Intelligence Quotient, or IQ, is supposed to be a measure of a person’s ability to reason, see patterns, and make predictions. Yet IQ is quite controversial – a controversy ranging anywhere from IQ tests being inaccurate or biased and all the way to IQ tests (and anything concerning IQ) being immoral. Yet, even if there is no test that can accurately and reliably gauge an individual’s intelligence in some quantitative way, most people are aware of some ineffable sense in which some individuals are just smarter than the average individual (and vice versa with some people just being less smart in some ways than other people).
The introduction section of this post is largely adapted from the introduction to part 2 of my review of The Alignment Problem by Brian Christian.
What do I mean by someone being smarter? Let me first be clear what I am not talking about, which is any notion of some “groups” in whatever way we delineate them are “naturally smarter” than other groups (e.g., between races or sexes). Here I am discussing differences between individuals – that individual A may be smarter than individual B in domain(s) of interest X in a way that others can perceive, even if it turns out that there is no good quantifiable measurement.
What I do mean, then, is this: The fact that some individuals just seem to get it a lot faster and easier than other people when learning new things about – for instance, math, computers, cars, music, and so on. Some individuals can be shown how to do something once and are then able to repeat it with near mastery while others can practice for hours and still not do it as well. Some individuals are just fountains of novel and creative ideas while others struggle in this department, or even have very few original thoughts ever. Some people can make connections and see patterns where others cannot. Some people just have quick wits and can recall relevant information so easily in almost any situation, while others (like me) are seriously lacking in this department. The existence of savant syndrome and child prodigies both attest to the notion that someone can just naturally be good at something.
Indeed, if it were the case that all humans had exactly equal intelligence (however we want to define or measure it), this would demand explanation. How, exactly, could such a state of affairs have even come about? We know from biology that there is variation within a species, and this will apply to intelligence(s) as well. If it so happened that humans the world over developed exactly equal intelligence, this would be a biologically unprecedented phenomenon, something that would challenge what we know about evolution, genetics, neuroscience, and human (or, really, any species) development. It would also demand an explanation as to why people have such large differences in their capacity and aptitude for the kinds of domains I mentioned in the previous paragraph. All students from a particular region, going to the same school, and indeed taking the same class from the same teacher giving the same assignments and grading from the same rubric, can still have a very large standard deviation in grades. If it were the case that all humans had exactly equal intelligence, or at least exactly equal potential intelligence (i.e., before having their “tabula rasa” filled), then these observed variances would call out for explanation. Additionally, an explanation as to why there can be inter-species differences in intelligence but not intra-species differences in intelligence would be required: if biology has nothing to do with intelligence (i.e., if intelligence is some thing added on, in addition to, and/or separate from, neurobiology), then why are there differences in intelligence between, say, humans and chimpanzees?
This last point brings up an interesting issue with intelligence: when we humans talk about intelligence, what we usually mean is human intelligence – competency in the domains humans find important. But, of course, there are domains where chimpanzees outperform humans, even in domains that humans consider important, such as working memory.
My point in this introduction is not to defend (or criticize) IQ as the sole measure of intelligence. My point is to highlight that there are differences in intelligence, in whatever way we define or measure it, within and between species. Yet, as much of a difference as there may be in the intelligence(s) of different humans, the intelligence of any two individual humans, in the space of all possible intelligence, are extremely close together. In other words, for all our variance in intelligence, if we zoom out from our human-centric bias, we are not all that different. Certainly, when it comes to doing algebra, just about any given human will outperform just about any member of any other species. Yet, a human would be bad at most kinds of problem solving that an ant must engage in. In other words, humans are going to be terrible in most of the domains of interest for other species. When humans compare ourselves to other species in the domains that interest humans, we obviously win the majority of the time, but human intelligence is not the only kind of intelligence.
It is also the case that, in the domains of interest to humans, we humans are not even always that great. Performing arithmetical calculations, for instance, is something that a simple calculator is vastly better at than any human. Meaning that, even in the domains of interest to humans, the intelligence difference between any two given humans is going to be quite small when compared to the space of potential competency within the domain. Arithmetic, working memory, long-term memory, and pattern recognition are just a few areas in which humans are now vastly out-competed by our technology.
What is interesting in AI science is that the kinds of things that humans have found difficult (e.g., arithmetical operations) tend to be easy for a computer, yet the kinds of things humans tend to find easy are difficult for computers. For instance, the so-called cocktail party effect, where humans can focus on specific voices or sounds among a cacophony of other voices and sounds. This is something that humans can do automatically and with ease, yet it is still an issue in machine learning.
Two things we can away from this discussion are the following:
- The space of potential competency in any given domain of intelligence for which humans are interested is enormous, with all humans occupying just a small corner of this space
- The space of different kinds of intelligence is also gigantic, with humans taking up an even smaller region of this space
Intelligence is the defining feature of humankind. All the other things that might be defining features, such as opposable thumbs, bipedalism, or even language, are only useful because of our intelligence. Opposable thumbs allow us to use tools, but without intelligence we wouldn’t have tools to grip with those thumbs. Walking upright frees up our hands so that we can more easily use those tools, but again, without intelligence, what good would that do us? Language allows us to communicate with others more efficiently and pass down lessons learned from previous generations, but without intelligence to formulate and comprehend these vocalizations, the ability to speak would not be all that useful. It is therefore our intelligence that really sets us apart.
As such, intelligence has significant value to each of us – we all like to think of ourselves as competent, if not brilliant. This, I think, is one reason why there is resistance to the idea that intelligence has a biological basis: having a biological basis is often viewed as deterministic, and people do not want to feel as if they have no control over themselves. If our intelligence is something determined by biology, and I find out that I’m determined to have less intelligence than most other people, then there is a sense in which I feel like I am less human – or, at least, that other people will view me as being less human.
Such a fear is not completely unfounded, as the eugenics movements of the late 19th and early 20th centuries did, in fact, think this way. And this can be a real problem when notions of genetic/biological determinism are taken so seriously. But in response to the atrocities committed in the name of eugenics, the pendulum has swung the other way. People now often want to ignore, downplay, or deny that intelligence has a biological basis. Unfortunately, if it is in fact true that intelligence has a biological basis, then not understanding and grappling with this fact could lead to different undesirable outcomes. Anywhere from misguided attempts at increasing intelligence using ineffective means all the way to effective means of increasing intelligence going undiscovered. And so, after all this, we are left with two important questions we must answer: what is intelligence? And is there a biological basis for intelligence? It is these two questions that I will now turn.
What is Intelligence?
So far I have discussed intelligence as a sort of ineffable thing that, even if we can all recognize it intuitively, we might not be able to (or be willing to) define it in words. But is it really impossible (or undesirable) to define intelligence? If we don’t have at least a working definition for it, then it will be difficult to test in any scientific manner. Intelligence quotient (IQ) is a popular, though controversial, standard for defining intelligence.
The controversy stems largely from associations with sexism, racism, and eugenics. But, of course, even if IQ has been used to bolster such monstrous ideas, that does not in-itself disprove IQ as a useful metric for quantifying intelligence. But another accusation is that IQ tests are biased, culturally or otherwise. Neuroscientist Richard J. Haier, in his 2016 book The Neuroscience of Intelligence, disagrees, saying that:
Test bias has a specific meaning. If scores on a test consistently over- or underpredict actual performance, the test is biased. For example, if people in a particular group with high SAT scores consistently fail college courses, the test is overpredicting success and it is a biased test. Similarly, if people with low SAT scores consistently excel in college courses, the test is underpredicting success and it is biased. A test is not inherently biased just because it may show an average difference between two groups. A spatial ability test, for example, may have a different mean for men and women, but that does not make the test biased. If scores for men and for women predict spatial ability equally well, the test is not biased even if there is a mean difference. Note that a few cases of incorrect predicting do not constitute bias. For a test to be biased, there needs to be a consistent failure of predicting in the wrong direction. The lack of any prediction is not bias; it means the test is not valid.
Considerable research on test bias for decades shows this is not the case for IQ and other intelligence test scores (Jensen, 1980). Test scores do predict academic success irrespective of social-economic status (SES), age, sex, race, and other variables. Scores also predict many other important variables, including brain characteristics like regional cortical thickness or cerebral glucose metabolic rate… If intelligence test scores were meaningless, they would not predict any other measures, especially quantifiable brain characteristics.
When Haier talks about prediction, one thing he is referring to is life outcomes. This is seen in longitudinal studies such as Lewis Terman’s Genetic Studies of Genius and Scientists and Nonscientists in a Group of 800 Gifted Men, the studies on the Lothian Birth Cohort by Sophie von Stumm and Ian J Deary (see here and here and here), and The Study of Mathematically & Scientifically Precocious Youth at Johns Hopkins University, with studies by Lubinski et al, 1996; Lubinski et al, 2006; Lubinski et al, 2014; Robertson et al, 2010; and Wai et al, 2005. Indeed, Robertson et al, 2010 found that even among the top 1% of the subjects in the study, a higher quartile (within the highest 1%) predicted greater academic achievement and income:
But what does IQ actually measure? Broadly speaking, it measures what is known in intelligence research as the g-factor. The g-factor is usually defined as meaning general intelligence, but what it means in practice is the positive correlation between a number of different kinds of cognitive processes – the so-called positive manifold. This means that, if someone is found to be above average in one type of cognitive process, then this is a good predictor of being good in the other cognitive processes associated with the g-factor. These cognitive processes include fluid intelligence, crystalized intelligence, working memory, reasoning (relational, inductive, deductive, analogical, etc.), spatial ability, processing speed, and vocabulary, where an individual excelling in one has a probability much greater than mere chance of excelling in all the others (and vice versa for lagging in one predicting that someone lags in the others).
