The “Science” of history

by Jon

At my undergraduate institution, I took a seminar on Early Modern England. The first day of class, we all introduced ourselves and mentioned our areas of study. When I said that I study the history of science, one particularly aggressive (and not overly bright) classmate attempted to correct me: “You mean the Science of History!” At the time, I didn’t really understand what he meant. After all, history isn’t a science, and no one these days even thinks that it should be.

Well, almost no one. It turns out that there are a group of scholars led by Peter Turchin, a professor of ecology and mathematics at the University of Connecticut, that do think that the discipline of history just isn’t scientific enough. Last week, a bunch of nerd-media outlets (Endgadget, io9, Neatorama, and Metafilter to name a few) picked up on the existence of “Cliodynamics”, as Turchin refers to his new-and-improved version of history. These accounts were all based on an article in Nature which provided a brief overview of the state of the emerging discipline.

The Nature article points out that historians are skeptical of this approach (the journal launched by Turchin in 2010 failed to get a the contribution of a single historian in its first issue), but it fails to engage with any criticisms of Cliodynamics beyond the most superficial (primarily the “lack of data”). The problems with Turchin’s approach, however go much deeper, and stem from the fact that he is trying to remake a discipline he is completely ignorant of.

In a 2008 essay also published in Nature, Turchin sets out his agenda. Bemoaning the “pre-scientific” state of history, he points out that there are hundreds of narratives for “what caused the collapse of the Roman Empire,” a state of affairs he considers “as risible as if, in physics, phlogiston theory and thermodynamics coexisted on equal terms.” This opening salvo against the discipline of history by itself demonstrates the profundity of Turchin’s failure to understand it. He fails to grasp that the questions he asks are themselves shaped by his understanding of history and not based on simple observations of the way things really are.

You can see this in his choice of the “fall of the Roman Empire.” Even setting aside the fact that the eastern half of the empire based in Constantinople kept going for another thousand years after high school textbooks claim that the Empire collapsed, the Roman Empire didn’t fall at all, at least not in the way that most people think when they see those words. It certainly changed over time. Its borders gradually contracted. The later Islamic rulers of much of the Western half eventually drifted away from a Roman identity, and the Roman governmental structures in the outer provinces evolved as they fell out of the centralized bureaucracy. But if you could go back in time and ask someone in 6th century Gaul or North Africa, they would tell you that they were Romans. Turchin’s question of why the Roman Empire collapsed arises from an understanding of what happened that is largely traceable to Edward Gibbon’s Decline and Fall, itself a book written with a political agenda. You can’t ask for a simple causal explanation for an event that only exists in hindsight (if it can be said to have happened at all).

This leads into an even worse failure: his implication that there can only be one correct explanation for a given historical occurrence, and that it is a failure of the historiography that there are many narratives of the same events. History deals with a complex, contingent world, and there are an infinite number of correct narratives and interpretations of any event. That’s not to say that there are no wrong narratives; there are infinitely more bad potential histories than there are good ones, but any attempt to arbitrarily limit the number of causal explanations and interpretations of chaotic, emergent occurrences is not only unhelpful, but stupid.

Also unhelpful and stupid, but in keeping with broader trends in our society, is the insistence that to be valid, an area of inquiry must be “scientific” and must be reducible to mathematics. Mathematics is a toolkit like any other, to be used when helpful and put away when not. Turchin might even be surprised to learn that many historians do employ mathematics in their research. What they don’t do, because it’s fucking stupid, is try to make mathematics the result of their research.

Of course, both the insistance that there can only be one true answer, and the idea that it has to be expressible in math aren’t even true in the sciences. Take the recent article in PNAS demonstrating that feathers were present in the ancestors of the theropod clade, which means it’s highly likely that most theropods (the group that includes Tyrannosaurus Rex) had feathers. The scientists working on this based their argument on the discovery of an extraordinarily well-preserved fossil and their conclusion was along the lines of “well, more dinosaurs had feathers than we’d originally thought.” These feathers can be explained in terms of evolutionary history or ecological advantages or biochemistry. All of those explanations would be equally valid (assuming they’re good explanations). The fact that you can talk about the emergence of feathers in terms of thermal regulation or camouflage or keratin deposits doesn’t mean that the field of paleontology is in disarray. It means that things are complicated and you can’t boil them down to simple monocausal narratives. The exact same thing is true if you’re talking about the decline of Roman power in terms of military incursions or poverty or lead poisoning from wine goblets.

