I'm building a setting in a fictional Latin American country which is under military rule. The junta starts a breeding program in the late 40's with help from unscrupulous escaped Nazi scientists.

Assuming the breeding program starts with a population of 10,000 people with an average IQ of around 130, how much could could they raise the IQ in 5 generations?

45-60 => 1st, 60-75 => 2nd, 75-90=> 3rd, 90-05 => 4th, 06-? =>5th

The constraints are as below:

  • There is a fund for keeping 10,000 people in the program only.
  • No genetic engineering or any other other biological experiments.
  • Children who don't fit the program's eugenic goals are given for adoption into selected couples of the general population.
  • Scientists running the program decide who will breed with who. For example they could keep 10% of the boys and 90% of the girls if that gives them best results. All the other children are given for adoption to selected couples.
  • The country has around 30 million people, the bred are very small part of it.

Note: (story related; might not be relevant to the question)

If it helps, my story happens in present age. The country finally democratizes and holds fair elections. The newly-elected president finds out about the secret breeding program. He also learns that the children given for adoption from the program are vastly over-represented among the country business & cognitive elite compared with their meager numbers.

  • $\begingroup$ Comments are not for extended discussion; this conversation about intelligence, IQ tests, and possible outcomes of this program has been moved to chat. $\endgroup$ Oct 9, 2018 at 2:54
  • $\begingroup$ I feel the need to point out that part of the problem with all the eugenics attempts made was that the people deciding what qualified as 'better' genes seemed to focus on relatively irrelevant details like hair color, eye color etc, or people couldn't agree. Getting people to keep to a definition for their eugenics for 5 generations and have it be a meaningful and useful measure is very hard, even before you consider that your super-smart designer babies may decide they don't want to be breeding lab rats and stop cooperating. $\endgroup$
    – dsollen
    Aug 25, 2020 at 23:03
  • $\begingroup$ Humans already have a huge selective pressure for intelligence (competition with other humans) how are you creating a greater selective pressure? Also if they are being bred based on IQ testes they may end up being very poor in real world situations since there is no focus on life management or social skills. sure they can solve a quadratic equation but they fall for con-men or alienate allies. They also may be very bad at learning from real world situations. $\endgroup$
    – John
    Aug 27, 2020 at 12:21

7 Answers 7


Since as far as I know nobody bred humans, we can't tell for sure, so take below answer as pop scientific. It's using the breeder equation and regression to the mean.

Assuming that that average IQ of the population that outliers are drawn is 100 IQ, heritability is 50%, and that with each generation it gets 20% harder to become outlier. Also you are selecting top 1-2% of males, and 50% of the brightest females.

In the first generation you started with equal number of 130IQ of males and females but in the next generation. You use top 50% of the females but only 1-2% of the males. In order to keep the population at 10,000 average number of children per women should be 4+ which is plausible.

  • Gen: 1 Avg 100 Growth: 0.30 50+ Female: 130 Top1-2% Male: 130
  • Gen: 2 Avg 115 Growth: 0.24 50+ Female: 132 Top1-2% Male: 150
  • Gen: 3 Avg 128 Growth: 0.19 50+ Female: 143 Top1-2% Male: 159
  • Gen: 4 Avg 139 Growth: 0.15 50+ Female: 153 Top1-2% Male: 166
  • Gen: 5 Avg 149 Growth: 0.13 50+ Female: 161 Top1-2% Male: 172

Tweak heritability (0.5) & decay (80%), and average IQ of your founding stock population for your story telling purposes needs.


Given that 3rd Generation average has enough IQ to graduate at MIT, and outliers could become professors there, the IQ is not a problem anymore. Your country is constrained by other factors such as culture, lack of capital, research facilities, high-tech industry etc. So most likely majority of your bred geniuses would end up leaving your middle income country and emigrate to some rich western country. Read Garret Jones Hive Mind: How Your Nation's IQ Matters So Much More Than Your Own to see the issue.


