IQ scores summarize performance on a battery of cognitive tasks into a single number, and that number turns out to predict a surprising range of real-world outcomes, from academic achievement to longevity. But treating an IQ score as a fixed, transparent readout of “how smart someone is” misses most of what makes the science interesting. The number is shaped by which test was used, how old the person was when they took it, what their childhood environment looked like, and even the social context of the testing room itself. Understanding what IQ analysis can and cannot tell you requires pulling apart the layers beneath that single number.
What IQ Tests Actually Measure
The core finding behind IQ testing is over a century old and still remarkably sturdy: if you give people a bunch of different cognitive tasks, their scores on those tasks tend to be positively correlated. Someone who does well on vocabulary also tends to do well on spatial reasoning, working memory, and pattern recognition. The statistical overlap among all these tasks is what researchers call the “g factor,” or general intelligence. It was first described by Charles Spearman in the early twentieth century and has remained central to intelligence theory ever since, despite steady criticism from multiple angles.1Learning and Individual Differences. A short history of g: Psychometrics’ most enduring and controversial construct The g factor holds up across different statistical methods, different test batteries, and different populations.2PubMed. The g factor: psychometrics and biology
That said, IQ is not a single measurement the way height is. It is a composite score derived from subtests, and the specific subtests vary depending on which instrument is used. The Wechsler scales, for instance, break cognition into verbal comprehension, perceptual reasoning, working memory, and processing speed, then roll those into a Full Scale IQ. The Stanford-Binet uses a slightly different set of domains. This matters in practice: when 74 adults with intellectual disability were given both tests, the Wechsler Full Scale IQ came in higher every single time, with an average gap of nearly 17 points.3PubMed Central. Stanford-Binet & WAIS IQ Differences and Their Implications for Adults with Intellectual Disability (aka Mental Retardation) A 17-point spread between two well-established tests, taken by the same people, is a reminder that the number on the report depends partly on which test was administered.
What Happens in the Brain
Researchers have spent decades trying to figure out what is physically different about brains that score higher on IQ tests. Two broad findings have emerged. First, there are structural correlates: people with thicker cortex across widely distributed brain regions tend to score higher on general intelligence measures.4PubMed Central. Associations between cortical thickness and general intelligence in children, adolescents and young adults The association is real but not huge, and it spans many areas rather than sitting in one “smart spot.”
Second, there is a functional story. The Parieto-Frontal Integration Theory proposes that intelligence depends heavily on how well frontal and parietal brain regions communicate with each other. A study of nearly 1,500 people confirmed that resting-state connectivity between these regions predicted individual differences in abstract reasoning.5Intelligence. Interindividual differences in matrix reasoning are linked to functional connectivity between brain regions nominated by Parieto-Frontal Integration Theory In other words, it is not just how much brain tissue you have; it is how efficiently different regions talk to each other.
This connects to the “neural efficiency” hypothesis, which holds that higher-scoring individuals use less brain activation to solve the same problems. Electroencephalography studies have found that brighter individuals show lower cortical activation and stronger short-range connectivity in the frontal cortex during cognitive tasks.6Intelligence. Intelligence and neural efficiency: Measures of brain activation versus measures of functional connectivity in the brain But that picture has an important caveat: the efficiency advantage disappears when the task is scaled to each person’s ability level. Once a problem is equally challenging for both a higher-scoring and a lower-scoring person, their brain activation looks similar. The efficiency effect seems to reflect how each brain calibrates its effort to the difficulty it perceives, not a fixed hardware advantage.7PubMed Central. Neural efficiency as a function of task demands
Genes, Environment, and Why the Split Is Not Simple
IQ is heritable, meaning that genetic differences among people account for a meaningful share of the variation in scores. Twin studies show that heritability increases with age in a surprisingly linear way: roughly 41% in childhood, 55% in adolescence, and 66% in young adulthood.8PubMed Central. The heritability of general cognitive ability increases linearly from childhood to young adulthood The rising heritability does not mean genes “turn on” later; it likely reflects the fact that as people gain more control over their own environments, they increasingly choose settings that amplify their genetic predispositions.