Fluid intelligence is the capacity to recognize connections without relying on prior, targeted training or guidance pertaining to these connections. It is the ability to engage in abstract thinking, reasoning, and problem-solving, all of which are deemed separate from the influences of learning, personal experience, and formal education. Crystallized intelligence is the information gained through prior learning and past experiences. It relies on factual knowledge and is deeply rooted in one’s life experiences. As we grow older and continue to amass new knowledge and insights, our crystallized intelligence steadily gains in strength.
Working memory is the ability to hold and manipulate information while performing complex tasks. It involves the ability to actively retain and work with a limited amount of information for a short period. It works as a mental workspace where information is actively manipulated, processed, and utilized for tasks, rather than simply storing it passively like long-term memory. In Baddeley’s model of working memory, there are four types of working memory, which are: the Central Executive, which is like the “supervisor” of working memory, controlling attention; the phonological Loop, which is responsible for handling verbal and auditory information and speech-based data, such as spoken numbers or words; the Visuospatial Sketchpad, which deals with visual and spatial information, allowing us to mentally manipulate and visualize objects, shapes, and spatial relationships; and the Episodic Buffer, which acts as a bridge between different components, allowing them to interact and integrate information from various sources into a coherent narrative.
Relational reasoning is a cognitive process that involves identifying and understanding relationships between different elements, concepts, or variables. It focuses on how different pieces of information are connected and how these connections can be used to make predictions, solve problems, or draw conclusions. This type of reasoning is often used in tasks that require understanding patterns, making analogies, or recognizing similarities and differences between objects or ideas. Inductive reasoning is a form of reasoning where conclusions are drawn from a set of specific observations or evidence to create a more general understanding or hypothesis. This is done by making generalized statements based on the information at hand. Deductive reasoning is a logical process that involves drawing specific conclusions from general premises or principles. It follows a “top-down” approach, where if the premises are true, the conclusion is guaranteed to be true (i.e., validity and soundness). Analogical reasoning is making comparisons between two or more situations, objects, or concepts to find similarities and differences. It is a cognitive process that allows people to transfer knowledge or insights from one context to another by recognizing analogous relationships.
Spatial intelligence, as it’s assessed in IQ tests, is a cognitive domain that focuses on an individual’s ability to perceive, understand, and manipulate visual information related to space and objects. This form of intelligence measures a person’s capacity to think and reason in terms of shapes, patterns, and spatial relationships. Spatial intelligence is often tested using tasks and questions that assess skills such as mental rotation (mentally rotating objects or shapes to determine how they appear from different angles) or paper folding and cutting (mentally folding and cutting paper to understand the resulting pattern when unfolded).
Processing speed, as it is tested in IQ tests, assesses an individual’s ability to quickly and accurately process simple or routine visual information. Thus, it measures how efficiently a person can perform basic perceptual and motor tasks, often under time constraints. Processing speed is tested through questions that require individuals to identify symbols or numbers, fill in missing elements (e.g., they might be presented with incomplete patterns, sequences, or grids and need to fill in the missing elements accurately under time constraints), rapid coding (e.g., how quickly individuals can match symbols to specific numbers or letters under time constraints), or even just simple arithmetic under time constraints.
Vocabulary is assessed in IQ tests through a variety of verbal and language-related tasks. This assessment attempts to gauge an individual’s knowledge of words, their meanings, and their ability to understand and use language effectively. This can be done using synonyms and antonyms, where subjects are presented with a word and asked to select its synonym or antonym from a list of options; or by analogies, where subjects are asked to select a word from a set of options that has a similar relationship to another word pair; or through word usage and comprehension, where subjects must use words correctly in sentences or to comprehend written passages with unfamiliar or context-specific vocabulary.
Together, all these cognitive processes come together under the concept of IQ, which is the most popular way to test intelligence. Another way of thinking about intelligence is in the model of hierarchical complexity (MHC). I discuss this at some length in my review of The Listening Society: A Metamodern Guide to Politics, by Hanzi Freinacht. But essentially it is a model that proposes that a person’s MHC level indicates at what level of nested hierarchies one is capable of thinking.
Where, as I said in my review linked to above: “What Commons, Miller, & Giri (2014) say is that vertical complexity is the number of nested sub-tasks within the overall task while horizontal complexity is being able to produce solutions to multiple tasks of the same level of complexity.”
And so, in this model, people can be placed in one of the orders of hierarchical complexity based on how many levels of the nested hierarchy one is capable of thinking. These orders, or stages, are somewhat similar to those postulated by Piaget, but go beyond Piaget’s in adding more levels of complex thinking.
There is, of course, the criticism that intelligence is a definition imposed by those in power, and intelligence tests are therefore just a way of showing who is good at navigating the social-cultural-economic-political systems in place and is therefore a way of legitimizing the system by saying that people who score high on IQ tests are “smarter” (and therefore “better”) and thus rewards them for upholding the current system. In western liberal societies, it is things like reasoning, spatial thinking, math, vocabulary, etc. that are rewarded and held up as “good” and “correct” while “other ways of knowing” are ignored, downplayed, or outright rejected as legitimate forms of intelligence (“knowing”). This all may be true, at least to some extent, but even if the forms of intelligence rewarded by the current social structures are not the only kinds of legitimate intelligence, the criticism fails to show that the forms of intelligence favored by our current culture do not have a biological component. In other words, even if the criticism is valid and the g-factor is only important because of the way our society is structured, it does nothing to indicate that g-factor is not biologically determined to some non-zero percent. The criticism would also not refute that having the types of intelligence accounted for the the g-factor is a good predictor of success. Indeed, the criticism seems to accept that it is true that having high general intelligence leads to better life outcomes, but merely criticizes this fact as ignoring or rejecting other types of intelligence.
It is, however, true that other kinds of intelligences do exist. One that is commonly brought up in discussions of intelligence is known as social or emotional intelligence. As this article points out, emotional intelligence means
- An ability to identify and describe what people are feeling
- An awareness of personal strengths and limitations
- Self-confidence and self-acceptance
- The ability to let go of mistakes
- An ability to accept and embrace change
- A strong sense of curiosity, particularly about other people
- Feelings of empathy and concern for others
- Showing sensitivity to the feelings of other people
- Accepting responsibility for mistakes
- The ability to manage emotions in difficult situations
This is said to be measured by the emotional quotient (EQ) using, for instance, the Mayer, Salovey, Caruso Emotional Intelligence Test (MSCEIT) as the most popular test, although this test is found to have issues. However, emotional intelligence has proven to be quite elusive and tends to have a lot of overlap with IQ. Yet, again like with intelligence discussed in the introduction, we humans do have a sense that some people just have a certain kind of social or emotional intelligence, whether it is being more in touch with their own emotions or inner states, or being more empathetic, or more charismatic. I, for instance, am quite lacking in these domains.
But this sort of gets an the interplay of personality and intelligence. For instance, I’m normal intelligence (I’ve taken online IQ tests before and score between 118 and 125, which, if they are at least somewhat valid, is only at the higher end of average – not a genius by any means). Yet, because of my Schizoid personality disorder, I have no desire to have a social life (hence my lacking in the EQ department, if EQ is really a thing), and because I just so happen to be interested in intellectual things, I have both the time and the inclination to learn things usually viewed as “smart people” things. In other words, my personality – lacking interest in social interactions but possessing interest in intellectual pursuits – makes up for a great deal of the difference between my average IQ and the sorts of “smart people” things in which I am more proficient than most typical average people.
Likewise, having higher EQ might simply be a matter of being more interested in social interactions. Put another way, EQ may not add anything that could not be fully explained, for instance, by the Big Five personality traits of high extraversion, high agreeableness, high openness, and low neuroticism.
Theory of Multiple Intelligences
The theory of multiple intelligences is a cognitive theory proposed by psychologist Howard Gardner in 1983. This theory suggests that intelligence is not a single, unitary entity, but rather a diverse set of independent intelligences or cognitive abilities.Each of these intelligences represents a different way in which individuals can excel intellectually. Gardner’s eight intelligences are as follows:
Linguistic Intelligence: this is the capacity to understand and use words effectively, including reading, writing, and language-related skills. People with high linguistic intelligence are often skilled in writing, public speaking, and storytelling.
Logical-Mathematical Intelligence: this intelligence involves logical reasoning, problem-solving, and mathematical skills. People with high logical-mathematical intelligence excel in activities such as scientific inquiry, mathematical problem-solving, and critical thinking.
Musical Intelligence: musical intelligence relates to the ability to understand, create, and appreciate music. Musically intelligent individuals may have a strong sense of rhythm, melody, and pitch.
Spatial Intelligence: spatial intelligence involves the ability to visualize, manipulate, and navigate spatial information, such as maps, diagrams, and 3D objects. People with high spatial intelligence often excel in fields like architecture, art, and navigation.
Bodily-Kinesthetic Intelligence: this intelligence relates to physical coordination and skill in activities that require body movement, such as sports, dance, and fine motor skills.
