Humans have probably always killed one another. Warfare is something narrower. It implies groups, planning, logistics, memory, leadership, and enough social organisation that violence stops being a personal act and becomes a collective institution. The interesting question is not when humans first killed. It is when killing acquired administration.
Ask that question and a small crowd of theories turns up, each with its own literature, its own favourite dig site, and its own conviction that the others have missed the point. The Hobbesians say war is ancient. The Rousseauians say civilisation invented it. The circumscription theorists say geography did. The economists point at surplus, the anthropologists at prestige, the international relations scholars at the security dilemma. Each theory has evidence. Each has counterexamples. The usual response is to pick a champion and treat the rest as noise.
There is a less exhausting option: to stop asking which theory is correct and start asking which substrate each theory is describing. Because they are not rivals. They are snapshots of different eras, each one accurate for the conditions it grew out of. Humanity’s circumstances kept changing, and the rational case for organised violence changed with them. The theories only look contradictory when thirteen thousand years are flattened into a single question.
By substrate I mean the relatively stable conditions that constrain behaviour: geography, climate, resources, technology, institutions, population density, information, and the cost of the alternatives to fighting. People choose within those conditions; they do not choose the conditions. And the claim is not that every theory of war is right, which would be a cheap sort of generosity. Plenty are wrong and stay wrong. It is that each of the major surviving theories names a real mechanism, one that tends to dominate only where the substrate makes it the decisive one.
War before walls
The obvious place to start is Hobbes, or at least the tradition that carries his name. In this view warfare is simply human competition scaled up: territory, mates, prestige, revenge. The cemetery at Jebel Sahaba in Sudan, roughly thirteen thousand years old, is the usual exhibit, and it has recently become a more interesting one for a reason its original excavators did not intend. A 2021 reanalysis of the remains set aside the old reading of a single massacre and argued instead for a long succession of small raids and ambushes. Most of the individuals carried injuries, and a substantial share showed both healed and unhealed wounds, which is the revealing detail: these were people who survived being attacked and were then attacked again, across time rather than in a single afternoon. The team tied the pattern to the upheaval at the end of the last ice age, reading this stretch of the Nile as a refuge where groups displaced by a shifting climate were pressed together and left competing for the same narrowing ground.
That is worth pausing on, because it quietly changes what the site is evidence of. It is probably not a war in the sense earlier writers imagined. It is recurrent raiding under pressure, and the pressure is environmental: the substrate compressed, exits closed, and violence became a standing feature of life in the refuge rather than an event. Territorial aggression is common in mammals, revenge cycles are observable in small societies today, and agriculture, in this reading, merely scaled something already present. But the older claim that Jebel Sahaba shows warfare deep in the hunter-gatherer past looks weaker than the claim it now supports, which is subtler and more useful: even foragers fight repeatedly when the ground stops letting them disperse. There is also Nataruk, a roughly 10,000-year-old massacre of foragers in Kenya, which sits beside Jebel Sahaba as a contested data point in that debate.
The wider record is real but uneven. Organised violence appears in patches, with long stretches and whole regions where it seems largely absent, and that unevenness suggests the Hobbesian picture is not a law of nature. It is a description of what happens under particular conditions: groups in contact, something contested between them, and no cheap way to avoid each other. Where those conditions held, people fought. Where they did not, apparently, people mostly did not bother.
Which is already a substrate observation, even if nobody involved would have used the word. The substrate here is thin: access to hunting grounds, water, perhaps grievance itself, since revenge is a resource of sorts once honour systems exist. Fighting over it stayed small because the substrate could not support anything larger. No stores to seize, no fields to occupy, no population that stays put long enough to be ruled. Victory bought very little, so nobody built institutions to pursue it. The Jebel Sahaba refuge is the exception that names the rule, and it names it early: when the substrate takes away the option of walking off, even people with almost nothing worth stealing will fight, and keep fighting.