Since you are selecting only small percentage 1-2% of males you are vulnerable to the founder effect. If your outlier males have some kind of quirky disease it gonna spread like wildfire if you don't cull it from your population. Think something like Tay–Sachs.

  • 3
    $\begingroup$ @Separatrix Breeding stock might be drawn from subpopulation with higher IQ, for example 95% are normal ratbertians and 5% are smart merchant minorty dogbertians.IQ for the country is 100, but since your founding stock is dogbertian based, you have to start with their average IQ. $\endgroup$
    – Taco
    Oct 5, 2018 at 14:07
  • 6
    $\begingroup$ Great answer but I think your math is a little optimistic because you aren't considering population size. Humans can't produce unlimited numbers of children so with each generation your original population of 10,000 will shrink significantly depending on how strict your exclusion criteria are. With your current numbers, you are only taking the top 1-5% of each generation. The second generation will only have a few hundred acceptable breeders in it so you won't have enough people to continue selection for a 3rd generation. You need more time, more population, or an artificial breeding program. $\endgroup$ Oct 5, 2018 at 14:52
  • 13
    $\begingroup$ People have bred humans before, and not just slaves either. Nobility in monarchies all over the world carefully chose who their kids would marry, in some cases based on physical traits. I don't know if you count this, but the Olympics are also somewhat of a breeding program. There are quite a few current olympians who are the children of previous olympians. The Olympics provides a way for the best athletes to meet and form romantic relationships, and since many traits that make good athletes are heritable, their kids end up being exceptional as well. $\endgroup$
    – Ryan_L
    Oct 5, 2018 at 16:28
  • 14
    $\begingroup$ This answer assumes that IQ is a measure of intelligence, and that it is a simple inherited gene like traits of pea plants. Instead, intelligence is ill-defined. Its inheritance and development is a tangled combination of many genetic and social factors. There is no single "smart gene". $\endgroup$
    – Schwern
    Oct 5, 2018 at 16:59
  • 7
    $\begingroup$ @Schwern The answer does not assume simple Mendelian inheritance of inheritance. The heritability estimate of 50% assumes that there are many small, additive factors that each contribute to the phenotype. In fact, if anything that 50% estimate errs on the low side. See Heritability of IQ. There is no single large "smart gene" but there are many small "smart genes". $\endgroup$ Oct 5, 2018 at 18:35

Intelligence is not a singular trait

There are many facets to intelligence. Let's start with some animal examples:


Everybody loves dogs and dogs love everybody. They have incredible social skills and are amazingly adapted to humans. They are the only domesticated animal that runs towards their "human" when frightened and not away. They can read human facial expression naturally. Even most apes have to learn that pointing towards something is not you just randomly raising your arms, but dogs instinctively look where you are pointing (if they trust you) as they understand the gesture.

When it comes to technical/mechanical understanding dogs are incredibly bad. Dogs are barely capable of solving mechanical puzzles to get their treats. They can follow commands to solve it, but they lack the capabilities of solving it on their own.

Dogs are very cooperative. They are very social amongst one another and with humans and can cooperate to achieve a common goal - like hunting. This is the main reason we bred dogs from wolves, originally.


Birds are the exact opposite of dogs regarding the two aspects I highlighted. Crows, for example, can solve mechanical puzzles of several steps to secure their prized treats, but they are absolutely incapable of cooperation. When two birds are sitting in front of a see-through box with treats and a lever they can pull connected to the box, they will quickly find out they can lift the box to reach the treats, but each bird realizes it can not reach the treats while it pulls the lever. So they stop. They could cooperate, one pulls the lever, the other one pulls out the treats, but they will not. Each bird is just thinking of "how can I get the treat?". Cooperation is not ingrained in their psychology.

IQ Tests

What we usually refer to as intelligence with humans is mathematical and geometrical understanding. That is what IQ tests are going for in most cases, too.

