8 Reasons for Doubting Claims of the Heritability of Intelligence

How much, if any, is intelligence inherited? We can imagine two extreme scenarios. Under the scenario of 0% intelligence heritability, two parents who both had an IQ of about 130 would have no reason at all for thinking that their children would have an IQ above 100. Under the scenario of 100% intelligence inheritability, two parents with an IQ of about 130 could be certain that their children would have an intelligence about the same as their parent's intelligence. It is often claimed that the heritability of intelligence is about 50%. If that were true, two parents with an IQ of about 130 would have a fairly strong reason for suspecting that their children's intelligence would be above average, but not have anything close to certainty about such a matter.

But the claim that intelligence is even 50% inheritable is unproven. I will explain some reasons for doubting such a claim. I will argue that we do not have very convincing evidence that the genetic heritability of intelligence is much higher than 0.

Reason #1: The Lack of Large Parent-Child IQ Databases Needed to Prove a Claim of Intelligence Heritability

What would you need to firmly establish a claim that intelligence is 50% heritable? You would need a very large database listing the IQ scores of very many thousands or millions of individuals, along with the IQ scores of their parents. But such a database is not known to exist anywhere. Consider a case such as my own family. I have never formally had my IQ tested, nor have my two children. When I was in high school, I was told that students had their IQ tested by the school. Conceivably somewhere in some school's database might be an IQ score for me, and in some other school's database might be IQ scores for my children. But there is no database that links these records together, allowing a researcher to use the information as part of a deduction about the heritability of intelligence. The few studies that have attempted to gather the IQ of both parents and children have usually involved samples of no more than a few thousand parents and children. It is impossible to draw very reliable conclusions about human intelligence in general from samples so small.

The study here claims to have used a large database of IQ scores from Norway, one that included parent and child relationships. I'll discuss in the next paragraph why this wasn't actually a database of IQ scores. The study concludes, “The correlation we observe between father’s IQ and that of his son is .38 .” That's an estimate below the typical estimate of 50% for the heritability of intelligence. But since the paper fails to specify how these parent and child relationships were derived, we should be skeptical of this result. The paper is sketchy in its details, and does nothing to tell us: exactly how did the authors know how to connect fathers and sons? We are not even told how many father-son pairs were used. In a large data set with lots of information, someone might use a computer program to scan the data and do a calculation of the correlation between father's IQ and son's IQ. But a single line of errant code in such a computer program would cause the program to produce the wrong answer.

Also, the paper tells us that a standard IQ test was not even the basis of these supposed “IQ scores,” and that the “IQ scores” were derived from a trio of tests. The paper tells us, “The IQ measure is a composite score from three speeded IQ tests -- arithmetic, word similarities, and figures...the word test is similar to the vocabulary test in WAIS.” That doesn't sound like a reliable way at all of testing IQ, and sounds largely like a way of testing education rather than intelligence. The source here says the vocabulary part of the WAIS test "measures word knowledge," but a good IQ test isn't supposed to measure knowledge. The paper seems to have wrongly described a trio of cognitive and educational tests as being an IQ score. We therefore should have no confidence that the paper has informed us about the heritability of IQ.

Reason #2: The Low Reliability of Twin Studies

Lacking what they need (very large databases with the IQ scores of both parents and children), scientists have tried to rely on much smaller databases involving the IQ scores of twins. The strategy of this type of study is as follows:

1. Collect data relevant to the IQ scores of twins.
2. Attempt to figure out whether the twins were monozygotic (so called “identical twins”) or dizygotic twins.
3. Look and see whether there is a closer correlation between the IQ scores of the monozygotic twins and the IQ scores of the dizygotic twins.

The largest study I have found of this type is one that uses something called the Genetics of High Cognitive Abilities (GHCA) Consortium, That term seems to be merely an umbrella used to collect the data from six different studies done in different countries. The number of twins in the database was about 11,000. Based on somewhat higher correlations in IQ between monozygotic twins than in dizygotic twins, the study estimated that intelligence is about 50% heritable.

There are several reasons why such a result is not terribly convincing.

1. Studies such as this are usually run by researchers interested in proving a genetic basis for intelligence. There are any number of ways in which bias might have influenced the studies, such as in decisions on which twins to include.
   
