In a scene from the 2013 BBC documentary series Africa, a giraffe approaches from a distance, ambling across the golden sand of the Kalahari. “A young male,” narrator David Attenborough announces. The newcomer heads toward another giraffe, Western showdown music warbling on the soundtrack. “The old bull won’t tolerate a rival,” Attenborough warns, as the giraffes begin to clash. “Pushing and shoving, they size each other up. The young rival seems to think he has a chance and attacks.” Moments later he slams his powerful neck into the old male’s, and the fight is on—a bloody battle for territory. “The stakes are high,” Attenborough explains. “To lose means exile in the desert.”

Wildlife documentaries commonly include such footage of animals engaged in aggressive contests. It’s not surprising, given the dramatic scenes that ensue. But have you ever wondered about the decision-making processes that underlie these encounters? We have been lucky enough to devote a large part of our research careers to this fascinating topic. And our work has generated some surprising insights into what animals are thinking when they face off.

Animals compete for resources, such as territory, food and mates. Sometimes these contests are mild and cause no physical harm. Other times they are violent and end in severe injury or death. Ultimately they result in unequal distribution of resources, have major effects on reproductive fitness and thus drive evolution. A creature that gathers information can benefit by avoiding potentially lethal fights with bigger, stronger opponents.

We humans are remarkably skilled at assessing the fighting ability of others and quickly learn to not pick fights with individuals larger than ourselves. In laboratory tests, human subjects are able to accurately gauge the power of males after briefly viewing photographs of their torsos or faces or listening to their voices. The judgment is spontaneous—members of both sexes reach it in less than 50 milliseconds. This ability reflects the importance of making accurate assessments of opponents during human evolution.

Are nonhuman animals as good as we are at evaluating rivals? Documentaries such as the ones Attenborough narrates so eloquently often describe the animals’ motivations in such terms. But relatively few of these species have actually been shown to make these kinds of assessments. In fact, our own research suggests that many creatures use different information when deciding whether or not to compete.

Display of force

Animals typically perform ritualized displays prior to engaging in combat. For example, male deer stags competing for access to females will engage in elaborate “roaring contests” and strut side by side in “parallel walks.” Researchers have commonly interpreted these behaviors as means by which each of the opponents can provide information for the other to assess. If the display can settle the contest, there would be no need to engage in a fight in which injury or even death is likely. It is better to spend energy for a short time so that the opponent that perceives itself as the weaker of the two can withdraw, so the thinking goes. We call this phenomenon mutual assessment, and it is central to a game theory model of fighting known as sequential assessment.

Siamese fighting fish assess one another's fighting ability and show greater aggression toward more formidable opponents. Credit: Jane Burton Nature Picture Library

Game theory is a branch of applied mathematics that was initially developed by economists to model human strategic decision-making. Biologists were quick to spot the utility of game theory for evolutionary biology, with John Maynard Smith and George Price being the first to use this framework for studying animal contests. The sequential assessment model proposes that contests should be easily settled by displays if the opponents differ widely in prowess, with fights occurring only when they are closely matched. As the contest escalates, it will become increasingly costly, but it will also provide increasingly accurate information, and so mutual assessment will continue throughout the contest. The model predicts that the greater the difference in fighting ability between the opponents is, the shorter the contest will be. And indeed for years biologists found exactly this negative relationship in the contests of virtually every species they studied. (To measure fighting ability in contests, biologists use a proxy measure, typically body size or weight.) As a result of this body of work, mutual assessment came to be seen as a fundamental ability of all animals.

In the rush to embrace the notion of a universal capacity for mutual assessment, however, some other interpretations of animal contests went unnoticed for the most part. With mutual assessment, we would expect large losers to persist longer than small ones in contests because the decision of the loser to quit is based partly on the animal’s own size or fighting prowess. And if the loser gathers information about the winner, then it should quit sooner if the winner is large. Although few studies examined these associations, some of them showed the predicted positive relationship between loser size and persistence. But there was a hitch: the link between winner size and fight duration was not different from random. This finding suggested that in these instances the loser had information about itself but not about the opponent. These animals were either unable to gather the information, or the information was too costly to gather, or they chose not to use information that would most likely enable them to make optimal fight decisions. In any case, they were exhibiting self-assessment rather than mutual assessment.