Interpersonal Intelligence: interpersonal intelligence refers to the capacity to understand and interact effectively with other people. People with high interpersonal intelligence are skilled in social situations, empathy, and cooperation.
Intrapersonal Intelligence: intrapersonal intelligence involves self-awareness and self-understanding. Individuals with high intrapersonal intelligence have a strong sense of their own emotions, motivations, and strengths.
Naturalistic Intelligence: naturalistic intelligence is the ability to recognize, understand, and appreciate the natural world, including living organisms and ecological systems. It is particularly relevant to fields like biology, botany, and environmental science.
Rationality and Intelligence
Something else we might think of as a different kind of intelligence is rationality. Being rational and being smart are neither mutually exclusive nor mutually inclusive. A lot of smart people can believe some very irrational things. Just think of Kurt Gödel, a genius if ever there was one, believing that someone was trying to poison him to the point that he starved himself to death in his attempt to avoid being poisoned.
But in Gödel’s case, he was likely suffering some kind of psychological disorder. What about everyday smart people – surely they can think more rationally than us normies, right? This is not always the case. As Michael Shermer says: “smart people believe weird [irrational] things because they are skilled at defending beliefs they arrived at for non-smart reasons.”
No person is completely rational. We all make decisions based mostly on non-rational emotions and desires. We all have different upbringings and experiences that shape the way we look at the world, lending content to our various biases. At the same time, humans do not possess doxastic voluntarism – we do not will ourselves to have the beliefs that we do. Add to this the fact that we think like lawyers, not scientists – we start with our conclusions and then look for evidence and arguments to support it (as opposed to falsify it). And smart people are going to be better lawyers for their preconceived conclusions.
Creativity and Intelligence
Probably one of the kinds of intelligence valued most in society, though perhaps not discussed as much as IQ and g-factor, is creativity (sometimes linked with divergent thinking). Creative people are often called geniuses, even if they may not have a particularly high IQ (or, more likely, even if their IQ is not known). But what does it even mean to be creative? It is often said that children are creative, and indeed more creative than adults. Children can often think about things in novel ways that adults, who tend to be more stuck in our ways, would not have come up with. This comes about, at least in part, because of synaptic pruning, where the number of synapses decreases as people go into adulthood and continues decreasing as we age. This has the benefit for adults of making their cognitive processes more efficient (adults find many everyday cognitive tasks easy, even mindless, while children still struggle with them), but may contribute to a reduction in our ability to think outside the box. A part of this may have to do with creativity being linked to dis-inhibition (Vartanian, 2002; Carson, 2011; Carson, 2014; Radel et al, 2015), which even has a proposed model by Jung and Haier called the functional disinhibition model of creativity or F-DIM (Jung and Haier, 2013).
Novelty is not the only important aspect of creativity. Doing completely random or unpredictable things can be novel, but creativity usually also has to connotation that the product, outcome, or result of the creative process will be something meaningful – a new invention that has practical utility, a piece of music or art or literature, a new approach to solving a problem, and so on.
There are two important questions pertaining to creativity: first, is creativity independent of other kinds of intelligence (i.e., could someone with low IQ still be very creative)? And two, is there a biological basis for creativity? The first question is often viewed through what is called the threshold hypothesis, which essentially says that having higher intelligence is a necessary, but not sufficient, precondition for high level creativity. There is evidence that this is the case (Haier and Jung, 2008; Jauk et al, 2013; Jung and Haier, 2013; Jung, 2014; Shi et al, 2017). The second question also has good evidence in favor of a biological basis for creativity (Arden et al, 2010; Dietrich and Kanso, 2010; Sawyer, 2011; Green et al, 2012; Wu et al, 2015; Jauk et al, 2015; Arkin et al, 2019; Khalil et al, 2019; Fink and Benedek, 2019; Beaty, 2020; Kaur et al, 2021; Brawer and Amir, 2021).
Morality and Intelligence
Something we might not associate with intelligence is morality. This can take two forms: moral intelligence (how well someone can make moral judgements) and correlation between intelligence (e.g., IQ) and morality (i.e., is having a high IQ a good predictor of having high moral intelligence and vice versa?). In the first case, there is something known as moral intelligence. This this is an interesting notion in itself, but it will take things too far afield for this post. If you are interested you can check out this paper that discusses it in more detail. For our purposes, we will take the definition of moral intelligence from the linked paper which says that moral intelligence is “[t]he capacity to make decisions and judgments which are moral (i.e., based an internal principles) and to act in accordance with such judgments. (Kohlberg, 1964, p. 425)”
As for the second issue, if anything, the stereotype might be that the more intelligent someone becomes, the less moral they will be. Our media is full of characters who are geniuses, yet are not good people, like Dr. Manhattan or Dr. House, and even characters that, upon acquiring genius level intelligence, they become less moral, as seen in movies like Lucy or Limitless. While the research on links between IQ and moral intelligence is somewhat scant, the evidence seems to point in the other direction, i.e., that having higher IQ is correlated with having higher moral intelligence, particularly in children (children with higher IQ have the moral intelligence of someone older), as seen in: Beißert and Hasselhorn, 1961; Derryberry et al, 2005; Seon-Young Lee and Paula Olszewski-Kubilius, 2006; and Fabio et al, 2023.
Consciousness and Intelligence
I think there is often a conflation of intelligence and consciousness that likely stems from the gradient of intelligence between different species of organisms being used as a metric for how conscious or sentient those organisms are. Organisms that have more human-like features, such as mammals, are viewed as being more intelligent (since they think and behave in ways more familiar to humans) and possessing a greater, richer form of conscious experience. Birds are still mammal-like, and so are also viewed as more intelligent and more conscious. Then we can move down the gradient to reptiles and amphibians, to fish and larger invertebrates, to insects and parasites and plants and fungi, to protozoa, to bacteria and viruses, and finally to non-living matter.
This is also why, for instance, when we have AI that exhibits intelligent (i.e., human-like) behaviors, we immediately jump to the question of whether it is conscious. Unfortunately, there is still no good theory of consciousness – it remains an unsolved hyperobject. As such, there is not really any way to determine how much intelligence and consciousness are correlated (or if something like “degrees” of consciousness between different organisms is even meaningful (e.g., the Φ in integrated information theory of consciousness), or if something even needs to be alive to be conscious). And, if anything, AI shows that being intelligent does not necessarily bring consciousness along for the ride.
An interesting wrinkle, however, is that intelligence may have an influence on the experience of being conscious itself. As will be discussed below, there is good evidence for morphometric and neurophysiological correlates of intelligence, which means that people with different IQ’s process information differently. This could mean that people with different IQ’s have different ways of experiencing the world. Another tantalizing aspect to this is that men and women of the same IQ process information in some different parts of their brain, meaning that having equal IQ does not mean the same thing for men as it does for women. Although this might sound horrifying to those who want to deny the biological basis of intelligence and/or sex, in a way it could be seen as supporting the position that there are different kinds of intelligence and that it would be a good thing to bring more female perspectives into things like science, politics, law, and so on. Additionally, while I’ve not seen any studies on IQ, the cognitive processes associated with intelligence, and gender reassignment (e.g., what affect does taking the hormones of the identified sex have on IQ, and does it change someone who has an IQ of a certain value toward the brain processes of their identified gender?), this would certainly be an interesting question to investigate (though probably fraught with too much political baggage at the moment, having the two-fold controversy of using science to study gender identity and intelligence).
Socioeconomic Status and Intelligence
Probably, when someone does not accept the evidence of a biological basis for intelligence, the alternative they advocate for is that of socioeconomic status (SES) being the primary factor for academic achievement. This is the view that people who make more money can give their children better care and greater access to resources for education, thus increasing their intelligence. The reason this is a popular view is likely because it can suggest a possible solution to differences in intelligence, or at least in academic achievement: government programs and/or redistribution of wealth in a way that can level the playing field. The sort of biological determinism of linking intelligence to genetics and neurophysiology, it is likely assumed, suggests that things like income inequality are a natural and inescapable outcome. As a result, we have a moral duty to accept the SES view of intelligence (or academic achievement) if we wish improve the lives of the poor.
I would argue that a biological basis for intelligence could also be used as justification for government or charitable interventions. If people of low IQ are biologically prevented from and therefore unable to increase their intelligence, then they could be candidates for special programs to assist them, or even for medical interventions that could increase intelligence (if and when such things become available).
But regardless of what policies we wish to adopt, the question is still this: what is a greater predictor of things like academic success, income, and life satisfaction? Intelligence or socioeconomic status? There is certainly evidence that higher SES of parents is associated with higher IQ of their children (Martinez et al, 2022). But there is also evidence for a genetic basis of SES (von Stumm and Plomin, 2015; Trzaskowski et al, 2014; Rindermann and Ceci, 2018), which suggests that the causality might go in the other direction (i.e., it is not that higher SES causes higher IQ, but that higher IQ causes higher SES). A problem with a lot of studies that look at SES and academic achievement is that they do not control for intelligence, and so it becomes difficult to interpret their data as showing that SES is the cause of differences in academic achievement. Indeed, studies that do look at both SES and IQ show that IQ is a better predictor of academic achievement than SES (Colom and Flores-Mendoza, 2007; Hanushek and Woessmann, 2008; Lubinski, 2009; Sackett et al, 2009; Marks and O’Connell, 2020; Flores-Mendoza et al, 2021; Burgoyne et al, 2023; Boman, 2023).