2. A large fraction of the data on whether twins were monozygotic twins was made from parent questionnaires rather than biological DNA testing. Such questionnaires are less reliable than DNA testing.
   
3. The study's paper tells us that the IQ of the twins was estimated by the people running the study (sometime from two or more other tests taken). But were the people doing such estimates blind as to whether the twins were monozygotic or dizygotic? If not, such people might have had a tendency to give similar IQ scores to monozygotic twins. The study makes no mention of a blinding protocol being used. 
4. We are not entitled to draw conclusions about the heritability of intelligence in the general population from studies involving only twins. There could be any number of reasons why the heritability of intelligence is a much different number when twins are involved, particularly since we do not at all understand where intelligence comes from and whether it is actually a product of the brain (there being many reasons for doubting claims that human minds are created by the brain).
5. It is clear from the scientific paper that IQ scores were often derived from written tests in which vocabulary knowledge played an important part. Such tests are largely measures of learned knowledge, something different from intelligence. 

Reason #3: The Sometimes Poor Research Record of Scientists Looking for Genetic Associations

In considering the question of whether intelligence is highly heritable, we should consider the research record of geneticists attempting to show genetic associations with human traits. Over the years such scientists have raised many false alarms, claiming that there were genetic associations when there wasn't good evidence for such a thing. One example of a blunder by such scientists was described in a recent article in The Atlantic entitled “A Waste of 1000 Research Papers.” The article is about how scientists wrote a thousand research papers trying to suggest that genes such as SLC6A4 were a partial cause of depression, one of the leading mental afflictions. The article tells us, "But a new study—the biggest and most comprehensive of its kind yet—shows that this seemingly sturdy mountain of research is actually a house of cards, built on nonexistent foundations." There have been many such cases in which scientists tried to convince us that some factor has a genetic link, but the evidence was weak.  An article says, "Prior to 2005, the field was largely a scientific wasteland scattered with the embarrassing and wretched corpses of unreplicated genetic association studies, with barely a handful of well-validated genetic risk factors peeking above the noise." 

It seems that when scientists are eager to show that there is some genetic basis for something, they can be guilty of analysis bias that causes them to think there is a good evidence for a genetic link when no such evidence exists. Such an example should make us wonder whether the same faulty biased analysis is at work in studies trying to show a link between genes and intelligence.

Reason #4: The Flynn Effect Contradicts Claims That Intelligence is Highly Heritable

The Flynn effect is a well-documented effect that involves a gradual increase in performance in intelligence tests. The increase seems to be about 3% per decade, and has seemingly been occurring since the 1930's (although in some countries in recent years we have failed to see evidence of such a thing). Given how often mothers in their thirties have children in recent decades, what this means is that a child born around 1990 may typically have an IQ 9% greater than the IQ of his parents. Such a thing is a point against claims that intelligence is heritable.

In some of the studies involving the heritability of intelligence, we see a kind of “Flynn effect subtraction” in which the data is massaged to factor out the Flynn effect, or a calculation formula is modified to remove the Flynn effect. But there is no justification for such a thing, which can be described as a kind of cover-up to get rid of a data effect that is inconsistent with the claim of a heritability of intelligence.

Reason #5: The Fact That There Is No Clear Evidence of Any Genes for Intelligence, and That No “IQ Gene” Has Been Found

To try to show a genetic basis for intelligence, scientists sometimes run what is called a genome-wide association study or GWAS. Some of these studies have claimed to have found genes that were associated with intelligence. But there are several reasons why such a thing does not itself show that genes determine intelligence.

Let us consider the important fact that a GWAS will usually find evidence for any thing that is being looked for, purely because of random variations unrelated to a causal effect.  For example, let's suppose you do a study in which you scan the genomes of 10,000 people and try to find correlations with the DNA in these people and their favorite songs.  You will inevitably be able to find some slight correlations here and there which could be cited as a genetic basis for your song preferences. But such an association would be spurious, and would be caused by mere random variations that did not have a causal effect. Your genes don't code your song preferences.

The same type of spurious association could be cropping up in studies trying to link genes to intelligence. So we may ask questions such as this:

(1) How much of a replication effect is there, with the same genes showing up in multiple studies trying to find genes for intelligence?

(2) How strong an association is reported? How much of intelligence variation might be attributable to a particular candidate for an intelligence gene?

