The Evolution of Cooperation, Especially in Humans

Submitted for the Degree of Doctor of Philosophy

Trinity Term 2011

presented by supervised by Claire El Mouden Prof. Stuart West Dr. Andy Gardner

dedicated in memory of my sparkly little cousin dedicated to the memoryCelia of my sp Brennerarkly little cousin

Celia Brenner

Table of Contents

Declaration 4

Acknowledgements 6

Abstract 8

Chapter 1. Introduction. 9

Chapter 2. Nice natives and mean migrants: the evolution of dispersal- dependent social behaviour in viscous populations. 38

Chapter 3. The enforcement of cooperation by policing. 66

Chapter 4. Group competition and the evolution of cooperation in humans. 98

Chapter 5. Reproductive levelling and the evolution of cooperation between non-kin. 118

Chapter 6. Promiscuity and the evolution of cooperative breeding. 133

Chapter 7. Spatial structure and inter-specific cooperation: Theory and an empirical test using the mycorrhizal mutualism 150

Chapter 8. Discussion. 178

Literature Cited 188

Appendix 205

Declaration

The literature review in Chapter 1 is adapted from a forthcoming book chapter: El

Mouden, C, West S, Gardner A, Burton-Chellew M. (2012). ‘What do humans maximise?’, in: Evolution and Rationality: Decisions, Cooperation and Strategic Behaviour,

Samir Okasha and Ken Binmore (eds) OUP. This is my own work; my coauthors provided comments and advice on draft manuscripts.

Chapter 2 is a published paper: El Mouden, C and Gardner, A. (2008) ‘Nice natives and mean migrants: the evolution of dispersal-dependent social behaviour in viscous populations’, Journal of Evolutionary Biology 21, 1480-1491. The original idea for the paper was AG’s, I led the theory and write-up, with significant assistance from AG at all stages; AG wrote the Appendix in response to reviewer comments.

Chapter 3 is a published paper: El Mouden, C., West, S. A. and Gardner, A (2010).

‘The enforcement of cooperation by policing’, Evolution. 64, 2139-2152. The original idea for the paper was developed together with SW and AG. I led the theory and write-up, with assistance from both coauthors. AG wrote the original Mathematica code for the eigenvalue analysis, which I extended.

Chapter 4 is being prepared for publication. I developed the original idea and led the theory and write up. AG and SW provided advice and ideas throughout.

Chapter 5 is in review at Evolution and Human Behaviour as a research paper. I had the original idea for the paper, developed the theory and performed the analysis. AG and SW contributed to the write up.

4 Chapter 6 is a published paper: Leggett, H., El Mouden, C., Wild, G., and West, S. A.

(2011) ‘Promiscuity and the evolution of cooperative breeding’, Proceedings of the

Royal Society, B doi:10.1098/rspb.2011.1627. HL and GW conceived the original idea, I helped formulate the inclusive fitness equation, I led the model analysis and wrote the Appendix, GW conducted the numerical investigation, HL led the write-up, to which SW, GW and I contributed.

Chapter 7 is in press: Verbruggen, E., El Mouden, C., Jansa, J. Akkermans, G.,

Bucking,̈ H., West, S. A. and Kiers, T. (2012) ‘Spatial structure and cooperation:

Theory and an empirical test using the mycorrhizal mutualism’, American Naturalist.

In collaboration with SW, I developed the theory, analyzed and wrote the theory section of the paper. The experimental work and remaining write-up was undertaken by my coauthors.

The Appendix contains my three other publications. Section 1 is a published book review: El Mouden, C. (2011) Life: social to its core. Evolution and Human Behaviour

33(1): 79-80. This is entirely my own work. Section 2 is a published experimental paper: Harrison, F. H. and El Mouden, C. (2011). ‘Money for nothing, cooperation for free: exploring the effects of working for endowments on behaviour in standard economic games’, PLoS ONE 6(11): e27623. FH had the original idea and designed the experiments. I collected one-third of the data for the public goods game and all the dictator game data (together with FH). FH performed the statistical analysis and we both wrote the paper. Section 3 is a published review paper: West, S.A., El

Mouden, C. and Gardner, A. (2011). ‘Sixteen common misconceptions about the evolution of cooperation in humans’, Evolution and Human Behavior 32(4): 231-262.

SW, AG and I developed the paper’s ideas together. SW led the write-up; AG and I wrote some sub-sections and provided extensive comments.