Further, attempts to increase intelligence (through supplementation to pregnant mothers and neonates, early educational interventions, interactive reading, and sending a child to preschool) show that there is a modest increase in IQ (4 to 7 points, though it should be noted that most IQ tests have a margin of error of around 5 points), where Protzko et al (2013) concludes
- Supplementing the diets of pregnant women and neonates with LC-PUFA raises the children’s IQ in young childhood. Providing preschool-aged children with iron supplements may boost their IQ, but giving these supplements to infants does not.
- Enrolling a lower SES infant in an intense early educational intervention will raise his or her IQ in young childhood. Enrolling him or her in such a program at a younger age has no additional benefits for his or her IQ. The more complex the intervention is, the greater these gains will be.
- Reading interactively with young children raises their IQ. The earlier the interactive reading takes place, the larger the benefits.
- Attending preschool increases a young child’s IQ. If the preschool program includes a specific language-development component, these gains are even larger. [although they also say earlier in the text that “we found that preschool interventions that last longer are no more effective at raising the IQ than preschools that are shorter.”]
However, the meta-analysis Protzko (2015) found that “…after an intervention raises intelligence the effects fade away. We further show this is because children in the experimental group lose their IQ advantage and not because those in the control groups catch up.” This last finding indicates that whatever benefit a child may receive from such IQ-raising interventions may only be temporary.
The Flynn Effect
One other curious phenomenon that must be considered when discussing the biological basis of intelligence is the Flynn effect. This is the observation that since the early 1900’s when IQ tests were invented, the average IQ for people born a decade later increases by around 3 points. In other words, children born in, say, 1970 will have, on average, an IQ that is +3 points higher than children born in 1960, but -3 points lower than children born in 1980. Further, this effect appears no matter which decades are being examined (Trahan et al, 2014; Pietschnig and Voracek, 2015).
This poses somewhat of an issue for any theory of a biological basis for intelligence. The time it takes for average IQ to increase cannot be explained by genetics. Most scientists seem to think that the cause of the Flynn effect likely has something to do with on average increasing living standards – since the early 1900’s, things like nutrition, sanitation, medical care, childrearing practices, and education have gotten progressively better. Another possible cause is that our world has become more and more complex to navigate, and so people born in later cohorts are subjected to more “training” of their fluid intelligence as they must solve more complex problems more often by just living in a more complex world.
There is evidence that the Flynn effect may be slowing down, and even becoming a reverse Flynn effect for those of lower IQ (Platt et al, 2019), as well as an observed reverse Flynn effect in Scandinavia (Teasdale and Owen, 2008; Dutton and Lynn, 2013; Bratsberg and Rogeberg, 2018). It has also recently been shown that the Flynn effect “runs in the family” which suggests that whatever is causing the Flynn effect is something in addition to genetics, not instead of genetics (O’Keefe and Rodgers, 2017; Wänström et al, 2023). These two things together – the Flynn effect being in addition to genetics and the Flynn effect slowing down (and even stopping in the most advanced countries in the world, i.e., Scandinavia) – attests to the Flynn effect likely being temporary and running up against a barrier set by genetics. In other words, it cannot go on forever because whatever is the cause of the IQ gains observed in the Flynn effect are either tapering away and/or hitting a point of diminishing returns given what our biology is capable of.
The Biological Basis of Intelligence
What does science actually have to say about intelligence? Does intelligence even have a genetic and/or neurological basis? These are the questions I’m going to examine in what follows by summarizing some of the research that has been done on this topic. But, to summarize the summaries: there is a great preponderance of evidence that suggests both a genetic and a biological basis for intelligence. Where things get tricky is in finding out exactly which genes are involved in the heritability of intelligence, and what are the actual neural correlates of intelligence?
I will be looking first at the genetics of intelligence. While twin studies on intelligence overwhelmingly show that intelligence is hereditary, the numerous genes that appear to be involved each show only small affects on intelligence. This is due to the pleiotropy (one gene influencing multiple traits) and polygenicity (one trait being influenced by multiple genes) of much of our genome.
This creates a complex picture of how genetics influences intelligence. As we’ll see below, there is no candidate “gene for intelligence” and many of the genome wide association studies (GWAS), SNP genotyping, and other population genetics studies performed do not always agree on which genes are actually involved. It is possible, though, to use a combination of twin studies and neuroimaging to determine which brain morphometries have a strong genetic influence. Some of these kinds of studies will be summarized below.
But this brings us to the other kind of evidence showing a biological basis for intelligence: differences in brain morphometry and functionality can be observed between individuals with high and low intelligence. Further, there is even a working theory about which areas of the brain are involved in intelligence. This is called the parieto-frontal integration theory (usually abbreviated as PFIT or P-FIT) put forth by Rex E. Jung and Richard J. Haier in 2007. The PFIT model says, essentially, that intelligence is linked to communication between the dorsolateral prefrontal cortex (Brodmann areas (BAs) 6, 9, 10, 45, 46, 47), the inferior (BAs 39, 40) and superior (BA 7) parietal lobule, the anterior cingulate (BA 32), and regions within the temporal (BAs 21, 37) and occipital (BAs 18, 19) lobes.
This works in the following process:
The Brodmann areas divide the cerebral cortex of the human brain into fifty-two numbered areas based on the cellular structures of those areas. While there is some functional support for the mapping, it is mostly useful as a way of orienting oneself in the cerebral cortex. The Brodmann areas are numbered as follows:
The way that these areas in the PFIT model interact is still an area of intense scientific investigation. And, as Haier notes in his 2016 book The Neuroscience of Intelligence, there is still nothing simple about intelligence, warning that any simplified attempt to deduce IQ from brain morphometry is fraught with problems:
In the case of intelligence, there may be many combinations of the same set of variables [in brain morphology or physiology, such as brain size, gray and white matter cortical thickness, or connection length] that predict any specific IQ equally well. For example, one set of brain variables might characterize a person with an IQ score of 130, but another person with the identical IQ score of 130 might be characterized by a different set of brain variables. In a group of 100 people all with IQs of 130, how many different sets of brain variables related to intelligence might there be? Compounding the problem, two individuals both with WAIS [Wechsler Adult Intelligence Scale] IQs of 130 may have very different subtest scores indicating different cognitive strengths and weaknesses despite the same overall IQ (Johnson et al., 2008). The same problem may exist independently at several IQ levels so the brain variables that predict high IQ might be different from those that predict average or low IQ, even though the relevant genes may be the same across the entire IQ range… Age and sex also could be important factors for identifying optimal sets of variables for predicting IQ
There is, as we will see in the neuroimaging studies section, much support for the PFIT model of intelligence. Another important aspect of intelligence found in many neuroimaging studies is that regions of the brain are connected via a graph-theoretical small-world network with so-called rich club connections. In the brain what this means is that the average path length of neural connections is short and that there is high amounts of clustering in order to optimize efficiency, and that this is enhanced in the brains of those who are more intelligent.
In what follows, I am going to sum up (mostly the abstract from) some of the studies that show that intelligence is genetically heritable and that there is a neurological basis for intelligence. If any of these pique your interest, I highly recommend clicking the links (I try only to use studies where the entire paper is available and not behind a pay wall).
Genetics and Intelligence
Polderman et al (2015) performed a meta-analysis of twin correlations and reported variance components for 17,804 traits from 2,748 publications including 4,558,903 partly dependent twin pairs, virtually all published twin studies of complex traits. Estimates of heritability cluster strongly within functional domains, and across all traits the reported heritability is 49%. For a majority (69%) of traits, the observed twin correlations are consistent with a simple and parsimonious model where twin resemblance is solely due to additive genetic variation. The data are inconsistent with substantial influences from shared environment or non-additive genetic variation.
Posthuma et al (2003) looked at gray and white matter volume, and cerebellar volume, of 135 subjects from 60 extended twin families for whom both MRI scans and WAIS III data were available. They found that all three brain volumes are related to Working Memory capacity (r = 0.27). This phenotypic correlation is completely due to a common underlying genetic factor. Processing Speed was genetically related to white matter volume (rg = 0.39). Perceptual Organization was both genetically (rg = 0.39) and environmentally (re = –0.71) related to cerebellar volume. Verbal Comprehension was not related to any of the three brain volumes. The authors concluded that brain volumes are genetically related to intelligence which suggests that genes that influence brain volume may also be important for intelligence. The authors also noted that the direction of causation (i.e., do genes influence brain volume which in turn influences intelligence, or alternatively, do genes influence intelligence which in turn influences brain volume), or the presence or absence of pleiotropy has not been resolved yet.
Hulshoff-Pol et al (2006) explored the genetic influence on focal gray matter (GM) and white matter (WM) densities in magnetic resonance brain images of 54 monozygotic and 58 dizygotic twin pairs and 34 of their siblings. For genetic analyses, they used structural equation modeling and voxel-based morphometry. To explore the common genetic origin of focal GM and WM areas with intelligence, they obtained cross-trait/cross-twin correlations in which the focal GM and WM densities of each twin are correlated with the psychometric intelligence quotient of his/her cotwin. Genes influenced individual differences in left and right superior occipitofrontal fascicle (heritability up to 0.79 and 0.77), corpus callosum (0.82, 0.80), optic radiation (0.69, 0.79), corticospinal tract (0.78, 0.79), medial frontal cortex (0.78, 0.83), superior frontal cortex (0.76, 0.80), superior temporal cortex (0.80, 0.77), left occipital cortex (0.85), left postcentral cortex (0.83), left posterior cingulate cortex (0.83), right parahippocampal cortex (0.69), and amygdala (0.80, 0.55). Intelligence shared a common genetic origin with superior occipitofrontal, callosal, and left optical radiation WM and frontal, occipital, and parahippocampal GM (phenotypic correlations up to 0.35). These findings point to a neural network that shares a common genetic origin with human intelligence.