"Here is a sketch of body X. Draw it rotated 90° to the right." Questions like this are aiming for a persons spatial understanding.

IQ tests are just a rough estimation of a person's skills. They are everything but a precise measurement of intelligence. Intelligence is an abstract concept with many facets that can not be expressed in a singular number. Simply do a second IQ test right after your first and your score will be higher. Just because you know what to expect of the questions you will be quicker and more secure in your decisions. You will not have been getting more intelligent by doing a single IQ test.

Now to the main question.

Could intelligence be bred?

Yes, basically any trait could be artificially selected for. It is more likely for two people with a very healthy and fit body to have children that are more prone to having healthy and fit body. (Same goes for unhealthy, btw.) If you have two people who are strong exhibitors of any trait it is likely that their offspring will be carrying that trait, too.

Intelligence is not a singular trait. There are a many influencing factors, genetically speaking. People with genetic defects like Down-Syndrome can still be perfectly functional humans, but it is significantly less likely for them to be above average in most things they do.

Other factors like high testosterone, estrogen and other hormone levels influence a person's tendencies and traits, too. Every hormone differently and there are always strong exceptions. Biochemistry is the most difficult thing to model precisely.

Does it suffice to just start a colony of intelligent people?

Not necessarily, no. Just because somebody is smart, does not mean they are a good parent. While many traits like physical prowess thrive on exercise, so does intelligence. It is extremely unlikely to grow up to be intelligent, when you have never had any nurturing of this skill. The human brain and body build what they need. When a child's mind is never given any intellectual input, why would it develop a skill to interpret such input?

The intelligent parents' DNA just makes it more likely that the child is going to be good at building up intelligence. If it is never given an opportunity to build it up it will not use that potential.

What you need

...is a program that selects compatible intelligent people and makes sure that they are raised well, including a healthy home environment and parental attention. Early education is important. Focus all of the education and attention on rewarding intellectual skill. This means lots of challenging games from a young age one in which the children can hone their skills. If you want them having a healthy psyche it should always be focusing on rewarding progress instead of demanding progress, the latter works to but comes with side effects I'll talk about below.

An important side note

While you could also create socially incapable, semi-autistic people with a very high IQ with less effort, by ignoring the "healthy" part I mentioned, I would personally advise against it as the people might have advanced skill sets, but their lack of social skills would make it more difficult for these skills to be applied in a productive context.

Besides that, it would also be child abuse. Not too fond of that.

  • 14
    $\begingroup$ Nice to see an answer explaining what IQ tests actually are and how intelligence cannot be measured with something as simple as an integer score. $\endgroup$ Oct 5, 2018 at 16:43
  • 6
    $\begingroup$ +1 for looking under the surface to realize that a human breeding program is inherently destructive to intelligence. $\endgroup$
    – MindS1
    Oct 5, 2018 at 17:43
  • 2
    $\begingroup$ @TracyCramer Not all genius scientists come from poverty. Yes, there are always some and there will always be, but statistically we have more intelligent people in the higher ranks of society. Just because someone is poor does not mean they have bad parents. And parenting is a very important factor in nurturing intelligence. But there are even rare cases where not even good parenting is given. And the children are extremely autodidactic regarding their field of interest. It happens, but those are extremely rare. $\endgroup$ Oct 6, 2018 at 8:36
  • 1
    $\begingroup$ I will say this in the most respectful way possible, but this answer reads like a typical layman's account of intelligence research and is in stark contrast to what actual experts would say. I don't understand why laymen always feel so confident in expressing 'facts' about intelligence when they clearly do not the scientific literature well enough. Strangely enough this doesn't seem to be happen to the same extend in other fields. Almost every single claim in this post, while generally having some mild truth to them, is very misleading. I say this a person who knows the literature inside out. $\endgroup$
    – Eff
    Oct 8, 2018 at 8:53
  • 1
    $\begingroup$ Unfortunately, these things require extensive discussion, so I should probably just recommend any decent book on intelligence research. For example, Richard Haier's "The Neuroscience of Intelligence" or Stuart Ritchie's "Intelligence" are two examples of decent books that are not too difficult to read. $\endgroup$
    – Eff
    Oct 8, 2018 at 9:00

The question as posed can't be answered in a meaningful way. "Intelligence" isn't a smoothly defined, observable trait in the same way that, say, height might be. So any effort to breed "smarter" humans is going to be part and parcel a response to how your government is trying to assess "smartness".