5 Acknowledgements

The work herein benefited enormously from the comments, discussion, collaboration and encouragement of many people, including: Alan Grafen, Alex Kacelnik, Anthony

Edwards, Ashleigh Griffin, Ben Roberts, Charlie Cornwallis, Dan Cornforth, Dan

Nettle, Daniel Rankin, Dave Queller, Denis Roze, Helen Leggett, Freya Harrison,

Geoff Wild, Heikki Helanterä, Joan Strassmann, Joao Alpedrinha, Ken Binmore,

Kevin Foster, Laurent Lehmann, Max Burton-Chellew, Paul Seabright, Peter Taylor,

Peyton Young, Rolf Kümmerli, Samir Okasha, Shakti Lamba and Tobi Kiers.

Ashleigh and Joan, thank you so much for the pro-femina motivational pep talks in

Denmark! You convinced me that I have a future in Science and I would not have applied to Nuffield had I not your voices in my head. I would not have survived the interview committee without Ashleigh, Alan, Kevin and Alex’s invaluable advice.

In particular, from the bottom of my heart, thank you to my supervisors Stu and

Andy. I cannot hope to convey how much your expertise, teaching, patience, understanding and unwavering encouragement these past four years has meant to me. You are, in your own very unique ways, outstanding role models and it’s an honour to consider you both friends. Respect.

I thank the Biotechnology and Biological Sciences Research Council and the

University of Edinburgh for providing me with my Doctoral Training Grant.

Finally, Anass, Mum and Dad thank you for all your love and constant support. I could not have done this without you.

6 7 Abstract

I develop social evolution theory to study the evolution of cooperation as follows: (1)

Many organisms undergo a dispersal phase prior to breeding; I demonstrate that knowing ones dispersal status aids the evolution of helping (by non-dispersers) and harming (by dispersers). (2) Policing driven by group-benefits may be selected to enforce cooperation in human and animal societies. I extend existing theory to show that policing may be harder to evolve that previously thought, but that it is maintained more readily than it evolves. (3) Archeological and anthropological evidence suggests that warfare was prevalent during our evolution. I show that, contrary to previous suggestions, between-group competition can favour any social behaviour (pro-social or anti-social) so long as it helps the group compete, and that such traits can be altruistic or mutually beneficial. (4) Reproductive leveling is analogous to policing; in the human literature there is doubt as to whether it can evolve. I extend my previous work to consider the coevolution of culturally and genetically inherited traits for reproductive leveling and selfishness. I find that cooperation can evolve between non-kin if they share the same culture. (5)

Monogamy is thought to favour the evolution of cooperative breeding. I show that in the simplest case, because of the cost of competition between non-dispersing siblings, the level of promiscuity has little or no effect on the evolution of cooperation. (6)

Spatial structure (limited dispersal) is thought to favour the evolution of inter- specific mutualisms as it aligns the partners’ interests. I consider the case of plant- fungi mutualisms and show that spatial structure can disfavour cooperation if it limits the potential fungal partners available to the plant.

Word Count (excluding references and appendices): 46,525 words

8 Chapter 1. Introduction.

This Chapter provides a background to the field of social evolution and how it is applied to study human behaviour (section 1.1) and an outline of the thesis’s structure (section 1.2).

1.1 Background

Research on human evolution and cooperative behaviour is undertaken in biology, psychology, economics, anthropology, sociology and archaeology. Whilst all these disciplines have important and relevant insights for my research, a comprehensive review would be far beyond the scope of this chapter. Therefore, the background section focuses on the question ‘what do humans maximise?’ as the answer introduces the key concepts of evolutionary theory, gives a flavour of the work being undertaken in other disciplines and neatly illustrates the breath and complexity of this field.

Section 1.1.1 explains the process of natural selection, discusses the purpose of adaptation and explains why cooperation is favoured in nature. Section 1.1.2 considers why organisms are not expected to be perfect or optimal inclusive fitness maximisers. Section 1.1.3 discusses what evolutionary theory predicts humans will maximize and, having established that humans may behave as inclusive fitness maximisers, section 1.1.4 discusses the reasons why people’s actions sometimes appear not to maximize inclusive fitness (or anything else). Finally, section 1.1.5 draws on the ideas developed in the earlier sections to discuss whether or when humans are expected to match the predictions of economic rational choice models.

9 1.1.1 What is the purpose of adaptation?

