Territoriality and Aggression

When male fig wasps, Idarnes spp., hatch inside the fig they attempt to decapitate their brothers that hatch in the same fig, attacking them with large and powerful mandibles (Hamilton 1967). Similarly, male elephant seals (Mirounga angustirostris) may kill rivals during fights over access to females (Hayley 1994) and male fallow deer (Dama dama) employ violent head-on "jump clashes" during the rut at the start of the breeding season (Jennings et al. 2005). In these three examples, aggressive behavior is being used by each rival in order to maximize its chances of success in a conflict over who gets to mate with the available females.

While mating opportunities represent the most highly-prized resources that animals may compete for, there are many other examples where individuals come into conflict over access to a limited resource. In addition to mates, animals may compete over ownership of food and shelter or territories containing these resources. Thus, natural selection has produced numerous adaptations, such as weapons or large body size, which enhance resource holding potential (RHP), the ability to defend territories and in many cases engage with opponents directly. What is perhaps most remarkable about "aggressive," "agonistic," "fighting," or "contest" behaviors is that the violent examples given above are exceptions rather than the rule. In most cases, contests are settled through the use of non-injurious aggressive behaviors such as displays and trials of strength. Examples of non-injurious fighting include roaring contests in red deer (Cervus elephas) (Clutton Brock & Albon 1979) claw waving in male fiddler crabs (Uca mjoebergii) (Morrell et al. 2005) during contests over territory, and "shell rapping" in hermit crabs (Pagurus bernhardus) during contests over the ownership of gastropod shells (Mowles et al. 2010) (Figure 1).

How can non-injurious aggressive behaviors — especially displays — induce one opponent to give up and relinquish a valuable resource? One argument is that "dangerous fighting would harm the species" (e.g., Huxley 1966) but if aggressive behavior is a result of natural selection, avoiding dangerous fighting "for the good of the species" is unlikely. Rather, we should expect to see individuals act to maximize their own fitness, regardless of the effect on the species. In order to understand how non-injurious fighting could evolve through natural selection, an approach called game theory, borrowed from the field of economics, has been used to construct models of the evolutionary processes that could result in aggressive behavior. Contests are regarded as "games" played out between two or more alternative strategies. The aim of the model is to determine which of the alternatives should be favored by natural selection. If one strategy, once adopted by most of the population, is immune to invasion by any of the alternatives in the game, it is called the evolutionarily stable strategy (ESS). The first game theory model of fighting is known as the Hawk-Dove game (although in its first iteration it was called the Hawk-Mouse game) (Maynard Smith & Parker 1976). This model pits a Hawk strategy (always try to injure your opponent and only withdraw from the contest if an injury is received) against a Dove strategy (always use a non-injurious display if the rival is another Dove and always withdraw if the rival is a Hawk). In order to find the ESS, the first stage is to specify the average payoffs (that is, the benefit in fitness, denoted as E) for individuals playing Hawk or Dove against opponents playing Hawk or Dove. This is best represented by a payoff matrix (Table 1).

In this very simple model, an individual playing the Hawk strategy against another Hawk is assumed to have a 50% chance of winning and a 50% chance of losing. The average payoff of playing Hawk against another Hawk therefore includes half the value of the resource, from the 50% of contests that they are expected to win. But since they only withdraw when injured we also have to add half of the cost of receiving an injury from the 50% of contests that they will lose. Clearly, an injury will have negative fitness consequences in contrast to the positive fitness consequences gained from winning the resource. The average payoff is therefore ½ of the positive value of the resource (V) plus ½ of the negative value of picking up an injury (C), which can also be written as: E = ½ (V - C). When a Hawk fights a Dove, however, the Hawk will always win with no risk of picking up an injury, so the average payoff to Hawks fighting Doves is V. Conversely, the average payoff to a Dove fighting a Hawk is 0 (zero), because the Dove will always lose but avoid an injury by just running away. For a Dove fighting another Dove, we again assume that there is a 50% chance of winning, but in this case there is no chance of receiving an injury as Doves never try to injure their opponent. The average payoff for playing Dove against another Dove is therefore ½V. Now that we have built the model, we can plug in some values for C and V, or "parameterize the model"). The next step is to calculate the payoffs in each cell of the matrix, given the values of C and V that we have chosen to use. Let’s say that V = 50 and C = -25 (so in magnitude C < V). For Hawk against Hawk E = +12.5, for Dove against Hawk E = 0, for Hawk against Dove E = 50, and for Dove against Dove E = 25. The last step is to find the ESS by "running the model," which involves doing a kind of evolutionary thought experiment:

In an ancestral population where most individuals played Dove, E = +25 but if a mutation for Hawk arose the average payoff for individuals bearing this mutation would be +50. Since this is better than +25, the mutation would quickly spread through the population meaning that Dove is not an ESS. On the other hand, a Dove mutation could not invade a population of Hawks because the average payoff to Hawks fighting Hawks is +12.5, which is better than the payoff of 0 received by a Dove that fights a Hawk. In this run of the model, where the value of the resource outweighs the cost of an injury, we find that natural selection is not expected to favor non-dangerous Dovelike behavior over dangerous Hawk-like behavior. In fact, the examples of injurious or even fatal fighting described above all occur when the benefits of winning the resource are extremely high in comparison to the cost of receiving an injury. In elephant seals, receiving an injury or being killed is obviously not good, but challenging the dominant male for a chance to mate is the only way to achieve any reproductive success. In fitness terms, failing to reproduce is worse than picking up an injury and just as bad as being killed.