Typically there are only weak associations found in such GWAS studies on the heritability of intelligence. For example, an article on one study says, "Even if a person had both copies of all the variants, she would score an average of 1.8 higher points in an IQ test."

As for replication, reports of "intelligence genes" generally do not replicate well. A scientific study attempted to replicate reported associations with some genes (DTNBP1, CTSD, DRD2, ANKK1, CHRM2, SSADH, COMT, BDNF, CHRNA4, DISC1, APOE, and SNAP25) using " using data sets from three independent, well-characterized longitudinal studies with samples of 5,571, 1,759, and 2,441 individuals." The replication attempt failed, and the study is entitled,  "Most reported genetic associations with general intelligence are probably false positives."

An article by a psychologist states the following:

"It’s the best kept secret of modern science: 16 years of the Human Genome Project suggest that genes play little or no role in explaining differences in intelligence. While genes have been found for physical traits, such as height or eye colour, they are not the reason you are smarter (or not) than your siblings. Nor are they why you are like your high-achieving or dullard parents, or their forebears."

Reason #6: The Intrinsic Implausibility of Claims That Genes  Determine Intelligence

Scientists have no understanding of how neurons can create thinking, imagination, abstract ideas or understanding. The claim that human thinking comes from the brain is a speech custom of scientists, rather than a fact established by observations.  There are actually strong reasons for doubting that the brain is the cause of human intelligence. One very strong reason is that when people undergo hemispherectomy operations (in which half of their brain is removed to stop epileptic seizures), it has little effect on their intelligence (as documented here).  In the same link just given, you can read about people with high intelligence despite losing the great majority of their brains.  Because we have such strong reasons for doubting that the brain is the actual source of human intelligence, it is intrinsically implausible that some genes relating to the brain may have a large effect on intelligence. 

Reason #7: Much of the Reported Heritability of Intelligence May Be a Heritability of Dyslexia (or Similar Perception Problems) Unrelated to General Intelligence

Dyslexia is a reading disorder that is believed to affect 5% to 10% of children. Dyslexia is not a defect in general intelligence, and a person with dyslexia will do just as well as a person without dyslexia on an intelligence test that does not require reading. "Dyslexia independent of IQ" is the headline of a story at the MIT News.   Dyslexia is believed to be a highly heritable. A scientific paper on dyslexia says, “Inherited factors are estimated to account for up to 80%.”

But almost all data on IQ comes from paper-and-pencil tests requiring reading. A large part of the data suggesting a heritability of intelligence may simply be caused by a heritability of dyslexia that has nothing to do with intelligence. This is what scientists call a “confounding factor,” and in the case of the heritability of intelligence, it seems to be a largely overlooked confounding factor.

Reason #8: The Fact That IQ Tests Are Imperfect Measurements of Intelligence

It has long been pointed out that IQ tests are far-from-perfect as measurements of intelligence.  One reason is that many things other than intelligence can affect scores on IQ tests. Here are some of the things:

(1) Visual perception. Since an IQ test uses a test booklet that requires a lot of reading, a person's vision clarity and visual perceptual abilities can affect the scores -- but such things are different from intelligence.  If a student comes from a poor family that lacks money to keep his vision at 20/20, so much worse will the student's IQ score be.
(2) Manual dexterity. A typical IQ test will require many different cases of filling in exactly the right oval on a test form, and a small percentage of students might be relatively slow at such a manual task, for reasons having nothing to do with intelligence. 

