My favorite bit from J.S. Mills' autobiography is when he falls into a massive depression at about age 18 and then after a year or two he finds something that cheers him up: music.
But then he decides that Beethoven and Schubert must have invented all possible new melodies, so he gets depressed again for years.
A former grad student of Leon Kamin told me that Kamin couldn't really believe that anybody could see pictures in their mind's eye. Kamin couldn't, so he didn't believe anybody else did. On the other hand, he said, Kamin had extraordinary kabbalistic savant capabilities at mentally working with letters and numbers.
*Is* your English ability much impacted by being a non-native speaker, d'you think? I am assuming Danes learn English quite early on, if it's anything like it is in Germany; and after so much writing & reading in English, for so long... I dunno; it would just sort of surprise me, I guess, if the vast majority of the handicap—from not "cradle-speaking" a language—is found to disappear in such a case.
Surely, those few early-childhood years can correspond to only a tiny bit of one's Anglinguistic competence—a missing ant-hill, at most, upon the Himalayan peak formed by decade(s) of professional-level work (& undoubtedly personal conversation, etc., too)...?
I admit that I've read no actual empirical investigation upon this sort of thing, though, and it's... uh... not exactly very germane to the post, really; just surprised me a bit to read it, 'sall–
I don't recall the details too well, but I do remember being unimpressed.
For example, and IIRC, Joseph makes some bad arguments about epigenetics: he predicts that epigenetic effects will soon be found to explain much of the (apparently inherited) variation in mental traits—which has not, as far as I am aware, actually happened, and indeed (I'd argue) betokens a misunderstanding of the mechanics at work in the first place. (This may have been in the book or may have been in a paper he wrote, can't recall for sure; but I *think* it was in the book.)
Similarly, he makes some unwarrantedly strong claims about shared environment, while apparently failing to realize (a) that the question has, in fact, been investigated, contra his charge that the "twin-studies camp" doesn't ever consider the matter & would be left flat-footed if presented it; and (b) that any systematic underestimation of the differential between MZ- & DZ- twin environment would have to be implausibly large to make any real difference. (He implies—again, IIRC—that this could conceivably explain away /all/ of the apparent heritability suggested by twin-studies... but this is both prima facie quite difficult to credit, and not supported by the evidence—I don't think such a hypothesis was very supportable even at time of publication, but the past couple decades haven't been very kind to it either.)
--------------------------
I want to accuse the book of having a habit of ignoring evidence which would cause difficulty for the author's thesis (say, adoption studies), and ignoring that "here's something that could bias the results of these studies!" has much less force when evidence from multiple angles converges (i.e., pretending that the hereditarian case stands or falls on MZ-vs.-DZ twin-studies alone)...
...but these might not actually be fair accusations—can't recall for sure which (types of) evidence he did/didn't consider, and maybe some of the papers I'm thinking about weren't even available back then; and the book /is/ (after all) entitled "The Trouble with Twin Studies" rather than "The Trouble with Evil Racist Hereditarians".¹
Our esteemed host is, of course, far more competent to review the book than I am—but if there's interest, and if Emil doesn't wish to spend any time upon such a task, I could probably go back through it & post a more detailed (and perhaps more accurate, for not relying upon possibly-conflated memory, heh) write-up.
.
-------------------------------------
¹: (interpreting things the latter way is something of a knee-jerk reaction of mine, I admit—but after a while, they all blur together: it's such a popular position, and nearly every time it /does/ turn out that the author of a work like the former turns out to hold an opinion best-described by the latter)
Thanks. I also didn't have enough competency to review the book but here were my notes:
I found his strongest point the lack of transparency by the MISTRA researchers in how they're not sharing their data (even anonymized), though that's sadly common with researchers. Another very strong point he makes is that some (though not most) of the twin pairs that are in the large reared apart studies were actually reared together for up to a decade or more, or that some were reared by close family and lived near each other and had a lot of contact; the researchers admit this, but it's still a good point that the data is very muddy. The researchers include this data because it's so rare for twins to be separated and adopted by different families so they were trying to get more statistical power and they tried to control and model these points, but controlling and modeling are always risky in statistics. I'd like to see the data just zoomed in on the "cleanest" reared apart cases.