However, in a contest over a resource other than a direct chance of securing limited mating opportunities, such as a fight over food, shelter, or territory, it is hard to see how the cost of picking up a serious injury would be as bad as losing the resource. If we re-parameterize the model such that C > V (say V still = 50, but C is now = -100) the results are different. By plugging these new values into the matrix, we now find that while Hawk could still invade an ancestral population of Doves, Dove can also invade an ancestral population of Hawks! We can try running the model with different ratios between C and V but as long as C > V the overall result is always the same, with neither strategy being immune to invasion by the other. This is called a "mixed ESS." This model predicts that evolution will proceed to an equilibrium where there is a mixture of Dovelike and Hawklike behavior in the population. This could involve some individuals playing Hawk all the time and others playing Dove all the time; or it could mean that all individuals use a mixture of Hawk and Dove behavior. Thus, in situations when C > V, a proportion of contests are predicted by the Hawk-Dove game to be settled through non-injurious fighting.

The extraordinarily influential Hawk-Dove model shows how, in a very simplified version of the world where there is no variation in things like how much each opponent values the resource (subjective Resource Value, RV), non-dangerous fighting could, in theory, evolve under natural selection. The next question, however, is how exactly can activities that don’t directly harm the opponent induce it to make a decision to withdraw from the contest, or "give up"? An intriguing feature of non-injurious aggressive behavior is that the encounters are often protracted. The contest may progress though a series of ever more demanding phases. For example, in red deer rival males start off by performing bouts of roaring vocalizations, then perform parallel walking, and only lock antlers and wrestle if the contest reaches a highly-escalated phase. Furthermore, within a phase contestants often perform the same behavior over and over again. For example, hermit crabs perform repeated bouts of "shell rapping," where an attacking crab repeatedly hits its gastropod shell against that of a defending crab. Allocating all this time and effort to the contest seems counter-intuitive, in that we might expect the weaker individual to give up quickly if it will eventually lose anyway. So various models have been proposed to explain how natural selection could produce such protracted and repetitive contests. These include the Sequential Assessment Model and the Energetic War of Attrition among others, but the differences among them really come down to different assumptions about how the weaker individual makes the decision to give up. One possibility is for each individual to gather information about the RHP of their opponent, compare this to their own RHP, and give up as soon as they know they are weaker. This is called mutual assessment and forms the core of the Sequential Assessment Model. Another, simpler, option is just to keep fighting until an individual threshold of costs, which could build up either as a result of performing non-injurious agonistic behaviors or receiving injuries, is reached. The first individual to cross its threshold will make the decision to give up through a process of "self assessment," as in the Energetic War of Attrition (for more details of these models, see Payne & Pagel 1997 or Briffa & Sneddon 2010). Although these models make different predictions, in practice determining the type of assessment rule that real fighting animals may be using can be fraught with difficulties; not least because some self-assessment models make predictions similar to mutual-assessment models! The best approach is therefore to obtain detailed behavioral data to enable an analysis of escalation patterns, data on costs such as lactic acid accumulation, and data on the relationships between contest duration and both the absolute RHP of the loser and the difference in RHP between winners and losers.

As with all theoretical models in biology, models of aggression are not intended as exact representations of the real world. Rather, the aim is to see whether an idea — "settling contests through signals can evolve due to the benefits to individuals, not the group" or "the loser gives up when it crosses an individual threshold of cost" — is a reasonable one that could at least work in a very simplified version of the real world. In the field of animal aggression and territoriality, as in other examples of animal behavior, the reality is likely to be far more complex than the simplified situations that can be modeled in a useful way. In killifish, for example, fighting males appear to switch between mutual and self assessment in different phases of the fight (Hsu et al. 2008) and in hermit crabs one of the individuals seems to assess its opponent’s RHP, while the other seems to assess only its own condition (Briffa & Elwood 2002). Nevertheless, aggression occurs in almost all animal taxa and takes a wide variety of forms encompassing aerial displays in butterflies (Kemp 2002), battles between rival armies in ants (Batchelor & Briffa 2010) (Figure 2) through to biting and gouging in elephant seals (Hayley 1994). The fact that many empirical studies of real animal contests are informed by a well-developed body of theory helps us to understand how natural selection can lead to the evolution of aggression in its varied forms.