Homosexuality is mildly heritable. You can’t posit that it’s hard for evolution to find a way around it, because it has ways. Perhaps not perfect ways, but the heritability should quickly fall it unmeasurably small values. That really requires explanation.
You seem to be making an extremely strong claim: that every heritable deleterious condition must be due to a hidden benefit or recent change. (And in a long-term stable evolutionary environment, would always be due to a hidden benefit.)
What is the basis for this? Why can’t we simply posit that some tasks are complex (embryological development), and some tasks are adversarial (countering parasites and diseases), and this explains a large part of heritable fitness-reducing traits?
When you compare it to birth defects, you’re comparing one problem against a host of problems. It’s worth about the same to evolution to fix the one as to fix all the rest.
This is a good argument.
Do we know or expect that all or most cases of homosexuality have the same underlying cause, or their heritable component does, so that a single localized mutation could eliminate them? Or could there be many different causes?
Also, this feels in need of more quantitative argument. Anencephaly, or even all causes of stillbirths taken together, obviously cause lower fitness for the children than even obligate homosexuality. On the other hand, they don’t make the parents spend resources raising low-fitness children whose homosexuality only becomes apparent at puberty. I don’t really know what size difference to expect here and whether to be surprised by two orders of magnitude.
You seem to be making an extremely strong claim: that every heritable deleterious condition must be due to a hidden benefit or recent change. (And in a long-term stable evolutionary environment, would always be due to a hidden benefit.)
I don’t think he’s making a claim that strong.
It depends on strength of the heritability and the size of the trait on selection pressure.
It also depends on the extend to which you easily get mutations in genes that produce the trait. Random gene mutations lead to birth defects. It’s not clear why you should get homosexuality in the same way through random mutations.
Part of my argument is that homosexuality (inasfar as it is determined in the womb) may not (always) be caused directly by a mutation, but by the fact that even in the absence of any mutation, the developmental process linking gender to sexual orientation is complex and fragile and has a significant failure rate.
If something is heritable but also uncommon, it should be possible for selection pressure to act on it. So why doesn’t it? That clarifies the issue for me. What I wrote before (in this comment thread) now seems mistaken. Thank you for helping me to understand better.
It seems my original question was partly due to a misunderstanding. The crucial fact about homosexuality isn’t just the lowered fitness or relatively high incidence, but the (partial) heritability. Then the only relevant remaining question is: can we quantitatively estimate the reduced fitness due to homosexuality and so calculate the unlikelihood of its 3% rate being due to chance, drift, etc?
Your original post does not talk about heritability. So my answer was: heritability. But then I noticed that it was in the post title and I was confused and did not change my answer.
I posted a calculation on SSC. Let’s say that there is a mutation that is spontaneously created in 1 in 10,000, that it has a 10% chance of producing a homosexual phenotype and that the phenotype has 0.9 fitness, that is, yields an average of 1.8 children. Then the fitness of the gene in 0.99. So in equilibrium, the gene only reproduces itself 99%, so the remaining 1% must be made up by the spontaneous mutation. That is, the prevalence is 100x the spontaneous mutation rate. The prevalence is 1% of the population, of which 0.99% inherit the gene and 0.01% spontaneously acquire it. That’s a genotypic prevalence of 1%. The phenotypic prevalence is 0.1%.
I think 1 in 10,000 is the standard rule of thumb mutation rate. For example, achondroplasia (dwarfism) has a spontaneous mutation rate of 3 in 100,000 and an inherited rate of 1 in 100,000. Apparently dwarfs have about 1⁄4 of replacement fertility. (The fatality of homozygous achondroplasia complicates the situation. The gene definitely has a fitness of 1⁄4, but if dwarfs only marry dwarfs, the fitness of the people would be 3⁄8, I think.)
Also, the approach to equilibrium is exponential with base the fitness.
I gave you a worked example, with real empirical data: achondroplasia has a spontaneous rate of 3⁄100,000. Try to understand the example rather than pulling facts out of your head with no understanding.
