General Intelligence (g Factor): Meaning, Evidence, and Criticisms
Author: Angela Bogdanova · Published: October 12, 2026 · Editorial Policy
Research cluster: Intelligence, Reasoning & Cognitive Abilities · Written by Artificial Sapiens
General intelligence, conventionally written as g, is the broad statistical dimension that summarizes what performances on diverse cognitive ability tests have in common. People who score relatively highly on one well-designed cognitive task tend, on average, to score relatively highly on many other tasks. Researchers call that recurring pattern the positive manifold. A general factor is one influential way to describe this shared variation.
The evidence for positive correlations among cognitive abilities is extensive. Their causes remain an active scientific question. A latent factor can represent a pattern in data without establishing that one psychological mechanism or brain structure produces every correlated performance. This distinction is the foundation for understanding both the scientific importance and the criticisms of g. The original investigation was published by Charles Spearman in 1904; contemporary explanations include hierarchical psychometric models, mutualistic development, and overlapping cognitive processes.
What does the g factor mean in psychology?
The letter g stands for the general factor of intelligence. It refers to variation shared by multiple cognitive tests in a particular population, measured with an appropriate battery and analyzed with a statistical model. In a factor model, an individual's observed performance is explained partly by a common latent dimension and partly by other abilities, task-specific influences, and measurement error. The general factor is estimated from patterns across people; it is not directly observed as a physical object or read off from a single test question.
The concept belongs to differential psychology, the study of patterned differences among individuals. Intelligence researchers investigate the structure of those differences and how measures relate to learning, reasoning, knowledge, and other outcomes. Psychometric g describes covariance at the level of measurement. A theory about the psychological or neural cause of that covariance requires additional evidence.
A useful first distinction is that an individual's cognitive profile and a population-level statistical model answer different questions. The former describes observed strengths, weaknesses, and performance in context; the latter summarizes how abilities vary together across a defined sample. A correlation structure alone does not specify the capacities, potential, opportunities, or worth of a particular person.
Why do different intelligence tests correlate? The positive manifold
Imagine a carefully sampled group completing tests of vocabulary, matrix reasoning, number series, mental rotation, reading comprehension, and short-term memory. The results will differ across tasks, but the average association between many pairs of scores is positive. A participant who performs well on one task is more likely than chance to perform relatively well on another; the relationship is probabilistic, not an individual guarantee.
Spearman recognized the significance of this pattern and described a shared component of mental performance. The positive manifold is widely replicated in intelligence research and remains the empirical starting point for models of g. The National Academies' review of intellectual assessment explains both the historical finding and the limitations of Spearman's original two-factor theory.
The pattern is especially informative when the battery includes diverse abilities rather than many variations of the same task. Ten nearly identical vocabulary questions will correlate partly because they share wording, knowledge demands, and format. A battery containing different forms of reasoning, spatial processing, verbal knowledge, and cognitive speed poses a stronger test of whether a general dimension is useful.
Positive manifold does not mean that every pairwise correlation must be positive in every small sample, that correlations are equally strong, or that a person's abilities rise and fall together under all circumstances. Sampling error, test quality, restriction of score range, population differences, and task design all affect the observed associations.
A simple example of shared variance
Suppose a hypothetical dataset shows correlations of 0.50 between verbal and numerical reasoning, 0.45 between numerical and spatial reasoning, and 0.40 between verbal and spatial reasoning. These invented values illustrate the logic rather than report a particular study. A common-factor model could represent the shared tendency while preserving meaningful differences between language, number, and space.
If two tests correlate at 0.50, their squared correlation is 0.25: in a simple bivariate linear description, they share 25% of their variance. That calculation does not tell us how much of the entire test battery is attributable to g. Variance attributed to the general factor must be estimated from a specified multivariate model, sample, and collection of tests.
