Hawk–dove game

The Hawk–dove game is a model of animal conflict in which alternative strategies fare differently according to their opponents and the costs and benefits of contests. Its central lesson is that selection need not favor either aggression or retreat absolutely: a stable balance can emerge, and more conditional strategies can produce a better account of behavior than the stark two-strategy contrast alone.

The model and its equilibrium

In the simple version, individuals behave either as Hawks or Doves. The resulting evolutionary stable state can be expressed as a stable ratio of Hawk genes to Dove genes in the gene pool, a genetic condition called stable polymorphism. The same equilibrium can instead arise if each individual can use either behavior on different occasions: in the example given, each contest begins with a random choice biased toward Hawk, with a probability of 7/12, or a 7:5 Hawk-to-Dove bias. The choice must be genuinely unpredictable to an opponent. A fixed sequence, such as playing Hawk seven times and then Dove five times, would let rivals exploit the pattern.

The model is explicitly an idealization, not a claim that nature contains only two fixed kinds of animal. Its value is explanatory: a simple model can isolate a principle, then be elaborated to represent more varied behavior. The broader reasoning is that fighting decisions involve costs as well as potential gains, including injury, lost time, and expended energy; what counts as a good strategy therefore depends on the circumstances of a contest and on the strategies already present in the population.

Conditional strategies

The model is expanded by adding strategies that respond to an opponent. A Retaliator begins like a Dove, using a conventional threat rather than immediately escalating, but retaliates if attacked. It therefore behaves like a Dove against a Dove and like a Hawk against a Hawk. A Bully initially behaves like a Hawk but retreats as soon as an opponent fights back. A Prober-retaliator resembles a Retaliator, except that it sometimes tests whether an opponent will tolerate a brief escalation; it persists if there is no response, retreats to conventional threatening if resisted, and retaliates if attacked.

These strategies make the model less binary. Behavior is no longer a fixed identity but a conditional policy: an animal’s response changes with the other animal’s conduct. The distinction matters because a strategy that appears advantageous against one opponent may be vulnerable against another.

Simulation and implications

When the five strategies are compared in a computer simulation, Retaliator is the only one that emerges as evolutionarily stable. Dove cannot resist invasion by Hawk and Bully; Hawk can be invaded by Dove and Bully; and Bully can be invaded by Hawk. In a population of Retaliators, no alternative strategy does better than Retaliator, although Dove does equally well. Dove numbers could consequently drift upward. If they became appreciable, Prober-retaliators, as well as Hawks and Bullies, would gain an advantage against them.

Prober-retaliator is nearly stable: in a population of Prober-retaliators, only Retaliator does better, and then only slightly. The model therefore suggests that Retaliators and Prober-retaliators might predominate, with a small Dove minority and gentle oscillations in their relative numbers. As with the simpler equilibrium, this need not mean that each individual permanently follows one strategy; an individual could use a mixture of strategies.

From abstraction to animal behavior

The model offers an explanation for the combination of restraint and force described as the “gloved fist” aspect of animal aggression. It does not predict that every contest will follow the same pattern: outcomes depend on the payoffs assigned to winning, suffering injury, and wasting time. In elephant seals, winning can bring near-monopoly rights over a large harem, making the prize valuable and serious injury a substantial risk. For a Great tit feeding nestlings, by contrast, time may be paramount: it needs to catch an average of one prey item every thirty seconds, so time spent in a contest has a different cost.

The model’s significance lies in showing how conflict can be analyzed without treating aggression as an unconditional virtue or retreat as an unconditional weakness. It provides a framework for asking how strategies perform in relation to one another and to their ecological costs, while its simplified assumptions remain a starting point for explanation rather than a literal description of every animal encounter.