Back to [Archive] Post-review discussions

[OBG] Sexual selection explains sex and country differences in fluid g
I think this paper has some extremely important insights and predictions and already deserves publication. Let me describe them in order below and then I would like to point out some relevant thesis by Nyborg (1994).

1. It has been long known well that men have more variability (larger SD) in IQ than women. However, not many scholars seriously tried to explain why this is so.

This paper assumes that IQ is a sexually selected trait that women have used to select men. This assumption helps understand why men have larger SD in IQ. Genes on sex chromosomes tend to be amplified in their expressions because sex chromosomes are haploid. This would make genes on sex chromosomes more easily expressed and selected than those on autosomal chromosomes. Hence, it seems reasonable for me that sexually selected traits tend to be on sex chromosomes.

I am not sure, though, why such sexually selected genes are on X-chromosome rather than on Y-chromosome, (as the author points out). Maybe this is something to do with segregation distortions due to the fact that there is no meiosis for sex chromosomes. (Certainly, this is not very different from the thesis of the paper.)

2. More novel prediction is that populations with higher IQ should show stronger sexual dimorphism in IQ. This is due to the fact that in higher IQ populations, sexual selection in terms of IQ must have been stronger.

Another prediction from the same assumption is that men in the higher IQ populations should show less SD because stronger selective pressure must have fixed more number of advantageous alleles in higher IQ populations.

Both of these are empirically supported in the paper. These predictions are highly novel and I have never heard of such or even similar postulates. I understand that there are some technical questions, such as possible existence of interactive factors between GDP and the sex difference, the usage of CPS as the best proxy for IQ. I agree that clarifying these points will improve the whole validity of the paper.


Some of the predictions may turn out to be false in the future. For example, East Asians are not only shorter in stature, but also their SD in height is lower than Europeans (5.5cm for Japanese and 10 cm for Europeans, Subramanian et al. 2006). Also Jelte Wicherts et al. (2009) seems to imply that the SD in African IQ is 12, which is less than Europeans, although this may be due to that fact that IQ tests were developed to measure Europeans and the measured IQ distribution of Africans tend to be narrower than the reality.


Aside from these technicalities, I strongly believe that the novelty and predictive power of the hypotheses presented in the paper, which are consistent with the human evolutionary theory such as the r/K theory, is worth enough to be published.

Lastly, let me connect Nyborg’s thesis to the "brain vs. brawn" hypothesis.

As IQ has been selected by women, male height and physical strength also should have been selected by women. This is almost apparent if we look at the omnipresent popularity of well-known professional male athletes as a mating partner.

However, the "brawn vs. brain" hypothesis seems to predict that women in higher IQ populations should rather prefer higher IQ men than women in lower IQ populations, and vice versa.

This prediction is in accord with Nyborg's (1994) thesis that there is a gradient of concentration of sex hormones such as Testosterone and Estradiol from Africans to Europeans to East Asians. Testosterone creates bigger muscle tissue and higher bone density and also Estradiol induces more pronounced female breast and buttocks, which make such individuals more sexually attractive for the opposite sex. Apparently, Asians are less sexually dimorphic in their appearance compared with Africans. So as the "brain vs. brawn" hypothesis postulates, the there may well be a trade-off between the development of stronger physique with more muscles and denser bones, and the neural development leading to higher IQ. This observed phenotypic difference is surely ultimately based on the difference of allele frequencies, but more proximately through the difference in the concentrations in sex hormones, at least partially.

So if we think about the mating strategy of humans, it is natural that in tropical climates where a mother can raise kids by herself, sexual competition is stronger for male to show physical strength, whereas in colder climates where fathers should help mothers raise kids, the competition should shift to more continuous and stable procurement of food by higher IQ. All of this makes a good sense over all, given the observed differences in marital stability between tropical populations to sub-Arctic ones.


Miscelanious
p.6 Table 2 periods instead of commas in Mean and SD.
p.10 References 1980 should be in parenthesis



References

Nyborg, Helmuth (1994) Hormons, sex, and society: The science of physicology.