(3) Learned vocabulary. IQ tests are supposed to use simple words, to avoid being tests of learned knowledge. But any test with many written question must inevitably be partially a test of learned vocabulary.  For example, when I went to the web site for the Wechsler IQ test, the first question (in the short sample test I tried) uses the word "contradictory" and the word "enhance." Very many school children would not know the meaning of these words.  
(4) Motivation. Scores on IQ tests are strongly affected by the degree of motivation of the test taker.  According to this page, when students are offered money rewards for high scores, it will affect the IQ scores by an average of 20 points. We read, "Thus rewards higher than $10 produced g values of more than 1.6 (roughly equivalent to more than 20 IQ points), whereas rewards of less than $1 were only one-tenth as effective."  So imagine the students in a low-quality school where students may tend to have lower motivation.  The average IQ result at such a school might be much lower than the average IQ result at a better school, even if the students at the schools have the same intelligence. 
(5) Dyslexia. This perceptual reading disorder (not an intelligence defect) can drastically affect scores on written IQ tests.
(6) Distractions in test rooms. This can have a large affect on IQ scores. Let's imagine a disorderly school, with quite a few "class clowns" or unruly students. Such students might create any number of distractions that might affect the concentration of students taking IQ tests. 
(7) Number of sharpened pencils of test takers. At a school with students from richer families, an average student may have two sharpened pencils for his IQ test. But at a school with students from poor families, an average student may have only one not-very-sharp pencil.  If the pencil breaks, so much worse for his IQ score. 
(8) Percentage of sick test takers.  A test taker will probably do worse on an IQ test if he is sick.  In affluent communities, parents might hire a "baby sitter" if their child is sick, preventing him from being present during an IQ test. In poorer communities, the sick student might be much more likely to be told to go to school even when he's sick. In some third-world countries,  we can easily imagine 5% or more of the IQ test takers being sick. 
(9) Number of recent traumas affecting test takers. Let's imagine student Sally sits down for a 90-minute IQ test at her school. She finds it rather hard to concentrate very hard for 90 minutes because her thoughts drift to the recent gunfire in her neighborhood, or the alcoholism or drug abuse of one or her family members.  The result may be a lower score on the test. 

The factors discussed above cast great doubt on claims of the heritability of intelligence. The same factors (particularly the ones discussed under Reason #8) are reasons casting great doubt on all claims that people of one race or nationality are (on average) more intelligent or less intelligent than people of some other race or nationality. 

We can imagine a radically different type of intelligence test. A person would be placed in a basement room with a small window three meters above the ground. There would be 20 large filled boxes in the room.  The person would be told that he is being locked in the room, and that he must escape. He would be told that the faster he escapes, the more money he will make. The test would be to see how fast the person can think of stacking the boxes to make a stair-like structure that can be used to allow escape through the window.  On such a test we can easily imagine students from poorer, troubled schools getting scores just as high as students from elite schools. 

I don't rule out the possibility that intelligence is to some degree heritable, although I think there is no strong evidence for a very large amount of such heritability.  We can certainly imagine some possibilities that might one day firm up that type of evidence, including better intelligence tests with fewer confounding variables, and larger databases tracking intelligence scores for parents and their children. 

Postscript: See the link here for a lengthy discussion of the work of Leon J. Kamin, who made a strong critique of twin studies used to bolster claims of a heritability of intelligence.  Below is an excerpt:

 In summing up his TRA study findings in The Science and Politics of I.Q., Kamin wrote, “To the degree that the case for a genetic influence on I.Q. scores rests on the celebrated studies of separated twins, we can justifiably conclude that there is no reason to reject the hypothesis that I.Q. is simply not heritable.” He reached a similar conclusion in relation to IQ genetic research as a whole. Kamin was not concluding that heredity had no influence on IQ test scores, but rather that IQ genetics researchers, who bore the burden of proof, had failed to provide scientifically valid evidence that it did. In a 1976 book review, Harvard evolutionary geneticist and Kamin’s future collaborator Richard Lewontin wrote that Kamin had discovered, in IQ genetic research, “a pattern of shoddiness, carelessness, miserable experimental design, misreporting, and misrepresentation amounting to a major scandal.”

One of the studies most enthusiastically cited by those claiming intelligence is highly heritable is a 1990 Minnesota study of twins reared apart, called MISTRA. The study is debunked by a lengthy 2022 scientifc paper by Jay Joseph entitled "A Reevaluation of the 1990 'Minnesota Study of Twins Reared Apart' IQ Study." We read this:

"In 1980, sociologist Howard Taylor described what he called 'The IQ game,' by which he meant IQ-genetic researchers’ 'use of assumptions that are implausible as well as arbitrary to arrive at some numerical value for the genetic heritability of human IQ scores on the grounds that no heritability calculations could be made without benefit of such assumptions' (Taylor, 1980, p. 7). The MISTRA IQ study can be seen as an exemplar of 'IQ game' bad science...The MISTRA IQ study failed to discover evidence that genetic factors influence IQ scores and cognitive ability across the studied population.".