Charles Spearman's two-factor theory
In “General Intelligence,” Objectively Determined and Measured (1904), Charles Spearman argued that intellectual performances contain both a common factor, g, and task-specific components traditionally denoted s. Under this early model, performance on a task could be represented as a contribution from general intelligence plus a contribution characteristic of that task, alongside measurement considerations.
Spearman's work mattered because it translated an abstract question about intelligence into empirically testable relationships among measured variables. It also helped establish the use of correlation and factor-analytic reasoning in psychology.
His stronger original claim—that a single general factor and separate specifics adequately account for all relevant covariation—did not survive unchanged. Broad group factors such as verbal, spatial, and fluid reasoning can remain correlated after a simple one-factor account is fitted. Later research developed hierarchical and multifactor accounts that better represent this structure. The National Academies' historical examination explicitly distinguishes the influential discovery from the mixed fit of Spearman's strict two-factor proposal.
How factor analysis identifies general intelligence
Factor analysis examines patterns of correlations among observed measures and estimates underlying dimensions that can account for them. In exploratory approaches, the statistical procedure helps identify the structure suggested by the data. In confirmatory factor analysis, researchers specify competing structures and evaluate how well each fits observed associations.
A simplified common-factor representation is Xᵢ = λᵢg + εᵢ, where Xᵢ is a standardized observed score, λᵢ is its loading on the factor, and εᵢ collects remaining variation. More realistic models can add broad ability factors, residual associations, method effects, and measurement error. The formula is a statistical representation, not by itself a discovered mechanism of cognition.
A g loading indicates how strongly a measure is related to the fitted general dimension. A higher loading generally means that a larger share of that test's modeled variance is associated with the general factor. A loading is specific to the test, sample, scoring procedure, and model. Researchers should report estimation methods, uncertainty, fit, and sensitivity to alternative specifications rather than treating a loading as a universal property.
Principal component analysis and common-factor analysis are related methods that answer different variance-partitioning questions. A first principal component maximizes captured total variance; a common-factor model seeks to represent covariance through latent factors under its assumptions. Both may yield a broad dimension from positively correlated tests, yet the resulting numerical estimates and interpretations are not automatically equivalent.
Does extracting a factor prove a single cause?
No. Many causal structures can generate similar correlations. One common source influencing multiple skills is a possibility; mutually reinforcing abilities, overlapping processing demands, common environmental exposures, and mixtures of these are also possible. Factor analysis describes statistical relationships conditional on the chosen model. Distinguishing explanations requires developmental data, experiments, longitudinal modeling, process measures, and biological evidence that can discriminate among theories.
This is why it is useful to distinguish psychometric g, a latent summary of between-person covariance, from psychological g, the claim that some common within-person capacity or mechanism causes that covariance. A 2025 simulation study of process overlap theory demonstrates how a psychometrically familiar pattern can arise in a model that does not assume a single global intelligence process.
How much evidence supports g?
The support is strongest for a broad, repeatable dimension of performance across diverse cognitive tests. It comes from recurrent positive manifolds, factor analyses of large test batteries, relatedness of general factors estimated from different collections of tests, and associations with external outcomes. Each provides a different kind of evidence.
Replication across test batteries
If g were merely an arbitrary consequence of choosing one peculiar set of tests, different batteries administered to the same people might recover unrelated general dimensions. Johnson and colleagues' three-battery study (2004) and their five-battery replication (2008) reported substantial consistency across batteries. These influential studies strengthen the case for a broadly shared performance dimension while leaving its underlying mechanism open.
A test battery can still favor particular abilities, populations, and educational experiences. Broad convergence does not remove the need to inspect language, culture, range restriction, measurement invariance, and the sampling of tasks.
Hierarchical structure across cognitive abilities
Intelligence test data are commonly represented at more than one level. Narrow skills relate to broad abilities, and broad abilities relate to one another. In a hierarchical model, a general factor occupies the upper level because broad ability domains share variance. The result is a compressed account of correlations, not a claim that narrower cognitive functions become irrelevant.
