In the first part of this post, I discussed a recent paper (Dutton, van der Linden, & Lynn, 2016) that attempted to empirically test the predictions of differential-K theory. To recap briefly, this theory proposes that racial groups differ in their preferred reproductive strategies and that as a result they differ in a wide range of physical and psychological characteristics including intelligence, personality, sexual behavior and attitudes, and even penis length. These differences are thought to derive from group differences in androgens (male hormones such as testosterone). According to the theory, African populations should have the highest androgen levels, followed by Caucasians, then by Asians. The study tested five presumed markers of androgens: CAG repeats on the AR gene; amount of androgenic body hair, specifically, mid-phalangeal hair (i.e. on the middle digit of the fingers); national prostate cancer incidence; and two measures of sexual behavior, specifically, number of partners, and annual frequency of sex. The results found were not entirely in accord with the theory’s predictions, as Africans were not significantly higher than Caucasians in CAG repeats on the AR gene, and were lower than Caucasians in androgenic body hair and prostate cancer incidence. The authors were not able to compare Africans with the other groups on sexual behavior, as they had no African data. They did find that Caucasians were higher than Asians in all androgen markers, which they took in support of their theory. In this post, I will discuss how the authors of the study interpreted their anomalous results, then respond to Dutton’s claims that his results support the validity of Lynn’s penis data and explain why his results actually contradict this.
Dutton et al. acknowledged that the results for androgenic hair and prostate cancer do not support their hypothesis, and offered some tentative explanations. In relation to prostate cancer, they suggest that differences in general diet might be a factor, as people in Western countries consume more dairy products and these have been linked to prostate cancer. Additionally, people in Western countries tend to be more obese, another risk factor for cancer. They point out that African American men have a 30% higher rate of prostate cancer than Caucasians even after taking into account obesity, while Asian Americans have lower rates than these two groups. They also cite a 1986 study that found that African American men have 15% levels of freely circulating testosterone than Caucasian Americans and that this might contribute to their higher prostate cancer rates. They acknowledge that differential access to medical care in the US might affect these results and that African American men might not be fully representative of current Sub-Saharan Africans.
Prostate cancer incidence and its relationship with testosterone is a complex subject that I am not able to review fully. However, I would like to briefly point out some research evidence that I am aware of that may be of interest. Environmental factors including conflict with the law can influence testosterone levels. Studies that take into account conflict with the law have found that African American and Caucasian American men have similar testosterone levels (Zitzmann & Nieschlag, 2001). This same paper stated that higher levels of prostate cancer in African Americans do not appear to be due to differences in testosterone levels. Additionally, although Dutton et al. report a substantial non-significant correlation between AR CAG length and prostate cancer incidence across nations, prior research has found little evidence for an association between CAG length and risk of prostate cancer at an individual level (Lange et al., 2008). Although prostate cancer may well be a marker of androgen levels, the causes of variations between ethnic groups in incidence rates are complicated by local environmental factors, which makes it difficult to interpret these ethnic differences. This raises doubts about the usefulness of treating national prostate cancer incidence as a marker of reproductive strategies.
Moving on to androgenic hair, Dutton et al. offer a highly speculative explanation of why Caucasian men have higher rates of body hair than other populations. This was actually my favorite part of the paper as I found it wryly amusing. They state:
The anomaly that Caucasians have the highest levels of androgenic hair and Africans the lowest can only be speculated upon. It has been found that Caucasians, in contrast to the other two populations, retain a small percentage (2–4%) of Neanderthal genes. It has been argued that this may be one of the reasons why Caucasians are unexpectedly hairy (e.g. Sankararaman, et al., 2014).
Note the reference cited in the last statement. There is a slight problem with this explanation, being that the facts stated are completely wrong. Here is what Sankararaman, et al. (2014) actually said about Neanderthal DNA:
… the proportion of the genome with confidently inferred Neanderthal ancestry has a mean of 1.38% in east-Asian and 1.15% in European populations consistent with previous reports of more Neanderthal ancestry in east-Asian than in European populations.
Here is their only statement concerning hairiness:
We do not detect tissue-specific expression patterns; however genes involved in keratin filament formation and some other biological pathways are significantly enriched in Neanderthal ancestry in European populations, east-Asian populations, or both. Thus, Neanderthal alleles that affect skin and hair may have helped modern humans to adapt to non-African environments.
Hence, this reference does not actually state that Caucasians are hairier than Asians because of Neanderthal ancestry. It is therefore not clear at all how the data for androgenic hair are supposed to fit in with differential-K theory.
