Although we all learn about the scientific process in courses and labs designed to introduce us to the intricacies of research, there are some elements that can slip by even the most detailed-oriented scientist without receiving proper consideration. The measurements that are used in any given analysis, especially in predominately observational sciences like anthropology, are often as much a matter of tradition and training as critical evaluation. The traits we choose to study need to be not only measured correctly, but also conceptualized correctly, and this includes understanding the importance of considering covariance that may exist between traits. Covariance refers to relationships between changing traits (if one gets larger, does the other get larger/smaller as a direct result?) that are caused by a combination of genetic, developmental, and environmental effects. Importantly, covariance between traits means that one trait is unable to change without affecting the other traits with which it covaries.
A good example of how covariance between traits may affect human evolution can be found in a recent publication on ecogeographic variation in human body form written by Charles Roseman (University of Illinois), Benjamin Auerbach, and myself (Savell et al. 2016). As has been noted by anthropologists and biologists for nearly a century, limb lengths and body shape (wide or thin?) changes in predictable patterns across climates and latitudes (hence “ecogeographic variation”). Traditionally, variation in these traits has been interpreted as evidence for climactic adaptation. This is based on patterns observed by Bergmann (1847) and Allen (1877), in which populations of the same or closely-related species living in colder environments demonstrated wider bodies and shorter appendages (limbs, tails, ears) than those living closer to the equator. Since these observations were made, the assumption has been that in colder environments, wider bodies and shorter extremities decrease the surface area exposed to the air which minimizes heat loss, and that the reverse is true for equatorial populations. These patterns have been called Bergmann’s and Allen’s ecogeographic “rules,” and have been generally supported in modern humans (Trinkaus 1981, Ruff 1994, Holliday 1997, Holliday & Ruff 2001; though see Auerbach’s 2012 paper on variation in the Americas for some interesting exceptions).
One major shortcoming of this classic understanding of human body evolution, however, is that each of the traits studied (usually humerus, radius, femur and tibia length, as well as various skeletal elements used to estimate body size) has been treated as if it is evolving independently. We more or less know this is not the case. Given their similar evolutionary history and development, we have every reason to believe that the long bone lengths at least will demonstrate some covariance with each other. In the paper I coauthored, we therefore examined the covariation that influenced limb evolution across latitudes.
Using groups that represented a range of limb segment lengths for each of four regions (arctic, temperate, North Africa, sub-Saharan Africa), we used a variant of Lande’s equation for multivariate response to selection (Lande 1979, Lande & Arnold 1983): Δz=Pβ, where Δz represents the change in mean trait values, P is the phenotypic variation-covariation matrix, and β is a vector of selection gradients. We used a phenotypic matrix (P), because it captures the covariation of a given trait with every other trait included in the analysis and is proportional to the genetic (G) matrix. After standardizing, we estimated the directional selection (if any) acting on each trait in a hypothetical “evolution” from sub-Saharan latitudes to other regions.
Just as predicted by Allen’s rule, we found that distal limb segment lengths (radius and tibia) were estimated to be under selective pressure to shorten at temperate and arctic latitudes. Interestingly, femoral length appeared to be evolving in response to neutral processes (that is, there is no directional selective pressure) – I’ll have more to write about this in some exciting upcoming studies! However, directional selection on humeral length was positive at higher latitudes – directly contradicting our expectations given Allen’s rule. We investigated this further and found that the pattern of covariation between the humerus and distal limb bones was so strong that the humeri not only failed to lengthen (despite being under positive selective pressure) but are essentially forced to shorten as a result of their relationship with the radius and tibia.
Though our study has a lot to say about evolution and covariation between limb segments, one major point here is that examining the evolution of traits in isolation, without considering their relationships, may result in misleading or false findings. Biological traits share evolutionary and developmental histories, and cannot be expected to evolve independently. As good biologists, anthropologists, and anatomists, we need to see each organism as a complex whole rather than a series of independent, arbitrarily determined variables, and to take that whole organism into consideration during research design. For more on this topic, I absolutely recommend Houle et al.’s 2011 piece on measurement and meaning in biology.