Dominance-Subordination

Abstract

Dominance-Subordination is a type of social dominance: the asymmetric relationship between two individuals in which one reliably takes priority and the other reliably yields. It is the dyadic unit from which group hierarchies are built, and this article treats it at that level — as a property of a pair, read off from patterned deference rather than from fighting. It sets out how ethology defined dominance as a prediction about who defers to whom, how linear hierarchies self-assemble from repeated dyadic contests through winner and loser effects, how rank maps onto glucocorticoid stress physiology, and how the brain learns and represents rank. Three interactive demonstrations let the reader watch a hierarchy crystallize from contests, trace the rank–cortisol gradient as social stability changes, and infer an unseen rank by transitive inference.

Keywords: dominance-subordination, dominance hierarchy, winner and loser effects

The word dominance is most often used of a group — a pecking order, a chain of command, a stratified society. But its primary referent is smaller and sharper: a relationship between two individuals. When an ethologist says that one baboon is dominant to another, the claim is not about aggression in the abstract but about a prediction — that in a contested encounter, the first will take the resource and the second will defer (Drews, 1993). Dominance-subordination is that paired asymmetry, and everything the broader construct does at the level of groups and traits is an aggregation of it.

Social dominance is the parent kind under which MeSH files this relationship; dominance-subordination is its dyadic core. MeSH defines the relationship in exactly these terms — the inter-relationship of two or more individuals wherein one is dominant over the other — and that relational, behaviorally read definition is the one this article keeps in focus. It is a deliberately narrow scope: not the preference for hierarchy that a personality scale measures, nor the two routes (dominance and prestige) by which humans climb a status ladder, but the elementary dominant-subordinate bond those larger phenomena presuppose.

What Dominance-Subordination Is

The relationship is defined by an asymmetry in a dyad. Of two individuals, one — the dominant — reliably gains priority of access to a contested resource (food, a mate, a resting site, space), and the other — the subordinate — reliably yields. The crucial move, made explicit in the ethological literature, is that dominance is not a quantity an animal possesses but a property of the pair: it is read off from the direction of deference in their encounters, and it predicts the outcome of the next contested encounter between them (Drews, 1993). An individual can be dominant to one partner and subordinate to another; the label is always relative to a specific relationship.

This has a counterintuitive consequence that recurs throughout the field. Because the relationship is defined by deference, not by aggression, a settled dominance-subordination bond typically shows less fighting, not more. Once the subordinate reliably yields, there is nothing left to contest, and the overt threat displays that marked the relationship's formation largely disappear (de Waal, 1986). Counting fights is therefore a poor way to find the dominant; counting who yields to whom is the right one. Frans de Waal's work on primates pressed the point further: the dominant-subordinate relationship is embedded in reconciliation, grooming, and coalition, so that dominance and social bonding are integrated rather than opposed.

The relationship is inferred from behavior, and ethologists built an inventory of the behaviors that reveal it: priority of access when a resource is contested, the direction of submissive signals (a subordinate's crouch, averted gaze, or formalized greeting), and displacement — the dominant approaches, the subordinate leaves. Ian Bernstein's critique sharpened the methodology by warning that dominance can become a circular label, invoked to explain the very behaviors used to diagnose it; the discipline's answer was to require that a dominance measure predict behavior in independent contexts rather than merely redescribe it (Bernstein, 1981).

The behaviors ethologists score to diagnose and read a dominance relationship are summarized in Table 1.

Table 1. Behavioral indicators used to read a dominance-subordination relationship.

The first demonstration makes the aggregation visible: it runs repeated pairwise contests among a set of individuals, lets the reader tune how strongly winning and losing feed back on future success, and shows the dominance matrix and its linearity settling out of the dyads.

Hierarchy Formation Lab

Individuals start equal. Each round a random pair contests; the winner’s future edge rises (winner effect) and the loser’s falls (loser effect). Watch a linear order crystallize out of pairwise contests, and read its linearity as Landau’s h.

Linearity (Landau’s h): 1.00  (1.00 = perfectly linear)

Rows and columns are sorted by number dominated, so a near-linear order appears as an upper-triangular block of filled cells. Turning the winner and loser effects up sharpens the order; turning them to zero leaves outcomes near chance and lowers h.

How Hierarchies Form From Dyads

If dominance were simply a readout of fixed individual strength, a group's order would be fully determined before any animal met another: rank the individuals by strength and the hierarchy follows. The striking finding of the experimental literature is that this prior-attributes account is incomplete. Chase and colleagues showed, in controlled studies where fighting ability was equalized or measured in advance, that the resulting hierarchy could not be predicted from individual attributes alone — the social dynamics of the interaction sequence did real work (Chase et al., 2002).