Additionally, even if we accept some effort to measure intelligence as meaningfully accurate, raw capacity alone isn't enough to get the results this government probably wants to see. Environmental factors are important in cognitive development as well (consider heavy exposure to lead, for example). So all of those need to be accounted for as well.

Intelligence (variably defined) also seems to be a very complex trait (in the sense that there doesn't appear to be any single gene, or small set of collocated genes, that are associated with it), so breeding to enhance it might well need far more than 10,000 individuals in a rotating stock. And even in the wider population intelligence (as typically measured today) tends to revert to the mean (two unusually smart parents might produce children of more average intelligence-- that is, geniuses do not exclusively beget additional geniuses).

Finally, there are serious ancillary matters that need to be considered. 10,000 is not a huge stock, and constantly breeding (and, presumably, inbreeding in this case) in pursuit of a single, complex trait is likely to cause unintended genetic issues as a function of that small population. These effects would be very difficult to predict in this abstract case.


There is so much handwaving required for such a program to exist and "work", and intelligence so vaguely defined in operational terms, that your scenario can produce humans with any degree of intelligence you like. You could represent this with an arbitrary scale (like a deflated IQ score that can account for the constant re-normalization of IQ scores, or some measure that you just make up), or in a less-quantified but easily observable way (children from the program tend to end up in high positions, as your story setup already suggests).

  • 1
    $\begingroup$ "Tameness" in foxes is a not-perfectly-well-defined concept, it is influenced by genetics and is a very complex trait. Environmental factors are probably important for tameness. It has regression to the mean, like all quantitative traits. Everything you've said in this post is also true for tameness in foxes. Despite all this, scientists could, without any difficulties whatsoever, select for tameness in foxes. What appears to "handwavy" is this argument against the possibility of it working which is directly contradicted by evidence from all selective breeding programs ever done. $\endgroup$
    – Eff
    Oct 8, 2018 at 9:31
  • $\begingroup$ @Eff I never said that intelligence can't be selected for, only that it's hard to do. The question in the OP is "how smart can they become", which is difficult to answer and depends on how much of the trait is down to genetic inheritance, how complex it is, and how the phenotype is measured. If foxes are tame enough by the standards you choose to measure the concept, and that works out as predicted in real life, then awesome. It's far from clear that those conditions hold or that such measurements exist for intelligence in humans generally. $\endgroup$
    – Upper_Case
    Oct 8, 2018 at 19:33

[Marilyn vos] Savant was listed in the Guinness Book of World Records under "Highest IQ" from 1986 to 1989[9] and entered the Guinness Book of World Records Hall of Fame in 1988.[9][10] Guinness retired the "Highest IQ" category in 1990 after concluding IQ tests were too unreliable to designate a single record holder.[9] The listing drew nationwide attention.[11]

Guinness cited vos Savant's performance on two intelligence tests, the Stanford-Binet and the Mega Test. She took the 1937 Stanford-Binet, Second Revision test at age ten.[5] She claims her first test was in September 1956 and measured her mental age at 22 years and 10 months, yielding a 228 score.[5] This figure was listed in the Guinness Book of World Records; it is also listed in her books' biographical sections and was given by her in interviews.