Some of these early examples of self-assessment came from the lab of one of us (Elwood). In 1990 he and his colleagues documented this tactic in amphipods, which are small, shrimplike animals. In this species, males engage in a tug-of-war for females, with one male literally grabbing a female from another male’s clutches. Unsurprisingly, they found that larger males are more successful than their smaller counterparts at making and resisting takeovers. And yet the competitors did not appear to be assessing one another: whereas loser weight and contest duration showed a strong positive relationship, winner weight and contest duration were not linked at all.

The biology community largely dismissed this finding as aberrant. But there were other examples, such as that of Metellina mengei, a species of orb-weaving spider. During contests between males for access to females, the spiders would stop grappling and stretch out their very long front legs, apparently comparing them. They looked for all the world like they were exchanging information. But here again winner size had no bearing on contest length, showing that this display did not affect the spiders’ decisions. The males were unable to evaluate one another, only themselves.

The discovery of self-assessment rather than mutual assessment in the orb-weaving spiders prompted zoologist Phil Taylor, now at Macquarie University in Sydney, to get in touch with Elwood. He was preparing a paper on fights in a species of jumping spider and was surprised to find self-assessment rather than mutual assessment in that animal, too. This contact led to a collaborative investigation into why, if the animals use self-assessment, the most common analysis predicted they would use mutual assessment.

Taylor and Elwood used a computer simulation to model a population of animals engaging in contests using self-assessment rules, in which the loser gathers no information about the winner’s ability. The results showed a negative relationship between size difference and contest duration—the more the opponents differed in size, the shorter the contest—exactly the same relationship predicted for mutual assessment. The reason is that with a large size difference the loser would necessarily be very small, whereas with a small difference the loser is more likely to be somewhat bigger. Thus, if the result is driven only by the loser, but the analysis uses the size difference, then it will appear to support mutual assessment. In other words, the tool that biologists had used for many years to study competing animals could give a false impression of their assessment abilities.

Studies of stalk-eyed flies—bizarre-looking insects whose eyes are situated on the tips of antlerlike stalks that stick out from their heads—illustrate the problem. Male flies compete for food and females. An early study that relied on size difference concluded that these animals compare their eye stalks to determine the winner. Researchers subsequently reanalyzed the original data using the winner and loser size separately with fight duration. This approach showed clearly that the loser uses information about its own size in deciding whether to continue competing but must not have information about winner size, because that factor has no effect on how long the contest lasts.

A positive or nonsignificant relationship between winner size and contest duration, coupled with a positive relationship between loser size or fighting prowess and contest duration, indicates what we call “pure self-assessment”—the participants are deciding whether to compete or retreat solely on the basis of the information they have about themselves. But if we detect a negative relationship of winner size to contest duration, that does not necessarily mean that the loser is gathering information about the winner. Instead another decision process, dubbed cumulative assessment, may be at work. With cumulative assessment, the animals can inflict costs on one another, and the larger the size difference, the greater the costs will be for the smaller contestant, which then gives up as soon as a threshold of costs is reached. It might seem like splitting hairs, but there is a major difference between cumulative assessment and mutual assessment. The former does not involve any direct assessment of the opponent; the contest is settled only after costs have accumulated. The latter does not involve a threshold; rather the information gathered about the opponent and self informs the decision to keep competing or throw in the towel.

Although cumulative assessment and sequential assessment produce the same negative correlation between winner size and contest duration, we have some tools for determining which of the two decision processes animals are using when they compete. First, we can set up contests in the lab wherein participants in each contest are matched for size, but average size varies from contest to contest. If the opponents are using cumulative assessment, the eventual loser knows only its own state and thus large losers should persist for longer. In this case, we would expect to see a positive correlation between average size and duration. In contrast, with sequential assessment the decision is based on relative size difference, and with size matching there is no difference regardless of the absolute pairs. We would thus expect to see no link between average pair size and contest duration if the opponents are using sequential assessment.

We can also use the nature of escalation and de-escalation of the contests to discriminate between the two decision strategies. Animals using cumulative assessment should exhibit phases of escalation interspersed by phases of lower-cost activities. Those using sequential assessment, on the other hand, should progress linearly from low- to high-cost activities.

Fight Club: House crickets (Acheta domesticus) (left) use cumulative assessment to make decisions about fighting. Beadlet anemones (Actinia equina) (middle) employ self-assessment. In contests between common hermit crabs (Pagurus bernhardus) (right), attackers use self-assessment, whereas defenders use mutual assessment. Credit: Maximilian Weinzierl Alamy (left); Sue Daly Nature Picture Library (middle); Alex Mustard Nature Picture Library (right)

Decisions, decisions

The revelation that animals use different forms of assessment when competing, along with the development of research protocols that can discriminate among these strategies, has led to a resurgence of interest in animal contests. Studies of a wide range of species have emerged in the past decade and from them many new examples of creatures that use one or the other of these three main strategies. Interestingly, most of them show self-assessment.