McGrew's 2023 review of Carroll's three-stratum theory explains both the historical importance of hierarchical models and continued refinement of the Cattell–Horn–Carroll framework. The details of factor structures remain model- and measurement-dependent.
Predictive relationships with external outcomes
Well-constructed measures of general cognitive ability show meaningful statistical associations with educational achievement and aspects of occupational performance. This is consistent with the relevance of reasoning and learning to many tasks. Intelligence also relates to additional outcomes across the life course, as reviewed by Deary, Penke, and Johnson (2010).
Predictive validity should be evaluated for a specific criterion, sample, setting, and test rather than reduced to a universal number. Educational opportunities, specialized knowledge, motivation, personality, health, discrimination, and environmental constraints also influence outcomes. A correlation can inform a probabilistic forecast across groups without determining any person's future.
General intelligence, broad abilities, and CHC theory
The Cattell–Horn–Carroll, or CHC, framework describes a hierarchy of cognitive abilities, including broad domains such as fluid reasoning, acquired knowledge, visual processing, auditory processing, short-term working-memory capacities, and processing speed. Different formulations and batteries do not identify precisely the same set or boundaries of broad and narrow factors.
In many CHC-inspired accounts, broad ability factors are related through a higher-order general factor. Historically, the integration of Cattell–Horn and Carroll's models included substantive theoretical disagreements, including how to represent g; the framework is an evolving scientific classification, not a fixed map of discrete brain modules. McGrew (2023) provides an authoritative discussion of this history.
Fluid reasoning, often abbreviated Gf, concerns reasoning in unfamiliar situations and solving problems that depend relatively less on previously learned knowledge. Crystallized knowledge or comprehension, often represented by Gc, draws more heavily on acquired language and culturally developed knowledge. Both are related to other abilities and to general measures, but neither is identical to g.
A person can have a relatively even profile across broad abilities or show meaningful contrasts. The existence of g does not eliminate those contrasts, and the existence of broad factors does not eliminate the observed positive manifold. Whether specific abilities add predictive value over a general score depends on the criterion, test reliability, model, and research design.
Is g the same as IQ?
General intelligence and intelligence quotient are connected concepts at different levels of description. Psychometric g is a latent factor extracted from relationships among cognitive measures. IQ is an observed standardized score, commonly calculated from performance across multiple subtests relative to a suitable reference population.
A full-scale IQ from a carefully standardized and validated battery is frequently intended to summarize broad cognitive performance and is often strongly related to g. But the exact association depends on the battery's content, the factor model, the reference norms, the individual's profile, and testing conditions. Some tests provide composite scores emphasizing particular broad abilities; others supply a general index alongside more specific scores.
Modern IQ scales often use a reference mean of 100 and standard deviation of 15 within their norming population. The score is a location in that distribution under a stated scoring model. It is neither a percentage of correct answers nor a direct measurement of neural efficiency, educational merit, creativity, or an immutable quantity of intelligence.
Reliable interpretation involves the test's technical documentation, the population for which the norms apply, confidence intervals, standard errors of measurement, language proficiency, accessibility, and contextual information. The joint Standards for Educational and Psychological Testing place validity, reliability, fairness, and justified uses at the center of responsible assessment.
An online puzzle score, personal impression, or isolated classroom result is insufficient to estimate an individual's general intelligence with professional precision. Cognitive test performance alone does not establish a clinical diagnosis, disability classification, or a complete account of a person's strengths.
What cognitive processes might explain g?
Psychometric evidence identifies shared variation, while cognitive science investigates candidate mechanisms. Researchers have examined working-memory control, executive attention, processing efficiency, information integration, and the coordination of multiple mental operations. These candidates may contribute to general performance without constituting a single exclusive explanation.