Subramanian, S. V, Ozaltin, E., & Finlay, J. E. (2011). Height of nations: a socioeconomic analysis of cohort differences and patterns among women in 54 low-to middle-income countries. PloS ONE, 6. doi:10.1371/journal.pone.0018962.

Wicherts, J. M., Dolan, C. V., & van der Maas, H. L. J. (2009). A systematic literature review of the average IQ of sub-Saharan Africans, Intelligence 38, 1-20. doi:10.1016/j.intell.2009.05.002
Admin
Do you have a copt of Nyborg's book? That one is really hard to find. Not even Nyborg has a PDF.
Duxide, have you tried to look at other IQ/achievement tests and see if the SD_score has the same relationship as the one you get with PISA CPS ? As you indicate, this test samples many 15-year-olds, and so maybe sex effect is under-estimated, but that should be (if possible) tested in older samples, and preferably, other tests as well. If your result is true only for PISA CPS, it's not robust. Concerning the modest correlation with math and none with reading, it may be reading that need to be replicated. Math is somewhat crystallized as well. But it, too, correlates with reading. I don't see why reading shouldn't correlate with your variable of interest, for example. [edit: for the last sentence, I haven't noticed one of your previous comment which states that female outsmart males on reading, however, I think in total score the correlation should be still different than zero, but granted that it should also be lower than fluid g if total score includes reading items.]

Chuck, thanks for trying it. This line of coding however :

/METHOD=ENTER Region_cat0 to Region_cat3

does not work because one value needs to be considered as the reference category. I have always coded my dummies so that cat0 is the reference. So, if you want SPSS not to drop them all but just the reference category, you should use :

/METHOD=ENTER Region_cat1 to Region_cat3

In model without dummies, you have +.406 and -.342 for Sexdiff and GenderEquality2. For model with dummies, you have +.339 and -.421.
Duxide, have you tried to look at other IQ/achievement tests and see if the SD_score has the same relationship as the one you get with PISA CPS ? As you indicate, this test samples many 15-year-olds, and so maybe sex effect is under-estimated, but that should be (if possible) tested in older samples, and preferably, other tests as well. If your result is true only for PISA CPS, it's not robust. Concerning the modest correlation with math and none with reading, it may be reading that need to be replicated. Math is somewhat crystallized as well. But it, too, correlates with reading. I don't see why reading shouldn't correlate with your variable of interest, for example. [edit: for the last sentence, I haven't noticed one of your previous comment which states that female outsmart males on reading, however, I think in total score the correlation should be still different than zero, but granted that it should also be lower than fluid g if total score includes reading items.]


Chuck explained this: "Math. reading, and science differences are confounded with broad factor (verbal/quantitative) sex differences. You could extract g-scores using math, reading, science, and problem solving, but you are right that this would be difficult, time consuming and generally not worth the effort.".

Concerning data, PISA CPS (and PISA in general) is better than any IQ data set that has been used so far, because it is the most representative, much more so than for example WAIS normative data that are routinely used in studies accepted on peer reviewed journals, even though the samples used for these norms in different countries are arguably not as representative and as amenable to cross-country comparison as PISA.

I agree that the data are not robust but this is not really a problem. The focus of my paper is mainly theoretical, and I could have written it out without even bothering to test my hypotheses with empirical data. Many prestigious journals allow authors to submit papers that contain hypotheses without backing them up with data, because it's not always possible to have data right away. In the history of science it's common that theory has precedence over empirical data. In physics for example, theories are first examined from a logical perspective, to ascertain their internal consistency and mathematical correctedness. Only later are empirical tests expected. This is necessary for innovative science. Neither Mendel nor Darwin had really robust data to back up their hypotheses and they published their papers before others could replicate with separate datasets. The same can be said about a lot of other theories (e.g. Wegener's theory of continental drift). Science always gives priority to a sound theory with poor data over robust data with a poor theory. You may be right that my paper has poor data, but if the theory is sound then I do not see why it should not be published so that other researchers will be allowed to validate my ideas with better datasets.

