So, I posted some curiosities here on the blog. And, got one of them satisfied! Within hours of wondering about how robust the digit-ratio marker is for exposure to androgen in the womb, Ruben Arslan tweeted a link to this meta-analysis by Martin Voracek. Conclusion: Not very.
The meta-analysis collects studies that has compared repeat polymorphism for two androgen receptor genes with digit ratio. I wasn’t sure this was getting to the point first. I guess I’m a bit wary after all those SNP searches for genes for traits that we know are heritable (intelligence, personality) which have turned up no results. But, I think I understand that this is different, more clearly established, and why finding no relationship is actually very suggestive that the finger-ratio is not a useful maker for the type of research I might be interested in.
Why would it be interesting with a marker such as the 2D:4D digit ratio? From what I understand, it is well established that exposure to androgens during gestation has enduring organizational effects on the mind. Exposure will alter the individual in a way that is permanent (for example, making them men, or not making them men as in those xy individuals who do not have androgen receptors). How well established this is, I don’t know. Again, I have to take researchers word for it (both Voracek, and Marc Breedlove from an earlier review). Mostly, I think, it has been established through animal studies, and studies of unusual humans, such as those xy females. Of course, it would be very interesting to study, in more detail, how androgen exposure may influence human traits in more detail.
But, as Voracek points out, testing this directly by checking for actual prenatal androgen exposure in humans is not usually feasible for many reasons. For one, amniocentesis is not risk free. I doubt anybody would allow this simply for research purposes. And, even if it was, it would be a rather massive undertaking to measure and follow a large enough sample of kids to see how this would work. But, if there could be a fairly reliable, non-invasive, and easily measured marker for androgen exposure, one could use it as proxy in research. Which is exactly what they have done for the 2D:4D marker.
Breedlove (who, overall, is a lot more convinced by evidence that it is useful) is careful to mention that the ratio does not have discriminant value. That is, you can’t look at someone’s fingers and determine from the ratio alone whether they are male or female, gay or straight, autistic or not, nice to their partner or not. (Like I mentioned in my other post, both my daughter and MIL have longer 4D than 2D – to the degree that my daughter once mentioned to me that the ring-finger was longer than the index finger, and I mentioned that this really varied with people in interesting ways). What it can be used for, according to Breedlove, is doing research on groups.
I should understand this, being the kind of researcher that I am. But, as Breedlove also notes, in the daily press and popular accounts, it is always talked about as something discriminant. Look at their hands, what does it say about them. (After all, this is something we would like to have. Some marker we can notice to make judgments about others without having to take the risk of getting to know them closer).
Still, even if it is not a marker you can use to directly say something about an individual, it could potentially be useful for research about how certain individual differences (possibly related to what we think of as masculinity) may arise from prenatal androgen exposure.
If it is a reliable marker, that is.
In 2003, Manning et al published a small study (50 people) that suggested that the length of a particular region that codes for Androgen Receptors is positively correlated with finger ratio (actually, it looks like it was positively correlated with the ratio on the right hand, and also with the difference between the ratios of the two hands). This was taken as evidence that, yes indeed, finger ratio is a marker for androgen exposure in the womb.
I think the chain goes somewhat like this. Cells will not respond to androgens if they don’t have receptors that bind to the androgens and then allow for changes to take place. (XY women lack androgen receptors). The genes coding for androgen receptors are evolutionarily old, and fairly well preserved across species. There are two different coding sites – the CAG and the GGC. Most of the comparison with finger-ratio has been done on the CAG site. Moreover, these are “repeat polymorphic” sites. That is, the snippet that codes for the receptor (or part of the receptor) comes in multiple copies. The measured ranges are 7-37 or 9-41 repeats. Moreover, the repeats are active. They code for the receptor. The effect is linear: the more repeats, the more receptors. And, the idea then is, the more receptors, the more sensitive to prenatal androgen. This can all be established without measuring fingers (and I assume it has). The next step to establish is that, indeed, the finger-ratios are related to the length of the coding-sites.
Length of coding ->prenatal androgen exposure -> 2D:4D finger ratios
And it doesn’t.
The finger measures are the Right hand 2D:4D, the Left hand 2D:4D and then the difference in ratio between the right hand and left hand.
For the 18 CAG studies (2909 individuals overall), the correlation for right hand is .005 [-,032 to .042]. for 16 (2803 individuals) looking at the left hand, r is -.003 [-.041 to .034]. Finally, for the 16 (2796 individuals) looking at the difference r is .013 [-.024 to .051].
It doesn’t look much different for the 5 studies (1497 individuals) that look at the GGC site: Right hand: r = .045 [-.006 to .095]; left hand r = .034 [-.017 to .085] and difference r= .019 [-.032 to .070].
The samples are from all over the world: UK, USA, Spain, Australia, China, Belgium, Slovakia, Tanzania. They look at men and women of different ages, as well as a couple of samples of male-to-female transsexuals.
It really doesn’t look convincing at all.
In the 4.6 section of the paper, Voracek considers where there may be alternative reasons that 2D:4D ratios would be indexing prenatal androgen exposure, even if the genes coding for receptors are not correlated with finger-length. The paths may be rather complex with unknown feedback-loops. (Could very well happen, but it is not clear how). But, as it stands now, I see no clear reason to believe that the ratio is related to prenatal androgen exposure.
Sure, there could still be interesting correlates between finger-ratios and behaviors related to masculinization, but this really weakens the explanatory power of this measure. I don’t think I would want to use it for explaining any traits in humans at this point.
Thank you Ruben, for satisfying my curiosity!
Voracek, Mark, (2014). No effects of androgen receptor gene CAG and GGC repeat polymorphism on digit ratio (2D:4D): a comprehensive meta-analysis and critical evaluation of research. Evolution and Human Behavior, 35, 430-437.
Breedlove, S. Marc, (2010). Minireview: Organizational hypothesis: Instances of the Fingerpost. Endocrinology, 151, 4116-4122.
February 18 edit: An artist friend made a comment that made me realize i need to explain 2D:4D a bit better. You can measure your index finger and your ring finger, and check which is the longest. An easy summary is to take the ratio of the lengths, so you divide length of index finger with the length of the ring finger. There are lots of sites about this, but I think this one was illustrative enough.
I checked my boys hands, they look like mine with longer index fingers. My daughter has a longer ring finger. I know it doesn’t have discriminant validity, but I still thought it was illustrative about how non-discriminant it is. They all seem like fairly normal boys and girls, without being extreme.
Richard Harper told me he has read something about it being a short window where there is a tug-of-war thing between estrogen and testosterone – I think that was brought up in the Voracek article towards the end, suggesting paths to go. But, now I’m thinking that the relative length of the digits have multiple causes, and thus are hopelessly confounded. The reverse-inference thing again.