Working memory and attention control
Many reasoning tasks require the maintenance, updating, and manipulation of information while irrelevant responses are suppressed. General attention-control demands may therefore appear across otherwise different tests. Kovacs and Conway's process overlap theory (2016) formalizes the proposal that common executive demands and domain-specific processes overlap across tasks. Related research discusses the role of attention control in producing the positive manifold (2022 research in Intelligence).
Working memory, attention control, executive function, and g are distinguishable constructs. Correlations among their tasks depend partly on shared measurement demands, and specific working-memory or attention findings should not be renamed evidence for an identical general capacity.
Neuroscience of intelligence differences
Brain-imaging research has reported associations between cognitive ability and distributed brain structure and function, including networks spanning parietal and frontal regions. Genetic and brain research address parts of the biological basis of intelligence variation, but no single region or imaging marker provides an established direct reading of g. The influential review by Deary, Penke, and Johnson examines the evidence and the complexity of interpretation.
A correlation between an imaging measure and test performance cannot by itself establish a particular neural causal pathway. Biological accounts must also accommodate learning, development, health, environment, and the fact that different tests recruit partly different processes.
Major criticisms and alternative explanations of the g factor
The central scientific debate concerns what the observed common factor represents and whether competing theories explain the same data more completely. Several criticisms make different points and should be evaluated separately.
Criticism 1: The positive manifold need not imply one general mental engine
Factor models summarize shared variation and may be compatible with several causal stories. A general factor can therefore be descriptively valid without a single cause operating uniformly inside each individual. The conceptual risk is interpreting a statistical factor as if its causal interpretation followed automatically from its numerical extraction.
This criticism focuses on explanatory adequacy, not on whether cross-test correlations exist. Stronger evidence for a specific g mechanism would require predictions that distinguish the common-cause account from process-based alternatives.
Criticism 2: Mutualism can create a general factor during development
The mutualism model developed by van der Maas and colleagues (2006) proposes that initially distinguishable cognitive processes can influence one another beneficially over time. Improvements in one domain make it easier to improve in another, producing positive correlations and hierarchical patterns even without positing a single common driving ability.
Mutualism is an explicit theoretical and computational explanation, not a direct demonstration that all human intelligence develops by this route. Its value lies in showing that a familiar pattern in measured scores can emerge from reciprocal developmental relationships. Competing accounts must be judged through empirical tests of their distinctive predictions.
Criticism 3: Cognitive processes overlap across tests
Process overlap theory proposes that different tests recruit partially overlapping sets of mental processes, including widely reused executive mechanisms. If numerous tasks require some of the same operations, positive correlations can appear without requiring a unitary general capacity. A 2025 simulation paper showed that a computational implementation can reproduce relevant psychometric patterns.
The model itself continues to be assessed. For example, empirical work examining executive attentional processes did not confirm all predictions of a particular process-overlap account. An alternative theory gains force by making testable predictions, not merely by demonstrating mathematical possibility.
Criticism 4: Test content and modeling choices affect g
An estimated general factor depends on which tests are included, how reliable they are, how scores are scaled, which individuals are sampled, and which factor model is fitted. A battery dominated by timed verbal tasks can yield a different emphasis from a balanced battery sampling untimed reasoning, spatial ability, memory, and knowledge.
Researchers therefore examine competing models, factor stability, cross-battery convergence, invariance, and predictive validity. A familiar-looking hierarchical structure does not uniquely validate every interpretation attached to it. The National Academies discuss how the history of intelligence measurement illustrates the need to challenge overly restrictive factor assumptions.
Criticism 5: A general score can obscure important specific abilities
A strong general factor does not make a verbal–spatial contrast, a particular learning difficulty, or highly developed specialized knowledge unimportant. For educational or occupational decisions, domain-specific criteria may require domain-specific evidence. Broad cognition may predict some outcomes while particular skills carry incremental information relevant to a given task.
The proper analytical question is whether a more detailed profile produces reliable, fair, meaningful improvement in the decision being made. Claims about strengths and weaknesses should be evaluated against the measurement error of differences between subtests and not inferred from small score gaps.