IQ is "supposed" to be normally distributed with mean 100 and sd 15. That puts an IQ of 228 as being 8.5 standard deviations from the mean. I tried to use Excel to calculate what percentile that would be, but it was so close to one that Excel rounded to exactly 1. Excel has 15 digits of accuracy, which corresponds to one quadrillion. So one should expect MvS levels of IQ in one in one quadrillion people. So it's understandable that Guinness has decided this is not a meaningful measure. At 130, IQ being normal would predict that we're already at the top 2.28%. That's about the limit at which IQ is valid.

For a somewhat quantitative measure, I took the estimate of 80% heritability, and assumed this is constant across IQ levels (it's probably not, but I'm not sure how to account for that).

I initialized a population with

genotypes = random.normal(loc=100,scale = 15,size=1000000)
phenotypes = [.8*genotype+.2* random.normal(loc=100,scale = 15) for genotype in genotypes]
indices = [index for index in range(len(phenotypes)) if phenotypes[index]>125]
genotypes = [genotypes[index] for index in indices] 
phenotypes = [phenotypes[index] for index in indices] 

(Note that I took a minimum of 125 so that the average would be around 130)

Next, I did:

for i in range(5):
    cutoff = sorted(phenotypes)[len(phenotypes)//2]
    genotypes = [genotypes[index] for index in range(len(genotypes)) if phenotypes[index]>cutoff ] *2   
    phenotypes = [.8*genotype+.2* random.normal(loc=100,scale = 15) for genotype in genotypes]

This simulates the top half of each generation having four kids per couple. I didn't bother modeling cross-breeding, assuming that it wouldn't affect the general trend much. The results I got were:

  1. 130.78537880060884
  2. 133.84842056079236
  3. 136.4018939835269
  4. 138.75734310708546
  5. 140.9813744502485

What if you have eight kids per couple? For that, I got

  1. 133.40673479773932
  2. 138.49951120551688
  3. 142.82479867356096
  4. 146.75034206831734
  5. 151.03096072228035

However, keep in mind that my numbers start after one generation of breeding. So, 133.4067 is the first generation to be bred. If "generation one" is just people taken from the general population with high IQ, then "generation five" corresponds to 146.75.

Also, 15 years is rather short for a generation. You have some leeway in that the people running this are, as you say "unscrupulous", and thus not concerned with age of consent, but even so, the whole reason (or at least, one of the reasons) that we have age of consent is that 15-year-olds' judgment is not the same as 20-year-olds. Some analyses have put the age at which the brain stops developing as late as 25. While you can take the 15-year-olds that are smarter relative to other 15-year-olds, and this would give you some information about their final intelligence, it would be of questionable validity. Furthermore, for a breeding program to work, you would have to have a lot of children; the more children you have, the more selective you can be. If each girl has four children by age 15, that puts the first conception at 12 at the latest, and that's only if they're continuously pregnant, which would be difficult to accomplish.

  • $\begingroup$ I tried to improve readability of the results by making it look better. I left an edit note. If you don't like it, feel free to revert it. $\endgroup$ Oct 7, 2018 at 11:48

You can definitely train for ability to score highly in IQ tests, so you should be able to breed for it. That's not going to make anybody "smarter" - all it will do is increase the ability to score highly on IQ tests. Even Mensa agrees that IQ is only an indicator of intelligence or ability to problem solve*

As pointed out by @F1Krazy, your group's median IQ will remain 100, but you will be able to raise median IQ as compared with outside the group, as even with the inclusion of environments less conducive to IQ, the improvement over time has been positive (see Flynn Effect) so providing an environment conducive to learning, personal development and support and an improvement in health should all contribute.

*I've been a member of Mensa for maybe 20 years and can categorically state that high IQ is not a consistent indicator of anything other than getting high scores in the Mensa test. Membership includes heads of business, unemployed plumbers, musicians, mathematicians...