Before the fields
One site refuses to line up on either side of the forager-farmer divide, and it happens to be the most consequential dig of the last thirty years. Göbekli Tepe, in south-eastern Anatolia, has monumental enclosures dating to roughly 9500 BCE, built by people who were still essentially hunter-gatherers. Its sister site Karahan Tepe, some forty-odd kilometres away, is part of the same Taş Tepeler world and was founded, on current readings, the better part of two thousand years before farming took hold in the region, though its monumental structures are largely later. This is the wrong way round for the standard story. The standard story has settlement and surplus arrive first, and monuments follow once there is wealth to spare. Here the monuments come first, and the wheat comes after. Einkorn’s wild ancestor sits on the Karacadağ slopes some sixty kilometres off, though both einkorn and emmer are now read as having more dispersed origins than the single-site story allowed, and one live hypothesis is that the pressure to feed large recurring gatherings is part of what drove the domestication, rather than the other way about.
That inversion is a problem for any account that treats stored grain as the origin of everything downstream, and the site does real damage to the sequence rather than merely adding to it. Whatever these people were accumulating, it was not a harvest. The substrate at Göbekli Tepe was open steppe grassland with abundant wild cereals and migrating gazelle, hunted alongside a wide range of other species rather than a single staple. What was being built up and defended, if anything was, was ritual capacity and the sheer mobilised labour to quarry and raise T-shaped megaliths. That is a kind of substrate the farming-first story has no slot for: fixed, valuable, non-portable infrastructure, arriving before agriculture and consisting of monuments rather than granaries.
It also unsettles the comfortable idea that mobile people own nothing worth taking. The reading of Göbekli Tepe has itself shifted, and instructively. The original excavator, Klaus Schmidt, framed it as a pure temple visited by wandering bands, the world’s first sanctuary. Since his death the project, under Lee Clare, has moved steadily away from that: there is now evidence of domestic structures, extensive cereal processing, a water supply, and everyday tools, pointing to a quasi-sedentary population living at or beside the site rather than pilgrims trekking in and out. If that reading holds, then here are people still largely foraging who are nonetheless staying put and piling up grinding stones, cisterns, and megalithic stone, none of which can be picked up and carried off. Stealability, in other words, did not wait for farmers. It began with foragers who stopped moving.
The violence signature is the telling detail. These sites show plenty of interpersonal violence and very little that looks like organised war. At nearby Körtik Tepe, from the same early horizon, researchers find raised levels of interpersonal violence and homicide but, so far, no evidence of fighting between groups. The German Archaeological Institute team’s own framing of the wider Taş Tepeler picture is careful and almost too apt: clear interpersonal and probably intergroup violence, more than Körtik Tepe alone attests and still well short of organised war, and the first stirrings of social inequality, running alongside visible strategies for defusing conflict and organising cooperation, with feasting and shared ritual acting as the incentives that pull people into working together. On that reading the enclosures are less a fortress than an institution, and the thing they institutionalise is not war but its avoidance. Nothing at these sites is fortified. The monumental architecture is emphatically not defensive.
The one detail that points the other way is a skull cult, and it is worth handling with tongs. Fragments of three carved and drilled human skulls from Göbekli Tepe carry deliberate post-mortem modification of a kind not seen elsewhere, and the excavators offered two readings without choosing between them: veneration of ancestors, or the display of dispatched enemies. The circumstantial weight sits slightly with veneration. One skull lay in a concentration of ochre, the drilled hole is placed so the skull would hang facing forward, and the surrounding Anatolian tradition of exhuming, displaying, and replastering skulls is overwhelmingly about ancestors rather than enemies. The enemy reading is real, and has its own comparanda in mutilated dead elsewhere in the Neolithic Near East, but even at its strongest it describes the display of individually dispatched dead, which belongs to the register of interpersonal violence and the treatment of the dead, not to organised between-group war. The more durable point is the meta one: the same institute that found the skulls has since argued the site was quasi-sedentary rather than purely ritual, which is a clean demonstration of an interpretation revising itself as the evidence accumulates, and a caution against reading any of this too confidently. A small fraction of Göbekli Tepe has been excavated, and whether it is best called residential or ritual is still openly contested.