In crossing a heath, suppose I pitched my foot against a stone, and were asked how the stone came to be there, I might possibly answer that, for anything I knew to the contrary, it had laid there for ever...But suppose I had found a watch upon the ground, and it should be enquired how the watch happened to be in that place, I should hardly think of the answer which I had before given, that, for any thing I knew, the watch might have always been there. Yet why should not this answer serve for the watch, as well as for the stone?

The opening lines of Natural Theology (Paley 1802).

What is the difference between a watch and stone? Unlike stones, Paley explains that watches consist of many parts, and that the form of these individual parts may only be understood when we realize they are all contrived for a common purpose (in this case to tell the time). In other words, unlike stones, watches appear designed. Paley argues that organisms are more like watches than stones by demonstrating that, like a watch, the parts of an organism are made up of related parts, contrived for common purposes. For example, consider the eye; the function of an eye’s retina, cornea, lens, muscles and nerves cannot be understood unless we appreciate these parts are contrived for the common purpose of seeing. So when Paley asked ‘what is the difference between a watch and a stone?’, he had identified the question which lies at the heart of evolutionary biology: how is it that the natural world appears designed?

(Darwin 1859; Williams 1966; Leigh 1971; Maynard Smith 1982; Gardner 2009). Fifty years later, Darwin provided the answer.

Darwin’s (1859) theory of natural selection accounts for both the process and the apparent purpose of adaptation (Gardner 2009). The process is simple: as more individuals are born than survive to reproductive age, any heritable traits associated with increased individual reproductive success will tend to accumulate in

10 populations. As a consequence of this process, Darwin argued, successive generations of organisms will appear increasingly well designed to maximize their reproductive success. Darwin therefore answered Paley’s question in two ways: first, he showed how complex design (adaptation) could result from a natural process; second, he explained what natural design was ultimately for – to maximize

Darwinian fitness.

The Origin of Species is a masterpiece, but its central argument is informal (non mathematical) and it was written before the mechanism of biological inheritance

(genes) was known. This meant Darwin’s ideas were not precisely defined. In the

1930s, Fisher formally linked the dynamics of population genetics to the process of natural selection in his seminal work The Genetical Theory of Natural Selection (Fisher

1930). He formally defined natural selection in terms of changes in gene frequencies and defined Darwinian fitness as an individual’s genetic contribution to future generations. His ‘Fundamental Theorem of Natural Selection’ (Fisher 1930) shows that genes associated with a greater individual fitness are predicted to accumulate in natural populations. Given this result, Fisher concluded that organisms would appear as if they are striving to maximize their Darwinian fitness (see also Grafen

2002, 2007).

Fisher’s theory contained an important caveat; whilst he knew that behaviours may be favoured because of their indirect effect on relatives who share genes, he explicitly chose to ignore this added complexity. It was Hamilton (1964, 1970) who formally incorporated the effect of relatives into fitness, making what is arguably, the most important contribution to evolutionary theory since Darwin. The result was inclusive fitness theory, which provides a better understanding of the process of natural selection, and is the modern view of what natural selection maximizes. An individual’s inclusive fitness may be divided into two components: the first

11 component is direct fitness, which is a measure of an individual’s genetic contribution to future generations via its own reproduction and the second component is indirect fitness, which measures the genetic contribution an individual achieves by affecting the reproduction of related individuals (Figure 1.1; Table 1.1).

Grafen (2002, 2006a, 2007) has confirmed Hamilton’s result more formally, by demonstrating that the phenotypes resulting from the dynamics of natural selection on gene frequencies in a related population, matches the result of an optimization program, where an agent is set to maximize its inclusive fitness within a phenotypic strategy set. Put simply, the fitness that organisms should appear designed to maximize is inclusive fitness.

Table 1.1. An evolutionary classification of social behaviour

Effect on actor Effect on recipient (indirect fitness, b)

(direct fitness, c) + -

+ Mutually beneficial Selfish

- Altruistic Spiteful

A social behaviour refers to any trait that has fitness consequences for another. A Hamiltonian (or evolutionary) classification is determined by the effect the behaviour has on the individual’s average lifetime fitness, not on short-term fitness consequences or the fitness consequences of single interactions. Both mutually beneficial (+/+) and selfish (+/-) traits may evolve between relatives or non-relatives. Altruistic behaviours (-/+) are favoured when the benefits to related individuals outweigh the cost to self. Spiteful traits (-/-) are very rare, because of the need for negative relatedness (where the actor and recipient on average share less genes than two individuals drawn at random from the population).