Joseph also makes a lot of interesting methodological claims but I think he did a poor job steelmanning the pro-twin-studies arguments, ending with what seems to me a hyperbolic conclusion that behavioral genetics is a pseudoscience. In particular, I just didn't see him struggle much with the point that most of the data seems to point in the same direction (both in reared together and reared apart studies). One proposal he makes is that there could be some huge confounds that cause this which leads to his more fundamental critiques (ultimately concluding it's a pseudoscience) and I think his arguments here just aren't that convincing though they're hard to fully evaluate without doing a deep dive steelman. I found it telling that he often used the phrase, "not scientifically acceptable evidence" in trying to explain some of the pieces of evidence that didn't reach the rigor that he wants: acceptable by whom? Sure, I'm a huge fan of scientific rigor and improving it, but if most evidence is pointing in the same direction, that would at least tend to make me put the burden of proof on the counter-arguments, especially with genetics which we know for sure has provable effects at least in some cases. Yes, it could be true that physical genetic effects are totally different than psychological genetic effects and psychological effects are mostly controlled by environment/parenting, but it seems at least plausible that the opposite is true since we know that genes ultimately create all of our body; so, if there is a huge mountain of evidence that points at the latter and it's biologically plausible, then I still go back to burden of proof.
With all that, I think the book is very thought-provoking and I think his methodological points are important, since of course we are not doing randomized controlled trials here: we're just looking at correlations of natural experiments and there are lots of confounding variables (which he validly points out), but I wasn't impressed with the way he tries to deal with the way that the data all seem to point in the same direction. He also has a bunch of woke-type arguments of guilt by association in that the eugenics movements and the Nazis also were looking into genetics a lot, so that was a big yellow flag. Clearly we know that genetics is a big factor in life, and the fact that some reprehensible people used the field in a bad way seems irrelevant to me. Finally, one of the researchers he cites which generally agrees with him at least proposes some potential ways to improve the twin studies that might resolve some of the methodological issues whereas he just ends the book by calling the whole field a pseudoscience which I thought was another yellow flag suggesting motivated reasoning.
So, overall, I'd say my confidence in the twin studies went down a little bit but overall I didn't find his core methodological critiques very persuasive and I think the burden of proof is on him to better substantiate that behavioral genetics is a pseudoscience.
The strongest argument appears to me to be, according to Joseph, in my summary:
1. The MISTRA general model supposes that the DZA (dizygotic raised apart) correlation should be one-half of MZA (monozygotic raised apart) (^r dza = 0.5h^2).
2. Originally, MISTRA researchers designated DZA pairs as the control group: "According to Segal's account, from the beginning of the study in 1979, Bouchard had designated DZA pairs as the MISTRA control group".
3. However, MISTRA researchers often didn't publish DZA correlations with various explanations; for example, from a MISTRA study: "Although intraclass correlations are often informative [...] for analytic purposes they can be misleading if MZ and DZ variances differ. Biometric geneticists [...] therefore, prefer analyzing variances over correlations" (Tellegen et al., 1988).
4. In other cases, DZA correlations exceeded MZA; for example, from a MISTRA study: "For 4 tests, DZ correlations actually exceeded MZ correlations, a situation we attribute to sampling variability" (Johnson et al., 2007).
5. As an example with IQ, Joseph writes, "There does not appear to be a statistically significant difference between the MISTRA near full-sample MZA and DZA correlations for either the WAIS IQ or Raven IQ scores, which runs counter to genetic theories which require the MZA correlation to be significantly higher than the DZA correlation". This is important because "IQ was the main focus area of the study, and that the researchers had designated DZA twins as the MISTRA control group they would compare with MZAs."
6. MISTRA researchers have been stingy with sharing data, even anonymized: "The researchers have, for decades, withheld important data produced by their own designated control group that may have cast doubt upon their conclusions in favor of genetics."
