In 1987, political scientist James Flynn of the University of Otago in New Zealand documented a curious phenomenon: broad intelligence gains in multiple human populations over time. Across 14 countries where decades’ worth of average IQ scores of large swaths of the population were available, all had upward swings—some of them dramatic. Children in Japan, for example, gained an average of 20 points on a test known as the Wechsler Intelligence Scale for Children between 1951 and 1975. In France, the average 18-year-old man performed 25 points better on a reasoning test in 1974 than did his 1949 counterpart.1
Flynn initially suspected the trend reflected faulty tests. Yet in the ensuing years, more data and analyses supported the idea that human intelligence was increasing over time. Proposed explanations for the phenomenon, now known as the Flynn effect, include increasing education, better nutrition, greater use...
A fundamental challenge in understanding the Flynn effect is defining intelligence. At the dawn of the 20th century, English psychologist Charles Spearman first observed that people’s average performance on a variety of seemingly unrelated mental tasks—judging whether one weight is heavier than another, for example, or pushing a button quickly after a light comes on—predicts our average performance on a completely different set of tasks. Spearman proposed that a single measure of general intelligence, g, was responsible for that commonality.
Scientists have proposed biological mechanisms for variations among individuals’ g levels ranging from brain size and density to the synchrony of neural activity to overall connectivity within the cortex. But the precise physiological origin of g is far from settled, and a simple explanation for differences in intelligence between individuals continues to elude researchers. A recent study of 1,475 adolescents across Europe reported that intelligence, as measured by a cognitive test, was associated with a panoply of biological features, including known genetic markers, epigenetic modifications of a gene involved in dopamine signaling, gray matter density in the striatum (a major player in motor control and reward response), and the striatum’s activation in response to a surprising reward cue.2
Understanding human smarts has been made even more challenging by the efforts of some inside and outside the field to introduce pseudoscientific concepts into the mix. The study of intelligence has at times been tainted by eugenics, “scientific” racism, and sexism, for example. As recently as 2014, former New York Times science writer Nicholas Wade drew fire for what critics characterized as misinterpreting genetics studies to suggest race could correlate with average differences in intelligence and other traits. The legitimacy of such analyses aside, for today’s intelligence researchers, categorization isn’t the end goal.
“The reason I’m interested in fluid intelligence tests”—which home in on problem-solving ability rather than learned knowledge—“is not really because I want to know what makes one person do better than another,” says University of Cambridge neuroscientist John Duncan. “It’s important for everybody because these functions are there in everybody’s mind, and it would be very nice to know how they work.”
In search of g
G, and the IQ (or intelligence quotient) tests that aim to measure it, have proven remarkably durable since Spearman’s time. Multiple studies have backed his finding of a measurable correlation among an individual’s performances on disparate cognitive tests. And g interests researchers because its effects extend far beyond academic and work performance. In study after study, higher IQ is tied to outcomes such as greater income and educational attainment, as well as to lower risks of chronic disease, disability, and early death.
Early studies of people with brain injuries posited the frontal lobes as vital to problem solving. In the late 1980s, Richard Haier of the University of California, Irvine, and colleagues imaged the brains of people as they solved abstract reasoning puzzles, which revved up specific areas in the frontal, parietal, and occipital lobes of the brain, as well as communication between them. The frontal lobes are associated with planning and attention; the parietal lobes interpret sensory information; and the occipital lobe processes visual information—all abilities useful in puzzle solving. But more activity didn’t mean greater cognitive prowess, notes Haier. “The people with the highest test scores actually showed the lowest brain activity, suggesting that it wasn’t how hard your brain was working that made you smart, but how efficiently your brain was working.”
In 2007, based on this and other neuroimaging studies, Haier and the University of New Mexico’s Rex Jung proposed the parieto-frontal integration theory, arguing that the brain areas identified in Haier’s and others’ studies are central to intelligence.3 (See infographic.) But Haier and other researchers have since found that patterns of activation vary, even between people of similar intelligence, when performing the same mental tasks. This suggests, he says, that there are different pathways that the brain can use to reach the same end point.