Against the substrate frame, the site stops being an anomaly, though prediction is the wrong word for it: nobody predicted Göbekli Tepe, and the reading arrived long after the digging. If fighting appears where the substrate makes it pay, then a substrate that rewards gathering and building over raiding could very well produce cooperation instead, and Göbekli Tepe is a case where it did. Which substrate is which is mostly read off what people did on it, so the frame explains after the fact rather than forecasting. A new substrate appears here, sedentism, monumental capacity, feasting-driven intensification of wild and then domesticated cereals, well before agriculture proper. And it does not produce war. It produces the machinery of cooperation: gatherings, shared building, ritual, the apparatus a society uses to keep interpersonal violence from escalating into the organised kind. The substrate, not the calendar, is what settles it. Change the substrate and what fighting is worth changes with it, in either direction. At Jebel Sahaba a compressing substrate made recurrent raiding rational. In the Taş Tepeler hills a substrate built around gathering and building made cooperation the better bet. Same species, a few thousand years and a few hundred kilometres apart, pulled opposite ways by what the ground and the moment made pay.
What the farmers made stealable
Then people settle down, and the Rousseau tradition gets its turn. Agriculture concentrates a list of things that were, before it, either absent or thin on the ground: land ownership, grain stores, permanent wealth, defensible settlements, and eventually taxation and hierarchy. A forager’s stores cannot easily be stolen: everything is carried, and the owner can walk away. A farmer’s harvest can very much be stolen, and the farmer knows it, which is one plausible reading of why early settlements acquire walls.
Göbekli Tepe complicates this, and the complication is worth keeping rather than smoothing over. The foragers of the Taş Tepeler hills were already accumulating fixed, non-portable things: cisterns, grinding installations, tonnes of raised stone. Agriculture did not invent stealability. It concentrated it, standardised it, and made it storable and countable. A harvest sits in one place, can be measured, taxed, hoarded, and carried off, and it comes back every year. That is a different order of target from a monument that cannot be moved or a herd that has to be chased. The Rousseauians are not wrong that something changed with settled farming. They are describing the moment the substrate first made organised taking a reliably better return than organised hunting, year-on-year.
Jericho’s stone wall and tower, from perhaps the ninth millennium BCE, is the monument everyone reaches for, and it turns out to be the wrong one, because the fortification reading of it has largely collapsed. Kathleen Kenyon dug it in the 1950s and called it a defensive system, which made Jericho the first city under arms. That reading has not held. Ofer Bar-Yosef’s objection is architectural: the tower stands inside the wall rather than projecting out from it, so it offers defenders no field of fire and makes no sense as a bastion. There are no warfare-specific finds, no battle damage, no trauma in the bones to match a besieged town, and the arrowheads read as hunting gear. What is left is a wall that most now take for flood and debris defence against the wadi to the west, where the ground is lowest, and a tower read variously as ritual, as communal statement, or, on one proposal, as a solstice marker aligned to the shadow of the summit behind it. The single most famous early wall in the popular story of war turns out to be, on the best current reading, probably not a war wall at all. Which is inconvenient for the tidy Rousseauian sequence, and entirely consistent with the substrate one: walls appear when there is something fixed and vulnerable to protect, and water, silt, and the labour of the community are things worth protecting long before the neighbours are.
What changed here was not human nature, which had presumably been available for violence all along. It was that the substrate became dense, storable, and reliably worth taking. Conquest turned profitable in a way it had not been, and stayed profitable season after season, which is the part that lets institutions grow up around it.
The exits close
Robert Carneiro’s circumscription theory adds the piece that turns raiding into states. Conflict only accumulates power when the losers cannot leave. In an open landscape, a defeated group walks somewhere else and the victory evaporates. In a constrained one, a fertile river valley hemmed by desert or mountains, losing means submission, because there is nowhere to walk to. Population grows, the good land fills, and suddenly defeat has a new meaning: incorporation.
This is why early states appear where they do, in the Nile valley, Mesopotamia, coastal Peru, parts of Mesoamerica, rather than everywhere agriculture existed. The theory is sometimes presented as being about warfare, but it is really about exit costs. The substrate feature that does the work is not the fertility of the valley. It is the desert around it. Circumscription is a theory of what happens when the substrate removes the option of leaving. Carneiro’s valleys are the political equivalent of Jebel Sahaba’s climatic refuge: in both, the exits disappear, and violence that would once have walked away has nowhere left to go.
Surplus buys specialists
Agriculture changes one more thing, quietly and permanently. Surplus allows specialisation, and among the specialists it allows are full-time warriors, alongside the priests, bureaucrats, and kings who direct them. Hunter-gatherer bands rarely support a standing military class; every fighter is also a hunter, and war has to fit around subsistence. States can keep men under arms through the harvest. War stops being an emergency and becomes a profession, with equipment, training, doctrine, and a payroll.