Dutton et al. argued that the pattern of correlations between the androgen indicators provides evidence of their validity. In his conference paper, Dutton also states that these androgen indicators are correlated with the penis length data used by Lynn (2013) in his paper. He claims that this supports the trustworthiness of Lynn’s data. Dutton’s argument might have some merit if certain assumptions were met. In principle, two variables that are closely related to each other should be correlated and this is known as convergent validity. Hence, a strong correlation between two variables is often taken as evidence of convergent validity. However, correlational analyses are based on the assumption that two variables have a linear relationship (i.e. the data points should form a roughly straight line pattern). When variables have a non-linear relationship, use of correlations may give a misleading account of how they are related. As I will show, Dutton’s data has some problems with non-linearity that affects its convergent validity.
Furthermore, Dutton groups national data into three racial categories which are supposed to form a distinct hierarchy of African > Caucasian > Asian. Hence, in order to argue that the data on androgen markers provides convergent evidence of the trustworthiness of Lynn’s data, both sets of data should follow the same hierarchical pattern. However, they clearly do not.
Dutton et al. report an impressively large correlation (r = .82) between androgenic hair and prostate cancer. However, an inspection of the scatterplot provided (reproduced below) shows that the relationship between the two variables appears decidedly non-linear.
Among Caucasian countries there is not much variance in the percentage of androgenic hair yet there is much more variance in prostate cancer incidence. This means that among these countries there is basically no correlation between the two variables. Among the African and Asian nations, prostate cancer incidence has a more restricted range, but there is somewhat more variance in the percentage of androgenic hair. So again, there is basically no correlation between the two variables among these countries either. Essentially, Caucasian countries have higher rates of both androgenic hair and prostate cancer compared to the other countries, but this does not mean that there is a linear correlation between these two variables, even though they are both supposed to be markers of androgens. This suggests that whatever underlies differences in national rates of prostate cancer is not related in a simple linear way to whatever underlies national differences in androgenic hair.
According to Lynn’s paper, data he derived from the world penis site indicated that men from African nations had the longest penises, followed by Caucasians, followed by Asians. Lynn argued that this was in accord with the predictions of differential-K theory. In my rebuttal, I argued that because the data on this site were compiled by an anonymous source and it is not clear how this information was derived or how valid it is, I could not see why a serious scholar would rely on such a source of information or draw any conclusions from it. In his paper, Lynn himself admits that the site is not a peer-reviewed source, so I can only wonder why he used it in the first place rather than drawing on primary research papers published in reputable journals. In his conference presentation, Dutton states that I pointed out "minor mistakes" on the website in order to "ridicule" Lynn. Perhaps, he and I have different opinions about what makes a mistake "minor". Without going into all my original criticisms here, among these "minor" mistakes I noted that some of the research papers the website cites as sources do not even exist. Hence, I find it natural to wonder what else on this site has been simply made up. Additionally, the penis sizes stated for particular countries do not always match the values provided in the actual papers cited as data sources. For many countries it is not clear what sources there even are.
Personally, I think these issues are cause for serious concern, but perhaps I am overstating their importance? Dutton stated that the data from his study on androgen markers correlates with the penis length data, indicating that the latter can be trusted. The argument here seems to be that penis length should be correlated with other androgen markers, and that if the penis length data can be statistically predicted from the androgen data then the former is probably valid, or at least in the right general direction. Dutton did find moderate statistical correlations between androgen markers and Lynn’s penis data, therefore, he argues, Lynn’s data passes muster.
However, there is a problem with this argument. As explained earlier, Lynn’s (2013) results and two of Dutton et al.’s (2016) results (regarding prostate cancer incidence and androgenic hair) are incompatible.
For penis length, Lynn found: African > Caucasian > Asian.
For androgenic hair, Dutton et al. found: Caucasian > Asian > African.
For prostate cancer incidence, Dutton et al. found: Caucasian > Asian = African.
I do not see how the pattern for the latter two results can be used to validate the first one. More specifically, if prostate cancer and androgenic hair accurately predicted penis length, then we would expect both African men and Asian men to have smaller penises than Caucasian men, contrary to what Lynn found. (Please note, I am not asserting anything at all about actual differences in penis length between races, because I do not have sufficient data. I am commenting on the methodology used to support such claims.) To put this another way, although Lynn’s penis data is correlated with the other two variables, the actual relationship between the former and the latter is non-linear, so the correlations do not allow valid predictions.
It might be helpful to illustrate this with a chart. Dutton does not provide relevant scatterplots, so I created one in Excel using prostate cancer data from Haas, Delongchamps, Brawley, Wang, & de la Roza (2008) and 2011 penis length data from the website Lynn used.