Brans et al (2010) performed a longitudinal magnetic resonance imaging study in twins, and found considerable thinning of the frontal cortex and thickening of the medial temporal cortex with increasing age and find this change to be heritable and partly related to cognitive ability. Specifically, adults with higher intelligence show attenuated cortical thinning and more pronounced cortical thickening over time than do subjects with average or below average IQ. Genes influencing variability in both intelligence and brain plasticity partly drive these associations. Thus, not only does the brain continue to change well into adulthood, these changes are functionally relevant because they are related to intelligence.
Chiang et al (2011) assessed white matter integrity voxelwise using diffusion tensor imaging at high magnetic field (4-Tesla), in 705 twins and their siblings (age range 12–29; 290 M/415 F). White matter integrity was quantified using a widely accepted measure, fractional anisotropy (FA). They fitted gene-environment interaction models pointwise, to visualize brain regions where age, sex, socioeconomic status (SES) and IQ modulate heritability of fiber integrity. They hypothesized that environmental factors would start to outweigh genetic factors during late childhood and adolescence. Genetic influences were greater in adolescence versus adulthood, and greater in males than in females. Socioeconomic status significantly interacted with genes that affect fiber integrity: heritability was higher in those with higher SES. In people with above-average IQ, genetic factors explained over 800% of the observed FA variability in the thalamus, genu, posterior internal capsule, and superior corona radiata. In those with below-average IQ, however, only around 40% FA variability in the same regions was attributable to genetic factors. Genes affect fiber integrity, but their effects vary with age, sex, SES and IQ. Gene-environment interactions are vital to consider in the search for specific genetic polymorphisms that affect brain integrity and connectivity.
Chiang et al (2012) report a novel approach that discovers which genes contribute to brain wiring and fiber integrity at all pairs of points in a brain scan. They studied genetic correlations between thousands of points in human brain images from 472 twins and their nontwin siblings (mean age: 23.7 ± 2.1 SD years; 193 male/279 female). They combined clustering with genome-wide scanning to find brain systems with common genetic determination. They then filtered the image in a new way to boost power to find causal genes. Using network analysis, they found a network of genes that affect brain wiring in healthy young adults. The gene network showed small-world and scale-free topologies, suggesting efficiency in genetic interactions and resilience to network disruption. Genetic variants at hubs of the network influence intellectual performance by modulating associations between performance intelligence quotient and the integrity of major white matter tracts, such as the callosal genu and splenium, cingulum, optic radiations, and the superior longitudinal fasciculus.
Koenis et al (2015) performed a longitudinal study of 162 healthy adolescent twins and their siblings (mean age at baseline 9.9 [range 9.0–15.0] years), and mapped local and global structural network efficiency of cerebral fiber pathways (weighted with mean fractional anisotropy (FA) and streamline count) and assessed intelligence over a three‐year interval. They found that the efficiency of the brain’s structural network is highly heritable (locally up to 74%). FA‐based local and global efficiency increases during early adolescence. Streamline count based local efficiency both increases and decreases, and global efficiency reorganizes to a net decrease. Local FA‐based efficiency was correlated to IQ. Moreover, increases in FA‐based network efficiency (global and local) and decreases in streamline count based local efficiency are related to increases in intellectual functioning. Individual changes in intelligence and local FA‐based efficiency appear to go hand in hand in frontal and temporal areas. More widespread local decreases in streamline count based efficiency (frontal cingulate and occipital) are correlated with increases in intelligence. The authors conclude that the teenage brain is a network in progress in which individual differences in maturation relate to level of intellectual functioning.
Koenis et al (2017) performed another longitudinal study, we mapped fractional anisotropy (FA) weighted efficiency of the structural brain network in 310 twins and their older siblings at an average age of 10, 13, and 18 years. Age‐trajectories of global and local FA‐weighted efficiency were related to intelligence. Contributions of genes and environment were estimated using structural equation modeling. Efficiency of brain networks changed in a non‐linear fashion from childhood to early adulthood, increasing between 10 and 13 years, and leveling off between 13 and 18 years. Adolescents with higher intelligence had higher global and local network efficiency. The dependency of FA‐weighted global efficiency on IQ increased during adolescence (rph=0.007 at age 10; 0.23 at age 18). Global efficiency was significantly heritable during adolescence (47% at age 18). The genetic correlation between intelligence and global and local efficiency increased with age; genes explained up to 87% of the observed correlation at age 18. In conclusion, the brain’s structural network differentiates depending on IQ during adolescence, and is under increasing influence of genes that are also associated with intelligence as it develops from late childhood to adulthood.
Panizzon et al (2015) tested multiple distinct models of the relationships among cognitive tests utilizing data from the Vietnam Era Twin Study of Aging (VETSA), a study of middle-aged male twins. Results indicated that a hierarchical (higher-order) model with a latent g phenotype, as well as specific cognitive domains, was best supported by the data. The latent g factor was highly heritable (86%), and accounted for most, but not all, of the genetic effects in specific cognitive domains and elementary cognitive tests. By directly testing multiple competing models of the relationships among cognitive tests in a genetically-informative design, we are able to provide stronger support than in prior studies for g being a valid latent construct.
Molecular Genetics Studies
Hill et al (2014) tested a network of 1461 genes in the postsynaptic density and associated complexes for an enriched association with intelligence. These were ascertained in 3511 individuals (the Cognitive Ageing Genetics in England and Scotland (CAGES) consortium) phenotyped for general cognitive ability, fluid cognitive ability, crystallized cognitive ability, memory and speed of processing. By analyzing the results of a genome wide association study (GWAS) using Gene Set Enrichment Analysis, a significant enrichment was found for fluid cognitive ability for the proteins found in the complexes of N-methyl-D-aspartate receptor complex; P=0.002. Replication was sought in two additional cohorts (N=670 and 2062). A meta-analytic P-value of 0.003 was found when these were combined with the CAGES consortium. The results suggest that genetic variation in the macromolecular machines formed by membrane-associated guanylate kinase (MAGUK) scaffold proteins and their interaction partners contributes to variation in intelligence.
Hill et al (2018) examined ~20,000 individuals in the Generation Scotland family cohort genotyped for ~700,000 single-nucleotide polymorphisms (SNPs), exploiting the high levels of linkage disequilibrium (LD) found in members of the same family to quantify the total effect of genetic variants that are not tagged in GWAS of unrelated individuals. In the authors’ models, genetic variants in low LD with genotyped SNPs explain over half of the genetic variance in intelligence, education, and neuroticism. By capturing these additional genetic effects our models closely approximate the heritability estimates from twin studies for intelligence and education, but not for neuroticism and extraversion. They then replicated their finding using imputed molecular genetic data from unrelated individuals to show that ~50% of differences in intelligence, and ~40% of the differences in education, can be explained by genetic effects when a larger number of rare SNPs are included. From an evolutionary genetic perspective, a substantial contribution of rare genetic variants to individual differences in intelligence, and education is consistent with mutation-selection balance.
DUF1220 protein domains (now known as the Olduvai domain) exhibit the greatest human lineage-specific copy number expansion of any protein-coding sequence in the genome, and variation in DUF1220 copy number has been linked to both brain size in humans and brain evolution among primates. And so, Davis et al (2015) examined associations between DUF1220 subtypes CON1 and CON2 and cognitive aptitude. They identified a linear association between CON2 copy number and cognitive function in two independent populations of European descent. In North American males, an increase in CON2 copy number corresponded with an increase in WISC IQ (R2 = 0.13, p = 0.02), which may be driven by males aged 6–11 (R2 = 0.42, p = 0.003). We utilized ddPCR in a subset as a confirmatory measurement. This group had 26–33 copies of CON2 with a mean of 29, and each copy increase of CON2 was associated with a 3.3-point increase in WISC IQ (R2 = 0.22, p = 0.045). In individuals from New Zealand, an increase in CON2 copy number was associated with an increase in math aptitude ability (R2 = 0.10 p = 0.018). These were not confounded by brain size. To our knowledge, this is the first study to report a replicated association between copy number of a gene coding sequence and cognitive aptitude. Remarkably, dosage variations involving DUF1220 sequences have now been linked to human brain expansion, autism severity and cognitive aptitude, suggesting that such processes may be genetically and mechanistically inter-related. The findings presented here warrant expanded investigations in larger, well-characterized cohorts.