The people with the highest test scores actually showed the lowest brain activity, suggesting that it wasn’t how hard your brain was working that made you smart, but how efficiently your brain was working.—Richard Haier, University of California, Irvine
Another problem with locating the seat of g via brain imaging, some argue, is that our instruments are still simply too crude to yield satisfying answers. Haier’s PET scans in the 1980s, for instance, tracked radiolabeled glucose through the brain to get a picture of metabolic activity during a 30-minute window in an organ whose cells communicate with one another on the order of milliseconds. And modern fMRI scans, while more temporally precise, merely track blood flow through the brain, not the actual activity of individual neurons. “It’s like if you’re trying to understand the principles of human speech and all you could listen to is the volume of noise coming out of a whole city,” Duncan says.
Models of intelligence
Beyond simply not having sharp-enough tools, some researchers are beginning to question the premise that the key to intelligence can be seen in the anatomical features of the brain. “The dominant view of the brain in the 20th century was anatomy is destiny,” says neurophysiologist Earl Miller of MIT’s Picower Institute for Learning and Memory; but it’s become clear over the past 10 to 15 years that this view is too simplistic.
Researchers have begun to propose alternative properties of the brain that might undergird intelligence. Miller, for example, has been tracking the behavior of brain waves, which arise when multiple neurons fire in synchrony, for clues about IQ. In one recent study, he and colleagues hooked up EEG electrodes to the heads of monkeys that had been taught to release a bar if they saw the same sequence of objects they’d seen a moment before. The task relied on working memory, the ability to access and store bits of relevant information, and it caused bursts of high-frequency γ and lower-frequency β waves. When the bursts weren’t synchronized at the usual points during the task, the animals made errors.4
Parsing SmartnessThe biological basis for variations in human intelligence is not well understood, but research in neuroscience, psychology, and other fields has begun to yield insights into what may undergird such differences. One well-known hypothesis, backed by evidence from brain scans and studies of people with brain lesions, proposes that intelligence is seated in particular clusters of neurons in the brain, many of them located in the prefrontal and parietal cortices. Known as the fronto-parietal integration, the hypothesis holds that the structure of these areas, their activity, and the connections between them vary among individuals and correlate with performance on cognitive tasks.
the scientist staff
Researchers have also proposed a slew of other hypotheses to explain individual variation in human intelligence. The variety of proposed mechanisms underlines the scientific uncertainty about just how intelligence arises. Below are three of these hypotheses, each backed by experimental evidence and computational modeling:
the scientist staff
Miller suspects that these waves “direct traffic” in the brain, ensuring that neural signals reach the appropriate neurons when they need to. “Gamma is bottom-up—it carries the contents of what you’re thinking about. And beta is top-down—it carries the control signals that determine what you think about,” he says. “If your beta isn’t strong enough to control the gamma, you get a brain that can’t filter out distractions.”
The overall pattern of brain communications is another candidate to explain intelligence. Earlier this year, Aron Barbey, a psychology researcher at the University of Illinois at Urbana-Champaign, proposed this idea, which he calls the network neuroscience theory,5 citing studies that used techniques such as diffusion tensor MRI to trace the connections among brain regions. Barbey is far from the first to suggest that the ability of different parts of the brain to communicate with one another is central to intelligence, but the whole-brain nature of network neuroscience theory contrasts with more established models, such as parieto-frontal integration theory, that focus on specific regions. “General intelligence originates from individual differences in the system-wide topology and dynamics of the human brain,” Barbey tells The Scientist.
General intelligence originates from individual differences in the system-wide topology and dynamics of the human brain.—Aron Barbey, University of Illinois at Urbana-Champaign
Emiliano Santarnecchi of Harvard University and Simone Rossi of the University of Siena in Italy also argue that intelligence is a property of the whole brain, but they see overall plasticity as the key to smarts. Plasticity, the brain’s ability to reorganize, can be measured via the nature of the brain activity generated in response to transcranial magnetic or electrical stimulation, Santarnecchi says. “There are individuals that generate a response that is only with the other nodes of the same network that we target,” he says.And then there are people in whose brains “the signal starts propagating everywhere.” His group has found that higher intelligence, as measured by IQ tests, corresponds with a more network-specific response, which Santarnecchi hypothesizes “reflects some sort of. . . higher efficiency in more-intelligent brains.”
Despite the hints uncovered about how intelligence comes about, Santarnecchi finds himself frustrated that research has not yielded more-concrete answers about what he considers one of neuroscience’s central problems. To address that shortcoming, he’s now spearheading a consortium of cognitive neuroscientists, engineers, evolutionary biologists, and researchers from other disciplines to discuss approaches for getting at the biological basis of intelligence. Santarnecchi would like to see manipulations of the brain—through noninvasive stimulation, for example—to get at causal relationships between brain activity and cognitive performance. “We know a lot now about intelligence,” he says, “But I think it’s time to try to answer the question in a different way.”