The rate of spontaneous mutation in humans is on the order of 10^-8 per haploid base-pair per generation. A few references: one (a blog post summarizing multiple academic papers), two (I think this is one of the papers cited by the foregoing), three (older; rate is ~2x higher), four (intermediate in age and also in estimated rate).
A given gene is many base-pairs long. Accordingly, the “per locus” mutation rate—which you can think of as how often a particular gene goes wrong, treating all its failure modes as equivalent—is on the order of 1000x higher. (This, I take it, is where the ~3x10^-5 spontaneous rate for achondroplasia comes from.) The usual cited per-locus rates are, accordingly, on the order of 10^-5, typically a bit lower; see e.g. one (a textbook, which specifically mentions achondroplasia and gives some reasons not to take its rate of spontaneous occurrence too literally as a per-locus mutation rate), two (old paper by J B S Haldane, mostly of historical interest now), three (journal article).
Which figure is more relevant in a given case depends on whether it’s one where messing up a single protein, no matter how, causes the effect you’re interested in (which seems to be the case for achondroplasia) or one where a very specific change is needed. Which is more likely for homosexuality? I’ve no idea; more likely than either, I’d have thought, is that sexuality is complicated, that there are lots of genetic changes that can affect someone’s propensity for same-sex attraction, and that treating homosexuality as the result of a single genetic change is just wrong.
Anyway, to summarize: no one, so far as I can tell, says that 10^-4 is “the standard rule-of-thumb mutation rate”, either per locus or per nucleotide; 10^-8 is roughly the right figure per nucleotide; the right figure per locus is more like 10^-5; which of these figures, if either, is most relevant when considering homosexuality is unclear but for any remotely credible figure it’s clear that most homosexuality is not the result of spontaneous mutations.
looks like 3⁄4 to me. not sure where you got 3⁄8. (also where two dwarves can have a normal child).
If you selectively look at the living (of which there are 3 options—Aa, Aa or aa) the gene has a 2⁄3 chance of being passed on. Assuming no other pressures apply.
First of all, you should distinguish between the fitness of the gene and the fitness of the people. Second, I am using as input the empirical observation that the fitness of the achondroplasia gene is 1⁄4. Third, and tangentially, you should distinguish between the fitness of the children and parents.
(1 gene vs parents) Let us consider the 3 surviving children. Out of the 6 copies of the gene, 4 are wild type and 2 are achondroplasia. But in the parents, half of the genes are achondroplasia. Thus, regardless of how many children the parents have, the fitness of the gene is 2⁄3 the fitness of the parents.
(2) Empirically, 1⁄4 of achondroplasia births are inherited and 3⁄4 are de novo. Assuming equilibrium, the gene is producing 1⁄4 of replacement fertility, so it has a fitness of 1⁄4. If dwarfs only reproduce with non-dwarfs, they, too, have a fitness of 1⁄4. But if they only reproduce with dwarfs, they have a fitness 3⁄2 of the gene, thus 3⁄8.
(3 parents vs children) The 3⁄4 you compute is the reduction in the proportion of pregnancies yield children. This is a kind of infertility, though more emotionally difficult. It is only relevant if the parents are trying to reproduce as fast as possible. In the modern world, parents usually target a small fixed number of children and infertility has little effect. In both farmer and forager societies, children were probably modulated to available food supply. Such a wasted pregnancy does not reduce the number of children by 1, but probably delays future children by a year. If the usual interval is 4, this might reduce fitness by 1⁄4. But the effect is probably significantly smaller. If people are reproducing at the optimal speed, taking into account risk of famine, a small perturbation probably has little effect.
So why doesn’t it? That clarifies the issue for me. What I wrote before (in this comment thread) now seems mistaken. Thank you for helping me to understand better.
You seem to be making an extremely strong claim: that every heritable deleterious condition must be due to a hidden benefit or recent change. (And in a long-term stable evolutionary environment, would always be due to a hidden benefit.)
