What Is Intelligence? Not Even the Experts Agree

They’ve been at it for more than a century. Entire university departments are devoted to it. And yet experts on intelligence can’t agree on something as basic as what exactly it is they’re studying.

The problem isn’t a lack of research or interest, but the word itself: “intelligence,” as used in everyday language, turns slippery the moment anyone tries to pin it down seriously.

Almost as many definitions as experts

Psychologist Arthur Jensen sums it up in The g Factor (1998) with a line that leaves little room for ambiguity: “no other term in psychology has proved harder to define than ‘intelligence.’”

And he adds something even more striking: there are almost as many different definitions of ‘intelligence’ as there are experts who’ve tried to define it.

It isn’t that each expert proposes a slightly different definition of the same idea. It’s that they focus on genuinely different aspects:

• Some define it around the capacity to reason and solve novel problems.
• Others, around the capacity to adapt to changing situations.
• Others, around the speed or efficiency of learning.
• Others include the capacity for abstraction, or handling symbols, or long-term planning.

Each of these proposals is reasonable on its own. The problem is that, together, they don’t outline a single concept but several partially overlapping ones, with no shared criterion for deciding which is “the” correct definition.

Two expert surveys, 65 years apart

The lack of agreement isn’t just an impression: it has been measured. In 1921, the Journal of Educational Psychology asked more than a dozen leading researchers to define intelligence. In 1986, Robert Sternberg and Douglas Detterman repeated the exercise with about two dozen experts and published the answers in What Is Intelligence?

Comparing the two rounds, they found some continuity: abstract reasoning, problem-solving and the capacity to learn came up often in both. But no single definition won over a majority, and the answers were so varied that they had to be sorted into dozens of categories. Jensen remarked at the time that there was barely more consensus than in 1921.

Two decades later, Shane Legg and Marcus Hutter (2007), two artificial intelligence researchers, gathered some seventy definitions from psychology, AI and expert committees. Their attempt at a synthesis was deliberately broad: intelligence measures an agent’s ability to achieve goals in a wide range of environments.

The consensus definition that does circulate

The most cited formulation in psychology comes from a collective text. In 1994, at the height of the controversy over the book The Bell Curve, psychologist Linda Gottfredson drafted a statement signed by 52 researchers, later published in the journal Intelligence (1997). Its definition says, in essence, that intelligence is a very general mental capability that involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.

The same context led the American Psychological Association to commission a report from a task force led by Ulric Neisser (1996), Intelligence: Knowns and Unknowns. Its tone was more cautious: intelligence test scores predict school performance well and job performance in part, but many questions remain open about what exactly they measure and how they develop.

Both definitions closely resemble the list of candidates above. The difference is that they don’t pick one; they gather them under a general capability. And that idea has a specific origin.

Spearman’s solution: stop arguing over the word and measure the correlation

In the early 20th century, British psychologist Charles Spearman made a methodological decision that would change the course of the entire field: instead of continuing to argue over what “intelligence” is in the abstract, he set out to observe something far more concrete and measurable.

Spearman started from an empirical finding that earlier thinkers like Herbert Spencer and Francis Galton had already sensed: scores on very different mental tests (vocabulary, numerical reasoning, spatial orientation) tend to be positively correlated.

Whoever scores high on one tends to score high on the others, even when they seem, on the surface, to have nothing in common.

To put this to a rigorous test, Spearman invented factor analysis, a statistical method capable of determining how much each of those tests relates to an underlying common factor.

That common factor, present in some form in any task that demands mental effort, he simply called g (for “general”).

An anecdote that sums up the idea better than any technical definition

To illustrate this same intuition long before factor analysis existed, Jensen recounts an anecdote involving the writer Samuel Johnson (1709-1784).

A well-known historian remarked to Johnson that it was thanks to very different talents that Caesar became a great commander, Shakespeare a great poet and Newton a great scientist. Johnson’s reply was blunt:

“No, it is only that one man has more mind than another. He may direct it differently, or prefer this study to that. Sir, the man of vigor may walk to the north as well as to the south, to the east as well as to the west.”

More than a century before Spearman, Johnson had already sensed, in a single line, what factor analysis would eventually show with numbers: that behind seemingly unrelated talents like military leadership, poetry and science there could be something in common, a kind of general mental “vigor,” applicable in any direction.