Putting the g in genes
As neuroscientists interrogate the brain for how its structure and activity relate to intelligence, geneticists have approached intelligence from a different angle. Based on what they’ve found so far, psychology researcher Sophie von Stumm of the London School of Economics estimates that about 25 percent of individual variation in intelligence will turn out to be explained by single nucleotide polymorphisms in the genome.
To find genes at play in intelligence, researchers have scanned the genomes of thousands of people. Earlier this year, for example, economist Daniel Benjamin of the University of Southern California and colleagues crunched data on upwards of 1.1 million subjects of European descent and identified more than 1,200 sites in the genome associated with educational attainment, a common proxy for intelligence.7 Because subjects in many types of medical studies in which DNA is sequenced are asked about their educational status to help control for socioeconomic factors in later analyses, such data are plentiful. And while the correlation between education and intelligence is imperfect, “intelligence and school achievement are highly correlated, and genetically very highly correlated,” says von Stumm, who recently coauthored a review on the genetics of intelligence.8 Altogether, the genes identified so far accounted for about 11 percent of individual variation in education level in Benjamin’s study; household income, by comparison, explained 7 percent.
Such genome-wide association studies (GWAS) have been limited in what they reveal about the biology at work in intelligence and educational attainment, as much remains to be learned about the genes thus far identified. But there have been hints, says Benjamin. For example, the genes with known functions that turned up in his recent study “seem to be involved in pretty much all aspects of brain development and neuron-to-neuron communication, but not glial cells,” Benjamin says. Because glial cells affect how quickly neurons transmit signals to one another, this suggests that firing speed is not a factor in differences in educational attainment.
Other genes seem to link intelligence to various brain diseases. For example, in a preprint GWAS published last year, Danielle Posthuma of VU University Amsterdam and colleagues identified associations between cognitive test scores and variants that are negatively correlated with depression, ADHD, and schizophrenia, indicating a possible mechanism for known correlations between intelligence and lower risk for mental disorders. The researchers also found intelligence-associated variants that are positively correlated with autism.9
Von Stumm is skeptical that genetic data will yield useful information in the near term about how intelligence results from the brain’s structure or function. But GWAS can yield insights into intelligence in less direct ways. Based on their results, Benjamin and colleagues devised a polygenic score that correlates with education level. Although it’s not strong enough to be used to predict individuals’ abilities, Benjamin says the score should prove useful for researchers, as it enables them to control for genetics in analyses that aim to identify environmental factors that influence intelligence. “Our research will allow for better answers to questions about what kinds of environmental interventions improve student outcomes,” he says.
Von Stumm plans to use Benjamin’s polygenic score to piece together how genes and environment interact. “We can test directly for the first time,” says von Stumm, “if children who grow up in impoverished families. . . with fewer resources, if their genetic differences are as predictive of their school achievement as children who grow up in wealthier families, who have all the possibilities in the world to grab onto learning opportunities that suit their genetic predispositions.”
The idea of manipulating intelligence is enticing, and there has been no shortage of efforts to do just that. One tactic that once seemed to hold some promise for increasing intelligence is the use of brain-training games. With practice, players improve their performance on these simple video games, which rely on skills such as quick reaction time or short-term memorization. But reviews of numerous studies found no good evidence that such games bolster overall cognitive abilities, and brain training of this kind is now generally considered a disappointment.
Transcranial brain stimulation, which sends mild electrical or magnetic pulses through the skull, has shown some potential in recent decades for enhancing intelligence. In 2015, for example, neurologist Emiliano Santarnecchi of Harvard Medical School and colleagues found that subjects solved puzzles faster with one type of transcranial alternating current stimulation, while a 2015 meta-analysis found “significant and reliable effects” of another type of electrical stimulation, transcranial direct current stimulation (Curr Biol, 23:1449–53).
While magnetic stimulation has yielded similarly enticing results, studies of both electrical and magnetic stimulation have also raised doubts about the effectiveness of these techniques, and even researchers who believe they can improve cognitive performance admit that we’re a long way from using them clinically.