- “Olduvai domains are, on average, 65 amino acids in length (ranging from 61 to 74) (Finn et al. 2014) and are encoded within a small exon and large exon doublet. The domains have been subdivided into six primary subtypes based on sequence similarity: Conserved 1–3 (CON1-3) and Human Lineage-Specific 1–3 (HLS1-3)” – see Heft et al (2020)
Davies et al (2012) conducted a genome-wide analysis of 3511 unrelated adults with data on 549,692 single nucleotide polymorphisms (SNPs) and detailed phenotypes on cognitive traits. The authors estimate that 40% of the variation in crystallized-type intelligence and 51% of the variation in fluid-type intelligence between individuals is accounted for by linkage disequilibrium between genotyped common SNP markers and unknown causal variants. These estimates provide lower bounds for the narrow-sense heritability of the traits. The authors partitioned genetic variation on individual chromosomes and found that, on average, longer chromosomes explain more variation. Using just SNP data they predicted approximately 1% of the variance of crystallized and fluid cognitive phenotypes in an independent sample (P = 0.009 and 0.028, respectively). Their results unequivocally confirm that a substantial proportion of individual differences in human intelligence is due to genetic variation, and are consistent with many genes of small effects underlying the additive genetic influences on intelligence.
People who score higher on intelligence tests tend to have larger brains. Twin studies suggest the same genetic factors influence both brain size and intelligence. This has led to the hypothesis that genetics influence intelligence partly by contributing to the development of larger brains. Elliott et al (2019) tested this hypothesis using four large imaging genetics studies (combined N = 7965) with polygenic scores derived from a genome-wide association study (GWAS) of educational attainment, a correlate of intelligence. The authors conducted meta-analysis to test associations among participants’ genetics, total brain volume (i.e., brain size), and cognitive test performance. Consistent with previous findings, participants with higher polygenic scores achieved higher scores on cognitive tests, as did participants with larger brains. Participants with higher polygenic scores also had larger brains. We found some evidence that brain size partly mediated associations between participants’ education polygenic scores and their cognitive test performance. Effect sizes were larger in the population-based samples than in the convenience-based samples. Recruitment and retention of population-representative samples should be a priority for neuroscience research. Findings suggest promise for studies integrating GWAS discoveries with brain imaging to understand neurobiology linking genetics with cognitive performance.
Neuroimaging and Intelligence
Here is the paper that first put forth the Parietal-Frontal Integration Theory (PFIT) back in 2007.
The following papers, in addition to showing a neurological basis for intelligence more generally, are also in support of the PFIT model for intelligence.
van den Heuvel et al (2009) examined the overall organization of the brain network using graph analysis, where they show a strong negative association between the normalized characteristic path length λ of the resting-state brain network and intelligence quotient (IQ). This suggests that human intellectual performance is likely to be related to how efficiently the brain integrates information between multiple brain regions. Most pronounced effects between normalized path length and IQ were found in frontal and parietal regions. Our findings indicate a strong positive association between the global efficiency of functional brain networks and intellectual performance.
van den Heuvel and Sporns (2011) set out to map out both the subcortical and neocortical hubs of the brain and examine their mutual relationship, particularly their structural linkages. They demonstrate that brain hubs form a so-called “rich club,” characterized by a tendency for high-degree nodes to be more densely connected among themselves than nodes of a lower degree, providing important information on the higher-level topology of the brain network. Whole-brain structural networks of 21 subjects were reconstructed using diffusion tensor imaging data. Examining the connectivity profile of these networks revealed a group of 12 strongly interconnected bihemispheric hub regions, comprising the precuneus, superior frontal and superior parietal cortex, as well as the subcortical hippocampus, putamen, and thalamus. Importantly, these hub regions were found to be more densely interconnected than would be expected based solely on their degree, together forming a rich club. The authors discuss the potential functional implications of the rich-club organization of the human connectome, particularly in light of its role in information integration and in conferring robustness to its structural core.
Recent analyses have shown that these putative hub regions are mutually and densely interconnected, forming a “rich club” within the human brain. van den Heuvel et al (2012) show that the set of pathways linking rich club regions forms a central high-cost, high-capacity backbone for global brain communication. Diffusion tensor imaging (DTI) data of two sets of 40 healthy subjects were used to map structural brain networks. The contributions to network cost and communication capacity of global cortico-cortical connections were assessed through measures of their topology and spatial embedding. Rich club connections were found to be more costly than predicted by their density alone and accounted for 40% of the total communication cost. Furthermore, 69% of all minimally short paths between node pairs were found to travel through the rich club and a large proportion of these communication paths consisted of ordered sequences of edges (“path motifs”) that first fed into, then traversed, and finally exited the rich club, while passing through nodes of increasing and then decreasing degree. The prevalence of short paths that follow such ordered degree sequences suggests that neural communication might take advantage of strategies for dynamic routing of information between brain regions, with an important role for a highly central rich club. Taken together, their results show that rich club connections make an important contribution to interregional signal traffic, forming a central high-cost, high-capacity backbone for global brain communication.
Song et al (2009) modeled the default network as undirected weighted graph, and then used graph theory to investigate the topological properties of the default network of the two groups of people with different intelligence levels. They found that, in both groups, the posterior cingulate cortex showed the greatest degree in comparison to the other brain regions in the default network, and that the medial temporal lobes and cerebellar tonsils were topologically separations from the other brain regions in the default network. More importantly, they found that the strength of some functional connectivities and the global efficiency of the default network were significantly different between the superior intelligence group and the average intelligence group, which indicates that the functional integration of the default network might be related to the individual intelligent performance.
Control of thought and behavior is fundamental to human intelligence. Evidence suggests a frontoparietal brain network implements such cognitive control across diverse contexts. Cole et al (2012) identify a mechanism—global connectivity—by which components of this network might coordinate control of other networks. A lateral prefrontal cortex (LPFC) region’s activity was found to predict performance in a high control demand working memory task and also to exhibit high global connectivity. Critically, global connectivity in this LPFC region, involving connections both within and outside the frontoparietal network, showed a highly selective relationship with individual differences in fluid intelligence. These findings suggest LPFC is a global hub with a brainwide influence that facilitates the ability to implement control processes central to human intelligence.
Langer et al (2011) set out to identify the functional brain network characteristics and their relation to psychometric intelligence. They examined whether the functional network exhibits efficient small‐world network attributes (high clustering and short path length) and whether these small‐world network parameters are associated with intellectual performance. High‐density resting state electroencephalography (EEG) was recorded in 74 healthy subjects to analyze graph‐theoretical functional network characteristics at an intracortical level. Ravens advanced progressive matrices (RAPM) were used to assess intelligence. The authors found that the clustering coefficient and path length of the functional network are strongly related to intelligence. Thus, the more intelligent the subjects are the more the functional brain network resembles a small‐world network. The authors further identified the parietal cortex as a main hub of this resting state network as indicated by increased degree centrality that is associated with higher intelligence. Taken together, this substantiates the neural efficiency hypothesis as well as the Parieto‐Frontal Integration Theory (P‐FIT) of intelligence in the context of functional brain network characteristics. These theories are currently the most established intelligence theories in neuroscience. These findings revealed robust evidence of an efficiently organized resting state functional brain network for highly productive cognitions.
Under the assumption that “stronger is better”, the exploration of brain properties has generally focused on the connectivity patterns of the most strongly correlated regions, whereas the role of weaker brain connections has remained obscure for years. Santarnecchi et al (2014) assessed whether the different strength of connections between brain regions may explain individual differences in intelligence. The authors analyzed functional connectivity at rest in ninety‐eight healthy individuals of different age, and correlated several connectivity measures with full scale, verbal, and performance Intelligent Quotients (IQs). Their results showed that the variance in IQ levels was mostly explained by the distributed communication efficiency of brain networks built using moderately weak, long‐distance connections, with only a smaller contribution of stronger connections. The variability in individual IQs was associated with the global efficiency of a pool of regions in the prefrontal lobes, hippocampus, temporal pole, and postcentral gyrus. These findings challenge the traditional view of a prominent role of strong functional brain connections in brain topology, and highlight the importance of both strong and weak connections in determining the functional architecture responsible for human intelligence variability.
The refinement of localization of intelligence in the human brain is converging onto a distributed network that broadly conforms to the Parieto-Frontal Integration Theory (P-FIT). Vakhtin et al (2014) looked at seventy-nine healthy subjects, performing fMRI scans while the subjects solved Ravens Progressive Matrices (RPM) problems and during rest. Functional networks were extracted from the RPM and resting state data using Independent Component Analysis. Twenty-nine networks were identified, 26 of which were detected in both conditions. Fourteen networks were significantly correlated with the RPM task. The networks’ spatial maps and functional connectivity measures at 3 frequency levels (low, medium, & high) were compared between the RPM and rest conditions. The regions involved in the networks that were found to be task related were consistent with the P-FIT, localizing to the bilateral medial frontal and parietal regions, right superior frontal lobule, and the right cingulate gyrus. Functional connectivity in multiple component pairs was differentially affected across all frequency levels during the RPM task. Their findings demonstrate that functional brain networks are more stable than previously thought, and maintain their general features across resting state and engagement in a complex cognitive task. The described spatial and functional connectivity alterations that such components undergo during fluid reasoning provide a network-wise framework of the P-FIT that can be valuable for further, network based, neuroimaging inquiries regarding the neural underpinnings of intelligence.