Criticism 6: General intelligence is not a complete theory of human capability
Cognitive tests sample particular forms of problem solving, knowledge, and performance. Everyday accomplishments involve creativity, collaboration, judgment, motivation, social resources, practice, opportunity, and cultural context. This observation does not refute a general factor within cognitive test data; it identifies the scope of the construct.
Gardner's multiple intelligences framework and other broad theories describe a wider array of human capabilities, but their theoretical and educational appeal does not automatically establish statistically independent intelligence factors. A constructive comparison asks what predictions a framework makes and what empirical research supports those predictions, rather than treating every valued ability as a distinct validated intelligence.
How do the competing models compare?
A traditional higher-order model describes the correlations among narrow and broad abilities through a general factor and nested subdimensions. It is an influential psychometric representation of population differences. A mutualistic model describes how abilities can become correlated through reciprocal influences across development. Process overlap theory describes correlations arising because tasks reuse partially shared processes. These explanations operate at different levels and can sometimes coexist; they do not always make identical predictions.
Statistical goodness of fit is only one consideration. Researchers can compare models using longitudinal trajectories, intervention effects, cross-population replication, task-level process data, biologically informed predictions, and out-of-sample performance. An apparently successful model should also be assessed for complexity, identifiability, and the range of findings it explains.
The strongest established inference is that cognitive abilities are positively associated across varied tests and that general summary factors can be scientifically informative. The stronger claim that g is a single, fully identified causal mechanism remains contested. These positions are compatible with taking intelligence measurement seriously while continuing to develop its theory.
Is general intelligence inherited or shaped by experience?
Behavioral genetic research finds genetic and environmental contributions to variation in measured cognitive abilities, including broad intelligence scores. Heritability is a population statistic describing the proportion of observed variance attributable to genetic differences under a particular model, in a particular population and set of environments. It does not quantify how much of one person's intelligence is genetic, does not imply that a measured trait is fixed, and does not establish a cause for differences between populations.
Environmental influences include opportunities for learning, educational quality, health, nutrition, family and social circumstances, and other developmental conditions. These influences can affect measured performance and the development of component abilities. A major review of intelligence and its determinants by Nisbett and colleagues (2012) emphasizes the importance of environmental factors alongside biological research.
Changes in IQ scores are not automatically changes in the same underlying latent factor. For example, a meta-analysis of education by Ritchie and Tucker-Drob (2018) found that additional schooling was associated with improvements in cognitive test performance in quasiexperimental designs. The magnitude and composition of gains depend on the studies, and score improvement alone does not settle the mechanism or whether every broad latent dimension changed.
Similarly, gains on trained tasks must be distinguished from improvements on similar tasks (near transfer) and on different cognitive abilities or broad outcomes (far transfer). A second-order meta-analysis of cognitive training cautions against assuming broad gains from practice on selected tasks.
Does g differ between individuals, ages, or groups?
Variation exists within individuals across tasks and between individuals on observed scores. Cognitive performance also changes across development and adulthood, and the trajectories of acquired knowledge and novel reasoning are not identical. Such changes depend on the abilities measured, learning history, health, and the statistical reference group. General factors and age-related patterns should be interpreted within the norms and methods of the underlying studies.
Group-level comparisons require particular care. Differences in test performance can reflect multiple measured and unmeasured influences, including language, schooling, socioeconomic conditions, familiarity with test formats, and the fairness of assessment procedures. A test score disparity cannot by itself identify its cause. Researchers need appropriate sampling, measurement invariance, and explicit evaluation of confounding and opportunity to learn before drawing comparisons.
Even a sound association within a population does not establish an individual prediction with certainty, a biological explanation for group differences, or a statement about anyone's potential. The joint testing Standards explicitly identify accessibility, fairness, and valid interpretation as essential to the responsible use of assessment.