12

Figure 1.1. Inclusive fitness is the sum of an individual’s direct and indirect fitness (Hamilton 1964). Social behaviours affect the reproductive success of self and others. The impact of the actor's behaviour (white hands) on its own reproduction (white offspring) is the direct fitness effect. The impact of the actor's behaviour (white hands) on the reproductive success of social partners (grey offspring), weighted by the relatedness of the actor to the recipient, is the indirect fitness effect. Inclusive fitness does not include all of the reproductive success of relatives (grey offspring), only that which is due to the behaviour of the actor (white hands). Also, inclusive fitness does not include all of the reproductive success of the actor (white offspring), only that which is due to its own behaviour (white hands). A key feature of inclusive fitness is that, as defined, it describes the components of reproductive success which the actor can influence, and therefore what they could be appearing to maximize. Adapted from West et al. 2007b.

This expanded view of fitness was instrumental in explaining how the diverse forms of sociality present in the natural world evolved (Hamilton 1996). A behaviour is cooperative if it has been selected for, at least in part, because of the beneficial effect it has on others (West et al. 2007b). Explaining the presence of such behaviours appears problematic given that – all else being equal – they reduce the relative fitness of the actor. However, inclusive fitness theory showed that natural selection could favour cooperation or the limitation of conflict under a wide range of conditions. One explanation is that seemingly disadvantageous genes can increase their transmission indirectly by helping other individuals (typically close relatives) that are likely to

13 share the same gene (Hamilton 1964). Yet cooperation also occurs between unrelated individuals and even between different species (Bourke 2011).

The inherent instability of cooperation between non-relatives is often conceptualized with the aid of the Prisoner’s dilemma (Axelrod and Hamilton 1981) or the tragedy of the commons (Hardin 1968), whereby individuals do best by not cooperating

(cheating), no matter what their partners do. When there is no scope for repeated interactions or sanctions, this results in an inevitable outcome (hence the ‘tragedy’) in which all rational actors cheat, even though they would have been better off if they had all cooperated (hence the dilemma). Thus for cooperative behaviour between non-relatives to be evolutionarily stable, it must be favoured by hidden direct-fitness benefits that outweigh any apparent costs and, when confronted with a new instance of sociality, it is the job of evolutionary biologists to elucidate these (Sachs et al. 2004;

Lehmann and Keller 2006; West et al. 2007b).

Over the past 40 years, an extensive literature has developed within evolutionary biology, building upon Hamilton’s work, to explain how natural selection may favour the formation and maintenance of cooperation at all levels of biological complexity, including between non-relatives and for different demographic and population structures (Figure 1.2). Broadly speaking, such cooperation may be favoured for one or other of two reasons: because it is also directly beneficial in the long run to the performer, more so than it is costly (as, for example, in group augmentation effects; Kokko et al. 2001); or because it is enforced, through policing

(Frank 1995, 2003 and see Chapter 3), punishment (Clutton-Brock and Parker 1995;

Gardner and West 2004a; Lehmann et al. 2007c) or sanctioning of cheaters (Kiers et al.

2003), and/or through rewards to cooperators, via mechanisms such as direct or indirect reciprocity (Trivers, 1971). This work illustrates that individuals are not born selfish (contra Dawkins 1976, p.3); they are born as inclusive fitness maximizers and

14 this may favour them to be selfish, spiteful, mutualistic or altruistic depending on the circumstances in which they find themselves (West et al. 2007a).

Figure 1.2. A classification of the evolutionary explanations for cooperation. Direct benefits explain mutually beneficial cooperation which may be between non-relatives, whereas indirect benefits explain altruistic cooperation, which is always between relatives (Hamilton 1964). Within these two fundamental categories, different mechanisms can be classified in various ways (Frank 2003; Sachs et al. 2004; Lehmann and Keller 2006; Bergmuller̈ et al. 2007; West et al. 2007b). The classification shown here is for illustrative purposes only. Not all explanations are included (e.g. group competition, group augmentation and the effect of demography is excluded). A single act of cooperation may require multiple explanations, for example it may evolve due to both direct and indirect fitness benefits, or interactions with relatives could be maintained by both limited dispersal and kin discrimination. Furthermore, it does not show how the relative importance of different mechanisms may shift through time. For example, group members may need to be related for a group to form so the benefits of cooperation are sufficient, but once established, cooperation may remain stable even if group members are unrelated due to enforcement mechanisms (Bourke 2011). In most social species, different explanations will apply in different circumstances. For example, in humans, shared environment mechanisms help explain cooperation between kin, whereas reputation and punishment are important for cooperation between non-relatives. Finally, we divide up facultative enforcement strategies here overly simplisticly; a detailed discussion is beyond the

15 scope of this chapter, and is provided elsewhere (Bergmuller̈ et al. 2007). Adapted from West et al. 2007b.

