Would People Screen Embryos for Low IQ? A New Study Answers

Polygenic embryo screening for disease risk is already a commercial reality at some fertility clinics. Screening for intelligence is the example every ethicist reaches for when they want to show where the technology should stop — a bright line between correcting a medical problem and engineering a trait. A study published on 15 September 2026 in Nature Human Behaviour tested whether that line exists anywhere outside of what people say out loud.

What the survey actually measured

Edmond Awad, Julian Savulescu and colleagues ran two separate exercises on a nationally representative sample of 1,467 US adults. One group of 735 people was simply asked whether they would want an embryo tested for each of four traits, framed as an elevated genetic risk score rather than a certainty. A second group of 732 people was shown two embryos side by side, each carrying a real risk profile roughly 5% above or below the population average, and asked which one they would choose to implant.

• Heart disease (medical): 79% said they'd want it tested for.
• Low IQ (non-medical): 65%.
• Antisocial behavior (non-medical): 58%.
• Vision impairment (medical): 55%.

Notice that the two “medical” traits do not move together: heart disease drew the highest willingness of the four, and vision impairment drew the lowest. Willingness to test does not sort neatly into medical versus non-medical — it tracks something closer to perceived severity, which the researchers say is also what drove the embryo-choice results.

The gap between what people say and what they do

The embryo-choice group did not answer a survey question. They looked at two real risk profiles and picked one to implant. Comparing how often participants chose the higher-risk embryo against how often they chose the lower-risk one gives a behavioral gap for each trait: 40 percentage points for heart disease, 36 for antisocial behavior, 32 for low IQ, and 19 for vision impairment.

Line the two sets of numbers up and the pattern is clear. In the abstract, low IQ (65% willing) sits closer to vision impairment (55%) than to heart disease (79%) — a moderate gap that fits the assumption that people treat intelligence as a special, harder-to-justify case. In an actual choice, low IQ (a 32-point avoidance gap) sits almost exactly between heart disease (40) and antisocial behavior (36), a long way clear of vision impairment (19). The ethical distance between screening for a disease and screening for a mind that people report in a survey mostly does not survive contact with a real decision.

Why intelligence is a harder case than it looks, genetically

Even setting the ethics aside, there is a technical reason to doubt this kind of screening delivers much. A 2019 simulation study in Cell modeled embryo selection using a polygenic score for cognitive ability built from a discovery sample of more than 234,000 people, one of the larger IQ-genetics datasets available at the time. Selecting the single top-scoring embryo out of a typical sibling group raised the expected result by only about 2.5 IQ points over picking an embryo at random — and the uncertainty around that estimate was wide enough that, for a comparable analysis of height, the embryo with the best polygenic score was often not the sibling who actually grew tallest.

The reason is structural, not a matter of needing a bigger study. Full siblings already share roughly half of the genetic variation that differs between people in a population, so the embryos produced in a single IVF cycle are far more genetically similar to each other than two unrelated people are. There is only so much daylight between them for any score to detect, no matter how large the underlying genetic study gets. That is a different and more fundamental limit than “the technology is not good enough yet,” and it is a large part of why several reproductive medicine and bioethics bodies have called embryo screening for a complex trait like intelligence premature rather than merely early.

The same 2019 paper ran the identical exercise for height, this time using a much larger discovery sample of roughly 700,000 people — nearly three times the size of the IQ dataset — and still found the same pattern of wide, often-wrong predictions about which sibling would end up tallest. That comparison is the clearest evidence the limitation is structural rather than a simple matter of the IQ-specific dataset being smaller than others in the field: even a much bigger genetic study did not fix it, because the ceiling here is set by how similar full siblings already are, not by sample size.

Where this sits next to the science of IQ genetics

None of this is news to behavioral geneticists. Our own coverage of what twin studies say about the balance of genetics and environment in test scores makes the same point from a different angle: a trait being substantially heritable across a population is not the same as being able to predict, let alone engineer, one specific embryo's outcome from a DNA sample. Heritability is a statement about a population's variance; a polygenic score applied to a single embryo is a far noisier instrument than that population-level number suggests, which is exactly what the wide prediction intervals in the Cell paper show in practice. For the genetics side of this in more depth, our piece on whether IQ is genetic lays out what heritability estimates can and cannot support. For how a modern test turns raw answers into a score at all, our IQ tools page walks through the calculators behind it.

The debate this reopens

The historical shadow here is not subtle, and researchers covering this study have not avoided naming it: selecting future people for intelligence was central to state eugenics programs in the early twentieth century, before the movement's coercive history made the word radioactive. Our history of IQ testing and eugenics traces how a legitimate measurement tool and a program of reproductive control got built by some of the same people, and are still routinely confused. A voluntary, individual embryo choice is a different act from state-mandated sterilization, and most bioethicists writing on this topic are careful to say so. What this study adds is evidence that the more comfortable, purely medical framing people reach for when discussing the voluntary version does not hold up once you watch what they actually choose.

The honest reading of this study is narrower than either “people are secretly eugenicists” or “the medical/non-medical line is fine.” It is that stated ethics and revealed behavior diverge by a wide margin on exactly the trait — intelligence — that regulation has treated as the clearest case for restriction, while the underlying genetic technology remains, on the study's own numbers, a blunt and unreliable instrument for choosing it either way.

Common questions