From the g factor to a hierarchy of abilities

Spearman’s model didn’t stand alone for long. Raymond Cattell and John Horn distinguished between fluid intelligence (reasoning through new problems) and crystallized intelligence (accumulated knowledge and vocabulary), and in 1993 John B. Carroll reanalyzed more than 460 factor-analytic datasets collected over decades.

His conclusion, the three-stratum model, places g at the top, about eight broad abilities beneath it (fluid, crystallized, memory, visual processing, auditory processing, speed, among others) and dozens of specific abilities at the base. The merger of the two proposals, known as the CHC model (Cattell-Horn-Carroll), is today the reference framework for most intelligence tests, such as the Wechsler scales or the Woodcock-Johnson battery.

The theories that challenge the framework

Not everyone accepts that intelligence boils down to what these tests measure:

• Howard Gardner (Frames of Mind, 1983) proposed multiple intelligences: linguistic, logical-mathematical, spatial, musical, bodily-kinesthetic, interpersonal, intrapersonal and, later, naturalist. The idea has had enormous influence in education, but little empirical support as a set of independent abilities: when they’re measured, several correlate with one another and with g.
• Robert Sternberg (Beyond IQ, 1985) defended a triarchic theory, later called the theory of successful intelligence, which adds creative and practical ability to analytical ability. His critics point out that the data supporting it are weaker than those behind the CHC model.

Our editorial take: the disagreement over the word is still alive, but on the structure of the abilities that can be measured there is considerably more agreement than the title suggests.

Why this isn’t just an academic word game

It might seem that swapping “intelligence” for “the g factor” simply replaces one hard-to-define word with another equally abstract one. That isn’t the case, and the difference matters:

• “Intelligence” is a word from everyday language, loaded with different connotations depending on who’s using it.
• “g” is a statistical factor, defined operationally from real correlation patterns between tests, not from a prior intuition about what being intelligent “should” mean.
• This makes possible something the philosophical debate over the word “intelligence” could never offer: measuring how much of that common factor each specific test reflects, comparing tests against each other, and accumulating verifiable empirical evidence about its biological correlates and its role in real-world outcomes.

Spearman didn’t resolve the debate over what “intelligence” is in the philosophical sense. He sidestepped it, wisely, and proposed instead a question that data could actually answer: is there a factor common to all mental abilities, and how can it be measured?

What’s left once you stop arguing over the word

More than a century after Spearman, the word “intelligence” still has no single definition accepted by all experts. It probably never will, because it is above all an everyday word, not a technical term with precise boundaries. As early as 1923, psychologist Edwin Boring settled it with a line as famous as it is unsatisfying: intelligence is what the tests test.

What does exist, with a century of accumulated research behind it, is a well-characterized statistical factor, with measurable biological correlates (from the brain’s energy efficiency to reaction time in milliseconds) and a demonstrated ability to predict outcomes in fields as different as academic performance and road safety.

We may never agree on exactly what it means to be “intelligent.” But that, as Spearman showed more than a century ago, was never really the question that needed answering.

If you want to see that same g factor in action, measured in fractions of a second, what your brain decides in a fifth of a second explains the experiment that has been putting it to the test for more than a century.

Sources

• Jensen, A. R. (1998). The g Factor: The Science of Mental Ability. Praeger.
• Sternberg, R. J., and Detterman, D. K. (Eds.) (1986). What Is Intelligence? Contemporary Viewpoints on Its Nature and Definition. Ablex.
• Legg, S., and Hutter, M. (2007). A collection of definitions of intelligence. Frontiers in Artificial Intelligence and Applications, 157, 17-24. arXiv:0706.3639
• Gottfredson, L. S. (1997). Mainstream science on intelligence: An editorial with 52 signatories, history, and bibliography. Intelligence, 24(1), 13-23.
• Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N., Ceci, S. J., Halpern, D. F., Loehlin, J. C., Perloff, R., Sternberg, R. J., and Urbina, S. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51(2), 77-101.
• Carroll, J. B. (1993). Human Cognitive Abilities: A Survey of Factor-Analytic Studies. Cambridge University Press.
• Gardner, H. (1983). Frames of Mind: The Theory of Multiple Intelligences. Basic Books.
• Sternberg, R. J. (1985). Beyond IQ: A Triarchic Theory of Human Intelligence. Cambridge University Press.
• Boring, E. G. (1923). Intelligence as the tests test it. New Republic, 36, 35-37.
• Spearman, C. (1904). “General intelligence,” objectively determined and measured. The American Journal of Psychology, 15(2), 201-292.