1.1.2 Adaptation does not imply perfection or optimality.

Watches vary. A stick in the ground can make a crude sundial, accurate to the nearest hour or so on a sunny day, whilst an atomic clock, cooled to near absolute zero, has a margin of error of about one second in 30 million years. Nevertheless, both sundials and atomic clocks are designed to tell the time. Even a broken watch, that does not work at all, shows evidence of design. According to Paley: ‘It is not necessary that a machine be perfect, in order to shew with what design it was made...the only question is, whether it were made with any design at all.’ (Paley 1802, p. 8). Applied to nature, Paley’s point is that living organisms need not be perfect or optimal fitness maximizers to exhibit adaptations for ‘fitness maximizing’ design (Gardner 2009).

The closest thing to perfect timekeeping is the atomic clock. If natural selection could produce its version of a ‘perfect’ watch, it would be a ‘Darwinian Demon’ (Law

1979). This imagined organism could simultaneously maximize all aspects of its inclusive fitness – it would live forever, in any environment, eat anything, be able to reproduce as soon as it was born and give birth to high quality offspring at an infinite rate whilst also helping its relatives produce their offspring at an infinite rate.

Darwinian demons are a fantasy for the same reason that we cannot wear atomic clocks on our wrists: constraints. In nature, phylogenetic, physical, time, developmental, information and resource constraints, mean that organisms are unable to be perfect fitness maximizers. In addition to constraints, organisms must efficiently allocate limited resources to competing concerns. Therefore, from an evolutionary perspective, organisms can never be perfect. However, it is worth asking: when organisms make their complex trade-offs between different aspects of

16 fitness (e.g. growth vs own reproduction vs relatives’ reproduction), can they do so optimally?

Optimality is the achievement of maximum fitness given the constraints an organism faces. Therefore, both atomic clocks and sundials may represent optimal designs in different environments. However, just as a broken watch still shows evidence of design, the presence of fitness-maximizing adaptation does not imply optimality of design. This is because in natural populations, in addition to genes, many other factors influence an organism’s chances of survival and reproduction (i.e. its fitness).

As the opening lines of Fisher (1930) famously read ‘Evolution is not natural selection’. This is because stochastic effects (e.g. unpredictable weather), mutations and population movements all shift gene frequencies, so are evolutionary forces too.

If selection pressures are weak (i.e. most individuals survive and reproduce), these non-selective evolutionary forces can be the main drivers of phenotypic change, not natural selection. As these forces do not maximize anything, they divert phenotypes away from natural selection’s goal of optimality, degrading any appearance of design. Therefore, even though organisms can be optimal, we expect measurably optimal behaviour to be extremely rare. There is no expectation that organisms will reach the goal of optimality, or even be anywhere near it, just that natural selection will direct genetic variation in that direction.

Although adaptation does not imply optimality, it is important to note that mechanisms under strong selection, such as those concerning life or death decisions or commonly encountered situations, will be more finely tuned toward optimality as here, natural selection will be the dominant evolutionary force. Thus we expect to see organisms performing ‘better’ in these situations. For example, as sex ratio theory predicts, a fig wasp will vary the sex ratio of her offspring depending upon the number of other females that lay eggs in the same fruit (Herre 1985, 1987). Studies

17 have demonstrated that the sex ratios the wasps produce are closer to the predictions of evolutionary optimality models in the situations they encounters more frequently

(Herre 1987).

1.1.3 What does evolutionary theory predict humans will maximize?

Many animals have evolved the capacity for phenotypic plasticity in order to cope with variable environments. There are various ways phenotypic plasticity can be achieved. Natural selection may favour a range of phenotypes (Smith and Skúlason

1996; Halama et al. 2001). For example, many fish species have evolved discrete morphs, which feed in different habitats or adopt different mating strategies

(Taborsky 1984; Ehlinger 1989). Alternatively, natural selection may favour the use of proximate cues to help determine optimal behaviour. For example, female great tits use increasing day length, temperature, food availability and social stimulation to time their egg laying with the peak emergence of caterpillars in spring. Studies of a great tit population in Oxford show that since the 1960s, great tits have used these cues to successfully track changes in the temporal availability of caterpillars (Dawson

2008). However, the main way animals achieve behavioural flexibility is through learning.