Modern genomics has tried to pin down the specific DNA variants responsible. Thousands of variants have been identified, each contributing a tiny effect. Polygenic scores, which sum up all these small effects, can predict some variance in IQ but not much. In one deep-phenotyping study of 557 adults, polygenic scores explained only about 2–5% of the variance in general, verbal, and numerical intelligence, and even less for nonverbal reasoning and memory.9PubMed Central. Polygenic Scores for Cognitive Abilities and Their Association with Different Aspects of General Intelligence-A Deep Phenotyping Approach When researchers look within families, comparing siblings who share a home environment but differ genetically, polygenic score predictions drop by a third to a half compared to population-level estimates. That reduction suggests that some of the “genetic” signal in population studies actually picks up family-level environmental factors like parenting quality, neighborhood, and the mating patterns of the parents themselves.10Intelligence & Cognitive Abilities. Polygenic Score Prediction Within and Between Sibling Pairs for Intelligence, Cognitive Abilities, and Educational Traits From Childhood to Early Adulthood
The gene-environment interplay is most dramatic when you look at socioeconomic status. In one influential U.S. study of young children, genetic contributions to IQ were close to zero in the poorest families, while shared environment accounted for about 60% of the variation. In affluent families, the pattern reversed almost exactly, with genes explaining most of the variation and shared environment shrinking to a small role.11PubMed. Socioeconomic status modifies heritability of IQ in young children A later study found a somewhat different pattern, with genetic influences growing proportionally with SES while environmental influences stayed constant, suggesting that genes may multiply environmental resources rather than just compete with them.12PubMed. Childhood socioeconomic status amplifies genetic effects on adult intelligence A UK-representative sample added a further wrinkle: greater variance in IQ appeared among low-SES families, but evidence for a strong gene-by-environment interaction across ages was minimal.13PubMed Central. Socioeconomic Status and Children’s Intelligence (IQ): In a UK-Representative Sample SES Moderates the Environmental, Not Genetic, Effect on IQ The takeaway from this mixed picture is that the relative weight of genes and environment is not a fixed constant; it shifts depending on where and when you look.
What IQ Predicts in the Real World
IQ scores correlate with a wide range of life outcomes. In education and the workplace, correlations around 0.5 between IQ and job performance have been widely cited. But those correlations often rely on statistical corrections for range restriction and measurement error that inflate the raw numbers, and the underlying data quality varies considerably. One critical review concluded that considerable caution is needed before treating those correlations as validation of the tests.14PubMed Central. Does IQ Really Predict Job Performance? The raw, uncorrected relationships are typically weaker, and other factors like motivation, conscientiousness, and domain-specific knowledge also matter.
The link between IQ and health outcomes is on firmer ground. A systematic review found that every study examining premorbid IQ and later mortality showed the same pattern: higher scores in childhood were associated with lower risk of death in adulthood.15PubMed. Premorbid (early life) IQ and later mortality risk: systematic review This relationship has spawned a subfield called cognitive epidemiology, which tracks how early-life cognitive test scores predict later morbidity and mortality.16PubMed Central. Cognitive epidemiology Part of the explanation is behavioral, since people who score higher tend to have better health literacy and engage in fewer risky behaviors. But a twin study found that the covariance between intelligence and lifespan was roughly 84–95% genetic in origin, suggesting that many of the same genetic variants that support cognitive function also support physical health.17PubMed Central. The association between intelligence and lifespan is mostly genetic
Cultural and Situational Bias in Testing
IQ tests were designed to be as culturally neutral as possible, and nonverbal, visuo-spatial tests like Raven’s Progressive Matrices have long been treated as the gold standard of “culture-fair” assessment. That reputation is increasingly contested. A review of cross-cultural research on visuo-spatial processing found that cultural assumptions are deeply embedded in all such tests, from how people parse visual arrays to how they interpret spatial relationships. The review concluded that these tests cannot be considered intrinsically culture-fair, and average score differences between ethnic groups cannot be cleanly attributed to differences in ability.18PubMed Central. Cross-cultural differences in visuo-spatial processing and the culture-fairness of visuo-spatial intelligence tests: an integrative review and a model for matrices tasks Even the Cattell Culture Fair Intelligence Test, explicitly named for its intended neutrality, shows item-level bias when administered to groups from different schooling traditions: about a quarter of items functioned differently for Nigerian and American students.19Applied Psychological Measurement. A Cross-Cultural Analysis of the Fairness of the Cattell Culture Fair Intelligence Test Using the Rasch Model
Bias is not only baked into test items; it can also be activated by the testing situation itself. When people belong to a group stereotyped as intellectually inferior, simply framing a test as a measure of cognitive ability is enough to depress their scores. In one study using heart rate variability as a physiological marker, this performance drop was linked to an increase in disruptive mental workload. Crucially, when the identical test was presented without the “cognitive ability” label, the gap between targeted and non-targeted groups disappeared.20PubMed. Stereotype threat undermines intellectual performance by triggering a disruptive mental load Score differences between groups, in other words, can reflect different situational burdens rather than different abilities.