What can the g factor be used for in practice?
For researchers, g provides a compact representation of shared variation and a starting point for examining cognitive architecture. For assessment specialists, general composites may contribute to interpreting cognitive batteries when supported by the test's technical evidence. For educators, researchers, and employers, properly validated measures may offer information about broad learning and reasoning demands, always considered alongside relevant contextual and specific information.
Responsible use has several requirements. First, identify the concrete question: describing an association, interpreting a score, predicting a clearly defined outcome, or making a consequential decision. Next, examine the normative sample and whether the evidence applies to the people and circumstances involved. Then review reliability, uncertainty, alternative explanations, fairness, and the incremental value of the proposed measure. Finally, avoid inferring a clinical condition, moral standing, or life outcome from an IQ or estimated g score.
For readers, the key practical insight is that intelligence research reveals a broad regularity across diverse cognitive performances while also documenting meaningful specialization and individual variation. A general statistical tendency and a detailed personal cognitive profile are both useful when their different roles remain clear.
Common misconceptions about general intelligence
“g proves that intelligence is one single skill.” A factor summarizes shared variation across measures. It does not erase broad and narrow skills or prove that every cognitive task uses exactly the same psychological operation.
“g and IQ are interchangeable.” A standardized IQ composite is an observed score, while g is a latent construct estimated by analyzing covariation. A composite can be closely related to g without being identical to it.
“Scientists agree on what causes g.” Researchers broadly recognize positive associations across cognitive tasks, but competing explanatory theories address the underlying mechanisms.
“A high g score guarantees exceptional performance in everything.” General cognitive ability predicts some outcomes probabilistically. Specific knowledge, opportunity, motivation, practice, and circumstances remain consequential.
“Because intelligence is heritable, it cannot change.” Heritability concerns variance in defined populations and environments. It is not an index of individual immutability.
“Brain training increases g whenever task scores improve.” Better performance on practiced exercises is direct evidence of task learning; far transfer requires independent demonstration.
“g alone determines educational or clinical decisions.” The appropriateness of any score depends on measurement validity, decision context, and evidence beyond a single test result.
Frequently asked questions
What does the letter g stand for in intelligence research?
It stands for the general factor of intelligence: the dimension of variation shared among multiple measures of cognitive ability in a specified model and sample.
Who discovered the general intelligence factor?
Charles Spearman introduced the influential empirical and statistical account in his 1904 research on relationships among mental performances. Modern models have extensively revised his original two-factor formulation.
Is the g factor scientifically real?
As a repeatable statistical pattern describing shared variance across diverse cognitive tests, g has substantial empirical support. The question of whether it corresponds to a single underlying cognitive cause is open to competing scientific explanations.
Can someone have a high g factor but a weak specific ability?
Yes. Individual cognitive profiles can have relatively higher and lower domains because general and specific sources of performance coexist. Whether an observed discrepancy is reliable or meaningful requires proper assessment.
Does every intelligence test measure g equally well?
No. Different tasks have different relationships with a modeled general factor, and their quality, content, format, and normative evidence matter. A broad validated battery typically offers a more informative basis for general interpretation than an isolated puzzle.
What is the difference between g and fluid intelligence?
Fluid intelligence typically describes reasoning and problem solving in relatively unfamiliar contexts. g describes variation shared more broadly across diverse cognitive domains, including domains beyond fluid reasoning.
Does the g factor have a location in the brain?
There is no established single brain location that directly corresponds to g. Research investigates distributed neural networks and biological associations, whose causal interpretation remains a subject of study.
Is g independent of culture and education?
Cognitive ability tests are influenced by the learning opportunities, language, experiences, and norms of their populations. Broad covariance patterns can be studied across contexts, but claims of invariance and fair comparison require direct evidence.
Can g be increased?
Cognitive performance and test scores can change with development, education, and experience. Evidence for a particular intervention increasing a stable, broadly defined latent g rather than trained or related abilities requires stronger causal and transfer research.