A large historical databank has since let this be tested directly. Weighing seventeen rival drivers of social complexity, Turchin and colleagues found the best-supported combination to be rising agricultural productivity and the spread of new military technologies, iron weapons and cavalry above all, with state-level features like territory size carrying surprisingly little weight. That is close to a restatement of the paragraph above: surplus feeds the specialists, and the specialists’ tools do much of the scaling. In their data the megaempires arrive three or four centuries after cavalry does, a lag the authors read as consistent with the speed of change their model predicts; one commentator, Monique Borgerhoff Mulder, reads the same lag the other way, arguing that a delay that long makes military technology “a very remote predictor of the outcome”. Both readings accept the sequence; they differ on what it licenses. Whether a gap that wide still counts as a cause is the actual question, and a substrate reading does not get to answer it for free.
Surplus stacks the deck
Once war is a profession, it acquires professionals, and professionals have interests. Which leads directly to the inequality hypothesis: surplus creates wealth differences, wealth differences create elites, and elites compete. A peasant may have little interest in conquering the next valley. A king with debts, ambitious sons, and an army that expects to be paid has rather more. It seems plausible that many wars in the historical record reflect elite incentive structures more than popular need, dressed afterwards in whatever justification the era found respectable. The substrate here is not land or grain. It is the surplus extraction machinery itself, which has to keep running, and which occasionally finds war the cheapest way to keep running.
The inequality hypothesis has its own dataset now. The GINI Project, coding house-size disparity across hundreds of sites, finds that war did not simply widen the gap between rich and poor or level it, but did both, depending on how collective the governance was and whether land or labour was the scarce input. An older strand of the same work adds a substrate detail worth keeping: Old World inequality kept climbing where societies could harness large draft animals, cattle and later horses, which let the already-rich work more land and pushed everyone else toward a landless class, while the New World, without such animals, plateaued. The animals available to be domesticated are themselves part of the substrate that sets how steep a hierarchy can grow.
Prestige, revenge, and other renewable resources
Not everything reduces to material return, and the anthropologists are right to insist on it. In many tribal societies raiding demonstrates courage, adulthood, leadership, and masculinity, with the loot almost incidental. East African cattle raids illustrate the blend: the animals have economic value, but they are also status, marriage wealth, and identity. That institution has not stayed fixed, though. Since automatic weapons spread through the rangelands in the 1980s, the raiding has been commercialising, the AK-47 turning a ritualised contest of young men into organised theft for cash markets, and the prestige logic thinning as it goes. The act looks similar from a distance; the substrate beneath it has changed, and the reason for the raid with it. War as social signalling is real and recurs, but even signalling answers to what the ground will pay for.
The revenge machine is real too. One killing demands another, which demands another, until nobody remembers the original dispute and the feud has become the relationship. Blood feuds show how violence sustains itself without any material benefit at all, and the mechanism scales. Tribes inherit grievances. So do states.
It is tempting to treat prestige and revenge as exceptions to a materialist account, but they fit the substrate frame comfortably enough. Honour and obligation are social substrates, not physical ones. They are stores of value, they can be raided, and they compound. The distinctive thing about them is that they are renewable. A grain store can be emptied. A grievance refills itself.
Wars nobody wanted
The security dilemma is the odd one out, because it needs no aggressor at all. Nobody intends conquest. Everyone intends safety. One village builds a wall; its neighbour, reading the wall, recruits fighters; the first village, reading the fighters, confirms its fears and arms further. Neither side wanted war. Both manufactured it out of mutual visibility and mutual uncertainty. The pattern recurs from village palisades to naval arms races, and it suggests that some fraction of warfare needs no explanation in terms of anyone’s interests at all. The substrate condition is simply proximity plus opacity: close enough to threaten, not transparent enough to reassure. There is a second substrate variable underneath it, what strategists call the offence-defence balance: when conquest is cheap the fear the dilemma runs on is rational, and when defence holds the advantage the same fear can subside. Experimental work suggests the thinking is not easily talked down once it has taken hold between rivals who already distrust each other.