One proven way researchers know to increase intelligence is good old-fashioned education. In a meta-analysis published earlier this year, a team led by then University of Edinburgh neuropsychologist Stuart Ritchie (now at King’s College London) sifted out confounding factors from data reported in multiple studies and found that schooling—regardless of age or level of education—raises IQ by an average of one to five points per year (Psychol Sci, 29:1358–69). Researchers, including University of British Columbia developmental cognitive neuroscientist Adele Diamond, are working to understand what elements of education are most beneficial to brains.
“Intelligence is predictive of a whole host of important things,” such as educational attainment, career success, and physical and mental health, Ritchie writes in an email to The Scientist, “so it would be extremely useful if we had reliable ways of raising it.”
Thinking about thinking
It’s not just the biology of intelligence that remains a black box; researchers are still trying to wrap their minds around the concept itself. Indeed, the idea that g represents a singular property of the brain has been challenged. While g’s usefulness and predictive power as an index is widely accepted, proponents of alternative models see it as an average or summation of cognitive abilities, not a cause.
Last year, University of Cambridge neuro-scientist Rogier Kievit and colleagues published a study that suggests IQ is an index of the collective strength of more-specialized cognitive skills that reinforce one another. The results were based on vocabulary and visual reasoning test scores for hundreds of UK residents in their late teens and early 20s, and from the same subjects about a year and a half later. With data on the same people at two time points, Kievit says, the researchers could examine whether performance on one cognitive skill, such as vocabulary or reasoning, could predict the rate of improvement in another domain. Using algorithms to predict what changes should have occurred under various models of intelligence, the researchers concluded that the best fit was mutualism, the idea that different cognitive abilities support one another in positive feedback loops.10
In 2016, Andrew Conway of Claremont Graduate University in California and Kristóf Kovács, now of Eötvös Loránd University in Hungary, made a different argument for the involvement of multiple cognitive processes in intelligence.11 In their model, application-specific neural networks—those needed for doing simple math or navigating an environment, for example—and high-level, general-purpose executive processes, such as breaking down a problem into a series of small, manageable blocks, each play a role in helping a person complete cognitive tasks. It’s the fact that a variety of tasks tap into the same executive processes that explains why individuals’ performance on disparate tasks correlates, and it’s the average strength of these higher-order processes, not a singular ability, that’s measured by g, the researchers argue. Neuroscientists might make more progress in understanding intelligence by looking for the features of the brain that carry out particular executive processes, rather than for the seat of a single g factor, Kovács says.
As researchers grapple with the intractable phenomenon of intelligence, a philosophical question arises: Is our species smart enough to understand the basis of our own intelligence? While those in the field generally agree that science has a long way to go to make sense of how we think, most express cautious optimism that the coming decades will yield major insights.
“We see now the development, not only of mapping brain connections in human beings . . . we’re also beginning to see synapse mapping,” Haier says. “This will take our understanding of the basic biological mechanisms of things like intelligence . . . to a whole new level.”
- J. Flynn, “Massive IQ gains in 14 nations: What IQ tests really measure,” Psychol Bull, 101:171-91, 1987.
- J.A. Kaminski et al., “Epigenetic variance in dopamine D2 receptor: A marker of IQ malleability?” Transl Psychiat, 8:169, 2018.
- R.E. Jung, R.J. Haier, “The parieto-frontal integration theory (P-FIT) of intelligence: Converging neuroimaging evidence,” Behav Brain Sci, 30:135–87, 2007.
- M. Lundqvist et al., “Gamma and beta bursts during working memory readout suggest roles in its volitional control,” Nat Comm, 9:394, 2018.
- A.K. Barbey, “Network neuroscience theory of human intelligence,” Trends Cogn Sci, 22:8–20, 2018.
- E. Santarnecchi, S. Rossi, “Advances in the neuroscience of intelligence: From brain connectivity to brain perturbation,” Span J Psychol, 19:E94, 2016.
- J.J. Lee et al., “Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals,” Nat Genet, 50:1112–21, 2018.
- R. Plomin, S. von Stumm, “The new genetics of intelligence,” Nat Rev Genet, 19:148–59, 2018.
- J.E. Savage et al., “Genome-wide association meta-analysis in 269,867 individuals identifies new genetic and functional links to intelligence,” Nat Genet, 50:912–19, 2018.
- R.A. Kievit et al., “Mutualistic coupling between vocabulary and reasoning supports cognitive development during late adolescence and early adulthood,” Psychol Sci, 28:1419–31, 2017.
- K. Kovács, A.R.A. Conway, “Process overlap theory: A unified account of the general factor of intelligence,” Psychol Inq, 27:151-177, 2016.