Learning occurs when an organism changes its behaviour in response to experience

(Pearce 1997). By understanding and remembering the consequences of their actions, animals can react better to the behaviour of others and can cope with changes in their local environment (Sullivan 1988; Hollis et al. 1997; Mahometa and Domjan 2005).

Many animals exhibit surprising abilities at learning. For example, goldfish, chickens, horses, cats and rhesus monkeys, learn equally well to discriminate between two stimuli to gain a reward (Warren 1965). Despite the fact many behaviours are not hard-wired, an animal’s actions will still appear designed to maximize inclusive

18 fitness as its learning is driven by cognitive mechanisms like hunger, libido, pain and fear which are adapted to direct behaviour toward fitness-maximizing goals.

Humans differ from other organisms by degree not by kind; we too derive pleasure from, and direct our efforts toward, fitness-maximizing outcomes (Darwin 1871). We enjoy food, sex, friendship and social recognition and we dislike hunger, pain, fear, failure and ostracism. To illustrate how adaptive preferences can influence our behaviour, we use the literature on sexual preferences as an example. This work shows that men prefer plump women to thin women (81% of people in 58 cultures surveyed; Brown and Konner 1987), and that this preference becomes more extreme when men are hungry (Swami and Tovee 2006), indicating the adaptive explanation may relate to the risk of food shortage (Sugiyama 2004). Men also prefer women with a low waist-to-hip ratio (Singh 1993), which may be adaptive as a low waist-to-hip ratio correlates with offspring with higher cognitive ability (Lasseka and Gaulin

2008). During the few days of the month when women are fertile, they are considered more attractive (Roberts et al. 2004) and male partners are more attentive (Haselton and Gagensted 2006) and possessive (Gangestad et al. 2002). Whilst fertile, women prefer more masculine males (Penton-Voak, et. al. 1999; Gangestad et al. 2005), make more effort to dress attractively (Haselton et al. 2007) and are more flirtatious and sociable (Haselton and Gangestad 2006). Finally, ovulating lap dancers earn, on average, $30 an hour more than menstruating lap dancers (Miller et al. 2007).

The diverse and growing literature on human behaviour shows our adaptive preferences are many and varied and affect all aspects of our lives. We summarize their effect to be, that overall, humans strive to increase personal ‘happiness’, and minimize ‘unhappiness’, although what makes us happy and unhappy changes all the time. We would predict that people are most happy when their conditions are improving, and are less happy when their conditions stagnate, even if they are doing

19 well. This is because the positive emotions associated with happiness are evolved, proximate mechanisms to guide our behaviour toward adaptive ends.

Humans show an array of adaptations to help us pursue our adaptive preferences. In addition to an unrivalled capacity for problem solving and reasoning, what sets humans apart from all other species, is the fact that we have access to an accumulated cultural knowledge. Culture is defined as information capable of affecting individuals’ phenotypes which they acquire from other conspecifics by teaching or imitation (Richerson and Boyd 2004). Many non-human species, including blue tits

(Perrins 1979), chimpanzees (McGrew 2004) and dolphins (Simões-Lopes et al. 1998), have simple systems of cultural transmission where non-genetic, non-species wide behaviours are passed across the generations. However, only humans have the cognitive capacity to benefit from a cumulative culture where skills are learnt, improved and transmitted to others via teaching or imitation (Boyd and Richerson

1985, 2005a; Henrich 2005, 2008). The positive fitness effect of cumulative culture is evident in even simple human technologies such as a 60,000 year-old stone-tipped spear (Pettit 2005). Its design is the product of multiple innovations to the shaft, hafting and point. No modern human could ever arrive in the Savannah and design such a spear on the spot. We heavily rely upon the accumulated cultural knowledge of past generations for our survival.

Humans show many adaptations for acquiring and using cultural knowledge. We are unusually docile (Simon 1990), highly sensitive to expressions of approval and disapproval by parents and peers (Baum 1994; Henrich and McElreath 2003) and most importantly, we excel at high-fidelity imitation (Tomasello 1999). These imitation skills enable humans to perform social learning via imitation and practice

(Whiten and Ham 1992; Heyes and Galef 1996), which ethnographic studies confirm is how we acquire the majority of our skills (as opposed to independently via

20 individual learning which is what animals typically do; Fiske 1999). Furthermore, our cognition adapts to the local environment as some behaviours learnt during our childhood can become hard-wired in the brain, permanently modifying our minds at the sub-conscious level (Quartz and Sejnowski 1997; Quartz 2002). This is evidenced by between-cultural variation in susceptibility to optical illusions, hand-eye coordination and male stress and aggression levels (Segall et al. 1966; Cohen et al.