Other studies have shown that some species use a combination of approaches to figure out when to back down from a contest and when to go to the mat. For example, in mangrove killifish, individuals compete over territory. Researchers led by Yuying Hsu of National Taiwan Normal University found that opponents decided whether to fight based on prefight displays. During this phase of the encounter, the larger one opponent was, the more likely the smaller contestant was to back down before the encounter escalated to fighting. Those rivals that were closer in size tended to escalate to fighting. They appeared to get no further information about their opponents after the fight began, however. This strategy, termed switching assessment, seems to be a mash-up of mutual assessment followed by self-assessment.

Our studies of hermit crabs revealed yet another form of decision-making. Hermit crabs salvage the shells of dead snails and use them to protect their delicate abdomen. The crabs will fight for access to a rival’s shell. We found that during these attempted takeovers the opponents get different information depending on their role. Attackers seemed to receive little or no information about defenders, whereas defenders were influenced by the way the attackers fought. Thus, within the same contest one role seemed to use self-assessment, whereas the other used mutual assessment.

The existence of all these forms of assessment raises an intriguing question: What determines which decision-making strategy an animal employs? One possible factor is cognitive ability. Some experts have argued that just knowing one’s own state is simple but that integrating or comparing it with the state of the opponent is more cognitively challenging. This idea remains to be systematically tested, but a quick survey of taxa that differ in their cognitive sophistication provides tentative support for it. For instance, sea anemones have a simple neural network, and analyses of their fights suggest they use self-assessment. At the other extreme, complex animals with refined perceptual systems, such as cuttlefish, have been found to use mutual assessment.

In line with this pattern, we expect that mammals, with their large, highly developed brains, will use mutual assessment. But few experiments of the kind needed to distinguish among the various assessment models have been carried out on mammals. A mammal for which we do have some experimental data on assessment is the domestic pig. One of us (Arnott) has been working with Simon Turner of Scotland’s Rural College and Irene Camerlink of the University of Veterinary Medicine Vienna to study pig aggression with an eye toward improving the welfare of farmed animals. Pigs naturally form dominance hierarchies. During pig farming it is routine practice to regroup pigs together at various stages of the production cycle. Whenever the animals are regrouped, a period of intense aggression ensues as the animals hash out a new hierarchy. These repeated bouts of aggression pose a major welfare issue.

When we took a closer look at this aggression, we determined that pigs use mutual assessment but require prior contest experience to become proficient at it. The next step was to see if we could provide the necessary experience in a manner that avoids costly aggression. To that end, we decided to experiment with manipulating the pigs’ early-life rearing environment. We found that piglets that were allowed to mingle with another litter prior to weaning subsequently developed enhanced social skills that enabled them to have shorter contests when introduced to an unfamiliar individual in later life. Our results suggest that simple early-life socialization may be an effective, practical intervention that farmers can adopt to curb fighting among adult pigs during regrouping.

One more aspect of contests warrants mention in the discussion here. Although cognitive capacity probably helps to determine which kind of assessment an animal uses, it is not the only factor at work. The value of the resource to be won or lost can itself influence decision-making. The shells of hermit crabs are a prime example. During contests over shells, one crab termed the attacker (usually the larger crab) approaches and grasps the shell of the defender, and the defender then withdraws into its shell. The attacker then vigorously hits its shell against the defender’s again and again. This shell rapping, as it is known, ends with either the defender being dramatically evicted or the attacker giving up and retreating empty-handed.

We have found that the crabs consider multiple aspects of shells when determining how hard to fight for them. A key variable is the size of the shell relative to the size of the crab—the ideal size is small enough to carry around with minimal energy expenditure but big enough to accommodate a certain amount of growth. The crabs modify their behavior depending on their assessment of their own shell and that of their opponent. When attackers have poor shells and their opponents have good shells, the attackers are more likely to escalate aggression and take their opponents’ shell; when defenders have poor quality shells, they will oppose the seizure less vigorously.

So next time you are watching a wildlife documentary with animals fighting, you will know there is a lot going on in that interaction. In many cases, though—as in that of the giraffes—whether the creatures are truly “sizing each other up” remains to be determined, despite what the narrator may tell you.