Why human cognitive ability is neither one undivided faculty nor a collection of unrelated talents—and what “both” means in practice.
Human intelligence is both general and differentiated.
There is a broad general tendency: people who perform relatively well in one cognitive domain tend, on average, to perform relatively well in others. But there are also meaningful domains of ability—such as fluid reasoning, acquired knowledge, visual-spatial ability, memory-related abilities and processing speed—that are not interchangeable.
So the best-supported picture is not one ability versus many abilities. It is a hierarchy in which narrower abilities cluster into broader abilities, while those broader abilities also share a general component.
What would “one ability” look like?
Imagine that every cognitive task were simply another thermometer measuring exactly the same hidden quantity. If that were true, a person’s ranking should be almost identical across vocabulary, spatial reasoning, memory, processing speed and abstract reasoning.
That is not what we observe.
A person may be excellent at verbal reasoning and only average at processing speed. Another may be strong in visual-spatial reasoning but less strong in acquired verbal knowledge. These differences can be reliable.
So intelligence cannot be reduced to the claim that every cognitive task measures one identical ability.
What would “many unrelated abilities” look like?
Now imagine the opposite: vocabulary, reasoning, memory and spatial ability are completely independent. If that were true, knowing that someone was strong at reasoning would tell us essentially nothing about how they tended to perform on any other cognitive domain.
That is also not what we observe.
Across large samples, cognitive tests tend to correlate positively. This is the positive manifold. The correlations are imperfect—which leaves room for differentiated abilities—but they are too systematic to treat the domains as unrelated.
That is why “both” is not a compromise answer. It is a description of the data.
A worked example: two people with the same overall level
Consider two fictional adults assessed on five broad domains. The scores below are illustrative standardized scores, not results from a real instrument.
| Ability | Person A | Person B |
|---|---|---|
| Fluid reasoning | 120 | 102 |
| Verbal knowledge | 114 | 120 |
| Visual-spatial ability | 116 | 105 |
| Working-memory-related performance | 103 | 113 |
| Processing speed | 97 | 110 |
| Broad composite | 110 | 110 |
Both have the same illustrative broad composite. Yet their patterns differ. The composite values are pedagogical and are not calculated from the displayed domain scores using the scoring rules of any real instrument.
Person A’s overall result is supported more strongly by reasoning and spatial performance. Person B’s is supported more strongly by verbal knowledge, memory-related performance and speed.
This does not mean every difference in a real profile is automatically meaningful. Real scores contain measurement error, and some discrepancies are common in the population. But the example shows why a single general score can be informative without exhausting the structure of cognitive ability.
How psychologists represent “both” statistically
A hierarchical factor model contains more than one level.
At the bottom are individual tasks: a vocabulary task, a matrix-reasoning task, a visual-memory task and so on.
Tasks that share more with one another may form broad ability factors. Several tasks involving novel relational reasoning, for example, may contribute to a fluid-reasoning factor. Several verbal knowledge measures may contribute to a crystallized-knowledge factor.
Those broad factors are themselves correlated. A higher-order general factor can summarize part of what they share.
A simplified picture looks like this:
Specific tasks → broad abilities → general factor (g)
The arrows should not automatically be read as proven causal pathways. They represent how covariance can be organized statistically.
Are broad abilities “real,” or are they just leftovers after g?
Broad abilities are measurable dimensions in their own right. They capture additional systematic covariance among more specific tests beyond what is summarized by a general factor.
For example, spatial tasks correlate with one another for reasons not completely captured by overall general ability. The same is true for acquired knowledge and other broad domains.
But the relative usefulness of a broad score depends on the question. If we want to predict performance on a highly spatial task, a spatial factor may provide relevant information. If the criterion is broad learning across many domains, general ability may be more informative.
This is a measurement issue, not a competition in which one level must defeat the other.
Does this validate every theory of “multiple intelligences”?
No.
The phrase “multiple intelligences” is used loosely in everyday discussion and specifically in some educational theories. Psychometric evidence that cognitive ability is multidimensional does not automatically validate any proposed list of independent intelligences.
To establish a distinct cognitive ability scientifically, researchers need evidence that it can be measured reliably, shows a coherent pattern across multiple indicators, has discriminant validity from other abilities and contributes explanatory or predictive information.
Simply giving a human talent a name does not make it an independent intelligence.
Why the hierarchy matters in real assessment
Suppose an intelligence battery gives a Full Scale IQ and several broad index scores.
The Full Scale IQ may provide the most reliable summary of overall cognitive performance because it aggregates many observations. Broad indices can describe domains. Individual subtests are narrower still and often less reliable.
The deeper the interpreter travels down the hierarchy, the more cautious they usually need to become. A five-point difference between two subtests may look psychologically interesting but may be ordinary measurement noise. A large and uncommon difference between well-reliabled broad composites, especially if it corresponds to independent evidence, may deserve more attention.
The correct question is not, “Which score is the real intelligence?” It is, “At what level of the hierarchy does the evidence support the inference we want to make?”
General intelligence describes what cognitive abilities share. Broad and specific abilities describe how they differ. Both are parts of the same empirical structure.
A useful analogy is geography. Saying that a location is in Europe can be informative. Saying it is in France gives more specific information. Saying it is in Paris is more specific still. None of those descriptions makes the others false; they answer at different levels.
The analogy is imperfect, because cognitive factors are statistical constructs rather than physical territories. But it captures the central point: human cognitive ability has levels of organization.
References
- Carroll, J. B. (1993). Human Cognitive Abilities. Cambridge University Press.
- McGrew, K. S. (2009). CHC theory and the human cognitive abilities project. Intelligence, 37, 1–10.
- McMillen, P., & Levin, M. (2024). Collective intelligence: A unifying concept for integrating biology across scales and substrates. Communications Biology, 7, 378. DOI
- Schneider, W. J., & McGrew, K. S. (2018). The Cattell-Horn-Carroll theory of cognitive abilities. In Contemporary Intellectual Assessment.
- van der Maas, H. L. J., et al. (2006). A dynamical model of general intelligence. Psychological Review, 113, 842–861. —. DOI