1996; Henrich 2008). It appears that such traits become permanently fixed by about 20 years of age, and remain fixed for life, even if the individual migrates into a new culture (Segall et al. 1966).

The cognitive adaptations and reservoirs of cultural knowledge which humans posses mean that, compared to other animals, we display a unique capacity for behavioural flexibility and specialization. People learn from their own life experiences and of those around them and use this experience to decide how best to seek out happiness and avoid pain in the circumstances they find themselves in.

Even with identical evolved preferences, as people have different life experiences, we expect them to learn very different behaviours and to possess widely divergent beliefs about how to achieve the same goal. Therefore, behaviours and preferences that appear drastically different on the surface may share a common evolutionary foundation. For example, whilst young men in the West may disrespect their parents, smoke and drive fast cars, young men from other cultures may choose to socialize with elder family members, work long hours and avoid contact with girls; both may be trying to fulfil their desire for social status, but they have learnt to achieve the same goal in very different ways. Similar explanations can sometimes be applied to those that vow to be celibate or to forego material wealth, behaviours often cited as a challenge to evolutionary explanations for human behaviour. Since members of the same family, class, culture or tribal group share many formative experiences in

21 common, it follows that the variation in beliefs within such groups should be far less than the variation found between them.

Humans, like other group-living vertebrates, show adaptations for sociality. This is evidenced by many elegant behavioural economic experiments, which show that we are often willing to cooperate and to punish those who do not cooperate.

Importantly, this includes cooperation or punishment toward strangers in one-shot experimental encounters where there is no way of benefitting from the act (Ledyard

1995; Fehr and Gachter 2002; Fehr and Fischbacher 2003; Gintis et al. 2005). Not surprisingly, these social behaviours have a neurological basis: the brain’s reward centres (e.g. the striatum; Stanfey 2007) activate and make us feel good when we donate money to charity (Moll et al. 2006), observe charitable behaviour (Harbaugh et al. 2007), engage in reciprocal cooperation (Rilling et al. 2002) and in costly punishment (Quervain et al. 2004). There are many possible evolutionary explanations for why we are motivated to perform these social behaviours (Figure

1.2) and a detailed discussion is beyond the scope of this chapter (see Sachs et al.

2004; Lehmann and Keller 2006; West et al. 2007b, 2011; Bourke 2011). The fact that these behaviours are seen as irrational (with respect to an income-maximizing agent) in economic experiments involving positive externalities has generated the false impression that humans are uniquely social organisms. Such an impression feeds from misconceptions about evolutionary theory, the difference between ultimate and proximate questions, and the ways that cooperation evolves (West et al. 2011).

Indeed, if the recent history of experimental social sciences had instead tested for income maximizing behaviour in experiments with negative externalities or direct benefits to cooperation, then the same logic may have resulted in the view that humans as irrationally anti-social (Kümmerli et al. 2010).

22 To explain the adaptive value of social behaviours, we must understand the evolutionary trade-offs and constraints the individual experienced in the environment where the behaviours were selected for and were maintained (Herre

1987). As we do not know exactly what our ancestral environment was like, it is not possible to know whether social adaptations we now see were favoured due to direct or indirect benefits. Therefore, we will probably never be able to quantitatively measure the extent to which these adaptations for sociality are (from an evolutionary perspective) mutually beneficial or altruistic. However we can glean insights into the key selective pressures that operated in the past with experiments that utilize implicit cues of key environmental features (Tooby and Cosmides 1990). For example, the use of eye-spots in experiments have proved to be a salient cue of reputational affects, and people condition their cooperation levels on the presence or absence of such cues

(Bateson et al. 2006; Haley and Fessler 2005; Rigdon et al. 2008; Ernest-Jones et al.

2011).