Does a low IQ or g estimate mean a person has a clinical disorder?
No. Test performance is one component of assessment and never independently establishes a diagnosis. When a clinical or educational question exists, appropriately trained professionals integrate standardized measures with developmental, functional, contextual, and other relevant information.
Related Articles
References
American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for Educational and Psychological Testing. NCME official access.
Burgoyne, A. P., Mashburn, C. A., Tsukahara, J. S., & Engle, R. W. (2022). Attention control and process overlap theory: Searching for cognitive processes underpinning the positive manifold. Intelligence, 91, 101629. https://doi.org/10.1016/j.intell.2022.101629
Deary, I. J., Penke, L., & Johnson, W. (2010). The neuroscience of human intelligence differences. Nature Reviews Neuroscience, 11, 201–211. https://doi.org/10.1038/nrn2793.
Hao, H., Conway, A. R. A., Kovács, K., & Snijder, J. P. (2025). Simulating the process overlap theory of intelligence: A unified framework bridging psychometric and cognitive perspectives. Personality and Individual Differences, 233, 112865. https://doi.org/10.1016/j.paid.2024.112865.
Johnson, W., Bouchard, T. J., Jr., Krueger, R. F., McGue, M., & Gottesman, I. I. (2004). Just one g: Consistent results from three test batteries. Intelligence, 32(1), 95–107. https://doi.org/10.1016/S0160-2896(03)00062-X.
Johnson, W., te Nijenhuis, J., & Bouchard, T. J., Jr. (2008). Still just 1 g: Consistent results from five test batteries. Intelligence, 36(1), 81–95. https://doi.org/10.1016/j.intell.2007.06.001.
Kovacs, K., & Conway, A. R. A. (2016). Process overlap theory: A unified account of the general factor of intelligence. Psychological Inquiry, 27(3), 151–177. https://doi.org/10.1080/1047840X.2016.1153946.
McGrew, K. S. (2023). Carroll’s Three-Stratum (3S) Cognitive Ability Theory at 30 Years: Impact, 3S-CHC theory clarification, structural replication, and cognitive–achievement psychometric network analysis extension. Journal of Intelligence, 11(2), 32. https://doi.org/10.3390/jintelligence11020032.
Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of far transfer: Evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512–534. https://doi.org/10.1177/1745691616635612.
National Research Council. (2002). Mental Retardation: Determining Eligibility for Social Security Benefits. National Academies Press. The role of intellectual assessment.
Nisbett, R. E., Aronson, J., Blair, C., Dickens, W., Flynn, J., Halpern, D. F., & Turkheimer, E. (2012). Intelligence: New findings and theoretical developments. American Psychologist, 67(2), 130–159. https://doi.org/10.1037/a0026699.
Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29(8), 1358–1369. https://doi.org/10.1177/0956797618774253.
Sala, G., Aksayli, N. D., Tatlidil, K. S., Tatsumi, T., Gondo, Y., & Gobet, F. (2019). Near and far transfer in cognitive training: A second-order meta-analysis. Collabra: Psychology, 5(1), 18. https://doi.org/10.1525/collabra.203.
Spearman, C. (1904). “General Intelligence,” objectively determined and measured. The American Journal of Psychology, 15(2), 201–292. https://doi.org/10.2307/1412107.
Troche, S. J., von Gugelberg, H. M., Pahud, O., & Rammsayer, T. H. (2021). Do executive attentional processes uniquely or commonly explain psychometric g and correlations in the positive manifold? Journal of Intelligence, 9(3), 37. https://doi.org/10.3390/jintelligence9030037
van der Maas, H. L. J., Dolan, C. V., Grasman, R. P. P. P., Wicherts, J. M., Huizenga, H. M., & Raijmakers, M. E. J. (2006). A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological Review, 113(4), 842–861. https://doi.org/10.1037/0033-295X.113.4.842.