General Intelligence vs. Multiple Intelligences: What Each Theory Gets Right

Is intelligence one thing or many? Compare the g-factor with Gardner's multiple intelligences. Read the full article and take the RIOT IQ test!

Dr. Russell T. WarneChief Scientist

General Intelligence vs. Multiple Intelligences: What Each Theory Gets Right

Few questions in the psychology of intelligence generate more everyday confusion than this one: is intelligence one thing, or many things? The debate is genuinely old — it predates the internet, predates modern neuroscience, and predates most of the psychometric tools we now use to study cognitive ability. What makes it worth revisiting isn't novelty but clarity: the popular framing of the debate tends to misrepresent both sides, and understanding what each position actually claims — and where the evidence lands — changes how you should think about IQ scores, cognitive profiles, and your own strengths.

This article walks through the three major frameworks in this debate, what each one gets right, where each one falls short, and what the research has settled on as the most defensible account of how human intelligence is actually organized.

The Question Each Theory Is Trying to Answer

Before comparing frameworks, it's worth being precise about the question they're all responding to. When researchers began systematically testing cognitive abilities in the early 20th century, they noticed something consistent and surprising: performance on radically different tests tended to correlate positively with each other. People who scored high on vocabulary tests also tended to score above average on spatial reasoning, arithmetic, and memory tasks. People who scored low on one tended to score below average on the others.

This pattern — called the positive manifold — is among the most replicated findings in all of psychology. Across almost any set of cognitive tests, people who do well on one tend to do well on the others. Every serious theory of intelligence is, at its core, an attempt to explain why this happens. The theories differ in what they conclude — but they are all responding to the same underlying observation.

The Case for General Intelligence

Charles Spearman's 1904 proposal was the founding answer: the positive manifold exists because a single general factor — which he called g — underlies performance on every cognitive task. People differ in how much g they have, and those differences show up consistently across any test that requires genuine mental effort. Factor analysis, the statistical technique Spearman helped develop, consistently extracts this general factor from cognitive test batteries, and it accounts for roughly 40–50% of the variance in performance across cognitive tasks.

The predictive validity of g is among the most robust findings in applied psychology. General intelligence predicts academic achievement, job performance, income, and even health outcomes more reliably than almost any other psychological variable. Its heritability increases across the lifespan, reaching 60–80% in adults. Its biological correlates — including neural efficiency, white matter integrity, and processing speed — have been documented through decades of neuroimaging research. In short: g is real as a statistical regularity, and its existence is among psychology's most replicated findings.

What general intelligence theory does not claim — though it is frequently misread as claiming this — is that intelligence is nothing but a single number, or that specific abilities don't exist and don't matter. Spearman's own framework acknowledged specific factors (s factors) alongside g, and every subsequent development in the psychometric tradition has recognized that the general factor sits above a layer of distinguishable broad abilities, not in place of them.

The Case for Multiple Intelligences

Howard Gardner's 1983 book Frames of Mind introduced the most influential challenge to the single-factor view. His argument was intuitive and, in some respects, clearly correct: a brilliant musician who struggles with mathematics, a gifted athlete with average verbal scores, an exceptional social communicator who finds abstract reasoning effortful — these are real people, and a single-number model feels inadequate to describe them.

Gardner proposed that intelligence consists of several independent modalities — linguistic, logical-mathematical, musical, spatial, bodily-kinesthetic, interpersonal, intrapersonal, and later naturalistic — each with its own neural substrate, developmental trajectory, and cultural expression. The framework gained enormous traction in education, where it offered teachers a vocabulary for honoring diverse student strengths and diversifying instructional approaches.

The empirical problem with MI theory is specific: when the abilities Gardner named are actually measured and factor-analyzed, they don't behave like independent intelligences. Factor analysis of tests corresponding to Gardner's eight domains revealed a large g factor having substantial loadings for tests assessing purely cognitive abilities — Linguistic, Logical/Mathematical, Spatial, Naturalistic, and Interpersonal — but lower loadings for tests of other abilities, especially Bodily-Kinesthetic. Within most domains, the two tests showed only weak non-g associations — meaning the correlation between the domains is far too high to support the claim that they are independent intelligences.

The more fundamental methodological problem is that Gardner resisted developing a validated psychometric instrument for his theory, arguing that standard test formats are biased toward analytical and linguistic ability. That resistance has a principled basis — the concern about format bias is legitimate — but it means the theory has never been exposed to the kind of falsifiable empirical test that would allow it to compete with the psychometric literature on equal terms. Without a validated instrument, there is no way to systematically confirm or disconfirm the central claim that the intelligences are genuinely independent. The fair summary is that g and the CHC hierarchy command broad empirical consensus, while multiple intelligences retain greater influence in educational contexts than in the research literature.

Sternberg's Third Way: The Triarchic Theory

Between the g-factor model and Gardner's MI theory sits Robert Sternberg's triarchic theory, introduced in 1985 and less widely discussed in public discourse despite being highly influential in academic psychology. Sternberg proposed that intelligence has three interrelated aspects: analytical intelligence — closest to what standard IQ tests measure; creative intelligence — the ability to cope with novelty and generate new ideas; and practical intelligence — the ability to adapt to, shape, and select real-world environments.

The triarchic theory makes a specific and testable claim: that analytical, creative, and practical intelligence are meaningfully distinct and that standard IQ tests miss the creative and practical dimensions in ways that matter for predicting real-world success. Sternberg developed the Sternberg Triarchic Abilities Test (STAT) to measure all three forms, and factor analysis of STAT data was supportive of the triarchic theory, measuring relatively separate analytical, creative, and practical abilities across 326 high school students in an initial validation study.