Two feedback mechanisms drive that self-organization. A winner effect raises an individual's probability of winning its next contest after a win; a loser effect lowers it after a loss. Because early outcomes bias later ones, a run of wins can lift an individual above others it did not initially outmatch, and the order can lock in. Layered on top are social dynamics — bystander effects, where animals observe contests they are not in and adjust accordingly, and the simple combinatorics of who encounters whom and in what order. Recent syntheses frame hierarchy formation as the joint product of these feedback and feed-forward mechanisms, which both generate the order and then stabilize it against reshuffling (Dehnen et al., 2022).

The degree to which a hierarchy is linear — a strict transitive order, with no intransitive triads in which A tops B, B tops C, yet C tops A — is itself a measured quantity. Landau's h captures it: it is 1 for a perfectly linear order and falls toward 0 as intransitive triangles accumulate. Linearity matters because a strictly transitive hierarchy is one an observer can represent compactly as a single rank axis, and, as the last section shows, both animals and humans appear to do exactly that.

Rank, Stress, and the Body

Because a subordinate yields priority of access and absorbs the costs of the relationship, it is natural to expect subordination to be physiologically expensive — and in many species it is. But the mapping between rank and stress physiology is not fixed, and untangling it is one of the clearer lessons of field endocrinology. Robert Sapolsky's long-term study of wild baboons established that basal glucocorticoid levels track social rank, and that the direction of the relationship depends on how rank is acquired and maintained and on how stable the hierarchy currently is (Sapolsky, 2005).

In a stable hierarchy where dominance is settled, subordinates — who have least control, least predictability, and least access — often carry the higher chronic glucocorticoid load. But the picture inverts under instability: when a hierarchy is reorganizing, or in species and troops where high rank must be continually enforced through costly aggression, dominants can show the higher stress load instead. The comparative review across species makes the general claim precise: it is not rank per se but the psychological and physical burdens attached to a given rank in a given social system that set the glucocorticoid profile (Sapolsky, 2004). This is why the health gradient associated with social status in humans cannot be read off from position alone without asking what that position entails.

The second demonstration traces this gradient: it lets the reader set an individual's rank and the stability of its hierarchy and reads off the relative basal glucocorticoid index, showing the inversion as a hierarchy moves from stable to contested.

Rank–Stress Gradient

An individual’s basal glucocorticoid (stress hormone) index as a function of its rank and how stable the hierarchy is. In a stable order subordinates carry the higher load; as the hierarchy becomes unstable and rank must be enforced, the gradient inverts onto dominants.

Glucocorticoid index: 112

At the 80th-percentile rank in a fully stable hierarchy (f = 0.80, s = 1) the index is 100 + 60 × 0.20 = 112; make the same rank fully unstable (s = 0) and it rises to 100 + 60 × 0.80 = 148 — the inversion the comparative literature describes.

Learning and Representing Rank

A dominance-subordination relationship is only useful to a social animal if it can be learned, stored, and generalized. An individual that must re-fight every partner to discover the order gains nothing from the hierarchy's existence; the payoff comes from inferring relationships never directly observed. Humans and other animals do this through transitive inference: having learned that A outranks B and B outranks C, they infer that A outranks C without ever seeing the pair (Qu et al., 2017).

Transitive inference about rank shows the same signatures as transitive inference about any ordered dimension. The symbolic-distance effect — judgments are faster and more accurate for pairs further apart in rank than for adjacent pairs — indicates that the discrete pairwise facts have been organized into a continuous mental representation of the ordering, along which distance is meaningful. Kumaran and colleagues used functional imaging to dissociate the neural computations involved, finding distinct mechanisms for updating a hierarchy as new outcomes arrive and for representing one's own and others' positions within it, and showing that a neural rank signal tracks the learned structure (Kumaran et al., 2016). The integrative account that has emerged treats social-hierarchy knowledge as structured knowledge handled by the same learning and representational machinery the brain uses for other relational structures, linked to reward circuitry and to the hormones that modulate status behavior (Qu et al., 2017).

The third demonstration isolates this inference: it trains the reader on adjacent pairs of a rank order, then asks about a never-seen pair, and shows how the symbolic-distance effect falls out of representing rank as a single axis.

Transitive Rank Inference

Training shows only adjacent pairs of the order A > B > C > D > E (A>B, B>C, C>D, D>E). Pick any two individuals — including a pair never shown together — and the learned rank axis yields the answer, faster and more accurately the further apart they are (the symbolic-distance effect).

