Evidence from Google analytics seems to suggest that people like examples of studies involving contextual analysis (or maybe its sex, hard to tell). With that in mind I will spend the next two weeks on two different studies. Next week I will discuss an example involving community ecology, however, this week I want to talk about Jake Moorad’s study of sexual selection in humans (Moorad 2013 Evolution 67: 1635).
Humans differ from other organisms in at least three respects. First, by in large manipulative studies are not possible, second, humans keep records about themselves, and third sometimes they get offended when their privacy is compromised. The first of these is a problem if for no other reason than it is difficult to remove confounding factors. The second can be a huge advantage. The Moorad study uses the genealogical records kept by the Mormons. The data set is very complete, and the sample size is 741,851 reproductive adults born between 1840 and 1970. The third is a problem in that the Mormon Church is very careful with these data, and is very cautious about whom they allow to have access to their records. We are very lucky that Jake was able to get approval to work with these data.
These data are interesting because 1840 is the year that the Mormons founded a major city (Nauvoo) in Missouri. At this time the Mormons were in the middle of a long migration that would eventually land them in Utah. Survivorship of children was relatively low and the Mormons actively practiced polygyny. By 1850 the Mormons had moved out to Utah, and finally started the process of settling down. By 1890 the Mormons had renounced polygyny, and had established themselves in what would end up being their permanent home. Thus these data span an important demographic transition from a period of high infant mortality and reproduction to a “demographic transition” to much lower reproductive rates and much higher survival.
We also see that rates of polygyny peak around 1860, the time of the demographic transition. It turns out that rates of polyandry also decline. I am told that Mormon women never had more than one husband at a time, thus, this decline probably reflects a decline in mortality of reproductive age men, that is women only remarry if their husbands die at a young age.
One other interesting sort of a graph that makes intuitive sense, but took me a bit by surprise. It turns out that there is much more variance in whether persons father was polygynous than whether an individual male is polygynous. I am pretty sure that this is because a fairly large fraction of males come from polygynous families, whereas only a relatively small fraction of males become polygynous.
One trait that Moorad examined was whether an individual ever mates. An individual can fail to mate either because they do not survive to reproductive age, or they can fail to mate because they never marry or are sterile (the records account for births, thus a barren marriage would count as non-mating).
That there is individual selection is hardly surprising: polygamous males and females had a higher chance of producing offspring than those with a single spouse or no spouse. Selection on the family of origin is more interesting. Here it can be seen that there is selection favoring individuals that come from polygynous families. That is individuals born into multiple female families were more likely to survive and reproduce.
So is this “group selection”? Well this is the problem with the old terminology. It is certainly safe to say it is not individual selection. The family of origin is a contextual trait, it is not a trait of the individual. Thus, we need to call it something. Traditionally that “something” would be group selection, however, it is probably better to go by the more recent term “multilevel selection” and to give that level a more descriptive term, such as family level selection or family of origin level selection.
The problem with this data set is that there is so much data, and it covers a time period in which different birth years cannot be considered comparable. For example, those with birth years prior to 1850 were born in a time when a significant proportion of families were headed by a single male with multiple wives, whereas very few if any of the 1890 birth year was born into such families. Making summary statements over this time period is difficult. Moorad’s solution was to define fitness as the reproductive value at birth, which is roughly the age and population size weighted lifetime reproductive success. He then used this measure of fitness to calculate the opportunity for selection, I, defined as the variance in relative fitness. Finally, he used contextual analysis in a regression of relative fitness on the individual traits (whether or not an individual reproduced, whether or not they were polygynous or polyandrous) and the group traits (whether or not their parents were polygynous or polyandrous). Rather than reporting the individual regression coefficients, he plots the proportion of the variance in relative fitness that explained by each of these factors. The problem with this approach is that there are an unimaginable number of statistical tests that can be done, and the multiple comparison problem gets out of control. Thus, it is best to just look at the plot and interpret it.
What this plot shows is that when looking at overall fitness, family level selection has relatively little impact on the overall variance in relative fitness. Interestingly, even whether or not a male was polygynous had much less effect on the variance in relative fitness than does whether or not an individual ever mated. This is actually a fairly typical result of sexual selection. The number of males that have a large number of offspring is small enough that it doesn’t contribute much to the variance in reproductive success. Far more important is the number of males that don’t produce any offspring. The importance of the polygyny on the proportion mating can actually be seen in the figure above. Note that up until 1860 the intensity of selection on males as to whether or not they ever mate is stronger than it is in females, whereas the two become equal in 1860 which is when polygyny was declining.
In sum this is a great example of the use of contextual analysis in an unmanipulated human population. Studies such as this show that the multilevel selection approach applies in situations that are very different than the traditional “group” of group selection. Even though these are not groups in the sense that Maynard Smith defined groups, they are part of the same mathematical continuum, which indicates that where you draw the line between what is and is not group selection is at best arbitrary.