Racialization through Testosterone

As part of my ongoing dissertation work, I am researching how testosterone has been used to racialize populations in biomedical and biosocial research. Drawing on a discourse analysis of 147 studies that evaluate population differences in testosterone, I find that, despite widespread claims that testosterone varies by race, these myths are unsupported by existing scientific literature. Using social network analysis to visualize these evidence networks, my paper shows how the “racial testosterone theory” proliferated throughout the literature since its formation in early eugenics research. I presented this work over the summer at the annual conference of the Society for the Social Studies of Science (4S). As a supplement to that presentation, I created some network visualizations. Below, I describe these visualizations in their static form (produced via Gephi) while interactive evidence networks are also available for both the network cut by outcomes and the network cut by age/gender (produced using igraph and visNetwork in R) on my Github site alongside the replication code. While the interactive networks do work on smartphones, I recommend using a tablet or computer to engage with data. For those interested, feel free to contact me about a draft of the working paper.

Population differences in testosterone (by study outcome)

In this image, nodes correspond to publications in my sample and ties stand for population-specific claims about testosterone (i.e. when a publication cites another study to claim there are racial differences in T). The size of nodes align with how many times that study is cited by others in the network. Red and orange nodes represent null and mixed outcomes respectively. Green nodes symbolize studies that find clear population differences. White nodes represent publications used as evidence by other studies despite offering no evidence of population comparisons of testosterone. While we see that the majority of studies find either no difference (52.3%) or mixed results (25.8%) when comparing T between groups, the largest nodes (i.e. most cited studies) are green and orange. This means that researchers tend to use papers that find population differences in testosterone despite over 75% of the literature offering null or mixed results.

Population differences in testosterone (by sex/age)

Like in the other network, nodes correspond to publications in my sample, ties stand for population-specific claims about T, and the size of nodes align with how many times that study is cited by others in the network. Here, blue nodes represent studies on men, magenta nodes correspond to studies on women, green nodes stand for studies on children (both boys and girls), and orange nodes have a mix of these three groups. The difference between studies on men (n=79) and women (n=63) is not that dramatic, which is surprising given that T is often said to be a "male sex hormone." However, the biggest finding is that studies conducted on men are much more densely connected than studies on women (i.e. cited more often). This is important because it feeds into how biomedical researchers frame health disparities. T is more often thought to contribute to racial disparities in male-specific health outcomes (like prostate cancer) than female-specific outcomes.