I will change my title to highlight that my paper presents a hypothesis that has not been validated by robust data.New title: "Sexual selection as a mechanism behind sex and population differences in fluid intelligence: an evolutionary hypothesis"
Admin
That seems better yes. Surely this paper is theoretically interesting. It may not be true of g, but perhaps it is true of the broad ability factors if they are extracted from a suitable dataset with Schmid-Leiman method.

The R package psych has Schmid-Leiman built in. The function is schmid().
Chuck:

You called a variable Gender Equality but it's actually Gender Inequality. This is misleading. Lower values indicate lower gender inequality/higher gender equality and higher values indicate higher gender inequality/lower gender equality. This I renamed it Gender Equality, which is the correct name: http://en.wikipedia.org/wiki/Gender_Inequality_Index

Even then, this variable behaves strangely. It's only weakly correlated to GDP (0.106), and it has a NEGATIVE correlation with Sexdifference in CPS (-.398)!! So, countries with higher gender inequality have less male advantage, contrary to expectations.

So I cannot include this variable in my regression because of its paradoxical effect: a variable that predicts gender inequality which reduces (instead of increasing) gender inequality in intelligence, contrary to the assumption (its reducing gender disparity in cognitive stimulation) that is required to justify its inclusion in the regression.

EDIT 1: It turns out that Chuck's gender inequality data was wrong. I had do redownload it from the UN website: http://data.un.org/DocumentData.aspx?id=332
It certainly was an involuntary mistake but if I had not checked his data's accuracy the acceptance of my paper could have been jeopardized.

The results now are much more straightforward: gender inequality is positively correlated to sex differences (male advantage) and negatively to CPS score. The partial correlation CPSxSex diff after partialling out Gender Inequality is much stronger (r=around 0.5).
Find attached the screenshots (my SPSS is not in English so forgive me).
Also attached the dataset.
I attach the updated manuscript and SPSS file with all the variables.
Even then, this variable behaves strangely. It's only weakly correlated to GDP (0.106), and it has a NEGATIVE correlation with Sexdifference in CPS (-.398)!! So, countries with higher gender inequality have less male advantage, contrary to expectations....EDIT 1: It turns out that Chuck's gender inequality data was wrong. I had do redownload it from the UN website: http://data.un.org/DocumentData.aspx?id=332 It certainly was an involuntary mistake but if I had not checked his data's accuracy the acceptance of my paper could have been jeopardized.The results now are much more straightforward: gender inequality is positively correlated to sex differences (male advantage) and negatively to CPS score. The partial correlation CPSxSex diff after partialling out Gender Inequality is much stronger (r=around 0.5). Find attached the screenshots (my SPSS is not in English so forgive me). Also attached the dataset.


I don't know what you're talking about. My data came from column 1 in table 4 here: http://hdr.undp.org/en/content/table-4-gender-inequality-index (Gender Inequality Index Value, 2013). If the variable in my SPSS file was misleading, it was because you didn't look at the original cite. I noted that when you partial out gender inequality, the correlation is more robust ("granted that when gender equality is partialed out, the national g x sex difference correlation is somewhat more robust"). I found an r ~ 0.5 myself. The correlation between GDP and gender (in)equality was ~ 0.50. I have no idea what you did. What numbers did you use? My SPSS file was here. http://www.openpsych.net/forum/showthread.php?tid=69&pid=983#pid983

It certainly was an involuntary mistake but if I had not checked his data's accuracy the acceptance of my paper could have been jeopardized.


Not checking would have been silliness on your part, since I was just pointing out a data set.
[hr]

Chuck, thanks for trying it. This line of coding however :

/METHOD=ENTER Region_cat0 to Region_cat3

does not work because one value needs to be considered as the reference category. I have always coded my dummies so that cat0 is the reference.


I didn't even look at the results; you asked for the data set, so I threw together and posted the SPSS file.
Do you have a copt of Nyborg's book? That one is really hard to find. Not even Nyborg has a PDF.