What is the basis for this? Why can’t we simply posit that some tasks are complex (embryological development), and some tasks are adversarial (countering parasites and diseases), and this explains a large part of heritable fitness-reducing traits?
This is a good argument.
Do we know or expect that all or most cases of homosexuality have the same underlying cause, or their heritable component does, so that a single localized mutation could eliminate them? Or could there be many different causes?
Also, this feels in need of more quantitative argument. Anencephaly, or even all causes of stillbirths taken together, obviously cause lower fitness for the children than even obligate homosexuality. On the other hand, they don’t make the parents spend resources raising low-fitness children whose homosexuality only becomes apparent at puberty. I don’t really know what size difference to expect here and whether to be surprised by two orders of magnitude.
I don’t think he’s making a claim that strong. It depends on strength of the heritability and the size of the trait on selection pressure.
It also depends on the extend to which you easily get mutations in genes that produce the trait. Random gene mutations lead to birth defects. It’s not clear why you should get homosexuality in the same way through random mutations.
Part of my argument is that homosexuality (inasfar as it is determined in the womb) may not (always) be caused directly by a mutation, but by the fact that even in the absence of any mutation, the developmental process linking gender to sexual orientation is complex and fragile and has a significant failure rate.
Then where does the heritability come from?
You’re right, this is a problem.
If something is heritable but also uncommon, it should be possible for selection pressure to act on it. So why doesn’t it? That clarifies the issue for me. What I wrote before (in this comment thread) now seems mistaken. Thank you for helping me to understand better.
It seems my original question was partly due to a misunderstanding. The crucial fact about homosexuality isn’t just the lowered fitness or relatively high incidence, but the (partial) heritability. Then the only relevant remaining question is: can we quantitatively estimate the reduced fitness due to homosexuality and so calculate the unlikelihood of its 3% rate being due to chance, drift, etc?
Your original post does not talk about heritability. So my answer was: heritability. But then I noticed that it was in the post title and I was confused and did not change my answer.
I posted a calculation on SSC. Let’s say that there is a mutation that is spontaneously created in 1 in 10,000, that it has a 10% chance of producing a homosexual phenotype and that the phenotype has 0.9 fitness, that is, yields an average of 1.8 children. Then the fitness of the gene in 0.99. So in equilibrium, the gene only reproduces itself 99%, so the remaining 1% must be made up by the spontaneous mutation. That is, the prevalence is 100x the spontaneous mutation rate. The prevalence is 1% of the population, of which 0.99% inherit the gene and 0.01% spontaneously acquire it. That’s a genotypic prevalence of 1%. The phenotypic prevalence is 0.1%.
I think 1 in 10,000 is the standard rule of thumb mutation rate. For example, achondroplasia (dwarfism) has a spontaneous mutation rate of 3 in 100,000 and an inherited rate of 1 in 100,000. Apparently dwarfs have about 1⁄4 of replacement fertility. (The fatality of homozygous achondroplasia complicates the situation. The gene definitely has a fitness of 1⁄4, but if dwarfs only marry dwarfs, the fitness of the people would be 3⁄8, I think.)
Also, the approach to equilibrium is exponential with base the fitness.
Its more like 10^-8.
http://www.sciencemag.org/content/328/5978/636.abstract
Nope, that’s the wrong rate.
I gave you a worked example, with real empirical data: achondroplasia has a spontaneous rate of 3⁄100,000. Try to understand the example rather than pulling facts out of your head with no understanding.
That seems needlessly inflammatory.
The rate of spontaneous mutation in humans is on the order of 10^-8 per haploid base-pair per generation. A few references: one (a blog post summarizing multiple academic papers), two (I think this is one of the papers cited by the foregoing), three (older; rate is ~2x higher), four (intermediate in age and also in estimated rate).