Can You Raise Your IQ?
The question of whether IQ can be trained has produced a frustrating back-and-forth. One widely cited 2008 study reported that training on a demanding working memory task transferred to gains on fluid intelligence tests, and that the gains were dose-dependent: more training sessions led to bigger improvements.21PubMed Central. Improving fluid intelligence with training on working memory That finding generated enormous excitement and a boom in commercial “brain training” products. But subsequent meta-analytic work tempered the enthusiasm considerably. A large meta-analysis of working memory training studies found no convincing evidence of improvements on measures of far transfer, which includes nonverbal ability, verbal ability, reading comprehension, and arithmetic. The training gains were specific and short-lived, confined to the trained tasks themselves. Even the degree of improvement on working memory did not predict whether any broader cognitive gains appeared.22PubMed Central. Working Memory Training Does Not Improve Performance on Measures of Intelligence or Other Measures of “Far Transfer”: Evidence From a Meta-Analytic Review
This does not mean IQ is totally immovable. Environmental improvements like better nutrition, reduced lead exposure, and expanded access to education have been associated with population-level score gains over decades. But the dream of a quick, targeted brain-training program that reliably boosts general intelligence has not panned out so far. Getting better at a specific cognitive exercise mostly makes you better at that exercise.
The Flynn Effect and Its Reversal
One of the most striking findings in IQ research is that average scores have risen substantially over the twentieth century. This generational climb, known as the Flynn effect, amounts to roughly 2–3 IQ points per decade. A meta-analysis of 285 studies found an overall gain of about 2.3 standard score points per decade, and the effect for modern Wechsler and Stanford-Binet tests specifically was about 3 points per decade.23PubMed Central. The Flynn effect: a meta-analysis Three points per decade sounds modest until you consider cumulative effects: over a century, that would be a 30-point shift, or two full standard deviations.
The most likely drivers include improved nutrition, greater exposure to abstract thinking through education and technology, and smaller family sizes allowing more cognitive stimulation per child. Genetic change over such short timescales is implausible as the main explanation. The Flynn effect has practical consequences for IQ-based decisions: because test norms become outdated as scores drift upward, a person tested with an older edition of a test will appear smarter than if tested with a newer edition using updated norms. This is especially consequential in high-stakes settings like intellectual disability diagnoses or special education placement.
More recently, though, the Flynn effect has shown signs of stalling or reversing in several countries. Data from Scandinavia showed test performance improving at a decelerating rate through the late 1990s, followed by some decline.24Personality and Individual Differences. A long-term rise and recent decline in intelligence test performance: The Flynn Effect in reverse Austrian data from 2005 to 2018 confirmed a similar pattern of stagnation and reversal.25PubMed Central. Increasing IQ Test Scores and Decreasing g: The Flynn Effect and Decreasing Positive Manifold Strengths in Austria (2005-2018) Why the gains are flattening is debated, and no single explanation dominates the field.
