Warfare and what makes groups hold together — candidate examples

Candidates only. Nothing here is written up for teaching yet.

Scope: this list is about cohesion inside a fighting group — what makes men stand in a line, and what happens when they stop. Cooperation across the lines — trench truces, the Christmas truce, rescue under occupation — is in the "Cooperation under war" section of examples-historical-successes.md and is not repeated here.

Two warnings that apply throughout.

Ancient warfare has unusually unreliable popular accounts and unusually contested scholarship. The mechanics of the hoplite push are a live academic fight, and the numbers at Thermopylae vary by more than an order of magnitude between ancient and modern sources. Disputes are flagged in the entry.

The modern combat-motivation literature has a contamination problem. S.L.A. Marshall's fire-ratio finding, which is the single most repeated claim in the field, was not researched the way he said it was. It is flagged below and should not be used.

The phalanx: the structure of a line that holds

The hoplite shield wall and the rightward drift

What happened: Greek heavy infantry from roughly the 7th to the 4th century BC fought shoulder to shoulder in a formation several ranks deep, each man carrying a round shield (aspis) about 90cm across on his left arm. Thucydides, describing the run-up to the battle of Mantinea in 418 BC, records that armies always drift rightward as they advance, "because fear makes each man do his best to shelter his unarmed side with the shield of the man next to him on the right".
Why it's a collective action case: This is the structural heart of the section, and Thucydides supplies the mechanism himself. The shield covers its bearer's left side and overlaps the man to his left, so protection is produced by the formation and consumed by your neighbor. Nobody can protect himself; the line protects you or nothing does. That makes cohesion a public good with a particularly brutal payoff structure — one man running does not reduce the good marginally, it opens a hole and can collapse the whole thing. The mechanism holding it together is mutual hostage-taking plus total observability: your own survival is in the hands of the two men either side of you, and they can see exactly what you are doing, because they are touching you. The rightward drift is the tell that individual incentives are pulling against the formation even while it holds.
Could serve: Introduction to collective action (S1), as the cleanest physical image of a public good in the course; War (S1); Olson (S3), on why small groups work.
Media: Extensive vase painting, particularly the Chigi Vase (c. 640 BC) which shows ranked hoplites. Reenactment groups (Hoplologia, the Hoplite Association) have produced experimental footage. The Bardo and Athens museums hold surviving shield facings.
Source: Thucydides, History of the Peloponnesian War 5.71; Victor Davis Hanson, The Western Way of War, 2nd ed. (University of California Press, 2009); Hans van Wees, Greek Warfare: Myths and Realities (Duckworth, 2004).

The othismos dispute

What happened: Greek sources use the word othismos — "pushing" — for a phase of hoplite battle. What physically happened is disputed and has been for about a century.
Why it's a collective action case: It matters for teaching because the two readings imply different mechanisms. On the orthodox reading, associated with Victor Davis Hanson, the rear ranks physically shoved the men in front, like a rugby scrum, so a man in the front rank literally could not run — cohesion enforced by the mass behind him. On the alternative reading, argued by Adrian Goldsworthy and George Cawkwell, othismos is metaphorical or a late-stage surge after a period of individual fighting, which makes staying in line a choice rather than a physical inevitability. Cawkwell called the literal scrum reading "wildest folly", on the grounds that the front ranks would have been unable to use their weapons at all.
Contested: This is the dispute, not a case with a dispute attached. No ancient source explicitly describes all ranks packing together and driving forward, so both readings are inferences. Use it as an illustration of how thin the evidence is for mechanisms that everyone thinks they know.
Could serve: War (S1), if the course wants to show that the mechanism is the contested part; Introduction to collective action (S1).
Media: Experimental archaeology footage exists from several reenactment groups; treat it as illustration, not evidence.
Source: A. K. Goldsworthy, "The Othismos, Myths and Heresies: The Nature of Hoplite Battle", War in History 4 (1997); Hanson, The Western Way of War; Peter Krentz, "The Nature of Hoplite Battle", Classical Antiquity 4 (1985).

Marathon and Plataea

What happened: At Marathon in 490 BC an Athenian and Plataean citizen army defeated a Persian invasion force on the coast of Attica. At Plataea in 479 BC a combined Greek army under Spartan command defeated the Persian field army in Boeotia, ending the invasion. Both are hoplite battles fought by men who were farmers and craftsmen the month before and after.
Why it's a collective action case: The mechanism worth naming is conscription from a face-to-face community into the unit that fights. Athenian hoplites served by tribe (phylē), so the man beside you in the line was a neighbor you would see in the assembly and the marketplace for the rest of your life, and your conduct would be reported there. That converts a one-shot terror decision into a repeated game with an audience — questions 2 and 4 of the diagnostic, supplied by the recruitment system rather than by anything done on the day.
Contested: Numbers for both battles, and particularly Persian numbers, are unreliable — see the Thermopylae entry.
Could serve: War (S1); Repeated interaction and the folk theorem (S3); The cooperative ape (S3).
Media: The Marathon tumulus (Soros) still stands and is visitable. The Serpent Column from Delphi, listing the Greek states that fought at Plataea, is in the Hippodrome in Istanbul — a good primary object.
Source: Herodotus, Histories 6.102–117 (Marathon) and 9.19–70 (Plataea); Peter Krentz, The Battle of Marathon (Yale UP, 2010).

Thermopylae, 480 BC

What happened: A Greek force held the pass at Thermopylae against the Persian army for some days before being outflanked by a mountain path. Leonidas dismissed most of the allied contingents and remained with the Spartans, the Thespians and the Thebans, and was killed.
Why it's a collective action case: The interesting part is the dismissal, not the last stand. A rearguard has an obvious free-rider structure — everyone is better off if someone stays and better off still if it isn't them — and the historical answer was not persuasion but selection in advance. Herodotus says Leonidas' 300 were chosen from men who had living sons, which removes the strongest private reason to survive and is as close to a designed selective incentive as the ancient record gets. The Spartan case also runs on the reverse side: see tresantes below, where the penalty for surviving a defeat was social death.
Contested: Heavily, and in the direction the popular account runs. The force was not 300 — modern reconstructions of the Greek total run to around 7,000, with Thespians and Thebans dying alongside the Spartiates, plus perioikoi and helots usually left out of the retelling; Diodorus gives 1,000 Lacedaemonians rather than 300. Herodotus' Persian figures (over 2 million) are not credible; modern estimates range from roughly 100,000 to 300,000 and are themselves contested. Say this plainly if the case is used, because the audience will arrive with the film version.
Could serve: War (S1); The bill (S5), on what cohesion is manufactured for.
Media: 300 (Zack Snyder, 2006) is the version everyone knows and is a comic adaptation, not a historical one; use only as the thing being corrected. The 1962 The 300 Spartans is closer to Herodotus. The Kolonos hill and the Simonides epigram are on site.
Source: Herodotus, Histories 7.201–233; Diodorus Siculus 11.4; Paul Cartledge, Thermopylae: The Battle That Changed the World (Macmillan, 2006).

Socrates as a hoplite

What happened: Socrates served in the Athenian heavy infantry at the siege of Potidaea (432 BC), at Delium (424 BC) and reportedly at Amphipolis (422 BC). In Plato's Symposium, Alcibiades says Socrates saved his life and his arms at Potidaea and then let the prize for valor be awarded to Alcibiades instead. Of Delium, Alcibiades says he watched the retreat from horseback while Socrates was on foot, and that Socrates withdrew so composedly — "looking about at friends and enemies", in the usual rendering — that pursuers went after the men who were running instead. In the Laches, the general Laches gives a corroborating eyewitness account.
Why it's a collective action case: The mechanism is orderly withdrawal as a coordination problem, and it is the exact inverse of the rout entries below. Men who retreat in formation are hard to kill; men who run are easy. Each individual's dominant move is to run, and the result of everyone taking it is that most of them die. Socrates' reported conduct is interesting precisely because Alcibiades reports that it worked — composure was legible to the enemy as continued cohesion, and pursuit went elsewhere.
Contested: These are literary portraits in philosophical dialogues written by a partisan, not military dispatches, and Alcibiades' speech in the Symposium is a set-piece encomium. Delium was an Athenian defeat and a genuine rout; that much is independently in Thucydides. Treat the detail of Socrates' bearing as Plato's characterization.
Could serve: War (S1); Introduction to collective action (S1) — a small, human-scale entry point to the rout problem.
Media: Plato's texts are short, public domain and readable aloud — Symposium 219e–221c is the passage. S. Sara Monoson's chapter is the scholarly treatment.
Source: Plato, Symposium 219e–221c and Laches 181a–b; Thucydides 4.89–101 on Delium; S. Sara Monoson, "Socrates' Military Service", in Socrates and Dionysus (2019).

The Sacred Band of Thebes

What happened: An elite Theban unit of 300, formed in the 370s BC and, on Plutarch's account, made up of 150 pairs of male lovers. It was central to the Theban defeat of Sparta at Leuctra in 371 BC and was destroyed at Chaeronea in 338 BC by Philip II of Macedon. Nineteenth-century excavation under the Lion of Chaeronea found 254 skeletons laid out in rows, generally identified as the Band.
Why it's a collective action case: An explicit, deliberate attempt to engineer cohesion by making the unit out of pre-existing bonds rather than hoping bonds form inside it. Plutarch reports the reasoning directly: a man will abandon a stranger and will not abandon a lover, because shame in front of that specific person is unbearable. This is the ancestor of the whole primary-group literature, stated as design intent 2,300 years earlier, and it is worth putting next to Shils and Janowitz for exactly that reason.
Contested: The Band is not mentioned in any source before Dinarchus in the 320s BC, and Plutarch writes some four centuries after Chaeronea, so the pairing detail rests on late testimony. The identification of the Chaeronea skeletons is generally accepted but is an inference from location.
Could serve: War (S1); Ritual and the manufactured tribe (S5), as a secular version of the same move; Olson (S3).
Media: The Lion of Chaeronea stands on the site and is photographed extensively. James Romm, The Sacred Band (Scribner, 2021) is the accessible recent account.
Source: Plutarch, Life of Pelopidas 18–19; Plato, Symposium 178e–179a, for the "army of lovers" idea Plutarch is drawing on; Romm, The Sacred Band (2021).

The Plataean breakout, winter 428/427 BC

What happened: Besieged Plataeans, walled in by a Peloponnesian circumvallation, planned an escape over the wall. They built ladders to match the wall's height by counting its courses of brick from a distance — many men counting at once, on the reasoning that although some would miscount, the most common answer would be right. Around 220 of roughly 480 men who had originally agreed went over on a stormy night; 212 reached Athens.
Why it's a collective action case: Two mechanisms in one short episode, and it is the best siege-and-sortie candidate on the list. First, aggregation as a substitute for measurement — an early and explicit use of independent errors cancelling, which is a coordination solution rather than a dilemma solution. Second, the participation decision: roughly half the men who had agreed to go lost their nerve at the last moment and stayed. The escape got safer the fewer went, since a smaller party is quieter, which makes this a rare military case where free riding does not degrade the good — and the split still fell near half.
Could serve: Introduction to collective action (S1); The games (S2), as a coordination problem; Olson (S3).
Media: None known in film. Thucydides 3.20–24 is short and vivid enough to read aloud.
Source: Thucydides, History of the Peloponnesian War 3.20–24.

Manufacturing cohesion: Sparta and Rome

The Spartan agoge

What happened: Spartan boys were removed from their households at around age seven into a state-run upbringing organized in age classes, with communal living, deliberate deprivation, competition, ritual theft and public beatings, continuing into adulthood. This is the description given by Xenophon and, much later and much more fully, by Plutarch.
Why it's a collective action case: The most thoroughgoing historical attempt to manufacture a primary group by destroying the competing one. Kin loyalty is the main rival to unit loyalty, so the system removed the child from the family and substituted an age cohort he would later fight beside. That is the same structural move Henrich attributes to the medieval Church's marriage rules — dissolve kin networks to make a larger unit possible — run by a state for military purposes.
Contested: Seriously. Nigel Kennell argued that the agoge as Plutarch describes it is largely a Hellenistic and Roman-era reinvention, archaizing features included, produced to support Sparta's claim to be a unique society. The word agoge itself is not securely attested for the classical system. The classical-period evidence is much thinner than the popular picture, and the popular picture is downstream of a source writing 500 years later. Stephen Hodkinson's work on the wider "Spartan mirage" applies.
Could serve: Ritual and the manufactured tribe (S5); Henrich and the trust of strangers (S5), as the comparison case; The bill (S5).
Media: Little reliable visual material; Sparta left almost no monumental remains, which is itself a teachable fact. Avoid 300 here.
Source: Xenophon, Constitution of the Lacedaemonians; Plutarch, Life of Lycurgus; Nigel M. Kennell, The Gymnasium of Virtue: Education and Culture in Ancient Sparta (University of North Carolina Press, 1995); Stephen Hodkinson, Property and Wealth in Classical Sparta (2000).

The syssitia

What happened: Spartan citizens ate their evening meal in small common messes (syssitia, also phiditia) rather than at home, for life. Each member had to contribute a fixed quota of barley, wine, cheese and figs from his own land; failure to contribute cost citizenship. Admission to a mess was by unanimous vote of the existing members, on Plutarch's account using a ball dropped into a vessel — one dissenter excluded the candidate.
Why it's a collective action case: The best single Spartan entry, because the mechanism is unusually legible and does several jobs at once. A daily attendance requirement is a cheap, continuous observability device: absence is immediately visible to the fifteen men who will stand beside you in the line. The contribution quota is a recurring costly signal with a hard penalty. And the unanimity rule on entry means every member has already been individually vouched for by every other member, which is a selection mechanism rather than a socialization one. Note the ugly corollary — the quota required land, and land required helots, so the mess system is directly financed by the subject population.
Contested: The blackball detail comes from Plutarch and is late. The wider economic argument — that syssitia contributions tied citizenship to landholding and so drove the long decline in Spartiate numbers — is Hodkinson's and is debated.
Could serve: Costly signals and Big Gods (S5); Ritual and the manufactured tribe (S5); Repeated interaction and the folk theorem (S3); The bill (S5).
Media: None known.
Source: Plutarch, Life of Lycurgus 10–12; Aristotle, Politics 1271a; Stephen Hodkinson, Property and Wealth in Classical Sparta (Classical Press of Wales, 2000).

The tresantes — Spartan "tremblers"

What happened: A Spartan who fled in battle, or survived a defeat his comrades did not, could be designated a tresas, a trembler. The reported consequences were civil and social rather than capital: exclusion from office and from the messes, ineligibility for marriage, an obligation to give way in the street, a duty to wear a patched cloak and shave half the beard, and no legal recourse against assault. Herodotus records Aristodemus, who left Thermopylae alive, being shunned on return and seeking death at Plataea the following year.
Why it's a collective action case: Punishment devolved to the whole population and made continuous. A death penalty for cowardice has to be administered by someone, costs the state a soldier, and is over quickly. Social death is administered by everyone at no cost to anyone, is visible daily, is imposed by exactly the people whose regard the sanction operates on, and does not end. It is the sharpest historical example of a second-order problem being solved by making punishment free.
Contested: Whether the status was permanent, temporary or heritable is not established. After Leuctra in 371 BC Plutarch reports Agesilaus being asked what to do with the many survivors and answering that the laws should be allowed to sleep for a day — which suggests the penalty was unenforceable at scale, and that the mechanism only worked on small numbers.
Could serve: Repeated interaction and the folk theorem (S3) — punishment; Coercion (S4); Ritual and the manufactured tribe (S5).
Media: None known.
Source: Herodotus, Histories 7.229–231 and 9.71; Xenophon, Constitution of the Lacedaemonians 9; Plutarch, Life of Agesilaus 30; Marcello Lupi, "Tremblers (tresantes) at Sparta", in The Encyclopedia of Ancient History (Wiley, 2013).

The Roman manipular legion

What happened: From roughly the 4th to the 2nd century BC the Roman legion fought in maniples — blocks of around 120 men arranged in three lines with gaps, hastati in front, principes behind, triarii in reserve — rather than as one continuous mass. Polybius, writing in the 2nd century BC, is the main source. The arrangement allowed a tired or pressed front line to fall back through the gaps and be relieved.
Why it's a collective action case: A structural answer to the problem the phalanx has no answer to. In an unbroken line, the only exit is to run, and one man running can start a rout. The maniple system supplies a legitimate, ordered exit, so exhaustion and fear do not have to express themselves as collapse. It converts an all-or-nothing cohesion problem into a rotation, which is the same move as shift work or crop rotation. If the course wants one institutional design lesson from ancient warfare, it is this: give people a permitted way to stop, or they will find an unpermitted one.
Contested: Exactly how line relief worked in practice — whether the maniples physically interchanged, and how the gaps were covered — is debated and not settled by Polybius' text.
Could serve: War (S1); Changing the payoffs (S4), read as institutional design; Introduction to collective action (S1).
Media: Trajan's Column (AD 113) is later and imperial but is the standard visual source for Roman infantry; the Ahenobarbus relief is closer in date. Reenactment groups (Ermine Street Guard) produce drill footage.
Source: Polybius, Histories 6.19–42; Adrian Goldsworthy, The Roman Army at War, 100 BC–AD 200 (Oxford UP, 1996).

Decimation and the fustuarium

What happened: Polybius describes the fustuarium: a soldier condemned for desertion, theft from the camp, false witness or repeated offenses was clubbed or stoned to death by his own comrades in front of the assembled army. For collective offenses — a unit abandoning its position — Rome used decimation, in which one man in ten was selected by lot and killed by the other nine. Crassus is reported by Appian and Plutarch to have revived it against troops who fled during the war with Spartacus, around 71 BC.
Why it's a collective action case: Two distinct mechanisms, and separating them is the value of the entry. The fustuarium solves the second-order free rider problem by conscripting the punishers: the men who would otherwise have to be paid or coerced into enforcing the norm are made to do the killing themselves, at no cost to the state, and are thereby implicated. Decimation is stranger and worth dwelling on — it punishes by lot, so it deliberately severs the link between individual guilt and individual punishment. What that buys is collective liability: since you may die for your neighbor's flight, you now have a reason to stop him running, which is precisely the incentive a line needs.
Contested: Decimation was rare and treated as archaic even by Romans; the reported instances are few and mostly attached to famous names. Do not present it as routine practice. The details of Crassus' decimation differ between Appian and Plutarch.
Could serve: Coercion (S4); Repeated interaction and the folk theorem (S3); The bill (S5).
Media: None reliable; illustrations are all much later imaginings.
Source: Polybius, Histories 6.37–38; Appian, Civil Wars 1.118; Plutarch, Life of Crassus 10; Sara Elise Phang, Roman Military Service (Cambridge UP, 2008).

The sacramentum and the contubernium

What happened: Roman soldiers swore an oath, the sacramentum, binding them to obey and not desert; under the empire it was renewed annually and sworn to the emperor personally. Below the century, the legion's smallest unit was the contubernium — eight men who shared a tent and a mule, cooked together and were punished together. Decorations included the corona civica, an oak wreath awarded specifically for saving the life of a fellow citizen in battle.
Why it's a collective action case: Three layers of mechanism stacked deliberately. The oath is a commitment device with a religious cost — breaking it is sacrilege, not just desertion. The eight-man tent group is the primary group as an organizational unit, built rather than left to emerge, and it is the direct ancestor of the modern section or squad. The corona civica is a selective incentive attached to exactly the behavior a line needs — not killing, which individual valor rewards anyway, but not abandoning the man next to you.
Could serve: Olson (S3), for the selective incentive; Costly signals and Big Gods (S5), for the oath; War (S1).
Media: Surviving military diplomas and oath texts; Vindolanda tablets for everyday unit life. Trajan's Column for tent groups on campaign.
Source: Polybius, Histories 6.21 and 6.39; Sara Elise Phang, Roman Military Service: Ideologies of Discipline in the Late Republic and Early Principate (Cambridge UP, 2008); Adrian Goldsworthy, The Complete Roman Army (Thames & Hudson, 2003).

Rout: the failure mode

Ardant du Picq on why lines break

What happened: Charles Ardant du Picq was a French colonel who circulated a questionnaire to serving officers asking what actually happened in battles they had fought in, and drafted a study of combat from the answers and from ancient sources. He was mortally wounded near Metz in 1870; the work was published posthumously as Études sur le combat.
Why it's a collective action case: The founding text of the whole subject, and it names the mechanism explicitly. Du Picq's argument is that battle is decided morally rather than physically — that two lines rarely meet, because one breaks first — and that what holds a line is not courage but mutual acquaintance and mutual surveillance. His formulation is worth quoting exactly: "Four brave men who do not know each other will not dare to attack a lion. Four less brave, but knowing each other well, sure of their reliability and consequently of mutual aid, will attack resolutely." That is Olson's group-size argument stated as tactics sixty years before Olson.
Contested: Du Picq's evidence is officer recollection and his reading of Polybius and Caesar, not systematic data. His influence on French doctrine before 1914 — where the emphasis on moral force fed the cult of the offensive — is a standard criticism, and is itself a useful lesson about a correct mechanism applied wrongly.
Could serve: War (S1); Olson (S3); Repeated interaction and the folk theorem (S3).
Media: Battle Studies is public domain and on Project Gutenberg. No film.
Source: Charles Ardant du Picq, Études sur le combat (1880), translated as Battle Studies (Macmillan, 1921), available via Project Gutenberg.

Where the casualties actually fall

What happened: Peter Krentz collected the casualty figures given by ancient sources for classical Greek hoplite battles between Plataea (479 BC) and Leuctra (371 BC). The winning side averaged around 5% dead; the losing side around 14%.
Why it's a collective action case: The number is the argument. If standing and fighting were what killed people, the two figures would be closer together. The gap exists because most of the dying happens after one line breaks, when men who have turned their backs are run down by men who have not. The individually rational move — run — is the one that gets you killed, and the collectively rational move — stand — is the one that keeps casualties at 5%. That is the payoff matrix of a fighting line, with real numbers in it, and it is the single most useful quantitative fact in this whole list.
Contested: Ancient casualty figures are reported by partisan sources and are often round numbers; Krentz is explicit that the figures are indicative rather than precise. The ratio, not the absolute levels, is the robust part.
Could serve: The games (S2) — build the payoff matrix from it; War (S1); Introduction to collective action (S1).
Media: None. This is a table, and it should be shown as one.
Source: Peter Krentz, "Casualties in Hoplite Battles", Greek, Roman and Byzantine Studies 26 (1985), 13–20.

Caporetto, October 1917

What happened: On 24 October 1917 German and Austro-Hungarian forces attacked the Italian line on the Isonzo using infiltration tactics, gas and concentrated artillery. The Italian front collapsed and the army fell back more than 100km to the Piave over about two weeks. Italian losses are usually given as around 10,000 dead and 30,000 wounded, against roughly 265,000 taken prisoner, with a further very large number of men separated from their units or dispersed.
Why it's a collective action case: The clearest modern illustration of the asymmetry du Picq describes, at industrial scale. The ratio of prisoners to dead is about 6 to 1, which is the signature of a cohesion collapse rather than a defeat by firepower — men stopped being a unit before they stopped being alive. Worth teaching alongside the Krentz figures as the same phenomenon two thousand years apart. The Italian command's response — Cadorna's public attribution of the collapse to the soldiers' own unwillingness, and a documented use of summary executions and decimation-style punishment in the Italian army — also makes it a test case for whether coercion restores cohesion or destroys it.
Contested: How much of the surrender was demoralization and how much was simple encirclement is argued; Cadorna's version was self-serving and was rejected by the subsequent commission of inquiry. The scale of Italian decimation practice is debated.
Could serve: War (S1); Coercion (S4), as a case against it.
Media: Extensive photography of the prisoner columns. Ernest Hemingway, A Farewell to Arms (1929), has the best-known fictional retreat sequence; Mario Monicelli's La Grande Guerra (1959) is the Italian film treatment.
Source: Mark Thompson, The White War: Life and Death on the Italian Front 1915–1919 (Faber, 2008); Vanda Wilcox, Morale and the Italian Army during the First World War (Cambridge UP, 2016).

The modern small-unit literature

Shils and Janowitz on primary groups in the Wehrmacht

What happened: Edward Shils and Morris Janowitz, both working in the Psychological Warfare Division of SHAEF, interviewed German prisoners and analyzed captured material to explain why the Wehrmacht kept fighting after the war was obviously lost. Their 1948 paper argued that the decisive factor was not Nazi ideology but the small face-to-face group — the squad and the section — and that units disintegrated when those groups were broken up.
Why it's a collective action case: The founding modern statement of the mechanism, and the reason "primary group cohesion" entered the language. The claim is that men fight for roughly ten people they know, that the ideological content of the war is largely irrelevant at that scale, and that army-level cohesion is therefore a property of thousands of small groups rather than of the army. Note what this predicts: destroy the small groups — by casualties, by rotation, by replacing men individually — and the army stops working regardless of morale campaigns.
Contested: See the next entry. The evidence base is prisoner interviews conducted by the enemy at the end of a lost war, which has obvious problems, and respondents had reason to attribute their service to comradeship rather than to ideology.
Could serve: Olson (S3); War (S1); Repeated interaction and the folk theorem (S3).
Media: None. The paper is short and is the material.
Source: Edward A. Shils and Morris Janowitz, "Cohesion and Disintegration in the Wehrmacht in World War II", Public Opinion Quarterly 12 (1948), 280–315.

Bartov's counter-argument: ideology and terror

What happened: Omer Bartov argued against the Shils–Janowitz reading on the Eastern Front specifically. His case is that casualty rates there were so high that primary groups could not survive — units were destroyed and rebuilt repeatedly, so there was no stable set of ten men to fight for — and that what held the army together instead was a combination of ideological indoctrination and an extremely harsh military justice system.
Why it's a collective action case: The direct challenge to the course's preferred mechanism, and it should be put in explicitly rather than avoided. If Bartov is right, cohesion under extreme attrition is produced by coercion plus shared belief rather than by repeated interaction, because repeated interaction is not available. The numbers support the coercion half: estimates of German soldiers executed by their own military justice system in the Second World War run to the order of 15,000, against one American soldier executed for desertion and none in the British forces, which had abolished the death penalty for desertion in 1930. The figures come from Manfred Messerschmidt's work on Wehrmacht military law and vary between studies.
Contested: The relative weight of ideology, terror and primary groups is not settled, and Bartov's argument is strongest for the Eastern Front and weaker elsewhere. Treat it as the live alternative hypothesis.
Could serve: Coercion (S4); War (S1); The bill (S5).
Media: Extensive; use with care given the subject matter.
Source: Omer Bartov, The Eastern Front, 1941–45: German Troops and the Barbarisation of Warfare (Macmillan, 1985) and Hitler's Army (Oxford UP, 1991); Manfred Messerschmidt, Die Wehrmachtjustiz 1933–1945 (2005).

S.L.A. Marshall's ratio of fire — discredited

What happened: In Men Against Fire (1947), Brigadier General S.L.A. Marshall claimed that only 15–25% of American infantrymen fired their weapons in combat, and said the figure came from immediate post-combat group interviews with some 400 rifle companies. In 1988 Roger Spiller, of the US Army's Combat Studies Institute, examined Marshall's notebooks and papers and concluded that the systematic data collection Marshall described did not happen — the company figure drifted between 400, 603 and "something over 500" in Marshall's own later accounts, and the fire ratio question does not appear in his interview notes. Spiller's conclusion: "the systematic collection of data that made Marshall's ratio of fire so authoritative appears to have been an invention."
Why it's a collective action case: Include it, but as the thing that is wrong. It was for fifty years the most cited claim about combat behavior, it drove real changes in US Army marksmanship training, and it is not supported. Robert Engen's later work on Canadian combat reports found nothing resembling Marshall's ratio. If the course wants a case about how a striking mechanism story propagates faster than the evidence for it, this is the best available one, and it is in our own subject area.
Contested: Not really — the methodology critique is broadly accepted. Some argue Marshall's underlying qualitative point about reluctance to fire had something to it even if the number was fabricated, and that is a defensible but much weaker claim.
Could serve: War (S1); as a methodological aside anywhere the course cites a striking number.
Media: Spiller's article is freely available online. Marshall's book remains in print, which is part of the point.
Source: S.L.A. Marshall, Men Against Fire (William Morrow, 1947); Roger J. Spiller, "S.L.A. Marshall and the Ratio of Fire", RUSI Journal 133:4 (1988), 63–71; Robert Engen, "S.L.A. Marshall and the Ratio of Fire: History, Interpretation, and the Canadian Experience", Canadian Military History 20 (2011).

Grossman, On Killing

What happened: Dave Grossman's On Killing (1995) argues that humans have a deep innate resistance to killing their own species, that armies have progressively overcome it through conditioning-based training, and that this carries psychological costs. The book was widely read inside and outside the military and is the main route by which Marshall's figure reached a general audience.
Why it's a collective action case: Include it chiefly to be honest about the chain of citation. Grossman's central empirical support is Marshall's fire ratio, so the critique above propagates directly into it. The thesis is not thereby disproved, but it is much less well evidenced than its popularity suggests, and parts of Grossman's later work on media violence have been criticized separately. If the course wants the "we are reluctant to kill" claim — which would be a striking counterweight to the arms-race framing in Session 1 — it needs a better source than this one.
Contested: Yes, on the evidence base. Say so if the book comes up, which it will, because it is widely read.
Could serve: War (S1), with the caveat attached; The cooperative ape (S3), if a defensible version of the claim can be sourced.
Media: Grossman lectures widely and video is plentiful.
Source: Dave Grossman, On Killing (Little, Brown, 1995); for the critique, Spiller (1988) and Engen (2011) above.

What happened: A body of later work questions whether interpersonal bonding is what produces combat effectiveness at all. Robert MacCoun's 1993 RAND review distinguished social cohesion — liking and attachment among members — from task cohesion, the shared commitment to doing the job, and found the performance evidence supported task cohesion. Anthony King argued from infantry drill and training that cohesion is produced by collective competence and rehearsed procedure rather than affection. Wong, Kolditz, Millen and Potter's 2003 interviews with US soldiers in Iraq found both — trust in comrades, and also trust in leaders' competence and stated ideological reasons.
Why it's a collective action case: The mechanism dispute that matters most for transferring any of this out of the military context. If cohesion runs on social bonds, it needs stable membership, time and small numbers, and it does not scale. If it runs on task cohesion — everybody knows the drill, everybody can see everybody else executing it, competence is observable — then it is portable, can be trained, and survives turnover. That is directly the course's scale question: Session 3's mechanisms all need repetition and acquaintance, and task cohesion is a candidate for something that works without them.
Contested: Live and unresolved; King and Guy Siebold conducted a sustained exchange over it. The policy stakes were real — the social-cohesion argument was used in debates over integrating women and openly gay personnel, and MacCoun's work was written into that context.
Could serve: Why this gets harder at scale (S3); Olson (S3); Cooperation among strangers framing generally (S4).
Media: None specifically.
Source: Robert J. MacCoun, "What Is Known about Unit Cohesion and Military Performance", in RAND, Sexual Orientation and U.S. Military Personnel Policy (1993); Anthony King, The Combat Soldier (Oxford UP, 2013); Leonard Wong et al., Why They Fight: Combat Motivation in the Iraq War (Strategic Studies Institute, 2003).

Replacement systems: van Creveld's comparison

What happened: Martin van Creveld compared German and American army organization in the Second World War. Among the differences he emphasized: the German system recruited regionally and rebuilt units by withdrawing them and refilling them from their own depot, keeping men together, while the US Army used an individual replacement system that fed men forward one at a time into units that stayed in the line.
Why it's a collective action case: A natural experiment on the Shils–Janowitz mechanism, run as policy. If cohesion is a property of stable small groups, then an individual replacement pipeline is a machine for destroying it — the new man knows nobody, nobody knows him, and casualty rates among such replacements were notoriously high. The mechanism named here is administrative: who arrives, with whom, and when. It is also the most directly transferable lesson in this section, because it is about organizational design rather than about combat.
Contested: Van Creveld's book is an argument with a thesis and has been criticized for admiring the Wehrmacht's professional qualities while underweighting its political character and its criminality. Other factors — German defensive posture, American material superiority — complicate the performance comparison.
Could serve: Why this gets harder at scale (S3); Olson (S3); Changing the payoffs (S4), as institutional design.
Source: Martin van Creveld, Fighting Power: German and U.S. Army Performance, 1939–1945 (Greenwood, 1982).

Mutiny: the collective action problem of the mutineers

The French army mutinies, 1917

What happened: After the failure of the Nivelle offensive on the Chemin des Dames in April 1917, large-scale refusals of duty spread through the French army from late April into June. Guy Pedroncini's 1967 archival study found 49 infantry divisions affected — nine gravely, fifteen seriously, twenty-five in isolated repeated incidents. He counted 3,427 men tried by court-martial, 554 death sentences and 49 executions. Pétain, appointed commander-in-chief in May, combined selective punishment with leave reform, better food, and a halt to offensives.
Why it's a collective action case: The cleanest case for the double structure, and the reason mutiny belongs in this course at all. Mutiny is a public good for the soldiers — everyone gains if the attacks stop — and the first man to refuse is shot, so the individual incentive is to let someone else start. The observed mechanism is threshold and simultaneity: refusals took the form of units declining to move up while agreeing to hold the line, which is a demand deliberately shaped to be defensible and to make punishment politically awkward, and they spread by contagion between units at rest, where men from different divisions met. The suppression mechanism is the mirror image: try a small number, publicly, so that the cost falls on identifiable ringleaders rather than on the many.
Contested: Figures vary between studies, and the 49 executions figure is Pedroncini's; other counts are lower. The word "mutinies" itself is contested — the refusals were mostly not insurrectionary, and French historians describe them as a strike more than a revolt.
Could serve: War (S1); Olson (S3); The reframe (S2) — the same structure read from the mutineers' side rather than the army's.
Media: Stanley Kubrick's Paths of Glory (1957) is set in 1916 and is fiction, but dramatizes the exemplary-execution mechanism precisely; it was effectively unshown in France for decades. Un long dimanche de fiançailles (Jeunet, 2004) also touches it.
Source: Guy Pedroncini, Les mutineries de 1917 (PUF, 1967); Leonard V. Smith, Between Mutiny and Obedience (Princeton UP, 1994), which reads the 5th Infantry Division's mutinies as negotiation.

The Russian army in 1917 and Order No. 1

What happened: On 1 March 1917 the Petrograd Soviet issued Order No. 1, directing that elected soldiers' committees be formed in every unit, that weapons be under committee control, that deferential forms of address to officers be abolished, and that orders from the Provisional Government be obeyed only where they did not contradict the Soviet. Discipline in the army deteriorated through the year; desertion ran at very large scale, particularly after the failure of the June offensive.
Why it's a collective action case: The mutineers' coordination problem solved from the outside, by a document. A scattered army cannot mutiny because no unit knows whether others will act; Order No. 1 supplied common knowledge and a standing institution in one move — every unit learned simultaneously that every other unit was to form a committee, which made refusal safe by making it general. It is the best historical illustration in the course of a focal announcement flipping an equilibrium, and it works in the direction of dissolving an army rather than building one, which makes it a useful counterweight to the rest of this list.
Contested: How far Order No. 1 caused the army's disintegration, as opposed to formalizing a collapse already underway, is a long-running historiographical argument. Soviet-era and émigré accounts differ sharply; treat monocausal versions with suspicion. Desertion figures for 1917 are unreliable.
Could serve: Schelling (S4) — focal points and common knowledge; War (S1); Coercion (S4).
Media: The text of Order No. 1 is short and translated; it reads well aloud. Extensive 1917 newsreel and photography.
Source: Petrograd Soviet, Order No. 1 (1 March 1917), translated text widely available; Allan K. Wildman, The End of the Russian Imperial Army, 2 vols (Princeton UP, 1980, 1987).

Spithead and the Nore, 1797

What happened: In April and May 1797 the Channel Fleet at Spithead refused to sail, demanding a pay rise, better provisions and the removal of specific abusive officers, while undertaking to put to sea if the French came out. The Admiralty conceded, around fifty officers were removed, and a royal pardon was granted. A second mutiny at the Nore from May into June, under Richard Parker as "President of the Delegates of the Fleet", added further demands, attempted to blockade the Thames, collapsed as ships defected back one by one, and ended with Parker and some twenty-nine others hanged.
Why it's a collective action case: Two mutinies a month apart with near-identical starting conditions and opposite outcomes, which makes it as close to a controlled comparison as this material gets. Spithead held because the demands were narrow and shared across every ship, the fleet's delegates met and coordinated in advance, and the crews publicly guaranteed to fight the French — a credible commitment that limited the scope of the action and kept the participants defensible. The Nore failed because demands expanded beyond what all crews had a common interest in, so the coalition's incentive-compatibility broke and individual ships found it worth slipping away. Defection from a mutiny is itself a collective action problem, and the Nore is what its unravelling looks like.
Could serve: Olson (S3); Schelling (S4), on credible commitment; War (S1).
Media: Royal Museums Greenwich holds the records and has a research guide; the delegates' printed demands survive. No reliable film treatment.
Source: Royal Museums Greenwich, Research Guide B8, "The Spithead and Nore mutinies of 1797"; N. A. M. Rodger, The Command of the Ocean (Allen Lane, 2004).

Kiel, 1918, and Invergordon, 1931

What happened: On 29–30 October 1918 crews of the German High Seas Fleet at Wilhelmshaven refused to raise steam for a sortie against the Grand Fleet ordered by the naval command without government sanction; the refusal spread to Kiel on 3 November, became a sailors' and workers' rising, and fed directly into the German revolution and the armistice. On 15–16 September 1931 around a thousand ratings of the British Atlantic Fleet at Invergordon refused duty over pay cuts of up to 25% for the longest-serving men; four capital ships could not sail. The cuts were reduced, and the resulting run on sterling contributed to Britain leaving the gold standard on 21 September.
Why it's a collective action case: Ships are the natural experiment for mutiny because a crew is a closed, permanently co-present group — high observability, no exit, forced repetition — which is exactly the condition list for sustained cooperation. Both cases show the same mechanism: refusal expressed as collective inaction rather than violence, which lowers the individual penalty for any single participant and makes it impossible to identify a first mover. At Invergordon the ratings organized in the canteen, cheered rather than issued orders, and committed no violence, which left the Admiralty with no ringleaders to try. Contrast with the Nore, where naming a president gave the state a target.
Contested: The causal weight of Invergordon in the sterling crisis is often overstated; it was one pressure among several. Kiel's relationship to the wider German revolution is a large literature in itself.
Could serve: Olson (S3); War (S1); The reframe (S2).
Media: Kiel is well photographed and is covered in the German revolution literature; the Kiel mutiny memorial exists. Invergordon: Admiralty records and contemporary press.
Source: For Kiel, Daniel Horn, The German Naval Mutinies of World War I (Rutgers UP, 1969); for Invergordon, Anthony Carew, The Lower Deck of the Royal Navy 1900–39 (Manchester UP, 1981), and Alan Ereira, The Invergordon Mutiny (1981).

Conscription, free riding and who fights

Substitution and commutation in the American Civil War

What happened: The Union Enrollment Act of March 1863 made men aged 20–45 liable to conscription but allowed a drafted man either to furnish a substitute or to pay a $300 commutation fee and be excused from that draft call. In the 1863 draft, of 88,171 men held to service, over 52,000 paid commutation and around 26,000 provided substitutes; fewer than 10,000 of those drawn actually entered the army. Commutation was repealed in July 1864. The Confederacy ran a parallel system of substitution and exemptions, including the 1862 provision exempting one white man on plantations with twenty or more slaves. "A rich man's war and a poor man's fight" was the contemporary slogan, and the New York draft riots of July 1863 followed the first drawing.
Why it's a collective action case: A state that has identified a public good and then legislated a price for opting out of contributing to it. The numbers are the entry's value: given a legal exit at $300, most of those eligible took it, and the share who simply served was under 12% of those held. Free riding is not a moral failure here — it is a checkbox on a form, which is exactly the course's point that defection needs no villain. It also shows the second-order cost: a buyout priced in money converts a shared obligation into a class sorting mechanism, and the legitimacy damage was severe enough to force repeal within sixteen months.
Contested: Historians disagree on how unequal the burden actually was — estimates of the wealth profile of substitutes and of who paid commutation vary, and some work argues the "rich man's war" claim is overstated for the Union as a whole.
Could serve: The games (S2) — free riding, with real numbers; Coercion (S4); Changing the payoffs (S4), as a case where pricing the obligation broke it.
Media: Draft wheels survive in museum collections (the Philadelphia draft drum is one). Extensive Harper's Weekly illustration of the draft and the riots.
Source: Enrollment Act, 12 Stat. 731 (3 March 1863); James W. Geary, We Need Men: The Union Draft in the Civil War (Northern Illinois UP, 1991); Iver Bernstein, The New York City Draft Riots (Oxford UP, 1990).

Mercenary versus citizen armies: Machiavelli and the condottieri

What happened: In The Prince (ch. 12–13) and the Art of War, Machiavelli argued that mercenaries are "useless and dangerous" — disunited, ambitious, undisciplined, brave among friends and cowardly against enemies — and that a prince must have his own citizen troops. He cites Francesco Sforza, a condottiere who took Milan from his employers in 1450. Machiavelli organized a Florentine citizen militia in 1506; it was routed at Prato in 1512.
Why it's a collective action case: A named historical argument about exactly the mechanism this section is about — whether cohesion can be bought or has to be grown — and one where the evidence goes largely against the famous version. The mechanism Machiavelli is pointing at is real: a wage is a private payment, so a mercenary's interest in surviving to collect it runs against the line's need for him to stay, while a citizen's stake in the outcome is inalienable. But the historical record does not support the conclusion. Michael Mallett's work shows the Italian condottieri as long-service professionals under increasingly durable contracts, in what became some of Europe's first standing armies, and the "bloodless battle" trope Machiavelli relied on is largely his own polemic. Swiss mercenary infantry in the same period were famous for holding when broken formations elsewhere did not. Machiavelli's own militia losing at Prato is the awkward fact.
Contested: Very. Machiavelli is a participant in the dispute, not a witness to it, and he was writing to advocate a policy he had implemented. Teach it as a claim and its rebuttal, not as a finding.
Could serve: War (S1); Coercion (S4); Changing the payoffs (S4), on the limits of paying for a good rather than organizing it.
Media: The Prince ch. 12 is short and in the public domain. Uccello's Battle of San Romano panels (National Gallery, Uffizi, Louvre) depict condottieri warfare and are excellent slide material.
Source: Niccolò Machiavelli, The Prince (1532), ch. 12–13, and Dell'arte della guerra (1521); Michael Mallett, Mercenaries and their Masters: Warfare in Renaissance Italy (Bodley Head, 1974; repr. with William Caferro, 2009); William Caferro, John Hawkwood (Johns Hopkins UP, 2006).

Gaps worth filling

Noted so the next pass doesn't rediscover them:

The Ottoman janissaries — a slave-soldier corps with barracks living, celibacy rules and a shared kitchen as its symbolic center, which then became a political veto player. Cohesion engineering that outlived its purpose.
The Swiss pike squares and the Landsknechte, where the mercenary-versus-citizen comparison could be tested properly rather than asserted.
Israeli and Finnish reserve systems, where the unit is the same group of men for decades — a modern deliberate test of the repetition mechanism.
Insurgent and irregular cohesion — why some guerrilla groups hold under far worse conditions than regular armies. Jeremy Weinstein's Inside Rebellion (2007) is the obvious starting point and would connect to the selective-incentives module.
Naval press-ganged crews: how the Royal Navy got cohesion out of men taken by force, which is an unusually hard case for every mechanism in this list.