The empirical reception of Sternberg's theory has been mixed, however. Practical intelligence — his contextual subtheory — has faced the most sustained criticism, with researchers arguing that what Sternberg calls practical intelligence largely reflects g operating in real-world contexts rather than a genuinely distinct capacity. The analytical scale of the STAT correlates strongly with traditional IQ tests, which Sternberg acknowledges — meaning at minimum, analytical intelligence is well captured by existing instruments. The creative and practical scales have shown less consistent evidence of independence from g in external replications. Sternberg's theory has been more productive as a conceptual framework that draws attention to what standard tests underemphasize than as a psychometric model that has been empirically validated at the same level as CHC.

What Sternberg gets right, and what matters regardless of the theory's empirical limitations, is the observation that intelligence is as much about knowing when to use which ability as it is about strength in any one area — and that the ability to adapt, generate novel solutions, and apply knowledge in practical contexts is genuinely important and genuinely undersampled by tests that prioritize abstract analytical performance.

What the CHC Framework Resolves

The framework that has achieved the closest thing to consensus in the research literature — and the one that underlies virtually every professionally designed IQ battery in use today — is the Cattell-Horn-Carroll (CHC) model. Understanding CHC is useful precisely because it shows how the competing intuitions behind general intelligence theory and multiple intelligences can be accommodated within a single, empirically grounded structure.

CHC is a three-stratum hierarchy: general intelligence at the apex (Stratum III), a set of roughly ten broad abilities beneath it (Stratum II) — including fluid reasoning, crystallized knowledge, visual processing, working memory, and processing speed — and more than 70 narrow abilities at the base (Stratum I). The broad abilities at Stratum II are meaningfully distinct from each other — they have different developmental trajectories, different biological correlates, different relationships with real-world outcomes, and different sensitivities to aging and neurological insult. A person can have a notably stronger crystallized knowledge profile than fluid reasoning profile, or exceptional visual-spatial ability alongside average verbal comprehension. Those differences are real, and a good cognitive battery is designed to detect and report them.

But the broad abilities also correlate with each other — which is why g sits above them at Stratum III. The CHC framework operationalizes general intelligence not as monolithic competence, but as coherent sufficiency across multiple cognitive domains. The general factor is not the only thing being measured; it is the apex of a hierarchy that includes genuinely distinct domains below it.

This structure explains why the debate between general intelligence and multiple intelligences involves a false dichotomy when framed as "one thing or many." The CHC hierarchy preserves Spearman's g while incorporating the more differentiated abilities that multiple-factor theorists emphasized — showing that the two positions are not mutually exclusive but are operating at different levels of the same hierarchy. What looks like "multiple intelligences" from below — the distinct broad abilities — looks like components of a general factor from above. Both observations are correct; they are just describing different strata of the same structure.

What Each Theory Gets Wrong

Being fair to this debate requires acknowledging the specific failures of each position, not just the successes.

General intelligence theory, in its strongest forms, tends to underemphasize the practical importance of ability differentiation. Collapsing all cognitive variation into a single g estimate — which is what a Full Scale IQ does — loses the profile information that matters most for educational placement, clinical diagnosis, and individual self-understanding. A person with exceptionally strong verbal comprehension and weaker processing speed, or strong fluid reasoning and average crystallized knowledge, is poorly served by a single composite that averages those differences into one number. The CHC framework is a corrective to this tendency within the general intelligence tradition.

Multiple intelligences theory, in its standard form, overcorrects in the opposite direction. By treating the intelligences as fully independent — and by resisting the development of validated measurement instruments — it has produced a framework that is influential in classrooms but lacks the empirical grounding that would allow it to generate testable predictions, accumulate disconfirming evidence, or compete with the psychometric literature on scientific terms. The genuine intuition behind MI — that human cognitive capacity is diverse and that standard tests undersample important abilities — is correct and important. The specific claim that the identified intelligences are independent rather than correlated is not empirically supported.

Sternberg's triarchic theory occupies a middle position: more empirically engaged than MI, less empirically established than CHC. Its greatest contribution is conceptual — drawing attention to creative and practical dimensions of intelligence that both the g-factor model and MI theory, in different ways, tend to miss.

What This Means for How You Read an IQ Score

The practical implication of this theoretical landscape is direct. A well-constructed IQ test — built on CHC principles — reports both a general factor estimate and a profile of distinct broad abilities. The general factor estimate gives you a comparison to the population on overall cognitive ability. The broad ability profile gives you the pattern of strengths and relative weaknesses that the general factor obscures.

Reading only the general factor estimate, as popular accounts of IQ tend to do, misses half the information the test was designed to provide. Reading only the specific indices, as an uncritical MI application might encourage, loses the context of how each domain relates to overall cognitive capacity and to each other.

Both levels of the hierarchy are informative. Neither tells the complete story on its own.

The Takeaway

General intelligence theory and multiple intelligences theory are not equally supported by the research literature — but they are not as far apart as the debate's most polemical versions suggest. The positive manifold is real. g is real as a statistical regularity. Multiple broad abilities that are distinct from each other and from g are also real. The CHC hierarchy is the framework that accommodates all of these findings simultaneously, and it is the framework that underlies virtually every modern cognitive battery precisely because it reflects what the data shows rather than what any single theory predicted.

If you want to understand where your own cognitive profile sits — across the broad ability domains that the CHC framework identifies — the RIOT gives you both a general intelligence estimate and a domain-level profile that neither a single number nor an unchecked list of eight "intelligences" can provide on its own.