7. Therefore, Joseph is suspicious of motivated reasoning. Overall, he concludes, "problem areas such as the similarity of the twins' social and cultural environments, the researchers' failure to make their data available for inspection by independent reviewers, the questionable assumptions underlying the researchers' model-fitting procedures, sample size issues, recruitment bias, the researchers' decision to bypass the step of determining that the MZA correlation is significantly higher than the DZA correlation, a reliance on twins' potentially unreliable accounts of their degree of separation and behavioral similarity, and issues related to the control group."
Hell, that's a better job that I could have done—thanks for this! Very interesting; perhaps I didn't give Joseph quite enough credit; possibly I have fallen into sort of an "arguments are soldiers" mindset, wherein admitting that he finds some legitimate confusions & possible issues = "betraying" "my" "side"...
...but, of course, I also think you're correct in that he doesn't clear the bar of "it's a pseudoscience; toss 'em all out!"—not even close; I don't believe that even a proper steelman of his arguments would warrant such a conclusion, really. People in general (including me, perhaps, heh!) tend to have trouble with the sort of nuance you display here: criticism along the lines of "here are some ways some results could be weaker than purported" need to be evaluated in light of the larger picture... but the temptation is, seemingly, to take "there is a possible confounder in this study" as meaning "therefore we have 'permission' to reject its conclusion!" (if one doesn't like it).
Now I'm thinkin' I'd like to try—or see you, or Emil, try!—to do a thorough, point-by-point consideration. (E.g., how many of the putatively-raised-apart twins were actually raised together; and how is, or ought this be, controlled for; and how much uncertainty ought this introduce? How do the results Joseph discusses square, or not square, with similar studies—esp. those he didn't, or couldn't, consider? Etc., etc.)
My favorite bit from J.S. Mills' autobiography is when he falls into a massive depression at about age 18 and then after a year or two he finds something that cheers him up: music.
But then he decides that Beethoven and Schubert must have invented all possible new melodies, so he gets depressed again for years.
Emil, thank you for reviewing the history of intelligence studies. Very interesting to see the evolution of thought over the decades.
Really interesting. I learned a lot from this - which was not hard since I knew so little before reading it.
Very helpful. Many thanks.
A former grad student of Leon Kamin told me that Kamin couldn't really believe that anybody could see pictures in their mind's eye. Kamin couldn't, so he didn't believe anybody else did. On the other hand, he said, Kamin had extraordinary kabbalistic savant capabilities at mentally working with letters and numbers.
The book gives examples of this numerical ability, at least, in his own words.
*Is* your English ability much impacted by being a non-native speaker, d'you think? I am assuming Danes learn English quite early on, if it's anything like it is in Germany; and after so much writing & reading in English, for so long... I dunno; it would just sort of surprise me, I guess, if the vast majority of the handicap—from not "cradle-speaking" a language—is found to disappear in such a case.
Surely, those few early-childhood years can correspond to only a tiny bit of one's Anglinguistic competence—a missing ant-hill, at most, upon the Himalayan peak formed by decade(s) of professional-level work (& undoubtedly personal conversation, etc., too)...?
I admit that I've read no actual empirical investigation upon this sort of thing, though, and it's... uh... not exactly very germane to the post, really; just surprised me a bit to read it, 'sall–
When people give numbers for heritability of IQ, do they do Spearman's correction for the reliability of the IQ tests of the parents and the child?
Not generally, but Jensen did, which is one reason his numbers are a bit higher.
Do you have any opinions on the book The Trouble with Twin Studies by Jay Joseph?
Didn't read any of his books.
I don't recall the details too well, but I do remember being unimpressed.
For example, and IIRC, Joseph makes some bad arguments about epigenetics: he predicts that epigenetic effects will soon be found to explain much of the (apparently inherited) variation in mental traits—which has not, as far as I am aware, actually happened, and indeed (I'd argue) betokens a misunderstanding of the mechanics at work in the first place. (This may have been in the book or may have been in a paper he wrote, can't recall for sure; but I *think* it was in the book.)