The bar shows illustrative accuracy above chance (50%). Adjacent pairs (distance 1) sit near chance even though they were trained directly; distant pairs such as A vs E (distance 4) are judged most accurately — the signature that rank is stored as one continuous axis.

Worked Example

Consider four individuals — A, B, C, D — and suppose their pairwise dominance relationships have settled as follows: A is dominant to B and D; B is dominant to C and D; C is dominant to A and D; D is dominant to none. Note the intransitive triad: A tops B, B tops C, yet C tops A. How linear is this hierarchy?

Landau's linearity index is

h = [12 / (n³ − n)] × Σ (Vₐ − (n − 1)/2)²,

where n is the number of individuals and Vₐ is the number of others that individual a dominates. Here n = 4, so the leading factor is 12 / (64 − 4) = 12 / 60 = 0.2, and the expected rank term (n − 1)/2 = 3/2 = 1.5.

Tally the wins: Vₐ = 2 (A beats B, D), V_b = 2 (B beats C, D), V_c = 2 (C beats A, D), V_d = 0. The squared deviations are (2 − 1.5)² = 0.25 for each of A, B, C, and (0 − 1.5)² = 2.25 for D. Their sum is 0.25 + 0.25 + 0.25 + 2.25 = 3.0.

So h = 0.2 × 3.0 = 0.60.

For comparison, a perfectly linear order over four individuals has wins V = {3, 2, 1, 0}, giving squared deviations (1.5)² + (0.5)² + (0.5)² + (1.5)² = 2.25 + 0.25 + 0.25 + 2.25 = 5.0 and h = 0.2 × 5.0 = 1.00. The single intransitive triad among the top three has pulled the hierarchy's linearity down from 1.00 to 0.60 — a concrete measure of how far this group departs from a clean rank axis, and the same quantity the first demonstration reports as contests accumulate.

Discussion

Holding dominance-subordination at the dyadic level clarifies several debates that become muddled when the term is used only of groups or traits. The ethological insistence that dominance is a relationship, not a possession, is what licenses the aggregation step: because each dyad yields a direction of deference, a group of dyads yields a matrix, and the matrix yields a hierarchy whose linearity can be measured. The trait and group-level constructs — who tends to win across many partners, or who prefers that their group rank above others — are summaries over these relationships, and they inherit their meaning from the dyadic level rather than replacing it.

The construct also carries a standing methodological warning. Bernstein's critique — that dominance can slide into a circular explanation — remains the right discipline: a dominance measure earns its keep only when it predicts behavior in contexts other than the ones used to score it, which is why the field distinguishes dominance rank (a position) from the behaviors used to infer it and from the resources it predicts. The self-organization results compound the caution: because hierarchies form partly through social dynamics rather than fixed attributes, the same set of individuals can settle into different orders, and rank is not a simple proxy for underlying quality.

Scope is the final point. This article has deliberately stayed below the level of its parent, social dominance, which treats the dominance–prestige distinction, social-dominance orientation, and the way power reshapes human cognition. Those are aggregations and elaborations of the dyadic relationship described here; the dominant-subordinate bond is the primitive they are built from.

Current Directions

The most active front is methodological and computational. Inferring a hierarchy from sparse, noisy, time-varying interaction data — the normal situation in the field — has moved from heuristic win-proportion scores toward explicit dynamic models; Strauss and Holekamp's framework for inferring longitudinal hierarchies formalizes how a dominant-subordinate order changes over time rather than treating it as a fixed snapshot, using decades of spotted-hyena data (Strauss & Holekamp, 2019). In parallel, the synthesis of feedback and feed-forward mechanisms has given hierarchy formation a mechanistic vocabulary that unifies winner/loser effects, bystander effects, and self-reinforcing resource advantages within a single generative account (Dehnen et al., 2022).

On the neural side, the interdisciplinary program set out by Qu and colleagues continues to link the cognitive representation of rank to its endocrine and reward substrates, treating social-hierarchy learning as a case of structured relational learning and asking how the brain reconciles a learned rank axis with the hormones that modulate status-seeking (Qu et al., 2017; Kumaran et al., 2016). The open questions are shared across levels: how stable a learned hierarchy representation is when the real order shifts, and how the dyadic computations scale to the large, overlapping networks of relationships that real social groups maintain.

Common Misconceptions

The dominant individual is the one that fights most.

The opposite is usually true of a settled relationship. Dominance is read from who yields, and once the subordinate defers reliably there is little to fight about, so overt aggression falls. High fighting rates often signal an unsettled or contested relationship, not a clearly dominant one (de Waal, 1986).