There is a linear correlation of r = .34 between prostate cancer incidence and penis length. This is not statistically significant due to the low sample size (only 21 countries), but Dutton has argued that correlations of this value are ‘substantial’ (slide 7). Does this mean that prostate cancer incidence can predict average national penis length? If there was a clear linear relationship between the two it might, albeit crudely. But look at the scatterplot I have provided. The relationship between the two variables is clearly non-linear. This is because the countries with the highest incidence of prostate cancer are Caucasian, yet these countries are mostly intermediate in penis length. A similar result would appear if one were to graph the relationship between androgenic hair and penis length, as Caucasians also had the highest rates of the former. At the risk of belabouring the obvious, it simply does not make any sense to say that because Lynn’s penis length data is correlated with these other variables that this means that the former can be trusted. Dutton’s and Lynn’s measures do not follow the same patterns so correlations between them can be explained as statistical artifacts of using linear tests on non-linear data. If scholars really want to know if there are differences between these groups in penis length, the sensible thing to do would be to find better data sources rather than trying to validate an anonymous database using crude predictions based on loosely related variables.
Differential-K is a grand theory that aims to explain a very wide range of human population differences. However, it appears to be out of step with the available data in some respects. One prediction of this theory is that racial groups should differ from each other in consistent ways in a number of factors that are supposed to be indicative of androgen levels. Yet, an empirical test of this prediction was not able to confirm the expected racial hierarchy. The relationship between things that are supposed to be indicators of androgen levels, such as androgenic hair and prostate cancer might be too complex to be represented in a simple linear way. Furthermore, how these things are supposed to be related to reproductive strategies is not entirely clear. Dutton et al. found differences in sexual behavior between Caucasian and Asian populations, but provided no data on Africans. The data source they used, an internet survey by Durex, is not a peer-reviewed source and may be of questionable validity. There are more scientific sources of information that could shed light on cross-national differences in sexual behavior, and I will make some relevant observations about these in a follow-up post. Richard Lynn (2013) argued that racial differences in penis length provide evidence of corresponding racial differences in sexual restraint, although critics consider this view to be naïve. Although Edward Dutton argued that his research results provide support for the validity of Lynn’s earlier claims about racial differences in penis size, a closer examination of his own data contradicts this. Differential-K theory has not fared well and may be an overly simplistic theory that attempts to explain too much with too little.
 Since writing my original article, the penis website has added measurement data from various countries to the sidebar. Most of these cite genuine research, but there is one on ‘Ecuadorian measurement data’, that cites a paper called ‘Study of penile dimensions in healthy Ecuadorian men of multiple ethnicities’ supposedly published in Andrologia. Included is a professional looking Abstract along with a pair of impressive looking graphs. However, Andrologia has never published anything by this name, and the paper does not appear to exist. The whole citation looks like an elaborate hoax. What does this say about the scholarly value of this website?
 See the blog Ethnic Muse for more detailed information.
Flight of the Witches by Francisco Goya. The Wikipedia article has an interesting explanation of the symbolism of this painting.
Personality, Intelligence and "Race Realism" - critiques J.P. Rushton's theory
The Pseudoscience of Race Differences in Penis Size - critiques Lynn's paper on the subject
Dutton, E., van der Linden, D., & Lynn, R. (2016). Population differences in androgen levels: A test of the Differential K theory. Personality and Individual Differences, 90, 289-295. doi: http://dx.doi.org/10.1016/j.paid.2015.11.030
Haas, G. P., Delongchamps, N., Brawley, O. W., Wang, C. Y., & de la Roza, G. (2008). The Worldwide Epidemiology of Prostate Cancer: Perspectives from Autopsy Studies. The Canadian journal of urology, 15(1), 3866-3871.
Lange, E. M., Sarma, A. V., Ray, A., Wang, Y., Ho, L. A., Anderson, S. A., . . . Cooney, K. A. (2008). The androgen receptor CAG and GGN repeat polymorphisms and prostate cancer susceptibility in African-American men: results from the Flint Men's Health Study. J Hum Genet, 53(3), 220-226.
Lynn, R. (2013). Rushton’s r–K life history theory of race differences in penis length and circumference examined in 113 populations. Personality and Individual Differences, 55(3), 261-266. doi: http://dx.doi.org/10.1016/j.paid.2012.02.016
Sankararaman, S., Mallick, S., Dannemann, M., Prufer, K., Kelso, J., Paabo, S., . . . Reich, D. (2014). The genomic landscape of Neanderthal ancestry in present-day humans. [Letter]. Nature, 507(7492), 354-357. doi: 10.1038/nature12961 http://www.nature.com/nature/journal/v507/n7492/abs/nature12961.html#sup...
Zitzmann, M., & Nieschlag, E. (2001). Testosterone levels in healthy men and the relation to behavioural and physical characteristics: facts and constructs. European Journal of Endocrinology, 144(3), 183-197. doi: 10.1530/eje.0.1440183
Be sure to read the following responses to this post by our bloggers: