Ultimatum Game¶

A two-player, one-shot bargaining test in which a proposer splits a fixed stake and a responder may accept or reject for nothing, isolating costly punishment of unfairness by pitting the selfish equilibrium against the offers and rejections people actually make.

Core Idea¶

The ultimatum game is a two-player, one-shot bargaining paradigm — introduced by Werner Güth, Rolf Schmittberger, and Bernd Schwarze (1982) — in which a proposer divides a fixed monetary stake between herself and a responder, who must either accept the proposed split (both receive their shares) or reject it (both receive nothing). The subgame-perfect equilibrium under the assumption of purely self-interested, forward-looking players predicts that the proposer offers the smallest positive amount and the responder accepts rather than receive zero — any positive share is better than none. The empirical finding, replicated across hundreds of laboratory studies and a wide range of cultural settings, is that proposers typically offer 40–50 percent of the stake and that responders reliably reject offers below roughly 20 percent even when doing so costs them a substantial absolute sum. The mechanism the game isolates is costly punishment of perceived unfairness: responders forgo a positive payoff specifically to deny the proposer a reward they deem inequitably large, and proposers — anticipating this — offer near-equal splits to avoid rejection. Ernst Fehr and Klaus Schmidt's (1999) inequity-aversion model formalised this by adding a disutility from payoff differences into the responder's utility function, capturing why rejection is rational given those preferences. The game became foundational in behavioural game theory because it provides the minimal laboratory instrument that separates the standard rational-choice prediction from the social-preference deviation with a single clear empirical test: if responders reject positive offers, the selfish model is falsified for that population. Joseph Henrich and colleagues (2001, 2005) extended the paradigm across small-scale societies — hunter-gatherer, pastoralist, and horticulturalist groups — and found that both the frequency of rejection and the level of offers co-vary with the degree of market integration and collective cooperation in each society, establishing that the fairness threshold itself is culturally calibrated rather than a fixed parameter of human psychology.

Structural Signature¶

Sig role-phrases:

the fixed stake — a known monetary sum to be divided between the two players
the proposer — the player who unilaterally offers a split of the stake
the responder — the player who must either accept (both get their shares) or reject (both get nothing)
the one-shot take-it-or-leave-it scaffolding — anonymous, single-round play that strips out reputation and repetition, leaving rejection as the only response move
the selfish-equilibrium benchmark — the subgame-perfect prediction under purely self-interested play: minimal offer, accept anything positive (something beats nothing)
the rejection-of-positive-offers deviation — the empirical finding that responders reliably reject offers below ~20% at real cost, falsifying the selfish model for that population and measuring its departure
the costly-punishment / inequity-aversion read — rejection modelled as a disutility-from-payoff-difference term that makes punishing an inequitable split rational, not irrational
the culturally calibrated threshold — the rejection bar treated as a quantity co-varying with a society's market integration and cooperative scale, not a fixed psychological constant

What It Is Not¶

Not a real-world mechanism. The ultimatum game is a controlled instrument — a stake, a proposer, a responder, one-shot anonymity — built to isolate a finding, not a force operating out in bargaining. What travels into wage disputes or treaties is the content it measures (fairness, reciprocity, costly punishment), cited as evidence for those patterns; "the ultimatum game" itself stays in the lab.
Not evidence that responders are irrational. Rejecting a positive offer looks like leaving money on the table only under a purely selfish utility function. Add a disutility-from-payoff-difference term and rejecting an inequitable split is the rational maximization of that utility. The result reframes the objective, not the rationality.
Not a measure of altruism or generosity. Proposers' near-equal offers are largely strategic anticipation of rejection, not kindness; the responder's veto is doing the work. Stripping that veto — the dictator-game variant — is the move that isolates pure altruism, and offers fall sharply when it is removed, showing the ultimatum game was never measuring generosity in the first place.
Not a fixed human constant. The rejection threshold is not hard-wired psychology: across small-scale societies it co-varies with market integration and cooperative scale, so the fairness bar is a calibrated quantity set by social conditions, not a universal parameter read off "human nature."
Not described by its subgame-perfect equilibrium. The "offer the minimum, accept anything positive" prediction is the selfish benchmark the game is read against, not an account of how people play. The whole point is the systematic departure from it; treating the equilibrium as a behavioral forecast inverts what the instrument was built to expose.

Scope of Application¶

Because the ultimatum game is a controlled experimental instrument rather than a causal mechanism, it is not bounded by a domain but runs wherever its precondition holds: a valuing human (or human-like) agent who can be seated in the proposer or responder role. The fields below are genuine literal deployments of the same paradigm, returning comparable offer/threshold/punishment numbers; what travels beyond an actual seatable agent is not the game but the findings it measures (fairness, reciprocity, costly punishment), which belong to those parent primes, not to this instrument.

Behavioural game theory / experimental economics — the home laboratory; the canonical minimal test that pits the selfish subgame-perfect benchmark against social-preference behaviour by measuring willingness to bear cost to punish unfairness.
Sibling-paradigm family — the instrument transfers cleanly among its near kin: the dictator game (strip the rejection right to isolate altruism from strategic fairness), the trust game (add a move to isolate reciprocity), and the public-goods game (test cooperation at group scale).
Cross-cultural anthropology — the Henrich et al. small-scale-society studies run the identical game across hunter-gatherer, pastoralist, and horticulturalist groups, operationalizing fairness-norm variation and showing the rejection threshold co-varies with market integration.
Neuroeconomics — fMRI of ultimatum rejection localizes the neural correlates of refusing an inequitable offer (insular-cortex activation), reading the same game's decisions off brain activity.
Evolutionary / cooperation theory — used to argue that strong reciprocity and altruistic punishment co-evolved with human-scale cooperation, the game supplying the behavioural signature.
Negotiation pedagogy — a vivid classroom demonstration that take-it-or-leave-it bargaining destroys surplus when the responder's outside option is to tear up the deal.

Clarity¶

The paradigm's clarifying force is that it makes a previously diffuse dispute — do people actually behave as the self-interested rational-choice model says? — into a single, unmissable, falsifiable test. Fairness-versus-self-interest had been a tangle of competing intuitions with no clean adjudicating instrument; the ultimatum game supplies the minimal scaffolding (one-shot, anonymous, take-it-or-leave-it, a stake to divide) under which the two accounts make contradictory predictions about one observable. The subgame-perfect equilibrium says responders accept any positive offer; the data say they reject offers below roughly a fifth of the stake at real cost. The gap is not noise to be explained away — it is the result the game was built to produce, and its size is a direct measurement of how far a population departs from the selfish benchmark.

In doing so the game sharpens distinctions that ordinary talk of "fairness" runs together. It separates the proposer's strategic anticipation (offering near-equal splits to dodge rejection) from the responder's costly punishment (forgoing a payoff to deny an inequitable reward), letting an experimenter localize where social preferences bite and parametrically vary it. It lets the field ask whether the deviation is best modelled as a taste — Fehr and Schmidt's disutility-from-payoff-difference, which renders rejection rational given those preferences — rather than as irrationality, reframing the anomaly as a question about the utility function instead of a failure of reason. And the cross-cultural extensions make the sharper question explicit: is the rejection threshold a fixed parameter of human psychology, or is it calibrated by a society's market integration and cooperative scale? Naming the instrument is what lets that be settled by measurement rather than asserted.

Manages Complexity¶

Fairness is otherwise a sprawling and slippery object of study: it surfaces in wage disputes, divorce settlements, treaty negotiations, allocations of windfalls, and countless everyday divisions of surplus, each laden with context, history, relationship, and stakes, and any attempt to ask "do people behave as the self-interested model predicts?" across that range drowns in confounds. The ultimatum game compresses the entire question to a single controlled instrument with four moving parts — a fixed stake to divide, a proposer's offer, a binary accept-or-reject, and one-shot anonymity that strips out reputation and repetition — under which the selfish rational-choice account and the social-preference account make flatly contradictory predictions about one observable. Behavioural game theory then no longer needs to track the full texture of each fairness situation; it tracks a small set of quantities the game exposes: the offer fraction the proposer puts on the table, the threshold below which responders reject, and the cost responders will absorb to punish. The qualitative verdict reads off these directly — if responders reject positive offers at real cost, the selfish model is falsified for that population, and the magnitude of the gap between the predicted penny-offer equilibrium and the observed 40–50 percent offers with rejection-below-a-fifth measures exactly how far that population departs from the self-interested benchmark.

That reduction also resolves the otherwise tangled space of explanations into a few separable, tunable parameters. The proposer's strategic anticipation and the responder's costly punishment, ordinarily fused in talk of "fairness," become distinct quantities an experimenter can localize and parametrically vary by changing stakes, information, anonymity, or adding third-party punishment. The deviation itself collapses from "people are irrational" to a single modelling choice — Fehr and Schmidt's disutility-from-payoff-difference term — that, once added to the responder's utility function, makes rejection rational and turns the anomaly into a one-parameter question about preferences rather than an open-ended failure of reason. And the rejection threshold, rather than being a fixed constant of human psychology, becomes a calibrated quantity the analyst can predict from a society's market integration and cooperative scale, after Henrich and colleagues found both offers and rejection co-varying with those variables across hunter-gatherer, pastoralist, and horticulturalist groups. So an enormous, context-saturated domain reduces to: run the minimal game, read off offer, threshold, and punishment-cost, and from that small set settle whether a population is selfish, how strong its inequity aversion is, and what social conditions set its fairness bar — a high-dimensional dispute rendered as a few measurable parameters with a clear falsification branch.

Abstract Reasoning¶

The ultimatum game licenses a distinctive set of moves in behavioural game theory, all generated by its status as the minimal instrument under which the selfish and social-preference accounts make contradictory predictions about one observable.

Diagnostic (falsify the selfish model for a population, and measure its departure, from rejection behavior). The defining move is to read a single observable — whether responders reject positive offers at real cost — as a decisive test between two theories. The subgame-perfect prediction under purely self-interested play says responders accept any positive offer (something beats nothing) and proposers offer the minimum; the observation that responders reliably reject offers below roughly a fifth of the stake falsifies the selfish model for that population. The reasoning then turns quantitative: the gap between the predicted minimal-offer equilibrium and the observed 40–50 percent offers measures how far the population departs from the self-interested benchmark, so the magnitude of the deviation is itself the datum. The inference runs from observed choices (offer fraction, rejection threshold, punishment cost absorbed) to a verdict on which utility function the population is operating under — a model test conducted by reading three numbers off a controlled instrument rather than by argument.

Interventionist (localize where social preferences bite by parametrically varying the game). Because the paradigm has few moving parts — a fixed stake, the proposer's offer, a binary accept/reject, one-shot anonymity — the experimenter can intervene on each and predict how behavior shifts, separating components ordinary talk of "fairness" fuses. Stripping the responder's rejection right (the dictator variant) is predicted to isolate pure altruism from strategic fairness, since the proposer no longer faces punishment; adding a stake increase, an information change, or third-party punishment is predicted to move the rejection threshold or offer in identifiable directions. The move is "change one feature of the minimal game, predict the change in offer or rejection," which lets the analyst localize where social preferences enter — the proposer's strategic anticipation of rejection versus the responder's costly punishment — and tune that locus deliberately rather than observing fairness as an undifferentiated lump.

Boundary-drawing (taste versus irrationality; fixed parameter versus calibrated threshold). The game draws two consequential boundaries. First, it separates a social preference from a failure of reason: rejection is not ruled irrational once a disutility-from-payoff-difference term is added to the responder's utility function, after which rejecting an inequitable offer is the rational maximization of that utility — so the move "responders are behaving irrationally" is out of bounds, replaced by "responders have a different objective." This reframes the anomaly as a question about the utility function, not a breakdown of optimization. Second, it draws the line between a fixed parameter of human psychology and a culturally calibrated one: the cross-cultural extensions establish that the rejection threshold is not a universal constant but co-varies with a society's market integration and cooperative scale, so the move "this fairness bar is hard-wired" is ruled out, and the threshold must be treated as a quantity set by social conditions. Deciding which boundary a given result sits on — taste or irrationality, universal or calibrated — is the interpretive judgment the instrument makes decidable.

Predictive (forecast a population's fairness behavior from its social structure, and a proposer's offer from anticipated rejection). The paradigm licenses two forward predictions. From a society's degree of market integration and collective cooperation, the analyst predicts the level of offers and the frequency of rejection — the fairness threshold becomes something forecast from social variables rather than assumed. And within the game, the proposer's behavior is predicted by backward reasoning from the responder's anticipated rejection: a proposer who expects offers below a fifth to be rejected is predicted to offer near-equal splits to avoid the joint-zero outcome, so the observed generosity is read as strategic anticipation, not altruism. The reasoning runs from social structure to fairness norms, and from anticipated punishment to equilibrium offers, turning a context-saturated phenomenon into predictions keyed to a few measurable parameters.

Knowledge Transfer¶

The ultimatum game is an experimental instrument — a controlled paradigm, not a causal mechanism out in the world — so its transfer has two distinct channels that must be kept separate, and "mechanism within / metaphor beyond" does not straightforwardly apply. The first channel is the instrument itself: the scaffolding (a fixed stake, a proposer's offer, a binary accept-or-reject, one-shot anonymity) transfers literally wherever its precondition holds — namely, wherever there are human (or human-like) agents who can be put in front of it. This is what lets the same minimal game be run across hundreds of laboratories and, in Henrich and colleagues' work, across hunter-gatherer, pastoralist, and horticulturalist societies with stakes scaled to months of income, returning comparable, calibratable numbers (offer fraction, rejection threshold, punishment cost). Within behavioural game theory and experimental economics the instrument also sits in a tight family of sibling paradigms it transfers cleanly among — the dictator game (strip the rejection right to isolate altruism from strategic fairness), the trust game (add a move to isolate reciprocity), the public-goods game (test cooperation at group scale) — and into cross-cultural anthropology (operationalizing fairness-norm variation), neuroeconomics (fMRI of rejection localizing insular activation), and negotiation pedagogy (a vivid demonstration that take-it-or-leave-it bargaining destroys surplus). Across all of these the construct is the same controlled test, meaning the same thing, wherever there is a valuing agent to seat in the proposer or responder role.

The second channel is the game's findings, and here honesty requires marking that what travels is not the paradigm but the structural content it measures — content already named by more general patterns. The load-bearing result ("humans will pay to punish allocations they perceive as inequitable, and anticipating this, allocators offer near-equal splits") really does reach into wage negotiations, divorce settlements, and international treaties, but it reaches as fairness, reciprocity, costly punishment, and the credible threat to destroy joint surplus — the genuine cross-domain mechanisms — not as "the ultimatum game." The tell is in the citation pattern: ultimatum-game results are cited as evidence for strong reciprocity and inequity aversion, not as those patterns themselves, which is exactly the signature of a paradigm rather than a portable structure. So the cross-domain lesson should carry the parent content (fairness preferences, reciprocity, costly punishment, the commitment logic of a credible rejection), while the game's own machinery — the proposer/responder roles, the one-shot take-it-or-leave-it scaffolding, the penny-offer subgame-perfect benchmark it is read against, the dictator/trust sibling variants — stays home as the apparatus that produced the evidence. The boundary to police is therefore twofold: instrument-reach (the game transfers wherever a valuing human agent can be tested, and over-reading begins when "ultimatum game" is invoked for systems with no such agent, no rejection option, and no surplus to destroy) and content-attribution (the durable lesson belongs to fairness/reciprocity/costly-punishment, not to the instrument that revealed them). The full split between the home-bound instrument and the traveling content is drawn in Structural Core vs. Domain Accent.

Examples¶

Canonical¶

Güth, Schmittberger, and Schwarze's original 1982 experiment is the defining run. Take the standard setup with a $10 stake divisible in whole dollars. Purely self-interested backward induction gives a sharp prediction: the responder should accept any positive offer because $1 beats $0, so the proposer, anticipating this, offers the smallest positive amount — $1, keeping $9. What Güth and colleagues found, and what hundreds of replications confirm, is nothing like it: proposers typically offer 40–50% of the stake (near $4–5 on $10), and responders reliably reject offers below roughly a fifth, walking away from $1 or $2 rather than let the proposer keep a split they judge unfair. Rejection leaves both with nothing, so the responder is paying real money to punish inequity.

Mapped back: The $10 to divide is the fixed stake; the player choosing the split is the proposer and the accept-or-reject player is the responder, playing under anonymous single-round one-shot take-it-or-leave-it scaffolding. The "$1 offer, accept anything positive" prediction is the selfish-equilibrium benchmark, and the observed rejection of $1–$2 offers at real cost is the rejection-of-positive-offers deviation that falsifies the selfish model for the tested population.

Applied / In Practice¶

Henrich and colleagues carried the identical instrument into the field, running the ultimatum game across fifteen small-scale societies (2001, 2004) with stakes scaled to a day or more of local wages. The results shattered any notion of a universal fairness constant. The Machiguenga horticulturalists of Peru made low offers, around a quarter of the stake, and almost never rejected them. The Lamalera whale-hunters of Indonesia, whose subsistence depends on large cooperative hunts, offered more than half — hyper-fair splits. In several Papua New Guinea groups, responders even rejected over-generous offers, which carry obligation in a gift-culture. Crucially, both offer levels and rejection rates tracked each society's degree of market integration and the payoffs to cooperation in daily life.

Mapped back: That the same game ran cleanly across foragers, herders, and horticulturalists shows the instrument transferring literally wherever a valuing agent can be seated in a role. The finding that the fairness bar shifted systematically with market integration and cooperative scale is the culturally calibrated threshold — the rejection point treated as a quantity set by social conditions rather than fixed psychology — and the varying rejection-of-positive-offers rates are the same deviation measured across populations.

Structural Tensions¶

T1: Isolating power versus external validity (the minimalism that makes it decisive strips out the world). The ultimatum game's force is that it removes reputation, repetition, relationship, and context — one-shot, anonymous, take-it-or-leave-it — so that the selfish and social-preference accounts make flatly contradictory predictions about a single observable. That surgical minimalism is exactly what makes it a clean falsification instrument. But the features it strips are precisely the ones that dominate the real bargaining situations it is cited to illuminate: wage disputes are repeated and non-anonymous, treaties are relationship-laden and reputational, divorce settlements carry history. The tension is that internal validity is bought at the cost of external validity — the more the instrument isolates costly punishment by deleting the world, the less directly its numbers speak to the world, so importing a lab rejection rate into a repeated, reputational negotiation over-reads a paradigm designed to exclude exactly those dynamics. Diagnostic: Is the claim confined to what the stripped-down one-shot instrument measures, or is it extended to a repeated, non-anonymous, reputational setting the game deliberately removed?

T2: A social preference versus a rescued rationality (making rejection rational by enlarging the utility function). Fehr and Schmidt dissolve the anomaly by adding a disutility-from-payoff-difference term, after which rejecting an inequitable offer is the rational maximization of that utility rather than a failure of reason — a genuine advance that reframes the objective instead of impugning the reasoner. But the move is a general solvent: almost any behavior can be made "rational" by adding the right term to the utility function, so rescuing rationality this way risks emptying it, sliding toward the revealed-preference tautology in which whatever is chosen is by construction utility-maximizing. The tension is that the inequity-aversion reframe both dignifies the behavior and threatens to make rational-choice unfalsifiable — the same enlargement that saves the responder from the charge of irrationality removes the model's capacity to be wrong. Discipline requires the added preference be pinned by independent variation (stakes, anonymity, third-party punishment), not fitted to the rejection it explains. Diagnostic: Is the inequity-aversion term independently constrained by other manipulations, or invoked solely to make the observed rejection come out rational?

T3: Strategic anticipation versus genuine fairness (what near-equal offers actually measure). The headline reading — "people are fair" — fuses two very different things: the responder's costly punishment of inequity, and the proposer's self-interested anticipation of that punishment. The dictator variant, which strips the veto, shows offers collapse, revealing that much proposer generosity is strategic prudence rather than altruism. The tension is that the instrument built to isolate fairness invites misattributing it: a naive read credits proposers with a social preference that is largely fear of rejection, so the same near-equal split can express genuine other-regard or pure strategic calculation, and the game cannot by itself tell which without the dictator comparison. The responder's veto is doing the work; the proposer's number is downstream of it. Diagnostic: Is the proposer's generous offer being read as fairness, when the dictator-game comparison shows it is anticipation of the responder's costly punishment?

T4: A universal instrument versus a culturally calibrated result (running the same game everywhere proves there is no single result). The paradigm transfers literally to any seatable agent — Henrich and colleagues ran the identical game across fifteen small-scale societies with stakes scaled to local wages. But that very portability is what exposes that the central finding is not portable: the Machiguenga made low offers and rarely rejected, the Lamalera offered hyper-fair splits, some Papua New Guinea responders rejected over-generous offers, and offers and rejections tracked market integration and cooperative scale. The tension is that the instrument's universality and the finding's non-universality are two faces of one fact — being able to run the game anywhere is precisely how the field learned there is no universal "ultimatum-game result," only a locally calibrated threshold. Treating any single population's numbers (typically WEIRD undergraduates) as "the human fairness bar" mistakes an instrument reading for a species constant. Diagnostic: Is a measured threshold being reported as a property of the tested population, or generalized to "human nature" the cross-cultural evidence shows to be locally set?

T5: Autonomy versus reduction (a laboratory instrument or the fairness, reciprocity, and costly-punishment primes its findings belong to). The ultimatum game is a specific, named experimental paradigm — proposer/responder roles, one-shot take-it-or-leave-it scaffolding, the penny-offer subgame-perfect benchmark, its dictator/trust/public-goods siblings — and as an instrument it transfers literally wherever a valuing agent can be seated. But it is not a mechanism out in the world: what reaches wage negotiations, treaties, and settlements is the content it measures — fairness, reciprocity, costly punishment, the commitment logic of a credible threat to destroy joint surplus — cited as evidence for those patterns, not as the game. The tension is between an apparatus that earns its own name and permanent place in the lab and the recognition that its durable cross-domain lesson belongs to the parent primes it revealed, not to the paradigm itself. Diagnostic: Resolve toward fairness / reciprocity / costly-punishment when carrying the finding into a real bargaining substrate; toward the ultimatum game when using the controlled instrument to measure a population's departure from the selfish benchmark in situ.

Structural–Framed Character¶

The ultimatum game sits on the framed side of the spectrum — best read as framed-leaning: a controlled experimental instrument constituted by the practice of behavioural experimentation, though one that transfers literally within its domain and whose findings reach the world through genuine parent primes (which keeps it off the pole, where a pure non-portable counterexample like the traveler's dilemma sits). The criteria run mostly framed. It is human_practice_bound in a strong and specific way: the entry is explicit that the game "is not a real-world mechanism" but "a controlled instrument," and it runs only "wherever there are human (or human-like) agents who can be put in front of it" — strip away the practice of seating a valuing agent in a proposer/responder role and there is nothing left; no counterpart operates observer-free in nature. Its institutional_origin is a made thing: Güth, Schmittberger, and Schwarze constructed the paradigm in 1982, and its scaffolding (fixed stake, one-shot anonymity, the penny-offer subgame-perfect benchmark it is read against, its dictator/trust/public-goods siblings) is apparatus invented by behavioural game theory. On vocab_travels it is bounded: the instrument transfers literally as an instrument wherever a valuing agent can be tested, but its operative vocabulary does not float free of the experimental setup, and beyond an actual seatable agent "the ultimatum game" has nothing to name. On import_vs_recognize the tell is precise and the entry names it: ultimatum-game results are cited as evidence for strong reciprocity and inequity aversion, not as those patterns — the signature of a paradigm rather than a portable structure, so cross-domain it is import-of-findings, not recognition of the game. The one criterion that pulls back is evaluative_weight: the instrument itself is evaluatively neutral — it measures a departure from a benchmark rather than convicting anyone — which is what keeps it leaning rather than at the pole (though the content it isolates, fairness, is itself normatively laden).

The portable structural skeleton is single: costly punishment of a distribution judged inequitable, backed by a credible threat to destroy joint surplus, which a forward-looking allocator pre-empts with a near-equal split. That skeleton is exactly what the ultimatum game instantiates from its parent primes — fairness, reciprocity, and costly-punishment (with the commitment logic of a credible rejection): the cross-domain reach — wage negotiations, divorce settlements, treaties — belongs to those umbrellas, which the game's findings are cited as evidence for, not to the paradigm. The instrument's distinctive content — the proposer/responder roles, the one-shot take-it-or-leave-it scaffolding, the selfish-equilibrium benchmark, the dictator/trust variants — is precisely the home-bound apparatus that produced the evidence and stays in the lab. Its character: an evaluatively neutral but thoroughly practice-constituted experimental paradigm, structural only in the costly-punishment-and-credible-threat skeleton it borrows from fairness/reciprocity and stages as a minimal falsification test.

Structural Core vs. Domain Accent¶

This section decides why the ultimatum game is a domain-specific abstraction and not a prime — a case with an unusual shape, because the object is an experimental instrument, so the split is between an apparatus that stays in the lab and findings that travel as their own parent primes.

What is skeletal (could lift toward a cross-domain prime). Strip the laboratory scaffolding and a thin relational structure survives in what the game measures: an agent will pay a real cost to punish a distribution it judges inequitable, backed by a credible threat to destroy joint surplus, which a forward-looking allocator pre-empts with a near-equal split. Stated abstractly that content is fairness, reciprocity, and costly punishment (with the commitment logic of a credible rejection). This skeleton is genuinely substrate-spanning: the willingness-to-pay-to-punish-inequity pattern really does reach wage negotiations, divorce settlements, and international treaties. But it reaches as those parent primes, not as "the ultimatum game" — and it is the content the game isolates, the core it shares, not what makes the paradigm distinctive.

What is domain-bound. Everything that makes the object the ultimatum game in particular is behavioural-experimentation apparatus, and it stays in the lab. The proposer/responder role structure; the fixed stake to be divided; the binary accept-or-reject move; the one-shot anonymous take-it-or-leave-it scaffolding that strips out reputation and repetition; the penny-offer subgame-perfect benchmark the behaviour is read against; the Fehr–Schmidt disutility-from-payoff-difference formalization; and the sibling-paradigm family (dictator, trust, public-goods) that isolate altruism, reciprocity, and group cooperation by controlled variation. The decisive test the entry supplies: the ultimatum game "is not a real-world mechanism" but "a controlled instrument," running only wherever a valuing agent can be seated in a proposer/responder role — remove the constructed apparatus and there is nothing operating in the world, only the fairness/reciprocity content it was built to measure. The instrument is furniture of a specific research practice, invented in 1982, not a mechanism nature runs.

Why this does not clear the prime bar. A prime's vocabulary travels and its cross-domain transfer is recognition of the same mechanism, not analogy. The ultimatum game's transfer runs on two separate channels, and neither makes it a prime. As an instrument it transfers literally — the same minimal game runs across hundreds of labs, across fifteen small-scale societies (Henrich et al.) with stakes scaled to local wages, into neuroeconomics and negotiation pedagogy — but only wherever a valuing human-like agent can be put in front of it, which is instrument-reach, not the substrate-spanning recognition of a mechanism. As findings the durable lesson does travel into real bargaining, but the citation pattern is the tell: ultimatum-game results are cited as evidence for strong reciprocity and inequity aversion, not as those patterns — the signature of a paradigm, not a portable structure. So when the bare structural lesson is needed in a real bargaining substrate — costly punishment of inequity, the commitment power of a credible threat to blow up joint surplus — it is already carried, in general substrate-neutral form, by fairness, reciprocity, and costly-punishment. The cross-domain reach belongs to those parents; the game's distinctive content — the proposer/responder roles, the one-shot take-it-or-leave-it scaffolding, the selfish-equilibrium benchmark, the dictator/trust variants — is exactly the home-bound apparatus that produced the evidence and stays in the lab. The ultimatum game clears the domain-specific bar comfortably as a canonical instrument of behavioural game theory, but its only substrate-spanning content is the fairness/reciprocity/costly-punishment the parent primes already carry; the paradigm is the apparatus that revealed them, not the mechanism itself.