Similarly, he makes some unwarrantedly strong claims about shared environment, while apparently failing to realize (a) that the question has, in fact, been investigated, contra his charge that the "twin-studies camp" doesn't ever consider the matter & would be left flat-footed if presented it; and (b) that any systematic underestimation of the differential between MZ- & DZ- twin environment would have to be implausibly large to make any real difference. (He implies—again, IIRC—that this could conceivably explain away /all/ of the apparent heritability suggested by twin-studies... but this is both prima facie quite difficult to credit, and not supported by the evidence—I don't think such a hypothesis was very supportable even at time of publication, but the past couple decades haven't been very kind to it either.)
--------------------------
I want to accuse the book of having a habit of ignoring evidence which would cause difficulty for the author's thesis (say, adoption studies), and ignoring that "here's something that could bias the results of these studies!" has much less force when evidence from multiple angles converges (i.e., pretending that the hereditarian case stands or falls on MZ-vs.-DZ twin-studies alone)...
...but these might not actually be fair accusations—can't recall for sure which (types of) evidence he did/didn't consider, and maybe some of the papers I'm thinking about weren't even available back then; and the book /is/ (after all) entitled "The Trouble with Twin Studies" rather than "The Trouble with Evil Racist Hereditarians".¹
Our esteemed host is, of course, far more competent to review the book than I am—but if there's interest, and if Emil doesn't wish to spend any time upon such a task, I could probably go back through it & post a more detailed (and perhaps more accurate, for not relying upon possibly-conflated memory, heh) write-up.
.
-------------------------------------
¹: (interpreting things the latter way is something of a knee-jerk reaction of mine, I admit—but after a while, they all blur together: it's such a popular position, and nearly every time it /does/ turn out that the author of a work like the former turns out to hold an opinion best-described by the latter)
Thanks. I also didn't have enough competency to review the book but here were my notes:
I found his strongest point the lack of transparency by the MISTRA researchers in how they're not sharing their data (even anonymized), though that's sadly common with researchers. Another very strong point he makes is that some (though not most) of the twin pairs that are in the large reared apart studies were actually reared together for up to a decade or more, or that some were reared by close family and lived near each other and had a lot of contact; the researchers admit this, but it's still a good point that the data is very muddy. The researchers include this data because it's so rare for twins to be separated and adopted by different families so they were trying to get more statistical power and they tried to control and model these points, but controlling and modeling are always risky in statistics. I'd like to see the data just zoomed in on the "cleanest" reared apart cases.
Joseph also makes a lot of interesting methodological claims but I think he did a poor job steelmanning the pro-twin-studies arguments, ending with what seems to me a hyperbolic conclusion that behavioral genetics is a pseudoscience. In particular, I just didn't see him struggle much with the point that most of the data seems to point in the same direction (both in reared together and reared apart studies). One proposal he makes is that there could be some huge confounds that cause this which leads to his more fundamental critiques (ultimately concluding it's a pseudoscience) and I think his arguments here just aren't that convincing though they're hard to fully evaluate without doing a deep dive steelman. I found it telling that he often used the phrase, "not scientifically acceptable evidence" in trying to explain some of the pieces of evidence that didn't reach the rigor that he wants: acceptable by whom? Sure, I'm a huge fan of scientific rigor and improving it, but if most evidence is pointing in the same direction, that would at least tend to make me put the burden of proof on the counter-arguments, especially with genetics which we know for sure has provable effects at least in some cases. Yes, it could be true that physical genetic effects are totally different than psychological genetic effects and psychological effects are mostly controlled by environment/parenting, but it seems at least plausible that the opposite is true since we know that genes ultimately create all of our body; so, if there is a huge mountain of evidence that points at the latter and it's biologically plausible, then I still go back to burden of proof.
With all that, I think the book is very thought-provoking and I think his methodological points are important, since of course we are not doing randomized controlled trials here: we're just looking at correlations of natural experiments and there are lots of confounding variables (which he validly points out), but I wasn't impressed with the way he tries to deal with the way that the data all seem to point in the same direction. He also has a bunch of woke-type arguments of guilt by association in that the eugenics movements and the Nazis also were looking into genetics a lot, so that was a big yellow flag. Clearly we know that genetics is a big factor in life, and the fact that some reprehensible people used the field in a bad way seems irrelevant to me. Finally, one of the researchers he cites which generally agrees with him at least proposes some potential ways to improve the twin studies that might resolve some of the methodological issues whereas he just ends the book by calling the whole field a pseudoscience which I thought was another yellow flag suggesting motivated reasoning.