How IQ Changes as You Age
IQ is often treated as a stable trait, and at a broad level it is: your score at age 11 is a decent predictor of your score at age 70. But this stability masks important divergences between different types of cognitive ability. Fluid intelligence, the capacity for novel problem-solving and abstract reasoning, peaks in early adulthood and declines with age. Crystallized intelligence, which reflects accumulated knowledge and vocabulary, tends to remain stable or even grow into later life.26PubMed Central. Fluid intelligence and gross structural properties of the cerebral cortex in middle-aged and older adults: A multi-occasion longitudinal study
These two trajectories used to be considered fairly independent, but recent longitudinal modeling tells a more intertwined story. In people around age 50, when fluid declines are already underway but crystallized ability is still rising, the individual rates of change in the two domains are strongly correlated. People whose fluid abilities decline faster tend to see their crystallized gains slow or reverse sooner as well.27PubMed Central. A strong dependency between changes in fluid and crystallized abilities in human cognitive aging Older research had noted that the correlation between fluid and crystallized scores is much lower in elderly samples than in young adults, suggesting the two abilities “decouple” with age.28Journal of Gerontology. Fluid and Crystallized Intelligence in Young Adulthood and Old Age The newer longitudinal data complicates that picture by showing that the rate at which one changes still tracks the rate of the other, even if their absolute levels diverge.
IQ at the Extremes
At the high end of the distribution, IQ scores predict outcomes that are genuinely remarkable in scale. A long-running Duke University identification program tracked 259 young adolescents with IQs at or above 160. By age 40, about 37% had earned doctorates, roughly 7.5% held academic tenure, and 9% held patents.29PubMed. When Lightning Strikes Twice: Profoundly Gifted, Profoundly Accomplished A 20-year Australian longitudinal study of 60 children with IQs of 160 and above found that educational acceleration made a substantial difference in outcomes. Those who were allowed to skip two or more years of schooling reported high life satisfaction, pursued research degrees at leading universities, and developed strong social relationships. Equally able children who were held back or minimally accelerated tended to enter less rigorous programs, reported lower life satisfaction, and often struggled socially.30Journal for the Education of the Gifted. Exceptionally Gifted Children: Long-Term Outcomes of Academic Acceleration and Nonacceleration That finding cuts against the common assumption that pushing gifted children ahead academically harms their social development. In this sample, the opposite was true.
What IQ Does Not Capture
IQ tests measure a specific set of cognitive abilities, but they leave out entire domains of thinking that matter in daily life. One of the most striking gaps involves cognitive biases. People with high IQ scores are not reliably better at avoiding common reasoning errors like anchoring, framing effects, or base-rate neglect. In one study, performance on cognitive bias tasks was largely independent of both fluid and crystallized intelligence, with correlations below 0.25. Bias measures defined their own distinct factors, suggesting that rational thinking and general intelligence are substantially separate capacities.31Intelligence. Individual differences in cognitive biases: Evidence against one-factor theory of rationality A person can be excellent at pattern recognition and verbal reasoning while still making systematically poor decisions because they are vulnerable to the same mental shortcuts as everyone else.
General Intelligence in Other Animals
The g factor is not unique to humans. Researchers have asked whether the positive correlation among different cognitive tasks found in people also appears in other species, and the answer is a qualified yes. In dogs, even within a single breed tested on detour tasks and choice tasks, faster and more accurate dogs on one type of task tended to be faster and more accurate on the others. The best statistical model included a general intelligence factor that accounted for about 17% of the variance in task scores.32Intelligence. A general intelligence factor in dogs A study of cotton-top tamarins, a New World monkey, used Bayesian modeling and similarly found that a general factor explained a small but consistent portion of variance across tasks, with every task loading positively onto it.33PLOS ONE. General Intelligence in Another Primate: Individual Differences across Cognitive Task Performance in a New World Monkey (Saguinus oedipus)
That said, a broad meta-analysis covering 555 bivariate correlations across many species found that the average correlation among cognitive abilities in non-human animals was low, around 0.19.34PubMed Central. How general is cognitive ability in non-human animals? A meta-analytical and multi-level reanalysis approach General intelligence exists in other species, but it is weaker and more domain-fragmented than in humans. Whether that reflects a genuine species difference in cognitive architecture or just the difficulty of designing equivalent tests across organisms with wildly different sensory worlds and motivational structures is an open question that the field has not resolved.