A given gene is many base-pairs long. Accordingly, the “per locus” mutation rate—which you can think of as how often a particular gene goes wrong, treating all its failure modes as equivalent—is on the order of 1000x higher. (This, I take it, is where the ~3x10^-5 spontaneous rate for achondroplasia comes from.) The usual cited per-locus rates are, accordingly, on the order of 10^-5, typically a bit lower; see e.g. one (a textbook, which specifically mentions achondroplasia and gives some reasons not to take its rate of spontaneous occurrence too literally as a per-locus mutation rate), two (old paper by J B S Haldane, mostly of historical interest now), three (journal article).
Which figure is more relevant in a given case depends on whether it’s one where messing up a single protein, no matter how, causes the effect you’re interested in (which seems to be the case for achondroplasia) or one where a very specific change is needed. Which is more likely for homosexuality? I’ve no idea; more likely than either, I’d have thought, is that sexuality is complicated, that there are lots of genetic changes that can affect someone’s propensity for same-sex attraction, and that treating homosexuality as the result of a single genetic change is just wrong.
Anyway, to summarize: no one, so far as I can tell, says that 10^-4 is “the standard rule-of-thumb mutation rate”, either per locus or per nucleotide; 10^-8 is roughly the right figure per nucleotide; the right figure per locus is more like 10^-5; which of these figures, if either, is most relevant when considering homosexuality is unclear but for any remotely credible figure it’s clear that most homosexuality is not the result of spontaneous mutations.
[EDITED to fix a typo.]
punnet square: Aa x Aa
…...............A...........a
. A...........AA.........Aa
. a............Aa..........aa
AA = death or infertile
Aa = like the parents
aa = normal
looks like 3⁄4 to me. not sure where you got 3⁄8. (also where two dwarves can have a normal child).
If you selectively look at the living (of which there are 3 options—Aa, Aa or aa) the gene has a 2⁄3 chance of being passed on. Assuming no other pressures apply.
First of all, you should distinguish between the fitness of the gene and the fitness of the people. Second, I am using as input the empirical observation that the fitness of the achondroplasia gene is 1⁄4. Third, and tangentially, you should distinguish between the fitness of the children and parents.
(1 gene vs parents) Let us consider the 3 surviving children. Out of the 6 copies of the gene, 4 are wild type and 2 are achondroplasia. But in the parents, half of the genes are achondroplasia. Thus, regardless of how many children the parents have, the fitness of the gene is 2⁄3 the fitness of the parents.
(2) Empirically, 1⁄4 of achondroplasia births are inherited and 3⁄4 are de novo. Assuming equilibrium, the gene is producing 1⁄4 of replacement fertility, so it has a fitness of 1⁄4. If dwarfs only reproduce with non-dwarfs, they, too, have a fitness of 1⁄4. But if they only reproduce with dwarfs, they have a fitness 3⁄2 of the gene, thus 3⁄8.
(3 parents vs children) The 3⁄4 you compute is the reduction in the proportion of pregnancies yield children. This is a kind of infertility, though more emotionally difficult. It is only relevant if the parents are trying to reproduce as fast as possible. In the modern world, parents usually target a small fixed number of children and infertility has little effect. In both farmer and forager societies, children were probably modulated to available food supply. Such a wasted pregnancy does not reduce the number of children by 1, but probably delays future children by a year. If the usual interval is 4, this might reduce fitness by 1⁄4. But the effect is probably significantly smaller. If people are reproducing at the optimal speed, taking into account risk of famine, a small perturbation probably has little effect.
sorry; point 2 again, (Aa x aa should product a 1⁄2 not a 1⁄4)
acondroplasia X normal
............A.............a
...a.......Aa..........aa
...a......Aa...........aa
50%Aa acondroplasia
50%aa normal
or am I confused somewhere? Is that not the punnet square?
Sure, that’s the punnet square. You should stop drawing punnet squares and ask yourself why you are drawing them and ask what role they play.
The number 1⁄4 is the empirical fitness. It is mainly about how many children dwarfs have. You cannot guess that number by looking at punnet squares.
Thanks for the calculation! I’ll probably revise the post tonight when I have some leisure time to integrate all the new info.
Congratulation for good updating.