Penke et al (2012) provide evidence that lower brain-wide white matter tract integrity exerts a substantial negative effect on general intelligence through reduced information-processing speed. Structural brain magnetic resonance imaging scans were acquired from 420 older adults in their early 70s. Using quantitative tractography, the authors measured fractional anisotropy (FA) and two white matter integrity biomarkers that are novel to the study of intelligence: longitudinal relaxation time (T1) and magnetisation transfer ratio. Substantial correlations among 12 major white matter tracts studied allowed the extraction of three general factors of biomarker-specific brain-wide white matter tract integrity. Each was independently associated with general intelligence, together explaining 10% of the variance, and their effect was completely mediated by information-processing speed. Unlike most previously established neurostructural correlates of intelligence, these findings suggest a functionally plausible model of intelligence, where structurally intact axonal fibres across the brain provide the neuroanatomical infrastructure for fast information processing within widespread brain networks, supporting general intelligence.
The Wechsler Adult Intelligence Scale (WAIS) assesses a wide range of cognitive abilities and impairments. Factor analyses have documented four underlying indices that jointly comprise intelligence as assessed with the WAIS: verbal comprehension (VCI), perceptual organization (POI), working memory (WMI), and processing speed (PSI). Gläscher et al (2009) used non-parametric voxel-based lesion-symptom mapping in 241 patients with focal brain damage to investigate their neural underpinnings. Statistically significant lesion-deficit relationships were found in left inferior frontal cortex for VCI, in left frontal and parietal cortex for WMI, and in right parietal cortex for POI. There was no reliable single localization for PSI. Statistical power maps and cross-validation analyses quantified specificity and sensitivity of the index scores in predicting lesion locations.
Gläscher et al (2010) investigated the neural substrates of g in 241 patients with focal brain damage using voxel-based lesion–symptom mapping. A hierarchical factor analysis across multiple cognitive tasks was used to derive a robust measure of g. Statistically significant associations were found between g and damage to a remarkably circumscribed albeit distributed network in frontal and parietal cortex, critically including white matter association tracts and frontopolar cortex. The authors suggest that general intelligence draws on connections between regions that integrate verbal, visuospatial, working memory, and executive processes.
Bowren et al (2020) analyzed behavioral and neuroanatomical data from 402 humans (212 males; 190 females) with chronic, focal brain lesions. Structural equation models (SEMs) demonstrated a psychometric isomorphism between g and working memory in their sample (which they refer to as g/Gwm), but not between g and other cognitive abilities. Multivariate lesion-behavior mapping analyses indicated that g and working memory localize most critically to a site of converging white matter tracts deep to the left temporo-parietal junction. Tractography analyses demonstrated that the regions in the lesion-behavior map of g/Gwm were primarily associated with the arcuate fasciculus. The anatomic findings were validated in an independent cohort of acute stroke patients (n = 101) using model-based predictions of cognitive deficits generated from the Iowa cohort lesion-behavior maps. The neuroanatomical localization of g/Gwm provided the strongest prediction of observed g in the new cohort (r = 0.42, p < 0.001), supporting the anatomic specificity of their findings. These results provide converging behavioral and anatomic evidence that working memory is a key mechanism contributing to domain-general cognition.
The [above, Figure 2] domain-specific cognitive abilities could be modeled from the observed data: crystallized intelligence (Gc), visuospatial ability (Gv), learning efficiency (Gl), processing speed (Gs), and working memory (Gwm). To account for method covariance, the unique variances of the Complex Figure Test Copy and Recall scores were allowed to covary in all models, as were the unique variances of parts A and B of the Trail Making Test, and the indices of the Rey Auditory Verbal Learning Test. A hierarchical model (Fig. 2A) was used to estimate g and to examine the variance in g that can be accounted for by each domain-specific ability. The square of the correlation between each ability and g served as an index of the variance in g explained by each specific cognitive ability; this is indicated by the factor loading of each ability on g (Kline, 2016). A confirmatory bifactor model (Fig. 2B) was used to evaluate the “distinctiveness” of each cognitive ability from g. In this model, all observed data were first set to load directly onto g, and then onto the appropriate subfactor. As with all bifactor models, the latent variables were set to be orthogonal to one another under the assumption that the covariance among the factors is captured by the general factor. Using this model, ω hierarchical (ωh) was calculated for each domain, which provided an index of the reliable variance associated with each domain-specific ability after removing the variance associated with g. Higher values indicate greater distinctiveness from g (i.e., greater reliable variance remaining after controlling for g).
Anderson and Barbey (2023) conducted a large‐scale connectome‐based predictive modeling study (N = 297), administering resting‐state fMRI and a comprehensive cognitive battery to evaluate the efficacy of modern neuroscientific theories of human intelligence, including spatially localized theories (Lateral Prefrontal Cortex Theory, Parieto‐Frontal Integration Theory, and Multiple Demand Theory) and recent global accounts (Process Overlap Theory and Network Neuroscience Theory). Their results demonstrate that general intelligence can be predicted by local functional connectivity profiles but is most robustly explained by global profiles of whole‐brain connectivity. Their findings further suggest that the improved efficacy of global theories is not reducible to a greater strength or number of connections, but instead results from considering both strong and weak connections that provide the basis for intelligence (as predicted by the Network Neuroscience Theory). Their results highlight the importance of considering local neural representations in the context of a global information‐processing architecture, suggesting future directions for theory‐driven research on system‐wide network mechanisms underlying general intelligence.
Cox et al (2019) report brain structure-intelligence associations on a large sample from the UK Biobank study. The overall N = 29,004, with N = 18,426 participants providing both brain MRI and at least one cognitive test, and a complete four-test battery with MRI data available in a minimum N = 7201, depending upon the MRI measure. Participants’ age range was 44–81 years (M = 63.13, SD = 7.48). A general factor of intelligence (g) was derived from four varied cognitive tests, accounting for one third of the variance in the cognitive test scores. The association between (age- and sex- corrected) total brain volume and a latent factor of general intelligence is r = 0.276, 95% C.I. = [0.252, 0.300]. A model that incorporated multiple global measures of grey and white matter macro- and microstructure accounted for more than double the g variance in older participants compared to those in middle-age (13.6% and 5. 4%, respectively). There were no sex differences in the magnitude of associations between g and total brain volume or other global aspects of brain structure. The largest brain regional correlates of g were volumes of the insula, frontal, anterior/superior and medial temporal, posterior and paracingulate, lateral occipital cortices, thalamic volume, and the white matter microstructure of thalamic and association fibres, and of the forceps minor. Many of these regions exhibited unique contributions to intelligence, and showed highly stable out of sample prediction.
[structural equation modelling] SEMs testing associations between g and the FA [fractional anisotropy] and MD [mean diffusivity] of each white matter tract fitted the data well (all CFI ≥ 0.995, TLI ≥ 0.987, RMSEA ≤0.015, SRMR ≤0.009); results are shown in Fig. 4 [above], and Tables S6 and S7. Associations with g were in the expected direction, such that higher g was related to higher FA and lower MD. Only a few pathways had non-significant associations with g (FA and MD in the left acoustic radiation, FA in the middle cerebellar peduncle, and MD in the right parahippocampal cingulum, Forceps Major, and bilateral medial lemniscus). The effect sizes were not homogeneous across tracts (FA range = 0.012 to 0.110; MD range = −0.100 to 0.007). Consistent with our hypothesis, the magnitude of associations with g were numerically largest within thalamic pathways (FA mean = 0.078, MD mean = −0.091), and in association fibres and Forceps Minor (FA mean = 0.062, MD mean = −0.049) than within projection fibres and Forceps Major (FA mean = 0.039, MD mean = 0.027).1 However, it is also notable that both aspects of the cingulum bundle showed among the weakest g relationship among association fibres, and that more generally there was a considerable amount of overlap between these classes of tract (for example, the right corticospinal tract MD was associated with g at levels comparable with most association fibres).
Associations between g and cortical regional volumes were all positive and all significant following FDR correction. The results are reported in Fig. 5 [above] and in Table S8; all models fitted the data well (CFI ≥ 0.995, TLI ≥ 0.985, RMSEA ≤0.018, SRMR ≤0.010). As with the white matter analyses above, there was regional heterogeneity in association magnitudes across the cortical surface. Substantial portions of the frontal lobe (frontal pole, frontal orbital, subcallosal) were among the numerically largest associations, bilaterally (range = 0.166 to 0.216), and these were significantly larger than other frontal regions (p < .001). Associations between the insula cortex and g (left = 0.194, right = 0.205) were also large compared to the average magnitude across all ROIs (M = 0.116, SD = 0.036). Notably, the temporal lobe (range = 0.152 to 0.062) exhibited a gradient of anterior > posterior for both lateral and medial portions, and the lateral surface also showed evidence of a superior > inferior gradient. Compared to the above-mentioned frontal, anterior temporal and insula volumes, parietal regions were consistently and significantly more weakly associated with g (range = 0.066 to 0.100, p < .001). With the exception of the lingual, precuneus, and lateral occipital cortex (range = 0.110 to 0.156), occipital volumes were among the most weakly associated with g (range = 0.065 to 0.093).
Corley et al (2023) modeled the relationship between common genetic variation, grey matter volume, early life adversity and education and cognitive ability in a UK Biobank sample of N = 5237 individuals using structural equation modelling. The authors tested the hypotheses that total grey matter volume would mediate the association between genetic variation and cognitive ability, and that early life adversity and educational attainment would moderate this relationship. Common genetic variation, grey matter volume and early life adversity were each significant predictors in the model, explaining ~15% of variation in cognitive ability. Contrary to their hypothesis, grey matter volume did not mediate the relation between genetic variation and cognition performance. Neither did early life adversity or educational attainment moderate this relation, although educational attainment was observed to moderate the relationship between grey matter volume and cognitive performance. The authors interpret these findings in terms of the modest explanatory value of currently estimated polygenic scores accounting for variation in cognitive performance (~5%), making potential mediating and moderating variables difficult to confirm.