What’s more, the fact that a basal theropod had feathers, and the inference that most theropods were thus feathered can’t be expressed in math. You can certainly use math in the process of constructing phylogenies (essentially family trees that show how different species are related to each other), but the end result isn’t an equation or formula; it’s an English-language narrative. Hell, even in math, verbal narratives can be more accurate than “mathy” formulas and equations. You can’t rationalize pi numerically, but “the ratio of the circumference of a circle to its diameter” is perfectly, 100% correct and accurate.

By insisting that good scholarship ends up with math, people like Turchin are trying to eliminate the “squishiness” or ambiguity inherent in narrative accounts. All that they manage to accomplish is to hide it. For example, in the chart used as an example of Cliodynamics in the recent Nature article, Turchin attempts to demonstrate the existence of 50-year cycles in violence in American history, broken down into categories like “political”, “labor or economic”, and “racially motivated.” The y-axis of the chart is “number of violent events per 5 years”, while the x-axis is time. In order to chart this, Turchin had to define what counted as a “violent event” and then how to categorize it, as though “political”, “economic”, and “racial,” are mutually exclusive and naturally occurring categories. That alone is sufficient to introduce enough ambiguity to allow his data to say whatever he wants it to, but even then, he doesn’t bother applying his categories in anything resembling a sensical way.

Take “racially motivated” violence. Turchin’s chart runs from 1800 to sometime in the 2000s. In it, he labels certain peaks, noting that the first big peak in racially motivated violence occurred during and in the aftermath of the civil war. Think about that. Generations of black people were held in slavery and subject to the whims of their owners, with virtually limitless documented examples of beatings, rapes, and murders, but “racially-motivated violence” didn’t begin until we went to war over it. Or to take a more recent example, in order to make his ~50 year cycle stick, he needed a peak in the ‘60s, so he points to the civil-rights movement has marking a high point in racial violence. The late ‘80s through the early ‘90s, however, mark a low point in his chart, both racially and overall. This, despite the fact that the ‘80s and ‘90s saw the worst race riots in American history after the acquittal of the police officers videotaped beating Rodney King, saw the incursion of violent drug cartels into cities like Miami, saw the ghettoization of huge swaths of cities like Los Angeles and Chicago, saw the rise of right-wing militia groups whose members included the likes of Timothy McVeigh and Terry Nichols, saw the American government sell weapons to a repressive regime and use the proceeds to fund Nicaraguan death squads, and saw the American military lead the invasion of a sovereign nation.

All of this is not to say that Turchin is deliberately cherry-picking evidence to fit his narrative. I strongly suspect that is not the case. He’s merely falling victim to the same trap that people have been falling into for as long as they’ve studied history. There is a reason that historians these days are skeptical of grand unifying theories and models; they’ve been tried, over and over again, and they don’t work. Contingency always rears its head and destroys the beautiful simplicity of the would-be rationalizer’s system. History is emergent, chaotic, contingent, and any attempts to reduce it to simple formulas (Asimov’s fictional science of history notwithstanding) will only produce distorted, potentially dangerous accounts of our past and our present.

I’m also not saying that Cliodynamics won’t produce anything useful. Large scale statistical analyses will undoubtedly be useful (although with all the caveats I alluded to above about deploying them). But just because it might be useful doesn’t mean it’s philosophically sound as a broader strategy. Marxist historiography produced many great studies and advanced our understanding in a lot of areas, but there’s a reason that there aren’t too many Marxist historians these days; the teleological aspects of the Marxist framework are and always were bullshit. Histories designed to show that violence occurs in predictable cycles will be just as flawed and biased as histories showing the inevitability of proletariat revolutions. Ultimately, historical data is squishy and if you look at it right you can see just about anything there. If you understand that, you can work with the evidence, as difficult and ambiguous as it is, to produce useful narratives that help us understand our past and our present. If you insist that there is only one truth to be read in the evidence, and you know how to obtain it, then there’s not much distinction between you and some asshole who finds revelation in a grilled cheese sandwich.