I took a break from writing this afternoon to read this new paper from Jonathan Donges and colleagues examining the role that patterns of climatic variability play in human evolution. The paper is the most recent in an increasingly interesting debate about the dynamics of climate change during the Pliocene/Pleistocene and its potential relationship in driving human evolutionary patterns. But what I found myself thinking about while reading the paper was prompted by this figure:
The paper itself is largely a methodological paper, arguing for the advantages of nonlinear analyses of complex time-series climate data, so the figure above contains a number of details specific to their analysis (much of which is relegated to the supplemental information). The conclusions of the authors regarding human evolution are introduced by this line from the abstract (my emphasis added):
A reexamination of the available fossil record demonstrates statistically significant coincidences between the detected transition periods and major steps in hominin evolution.
Separate from their analysis of the climate data, the authors conclusion regarding human evolution is based on the correlation between their climate analyses and a specific human evolutionary model. Given a different model (part E of the figure above), this correlation wouldn’t necessarily exist. And yet this kind of approach, combining information from different (hopefully somewhat independent) lines of evidence to address single hypotheses, is what paleoanthropology is primed for. It is certainly something I have been thinking about a lot lately. But it is a real challenge when we have important parameters of uncertainty in all of our data. Fossil and archaeological data are time-constrained only to the degree dating methods and high-quality excavations allow. Even with good excavation dates, uncertainty will remain about first-appearance and last-appearance of key technological or biological features. Meanwhile we have a huge trove of genetic data that directly addresses human evolutionary history, but it too is subject to uncertainty. The confidence intervals around genetic coalescent dates reflect assumptions about demographic models and basic mutation rates. Finally, we have climate data like those presented in this paper, but with uncertainty around the temporal and geographic scale of the data.
I am a very visual person. When I meet with students in office hours, and even when I am working alone on my own research, I regularly stand in front of the white-board in my office and draw or diagram my thinking. It would be great to be able to do that with the above sets of data – place each set on top of the other and see the different datasets coalesce into a single, clear picture. Sadly, we don’t have the 3D glasses necessary to make this happen. And it is often tempting to fall into the trap of seeing what we want to see, prompting John Hawks to start his “broadly consistent watch” (way back in 2005!). It does pose an interesting probabilistic problem, though…
******
1. Jonathan F. Donges, Reik V. Donner, Martin H. Trauth, Norbert Marwan, Hans-Joachim Schellnhuber, and Jürgen Kurths. Nonlinear detection of paleoclimate-variability transitions possibly related to human evolution PNAS 2011 ; published ahead of print December 5, 2011, doi:10.1073/pnas.1117052108
I am *so* happy you brought up the “Broadly Consistent Watch”. Man, I need to do more of those!
I don’t think that first/last appearances are very good markers of biological change, especially since the record gets better and better toward the present (and of course nobody agrees on species concepts). I wonder what would be better?
Glad you enjoyed the shout-out. I was kind of shocked that your first one of those was all the way back in 2005. That seems quite a long time ago.
I also think “firsts” and “lasts” are poor markers, but they are nevertheless commonly used. The paper linked above being a perfect example of this. Even our molecular clocks are calibrated to these kind of data (though less than they used to be). The alternative, I suppose, is to dig more, and do more advance prep to be able to quantify what could be dug, what has been dug and what has been lost to begin to differentiate absence of evidence from evidence of absence. I think we are getting better at this as a discipline (there is certainly no lack of enthusiasm for field work). My question is why don’t those molecular folks reach some consensus on mutation rates?
Hey Adam – good stuff. I assign my students a paper every semester asking “what is the rate of genetic mutation?” My students echo your lament on the lack of consensus on mutation rates Not surprising with non-coding sequences turning out to be less junky than we imagined and without a better understanding of selection for other putatively “selection-neutral” markers (eg., mitochondrial, etc). All of this complicated by extreme variation in substitution rates between sequence sites, plus the problem of parallel mutations confounding genetic distance estimation. This is probably why most estimated mutation rates are higher than those based on phylogenetic comparisons. Anyhoo, the famous case of the Romanov family (mtDNA) alone instantiates the difficulty with a standardized mutation rate. Lastly, since the genus Homo spans across ecological and climatic niches, it’s hard to trust rate estimates derived from more environmentally constrained nonhuman data which are not subject to same possible range of selective pressures. Just a few thoughts from a your friendly neighborhood vertebrae guy.
What we really see is not the mutation rate but the rate at which mutations become wide spread: evolution in action.
The speed of evolution is much faster, not in well adapted populations, but those under stress. Being at least mildly isolated in a location in which the population can survive but in some regards is poorly adapted to the local environment would seem to be ideal for bringing about rapid evolution.