Educational attainment (EA) is often used as a proxy for cognitive ability since it is easily measured, resulting in large sample sizes and, consequently, sufficient statistical power to detect small associations. Mitchell et al (2020) investigated the association between three global (total surface area (TSA), intra-cranial volume (ICV) and average cortical thickness) and 34 regional cortical measures with educational attainment using a polygenic scoring (PGS) approach. Analyses were conducted on two independent target samples of young twin adults with neuroimaging data, from Australia (N = 1097) and the USA (N = 723), and found that higher EA-PGS were significantly associated with larger global brain size measures, ICV and TSA (R2 = 0.006 and 0.016 respectively, p < 0.001) but not average thickness. At the regional level, the authors identified seven cortical regions—in the frontal and temporal lobes—that showed variation in surface area and average cortical thickness over-and-above the global effect. These regions have been robustly implicated in language, memory, visual recognition and cognitive processing. Additionally, they demonstrate that these identified brain regions partly mediate the association between EA-PGS and cognitive test performance.
Lombardo and Kaufmann (2023) analyzed within and between network connectivity patterns from resting-state functional MRI of 2707 children between 9 and 10 years from the ABCD study. The authors hypothesized that differences in functional connectivity at the default mode network (DMN), ventral, and dorsal attentional networks (VAN, DAN) explain differences in fluid and crystallized abilities. They found that stronger between-network connectivity of the DMN and VAN, DMN and DAN, and VAN and DAN predicted crystallized abilities. Within-network connectivity of the DAN predicted both crystallized and fluid abilities. Their findings reveal that crystallized abilities rely on the functional coupling between attentional networks and the DMN, whereas fluid abilities are associated with a focal connectivity configuration at the DAN. This study provides new evidence into the neural basis of child intelligence and calls for future comparative research in adulthood during neuropsychiatric diseases.
Cipolotti et al (2023) assessed 165 healthy controls and 227 frontal or non-frontal patients with unilateral brain lesions on the best-established test of fluid intelligence, Raven’s Advanced Progressive Matrices, employing an array of lesion-deficit inferential models responsive to the potentially distributed nature of fluid intelligence. Non-parametric Bayesian stochastic block models were used to reveal the community structure of lesion deficit networks, disentangling functional from confounding pathological distributed effects. Impaired performance was confined to patients with frontal lesions [F(2,387) = 18.491; P < 0.001; frontal worse than non-frontal and healthy participants P < 0.01, P <0.001], more marked on the right than left [F(4,385) = 12.237; P < 0.001; right worse than left and healthy participants P < 0.01, P < 0.001]. Patients with non-frontal lesions were indistinguishable from controls and showed no modulation by laterality. Neither the presence nor the extent of multiple demand network involvement affected performance. Both conventional network-based statistics and non-parametric Bayesian stochastic block modelling heavily implicated the right frontal lobe. Crucially, this localization was confirmed on explicitly disentangling functional from pathology-driven effects within a layered stochastic block model, prominently highlighting a right frontal network involving middle and inferior frontal gyrus, pre- and post-central gyri, with a weak contribution from right superior parietal lobule. Similar results were obtained with standard lesion-deficit analyses.
Smith et al (2015) found one significant mode of population variation that links a specific pattern of brain connectivity to a specific pattern of covariance between many behavioral and demographic subject measures. The vast majority of the subject measures (SMs) that correlate positively with this mode are “positive” subject traits/measures (education, income, IQ, life-satisfaction); those that correlate negatively are mostly “negative” subject measures. However, while strongly resembling the known general intelligence g factor for many of the subject measures, this mode does not trivially map onto just the strongest single principal component of the subject measures (see Supplementary Fig. 4; the canonical correlation analysis (CCA) mode maps strongly onto the top three SM principal components and not just the first). It is plausible that the CCA mode includes a neural correlate of g, but is a more general mode of positive brain function, and is more directly tied into to the underlying biology (specifically, connectivity between brain regions), given that it is driven both by structured population covariance in behavioral measures and by intrinsic brain connectivity. The authors note a common criticism of the g factor, that there could be many distinct uncorrelated neural systems underlying high-level cognitive function, and that different cognitive tasks will involve different but overlapping sets of these latent processes, resulting in “artificial” correlation between subject measures, and hence the appearance of a g factor. In future work, the authors say, it will be important to determine whether this known “unresolvable” ambiguity in g factor interpretation might be resolved through more fine-grained analysis of the data source newly available – the subject-specific functional connectomes – potentially even allowing direct investigation of the latent neural systems, which may help us understand the coordinated interactions among brain systems that give rise to a general mode of positive function in humans.
While fMRI studies typically collapse data from many subjects, brain functional organization varies between individuals. Finn et al (2015) establish that this individual variability is both robust and reliable, using data from the Human Connectome Project to demonstrate that functional connectivity profiles act as a “fingerprint” that can accurately identify subjects from a large group. Identification was successful across scan sessions and even between task and rest conditions, indicating that an individual’s connectivity profile is intrinsic, and can be used to distinguish that individual regardless of how the brain is engaged during imaging. Characteristic connectivity patterns were distributed throughout the brain, but notably, the frontoparietal network emerged as most distinctive. Furthermore, we show that connectivity profiles predict levels of fluid intelligence; the same networks that were most discriminating of individuals were also most predictive of cognitive behavior. Results indicate the potential to draw inferences about single subjects based on functional connectivity fMRI.
Here we show that an individual’s functional brain connectivity profile is both unique and reliable, analogous to a fingerprint. We demonstrate that it is possible, with near-perfect accuracy, to identify individuals from a large group of subjects based solely on their connectivity matrix. While inter-individual consistency in functional brain networks has been well characterized across both task and rest conditions24,25, and even across states of consciousness26, the remarkable intra-individual reliability observed here suggests that while the general blueprint may be shared, functional organization within individual subjects is idiosyncratic, relatively robust to changes in brain state, and provides meaningful information above and beyond the common template27.
We also demonstrate that this individual variability is relevant to individual differences in behavior, in that connectivity profiles can be used to predict the fundamental cognitive trait of fluid intelligence in novel subjects. These results underscore the potential to discover fMRI-based connectivity “neuromarkers” of present or future behavior that may eventually be used to personalize educational and clinical practices, improving outcomes28–30.
Even if a detractor to the biological basis of intelligence is convinced by all this evidence, they might still wonder: what is the point of this? Maybe, even if it is true that there is a biological basis for intelligence, that this is the kind of question we just shouldn’t even be asking? The fact that anyone is asking the question just shows that they must have some nefarious motive, perhaps wanting to slyly bring back eugenics or some form of racial hierarchies.
I think this line of thinking would be erroneous for three reasons. The first reason is that the truth has intrinsic value. While I think there is a point at which the potential for harm outweights whatever intrinsic value the truth has, such that we impose restrictions on what science is allowed to do, I do not think intelligence research passes that threshold. And, indeed, as I will discuss in my third reason, I think there are great benefits to studying intelligence. As such, from both a deontological and a consequentialist point of view, studying the genetics and biology of intelligence is valuable.
The second reason is that we cannot get an ought from an is: even if it is the case that intelligence is largely determined by biology, and that there are no social or educational interventions that could ever turn an average intelligence individual into a genius or a below average intelligence individual into an average intelligence individual (i.e., a strong sort of biological determinism), this says nothing about how people ought to be treated morally. In other words, the syllogism
P1: intelligence largely has its basis in genetics/biology
C: therefore we ought to install a regime of eugenics
is not valid – the conclusion does not follow from the premise. Even if someone were to try deriving this conclusion from the premise, they would require at least one other premise, and such a premise would need to smuggle in some kind value judgement that does not follow from P1. Policies must always be informed by what is objectively true (or as close to objectively true as we can get), but the truth alone does not necessitate a particular policy. It is where objective truth and (universal) human values converge that we must look to find the best course of action.
The third reason is that understanding intelligence could potentially unlock ways of increasing intelligence through genetic or biological intervention. Of course, some people will see this as a potential downside, but I am a cautious advocate of extropianism. A common refrain from me on this blog is that we humans have not evolved to live in the world we have created for ourselves. In my thinking, if we want humankind (in some form) to persist, we will need to alter ourselves in a way that make us more suitable for the world we’ve created. Something like transhumanism is likely the only long-term solution to all the myriad problems we’ve created for ourselves (and problems we’ve inherited from simply being biological creatures that evolved in a hostile universe). And while transhumanism will almost certainly create new problems for ourselves (or our future transhuman descendants), it is still worth trying, since not pursuing it (and continuing with our short-term fixes) will inevitably lead to our extinction (or worse, our continued existence but in an ever worsening state of affairs). If we value the long-term existence of humanity, we cannot afford to continue wallowing below type 1 on the Kardashev scale. Finding a way to enhance our intelligence is a necessary step toward ensuring our continued existence, and understanding the basis of intelligence (whether it is biological or otherwise) is a necessary step to enhancing our intelligence.