So, overall, I'd say my confidence in the twin studies went down a little bit but overall I didn't find his core methodological critiques very persuasive and I think the burden of proof is on him to better substantiate that behavioral genetics is a pseudoscience.
The strongest argument appears to me to be, according to Joseph, in my summary:
1. The MISTRA general model supposes that the DZA (dizygotic raised apart) correlation should be one-half of MZA (monozygotic raised apart) (^r dza = 0.5h^2).
2. Originally, MISTRA researchers designated DZA pairs as the control group: "According to Segal's account, from the beginning of the study in 1979, Bouchard had designated DZA pairs as the MISTRA control group".
3. However, MISTRA researchers often didn't publish DZA correlations with various explanations; for example, from a MISTRA study: "Although intraclass correlations are often informative [...] for analytic purposes they can be misleading if MZ and DZ variances differ. Biometric geneticists [...] therefore, prefer analyzing variances over correlations" (Tellegen et al., 1988).
4. In other cases, DZA correlations exceeded MZA; for example, from a MISTRA study: "For 4 tests, DZ correlations actually exceeded MZ correlations, a situation we attribute to sampling variability" (Johnson et al., 2007).
5. As an example with IQ, Joseph writes, "There does not appear to be a statistically significant difference between the MISTRA near full-sample MZA and DZA correlations for either the WAIS IQ or Raven IQ scores, which runs counter to genetic theories which require the MZA correlation to be significantly higher than the DZA correlation". This is important because "IQ was the main focus area of the study, and that the researchers had designated DZA twins as the MISTRA control group they would compare with MZAs."
6. MISTRA researchers have been stingy with sharing data, even anonymized: "The researchers have, for decades, withheld important data produced by their own designated control group that may have cast doubt upon their conclusions in favor of genetics."
7. Therefore, Joseph is suspicious of motivated reasoning. Overall, he concludes, "problem areas such as the similarity of the twins' social and cultural environments, the researchers' failure to make their data available for inspection by independent reviewers, the questionable assumptions underlying the researchers' model-fitting procedures, sample size issues, recruitment bias, the researchers' decision to bypass the step of determining that the MZA correlation is significantly higher than the DZA correlation, a reliance on twins' potentially unreliable accounts of their degree of separation and behavioral similarity, and issues related to the control group."
(P.S.—you are likely to have seen it already, but Scott Alexander's post on "missing heritability" is pretty interesting & sort of related: www.astralcodexten.com/p/missing-heritability-much-more-than )
Hell, that's a better job that I could have done—thanks for this! Very interesting; perhaps I didn't give Joseph quite enough credit; possibly I have fallen into sort of an "arguments are soldiers" mindset, wherein admitting that he finds some legitimate confusions & possible issues = "betraying" "my" "side"...
...but, of course, I also think you're correct in that he doesn't clear the bar of "it's a pseudoscience; toss 'em all out!"—not even close; I don't believe that even a proper steelman of his arguments would warrant such a conclusion, really. People in general (including me, perhaps, heh!) tend to have trouble with the sort of nuance you display here: criticism along the lines of "here are some ways some results could be weaker than purported" need to be evaluated in light of the larger picture... but the temptation is, seemingly, to take "there is a possible confounder in this study" as meaning "therefore we have 'permission' to reject its conclusion!" (if one doesn't like it).
Now I'm thinkin' I'd like to try—or see you, or Emil, try!—to do a thorough, point-by-point consideration. (E.g., how many of the putatively-raised-apart twins were actually raised together; and how is, or ought this be, controlled for; and how much uncertainty ought this introduce? How do the results Joseph discusses square, or not square, with similar studies—esp. those he didn't, or couldn't, consider? Etc., etc.)