Dominance is a fixed quality an animal has.

Dominance is a property of a pair, not of an individual in isolation. The same animal can be dominant to one partner and subordinate to another, and the group order can emerge partly from the dynamics of who meets whom, not from fixed attributes alone (Drews, 1993; Chase et al., 2002).

Subordinates always carry the most stress.

Which rank bears the higher glucocorticoid load depends on the social system. In stable hierarchies subordinates often do, but under instability, or where high rank must be continually enforced, dominants can show the higher chronic stress instead (Sapolsky, 2004).

A dominance hierarchy is just an aggression ranking.

A hierarchy is an order over relationships defined by deference, and its usefulness is that it can be learned and generalized by transitive inference. It need not be perfectly linear, and its linearity (Landau's h) is a measured property, not an assumption (Dehnen et al., 2022).

Key Researchers

Ivan D. Chase

(Stony Brook University). Sociologist whose controlled experiments showed that dominance hierarchies emerge from the social dynamics of sequential contests — winner and loser effects, bystander effects, triad formation — rather than from pre-existing individual differences alone.

ORCID - Faculty page

Jean-Claude Dreher

(CNRS, Institut des Sciences Cognitives Marc Jeannerod, Lyon). Cognitive neuroscientist whose integrative work maps the brain systems that learn and represent social-dominance relationships, linking hormones, reward circuitry, and hierarchy processing.

ORCID - Google Scholar

Kay E. Holekamp

(Michigan State University). Behavioral ecologist whose decades-long study of spotted hyena clans advanced methods for inferring the dynamics of dominance relationships over time, including maternal rank inheritance in matrilineal hierarchies.

ORCID - Wikipedia

Dharshan Kumaran

(University College London / Google DeepMind). Cognitive and computational neuroscientist who identified distinct neural mechanisms for updating and representing rank when humans learn a social hierarchy, bringing dominance-subordination into the reinforcement-learning framework.

Google Scholar - Wikipedia

Robert Sapolsky

(Stanford University). Neuroendocrinologist whose long-term field studies of wild baboons established how an individual's rank in its dominance relationships maps onto glucocorticoid stress physiology and health.

Wikipedia - Faculty page

Frans de Waal

(Emory University; 1948–2024). Primatologist whose studies of chimpanzees and bonobos reframed the dominant-subordinate relationship as integrated with reconciliation, coalition, and social bonding rather than reducible to aggression.

Wikipedia

Glossary

Bystander effect (dominance).

The adjustment of an animal's future contest behavior after observing a contest it was not party to; one of the social dynamics by which hierarchies self-organize.

Deference.

The yielding of priority by one individual to another; the direction of deference across encounters is what defines which member of a dyad is dominant.

Displacement.

An interaction in which a dominant individual's approach causes a subordinate to leave a resource or location, used as a behavioral index of the dominance relationship.

Dominance hierarchy.

The group-level order that aggregates from many dominant-subordinate dyads, often approximately linear and summarizable as a single rank axis.

Dominance rank.

An individual's position in a dominance order, distinct from the behaviors used to infer it and from the resources it predicts; valid only when it forecasts behavior in independent contexts.

Dominance-subordination.

The asymmetric relationship between two individuals in which one reliably takes priority of access and the other reliably yields.

Glucocorticoid.

A class of steroid stress hormone (cortisol in humans, corticosterone in many other species) whose basal level tracks the burdens attached to an individual's social rank.

Landau's h.

An index of a hierarchy's linearity, equal to 1 for a perfectly transitive order and falling toward 0 as intransitive triads accumulate.

Loser effect.

The reduction in an individual's probability of winning its next contest following a loss; with the winner effect, a driver of hierarchy self-organization.

Priority of access.

The reliable first claim on a contested resource that the dominant member of a dyad enjoys; the operational payoff by which dominance is often measured.

Submissive signal.

A crouch, averted gaze, or formalized greeting by which a subordinate expresses deference directly, often the most reliable single marker of the subordinate role.

Symbolic-distance effect.

Faster, more accurate rank judgments for pairs further apart in an order, taken as evidence that discrete pairwise facts are represented along a continuous mental axis.

Transitive inference.

Deducing an unobserved relationship (A outranks C) from observed ones (A outranks B, B outranks C); the mechanism by which an individual generalizes a learned hierarchy.

Winner effect.

The increase in an individual's probability of winning its next contest following a win; with the loser effect, a driver of hierarchy self-organization.

Frequently Asked Questions

What is the difference between dominance-subordination and social dominance?