It is also a substrate made almost entirely of information. Nothing material changes hands and nothing shifts on the ground; what moves is only what each side believes about the other. The mechanism that runs on belief alone is the one that reaches furthest into the present, where information has become the contested ground itself.
Ideology arrives late, and mostly writes the press release
Religion, nationalism, and political ideology dominate how wars get narrated, which makes it easy to assume they dominate why wars get fought. The record suggests something more modest. Many medieval wars filed under religion also involved taxation, succession, trade routes, or dynastic manoeuvring, and modern ideological wars still run on industry, logistics, and territory. Ideas seem to decide who counts as friend and who as enemy. Material conditions tend to decide whether fighting is feasible, and often whether it is worth anyone’s while.
The honesty tax on this argument is the Crusades, the very case a materialist would most want to claim. The older reading of them as land-hunger in pious dress, younger sons riding east for estates, has largely been abandoned: charter evidence shows the crusaders were drawn from wealthy families, with the eldest sons going about as often as younger ones, and typically mortgaged or bankrupted themselves to go, while the current weight of scholarship puts genuine religious impulse near the centre, at least for the First Crusade. The substrate still set what was feasible, who could be fed and moved and paid, but here it does not look like the thing that lit the fire.
None of this makes ideology mere decoration. Sorting the world into friends and enemies is real work; it just tends to sit on top of the machinery rather than drive it.
Clausewitz called war politics by other means, and the line survives because it is half of a truth. The other half inverts it: politics may exist largely to avoid fighting. Negotiation is vastly cheaper than mobilisation, and diplomacy becomes attractive precisely because war is so expensive. Read that way, the history of civilisation is not merely the history of war. It is also the history of inventing alternatives to it, and pricing them.
Reading the substrate
Lined up together, the theories share a spine. Each one answers the same underlying question for a different era: what substrate made fighting a rational strategy here?
For hunter-gatherers, the substrate was access, territory, water, and the renewable resource of grievance, and it supported only small, episodic violence, except where a compressing climate closed the exits and made raiding recurrent, as at Jebel Sahaba. For the quasi-settled foragers of the Taş Tepeler hills, it was monumental and ritual capacity, and it pulled the other way, toward gathering and building and the machinery of cooperation, rather than toward war. For early farmers, it was stored grain and fixed land, and it made hoarding, walls, and taking all sensible at once. In circumscribed valleys, it was the geography of no exit, and it turned victory into accumulation and defeat into incorporation. For kingdoms, it was surplus and the extraction machinery, which bought professional armies and gave elites both the means and the motive. For tribal honour systems, it was prestige and obligation. For neighbours in general, it was proximity without transparency.
The Taş Tepeler case is the one that keeps the lens honest, because it is a counterexample the frame accommodates rather than trips over, though accommodation is the easier of the two tricks. If the substrate is what makes fighting rational, then some substrates should make it irrational, and reward cooperation instead. A world organised around large recurring gatherings and shared building is such a substrate, and it appears to have produced institutions for defusing violence rather than conducting it. The claim was never that the substrate drives people to war. It is that the substrate sets the price of war against its alternatives, and people, then as now, tend to read the price.
The long quarrel between war-is-ancient and war-is-invented now has a careful adjudication. Reviewing the global archaeological record region by region, Hugo Meijer finds neither camp vindicated: lethal violence runs deep in the human lineage, but whether it surfaced as war varied enormously across time and place, a plasticity rather than a constant. The deep-roots and shallow-roots camps are each describing a real part of the range. Naming the substrate as what decides between them is a gloss on that finding rather than part of it, though it is a gloss the record seems to invite.
The ideology changes dramatically across this history. Gods, crowns, nations, classes, civilisational missions: the narrative layer turns over every few centuries and sometimes faster. The substrate changes much more slowly. Arable land was worth fighting over for ten thousand years. Trade routes for five. Industrial capacity for two hundred and counting. Anyone trying to read a conflict, past or present, could do worse than to ignore the stated reasons for a moment and ask what, underneath, is actually contested, and what made contesting it by force look like a reasonable bet to somebody.
That is the same move as reading any substrate: the narrative layer records what people say, the substrate records what constrains them, and the two rarely agree. The origin of war is not a single event awaiting discovery in the right stratum of the right dig. It is a pattern that reappears, era after era, whenever the substrate makes organised violence cheaper than its alternatives.