Yes I do. It was very expensive as I remember. It may be a good idea to PDF the book so that you guys can read it. Tell me if you want.
Admin
Generally, studies on fertility x education/g find that dysgenics is stronger in women than men. So, if there was sex-linked heritability, then one would expect to see an increasing sex difference. FUrthermore, populations which have had dysgenics for a longer time should have larger sex differences in these traits. There are historical estimates for this, perhaps simply using GDP as a proxy for early onset of dysgenics. This leads to the prediction that countries with higher GDP have larger sex diffs in whichever traits been under negative selection that differed by sex.
Generally, studies on fertility x education/g find that dysgenics is stronger in women than men. So, if there was sex-linked heritability, then one would expect to see an increasing sex difference. FUrthermore, populations which have had dysgenics for a longer time should have larger sex differences in these traits. There are historical estimates for this, perhaps simply using GDP as a proxy for early onset of dysgenics. This leads to the prediction that countries with higher GDP have larger sex diffs in whichever traits been under negative selection that differed by sex.


That's an interesting idea. However the correlation between GDP and sex differences in CPS is negative.
I noted that when you partial out gender inequality, the correlation is more robust ("granted that when gender equality is partialed out, the national g x sex difference correlation is somewhat more robust"). I found an r ~ 0.5 myself. The correlation between GDP and gender (in)equality was ~ 0.50. I have no idea what you did. What numbers did you use? My SPSS file was here. http://www.openpsych.net/forum/showthread.php?tid=69&pid=983#pid983


I do not know what file I was using, this is really strange.


It certainly was an involuntary mistake but if I had not checked his data's accuracy the acceptance of my paper could have been jeopardized.


Not checking would have been silliness on your part, since I was just pointing out a data set.

It's always better to use the proper names and not a different one, so that the users do not have to reverse score the variables (or they could even assume that you've done so) to make the results interpretable. Gender Inequality is Gender Inequality, not Gender Equality unless it's reverse scored.
Nonetheless I am happy you pointed out the Gender Ineq. index, because now my results are much stronger, even stronger than after controlling for GDP!
"Gross Domestic Product (World Bank, 2014) at purchasing power per capita (GDP (PPP)) was used as an independent variable due to its potential relationship with g and sex differences in country scores. That is, GDP is known to be positively related to country IQ (Lynn and Vanhanen, 2006, 2012; Rindermann, 2012; Sailer, 2012) and could predict sex difference in g, possibly with more economically developed countries showing lower sex difference (lower male advantage) due to females enjoying more enriched environments, since rich countries probably undertake efforts to raise female schooling and intelligence in an attempt to make men and women more equal."


Note the use of the gender inequality variable in the section above; provide a reference for the source. Also, correct the spacing problems. After, I approve. (You might also included the gender inequality variable in your table 1.)
I've noted the use of gender inequality in the section above and provided a reference for the source in the methods section. I included the gender inequality variable in table 1.
I've noted the use of gender inequality in the section above and provided a reference for the source in the methods section. I included the gender inequality variable in table 1.


You still have spacing problems e.g., between paragraphs. I'm certain that Emil will want you to fix those.
I've noted the use of gender inequality in the section above and provided a reference for the source in the methods section. I included the gender inequality variable in table 1.


You still have spacing problems e.g., between paragraphs. I'm certain that Emil will want you to fix those.


I've now eliminated the spacing problems between paragraphs. If there are more spacing problems, I guess Emil will find them (I am not really good at catching these).
Admin
There are lots of spaceing problems. Just search for a double space with search in word. First one is on line 52.
There are lots of spaceing problems. Just search for a double space with search in word. First one is on line 52.


Found quite a few and replaced them. Looks like I'll have to buy a new keyboard...
Chuck, I've corrected the spacing problems. The updated file is in the previous post. Can you approve now? Thanks again for your review, especially for pointing out the Gender Inequality Index.
Chuck, I've corrected the spacing problems. The updated file is in the previous post. Can you approve now?


I approve.