The state reduces violence, though unevenly so
Bioarcheological evidence both undermines and corroborates Pinker's narrative
I’ve already written on the historic trajectory of human violence and war, and Steven Pinker’s famous take on it. There, I noted that he’s both correct and wrong, because although the general spirit of what he says with respect to the state and modernity is (with some qualifications) correct, he is mostly wrong about human pre-history.
Now, there’s a new review paper on the matter, bringing together lots of the latest bioarcheological evidence, most of which was not even available at the time of Pinker’s writing.
As before, this body of literature both affirms and complicates what you think about human violence, if you’re coming at it from the Pinkerian/Hobbesian perspective. Put very briefly, centralized states and well-functioning institutions do appear to reduce violence, which is broadly consistent with the thesis. But the biorecord is also quite clear that violence does not follow a single downward trajectory from savage prehistory to peaceful modernity. Of course, there’s lots of methodological uncertainty, as is always the case with the anthropology and archaeology of violence.
How much violence was there?
A recent meta-analysis estimated that in the modern United States, there’s about 18% prevalence of brain trauma, of course primarily due to non-violence. That turns out to be roughly comparable to the cranial trauma rates found in many historic skeletal samples, though the causes are presumably very different! See the table below for detailed percentages across historic samples. Note the separate columns for any/total trauma shares and perimortem trauma, the latter referring to cranial damage immediately prior, during, or after death.
Importantly, the average perimortem trauma rate of 1–3% (plausibly indicating homicide, not just physical violence) in normal historic burial populations is nowhere near the commonly cited Pinker claim that 15–25% of all deaths in hunter-gatherer societies resulted from violence. Not really news, because we’ve known that already, but still important. I wouldn’t place to much emphasis on exact percentages (is it 2% or 5%?), but the 15–25% is definitely off, which we know from other archaeological and anthropological data.
Unsurprisingly, across virtually all regions and time periods, males show higher rates of cranial trauma than females. In the archaeological record, male prevalence is typically about two to four times that of females, depending on the region. This is more or less like today. From the paper:
Broadly, the sex differences in archaeological samples are consistent with—but somewhat smaller than—the patterns in modern forensic and public health records. Today, males account for 81% of adult homicide victims globally and the vast majority of nonlethal violent injuries: 87% in the United States, 84% in the United Kingdom, and 76% in Brazil.
Eisner and Peng raise several points about interpreting these patterns. One is that the smaller gap in archaeological samples may partly reflect the inclusion of accidental injuries, which shouldn’t be that different across the sexes. Another is that the gap between male and female trauma rates may have actually widened over time, as states and armies formalized warfare as a predominantly male activity. A large meta-analysis of pre-contact South American data found that in egalitarian societies, violent trauma rates did not differ significantly between sexes. It was in state-level societies with armies that the male-female gap became pronounced.
No simple decline
To my mind, the most important thing is the data showing that violence doesn’t evince a clear or even linear decline from prehistory to the present. Instead, what emerges across multiple regions are just up and down fluctuations (which I covered in my peer-reviewed work). But here, fluctuations are even more fine-grained or frequent than in my talk of an inverted-U curve. Times of relative peace alternate with periods of heightened violence multiple times, though, as the authors say, this “often correspond[s] to the consolidation, collapse, or internal quality of state institutions.”
This is a win and a loss for Pinker. For instance, in a few regions the lowest rates of cranial trauma are found not in state-level societies but in settled pre-state communities (see figure below for the South Andes case). So neither nomadic hunter-gatherers nor state societies but a secret third thing (early agriculture) that we usually paint as the most violent.
At the same time,
some of the highest levels of trauma appear not in the absence of states but during periods of civil war, imperial collapse, or contested sovereignty.
Depending on how you look at it, this is roughly what one would expect on the state-pacification thesis. But, the authors themselves also legitimately describe these as the “turbulent aftermaths” of centralized systems in crisis. Not a return to some pre-state baseline, but a specifically modern kind of disorder:
Overall, the evidence suggests that although states may reduce violence under certain conditions, their mere presence does not guarantee pacification—and may in some cases facilitate new forms of coercion and conflict.
Fair enough.
Here’s data for Europe: from the pre-medieval to the industrial period.
The Fertile Crescent: An S-shaped curve over 11,000 years
Baten and colleagues (2023) have an interesting case study, which is summarized in the review paper.
In the Fertile Crescent, during the Pre-Pottery and Pottery Neolithic (roughly 10,000–4500 BCE), cranial trauma rates were low. The review claims this suggests relative peace and absence of violence in these early farming communities, which (again) contradicts assumptions that the transition to pre-state agriculture was marked with exorbitant increases in violence, although of course we cannot be sure. Trauma then rose substantially during between 4500–3300 BCE, which was a time of growing settlement density and increasing inequality. That’s in line with the standard narrative.
With the creation of centralized states, violence then declined again during the Early and Middle Bronze Ages (3300–1500 BCE). So here again we have the standard institutional mechanisms (states, legal codes, armies) that should have reduced interpersonal violence by monopolizing the legitimate use of force.
But then trauma rates rose again in the Late Bronze and Iron Ages (1500–400 BCE). The authors say this is likely due to the collapse of centralized states, combined with increasing environmental pressures. And this again makes a lot of sense to me, and it seems to be in line with the standard narrative. Obviously, the emergence of the state at a particular point in history is not enough for violence to permanently go down. You need to maintain those institutions. If they collapse, violence resurges.
Baten and colleagues also mention the doux commerce thesis, which is the Montesquieu-related idea I wrote about some time ago, according to which expanding trade and economic intertwining between groups provide clear material incentives (and potentially cultural values) to reduce violence. Again, nothing earth-shattering here, but good to have the evidence.
There are other very instructive and similar case studies in the review (indicating a relative absence of violence in the period of early agriculture, which is weird) that you should check out, including the Southern Andes case that is depicted above in green.