  • 2
    $\begingroup$ I've known highly competent MENSA members, and one I wouldn't trust with a burnt-out match. $\endgroup$ Oct 5, 2018 at 18:21
  • $\begingroup$ Yup - exactly my point. $\endgroup$
    – Rory Alsop
    Oct 5, 2018 at 19:38

Intelligence is heritable


Scientists have investigated this question for more than a century, and the answer is clear: the differences between people on intelligence tests are substantially the result of genetic differences.


We are talking about average differences among people and not about individuals.


Any one person's intelligence might be blown off course from its genetic potential by, for example, an illness in childhood.


we know, for example, that later in life, children adopted away from their biological parents at birth are just as similar to their biological parents as are children reared by their biological parents.

Bottom line

You'll need to perform selective breeding, "just" as farmers breed plants, ranchers breed cattle, and rich people breed dogs.


we can't answer how much smarter they'd be.


High intelligence correlates to higher rates of mental illness.


The biggest differences between the Mensa group and the general population were seen for mood disorders and anxiety disorders. More than a quarter (26.7 percent) of the sample reported that they had been formally diagnosed with a mood disorder, while 20 percent reported an anxiety disorder—far higher than the national averages of around 10 percent for each.

  • $\begingroup$ I am intrigued by your caveat that high intelligence correlates to higher rates of insanity. Have you got a source for this? If so could you add it to your answer? $\endgroup$
    – mwarren
    Oct 5, 2018 at 14:01
  • $\begingroup$ @mwarren adding the citation now. $\endgroup$
    – RonJohn
    Oct 5, 2018 at 14:07
  • $\begingroup$ @mwarren and changed "insanity" to "mental illness". $\endgroup$
    – RonJohn
    Oct 5, 2018 at 14:10
  • 1
    $\begingroup$ @RonJohn sure you can, the brain has plasticity the parts that see more use development more, no matter how you measure intelligence it can change and improve during a persons life. The ability to improve is not infinite but it definitely can improve. $\endgroup$
    – John
    Oct 6, 2018 at 4:27
  • 2
    $\begingroup$ "High intelligence correlates to higher rates of mental illness" - note that the study is based on members of Mensa (who are not necessarily representative of "high intelligence") and that it's really reporting on diagnosed mental illness. An alternate explanation would be "the kind of people who join Mensa are also more likely to be getting the medical attention required to detect these conditions". $\endgroup$ Oct 7, 2018 at 13:39

If instead of traditional breeding methods one were to use a method called iterated embryo selection [1][2], the potential IQ gain values could be much higher than the ones indicated by the other answers.

In their paper [1], Shulman & Bostrom note that

The effectiveness of embryo selection would be vastly increased if multiple generations of selection could be compressed into less than a human maturation period. This could be enabled by advances in an important complementary technology: the derivation of viable sperm and eggs from human embryonic stem cells. Such stem-cell derived gametes would enable iterated embryo selection (henceforth, IES):

  1. Genotype and select a number of embryos that are higher in desired genetic characteristics;
  2. Extract stem cells from those embryos and convert them to sperm and ova, maturing within 6 months or less [3];
  3. Cross the new sperm and ova to produce embryos;
  4. Repeat until large genetic changes have been accumulated.

Assuming a normal distribution of IQ scores with $\sigma = 7.5$, they [1] estimate the upper bound of the gain values from this method in a single generation as follows:

| Selection | Average IQ Gain |
| 1 in 2    | 4.2             |
| 1 in 10   | 11.5            |
| 1 in 100  | 18.8            |
| 1 in 1000 | 24.3            |

The predicted gain over 5 and 10 generations using the relatively modest 1-in-10 scheme have upper bounds of 65 and 130 points, respectively, after discounting the loss from diminishing returns, and the theoretical maximum gain over multiple generations has an upper bound in the interval between 30 and 40 standard deviations (225 and 300 points), likewise discounted.

[1] C. Shulman and N. Bostrom, 2014
[2] G. Branwen, 2016
[3] R. Sparrow, 2013


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .