decades. Herrnstein and Murray's The Bell Curve (1994) argued a genetic component was likely and predicted a cognitively stratified society. The American Psychological Association convened a task force, reported by Ulric Neisser and colleagues in 1996, that found no support for a genetic explanation of the gap, noted that environmental candidates such as socioeconomic conditions, schooling quality, and stereotype threat (documented experimentally by Claude Steele and Joshua Aronson in 1995) remained far more plausible, and stressed that variation within any group dwarfs the average difference between groups. Later work by William Dickens and James Flynn (2006) found the Black-White gap had narrowed by several points between 1972 and 2002, evidence that critics of a genetic explanation cite directly, since a genetically fixed gap should not move on a generational timescale.

Where it stands now

The working consensus among differential psychologists is that g is real and well replicated, best organized today by hierarchical models like CHC theory, and that IQ tests predict school, job, and health outcomes with genuine, if partial, validity. What they measure is not the whole of valuable human ability, a point Gardner and Sternberg pressed hard even if their own alternatives remain less validated than the model they challenge. The causes of group differences in scores remain actively studied and are not settled, and the Flynn effect's rise and partial reversal is a live reminder that these scores respond to environment on a scale that outpaces any plausible genetic shift.

Test yourself

Think of someone you consider highly intelligent who would score unremarkably on a standard IQ test, and someone who tests brilliantly but whom you would not trust with a difficult real-world problem. Notice which theory in this entry, general g, fluid versus crystallized ability, multiple intelligences, or the triarchic model, best explains the gap between the label and the person.