Dominance-subordination is the dyadic relationship between two individuals, in which one takes priority and the other yields. Social dominance is the broader parent construct that also covers the two routes to rank (dominance versus prestige), the personality-level preference for group hierarchy, and the effects of power on cognition. The dyadic relationship is the unit the broader constructs aggregate.

How do researchers decide which individual is dominant?

By the direction of deference, not the amount of aggression. The dominant is the one that reliably gains priority of access to contested resources and that the other yields to or displaces; submissive signals and displacement interactions are scored across many encounters. A dominance measure is considered valid only when it predicts behavior in contexts other than the ones used to score it.

Why does a stable dominance relationship show less fighting?

Because the relationship is defined by the subordinate yielding. Once deference is reliable, contests become unnecessary, because the subordinate concedes priority in advance, so overt threat and fighting fall. High rates of aggression usually mark a relationship that is still being settled, not a clearly dominant one.

Are dominance hierarchies determined by physical strength?

Not entirely. Controlled experiments show that hierarchies form partly through the social dynamics of sequential contests (winner and loser effects, the order in which individuals meet, and observation of others' contests), so a group can settle into an order not predictable from individual attributes alone.

What is Landau's h?

A numerical index of how linear a hierarchy is. It equals 1 when the order is perfectly transitive (no intransitive triads) and falls toward 0 as triangles in which A tops B, B tops C, yet C tops A accumulate. It lets researchers quantify the difference between a clean rank axis and a tangled one.

Do subordinate animals always have higher stress hormones?

No. In stable hierarchies subordinates often carry the higher basal glucocorticoid load, but the relationship inverts when a hierarchy is unstable or when high rank must be enforced through constant aggression, in which case dominants can show the higher stress. It is the burdens attached to a rank, not the rank itself, that set the profile.

How do animals and people use a hierarchy they have learned?

By transitive inference: having learned some pairwise rankings, they infer the rankings they have never directly observed. Judgments about pairs far apart in rank are faster and more accurate than those about adjacent pairs (the symbolic-distance effect), indicating that the hierarchy is stored as a continuous rank axis rather than as a list of separate facts.

Is dominance the same as aggression or as leadership?

No. Aggression is one behavior that can establish a relationship but is not the relationship itself, which is defined by deference. Leadership, meaning influence over a group's movement or decisions, can coincide with dominance but need not; an individual can be dominant without leading or lead without being the most dominant.

References

Bernstein, I. S. (1981). Dominance: The baby and the bathwater. Behavioral and Brain Sciences, 4(3), 419-429. https://doi.org/10.1017/S0140525X00009614

Chase, I. D., Tovey, C., Spangler-Martin, D., & Manfredonia, M. (2002). Individual differences versus social dynamics in the formation of animal dominance hierarchies. Proceedings of the National Academy of Sciences, 99(8), 5744-5749. https://doi.org/10.1073/pnas.082104199

Dehnen, T., Arbon, J. J., Farine, D. R., & Boogert, N. J. (2022). How feedback and feed-forward mechanisms generate and stabilize dominance hierarchies. Biological Reviews, 97(3), 1210-1230. https://doi.org/10.1111/brv.12838

de Waal, F. B. M. (1986). The integration of dominance and social bonding in primates. The Quarterly Review of Biology, 61(4), 459-479. https://doi.org/10.1086/415144

Drews, C. (1993). The concept and definition of dominance in animal behaviour. Behaviour, 125(3-4), 283-313. https://doi.org/10.1163/156853993X00290

Kumaran, D., Banino, A., Blundell, C., Hassabis, D., & Dayan, P. (2016). Computations underlying social hierarchy learning: Distinct neural mechanisms for updating and representing self-relevant information. Neuron, 92(5), 1135-1147. https://doi.org/10.1016/j.neuron.2016.10.052

Qu, C., Ligneul, R., Van der Henst, J.-B., & Dreher, J.-C. (2017). An integrative interdisciplinary perspective on social dominance hierarchies. Trends in Cognitive Sciences, 21(11), 893-908. https://doi.org/10.1016/j.tics.2017.08.004

Sapolsky, R. M. (2004). Social status and health in humans and other animals. Annual Review of Anthropology, 33, 393-418. https://doi.org/10.1146/annurev.anthro.33.070203.144000

Sapolsky, R. M. (2005). The influence of social hierarchy on primate health. Science, 308(5722), 648-652. https://doi.org/10.1126/science.1106477

Strauss, E. D., & Holekamp, K. E. (2019). Inferring longitudinal hierarchies: Framework and methods for studying the dynamics of dominance. Proceedings of the National Academy of Sciences, 116(18), 8919-8924. https://doi.org/10.1073/pnas.1810384116