Relationships to Other Abstractions¶

Current abstraction Ultimatum Game Domain-specific

Parents (2) — more general patterns this builds on

Ultimatum Game is part of Subgame Perfect Equilibrium Domain-specific

The Ultimatum Game contains the minimal-offer, accept-any-positive-share subgame-perfect solution as the benchmark its behavioral reading falsifies.
Ultimatum Game is a decomposition of Other-Regarding Preferences Prime

Stripping the one-shot bargaining protocol leaves utility with a signed cross-argument on the other player's payoff, expressed as inequity aversion.

Hierarchy paths (4) — routes to 3 parentless roots

Ultimatum Game → Subgame Perfect Equilibrium → Nash Equilibrium → Equilibrium → Fixed Point

Show alternative paths (3)

Not to Be Confused With¶

Dictator game. The sibling paradigm that strips the responder's rejection right — the proposer simply dictates a split the recipient must accept. Removing the veto isolates pure altruism from the strategic fairness the ultimatum game measures, and offers fall sharply, revealing that much ultimatum-game generosity was anticipation of punishment, not kindness. Tell: can the recipient reject and impose the joint-zero outcome (ultimatum game), or must they accept whatever is given (dictator game)?
Trust game. The sibling paradigm that adds a move to isolate reciprocity: an investor sends some amount, which is multiplied en route, and the trustee then chooses how much to return. It measures reciprocation of trust, not punishment of unfairness; there is no take-it-or-leave-it veto over a fixed stake. Tell: is the second player punishing an inequitable split by rejecting (ultimatum game), or deciding how much of a multiplied transfer to reciprocate (trust game)?
Public-goods game. The sibling paradigm that tests cooperation at group scale: several players each contribute to a common pool that is multiplied and shared, tempting free-riding. It studies multi-player cooperation and (with punishment options) collective sanctioning, not one-shot two-person bargaining over a fixed stake. Tell: is it a two-player divide-a-stake-with-veto test (ultimatum game), or an N-player contribute-to-a-shared-pool test of cooperation (public-goods game)?
Subgame-perfect (selfish) equilibrium. The game-theoretic prediction the ultimatum game is read against — offer the minimum, accept anything positive — not an account of how people play. The whole point of the paradigm is the systematic departure from this benchmark; treating the equilibrium as a behavioral forecast inverts what the instrument was built to expose. Tell: is it the theoretical penny-offer/accept-anything solution under purely selfish players (the equilibrium), or the empirical 40–50% offers and sub-20% rejections that falsify it (the game's finding)?
fairness / reciprocity / costly-punishment (parent primes). The substrate-neutral content the game measures — costly punishment of an inequitable distribution, backed by a credible threat to destroy joint surplus. This is what actually travels into wage disputes, settlements, and treaties (cited as evidence for these patterns), whereas the proposer/responder scaffolding stays in the lab. It is the umbrella, not a peer confusable. Tell: is the lesson the fairness/reciprocity/costly-punishment content operating in a real bargaining substrate (the parents), or the controlled proposer/responder instrument that measures a population's departure from the selfish benchmark (the named game)? (Treated fully in a later section.)

Neighborhood in Abstraction Space¶

Ultimatum Game sits in a crowded region of the domain-specific corpus (1^st percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Mechanism Design & Strategic Bargaining (9 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12