1.1.4 Why do humans sometime appear not to maximize inclusive fitness (or anything else)?

Behavioural ecologists commonly use the economic tools of optimization theory and game theory, which assume organisms behave as optimal fitness maximizing agents

(Maynard Smith 1982; Davies et al. 2011). Constructing such models entails making assumptions about the constraints and trade-offs facing the individual in order to define the available choices (the strategy set) and the potential payoffs (in terms of fitness) they may receive. Crucially, in order to be testable, the model requires feasibly measurable parameters. This means proxies for fitness (e.g. number of offspring, food intake, number of matings or income) and fitness-based payoffs (e.g. cost = time spent on task, calories used or price; benefit = calories gained, matings, change in social rank or profit) must be used (Parker and Maynard Smith 1990). If the

23 data and model match, the model may correctly capture the adaptive purpose of the behaviour being studied. However, often observations do not fit model predictions and it appears that the individuals are not maximizing their inclusive fitness (or anything else). Indeed, whilst data and models can match well, meta-analyses show that on average, evolutionary or ecological models explain only 2-5% of the variation between natural populations (Møller and Jennions 2002). There are many reasons for this - here we discuss six of them.

First, natural selection acts on average consequences of particular traits. This means that a particular trait may entail some costs, but as long as the cost/benefit analysis is favourable (i.e. on average increases inclusive fitness over the individual’s lifetime), the trait can be favoured. For example, in Hedge Sparrows (Prunella modularis), females often mate with multiple males, so a male cannot be sure which eggs he sired

(Davies et al. 1992). As the males invest in parental care, he must estimate his paternity. This potentially difficult task is achieved via a simple approximation; the amount of uninterrupted access he had with the female (Burke et al. 1989). This rule is favoured despite the fact that about 15% of males estimate their paternity incorrectly and raise chicks of another male (Burke et al. 1989). This fitness cost is carried, as it seems the males are unable to achieve any better with the information they have. As we mentioned above, humans react to a subconscious cue of being watched by behaving more cooperatively. Whilst the cues we respond to usually correlate with being watched, we do make mistakes. For example, if pictures of eyes, or stylised eye-like drawings are displayed, we will contribute more to public goods games, donate more to honesty boxes or litter less (Haley and Fessler 2005; Bateson et al.

2006; Burnham and Hare 2007; Ernest-Jones et al. 2011). Even seeing three dots configured as a down-pointing triangle (like two eyes and a mouth) instead of an up-

24 pointing triangle, is enough to make males offer significantly more in a dictator game

(Rigdon et al. 2008).

The lesson here is that if a few individuals in a population act irrationally, it does not automatically imply that their behaviour is maladaptive or that it needs a specific evolutionary explanation. We expect natural selection to minimize the average cost of errors, but we do not expect it to eliminate them. This may mean an odd individual makes a very costly error, or that everyone who performs the behaviour may incur occasional costs. As these errors in themselves are not adaptive, to seek an explanation for them in isolation is not possible. Psychologists refer to this idea as

‘error management theory’ (Funder 1987; Haselton and Buss 2000; Haselton et al.

2005; McKay and Efferson 2010).

Second, natural selection prefers cheap solutions. In response to the varied constraints organisms face, natural selection tends to favour traits of sundial-like design, by which we mean adaptations that are quick and cheap, but do the job, rather than time consuming, costly and precise. Biologists call these adaptations rules-of-thumb; psychologists and social scientists refer to them as heuristic biases

(Kahneman and Tversky 1972; Tversky and Kahneman 1974; Gilovich et al. 2002;

Gigerenzer and Gaissmaier 2011). For example, the parasitoid wasp Trichogramma uses a rule-of-thumb to estimate the volume of a sphere. This sphere is the egg of its host species. Its estimate must be accurate so that it lays the right number of its own eggs inside – too many and the larvae will run out of food, too few and it reduces its reproductive success. The wasp has reduced this complex calculation to a single proxy measure, the angle her head makes with her front leg when balanced on the egg (Wehner 1987). Since the 1970s, psychologists have discovered that humans rely on a diverse array of heuristic biases for many aspects of our decision-making

(Gilovich et al. 2002; Pohl 2004; Haselton et al. 2005). We use heuristics to make

25 estimates of relatedness, which allows us to avoid incest and direct cooperative behaviours toward those we share genes with. For example, individuals are treated as closer relatives if there was a longer period of association during their childhood

(Westermark 1921; Lieberman et al. 2003; Corriveau and Harris 2009); if they bear a facial resemblance (De Bruine et al. 2008); if they speak the same dialect (Shutts et al.

2009) or smell familiar (Russel 1976; Russel et al. 1983; Porter and Cernoch 1983;

Olsson et al. 2006).

Heuristics are cheap, but they are only adapted to maximize an aspect of fitness in a particular situation. They allow fast decision-making and are accurate in commonly encountered situations, but they can cause us to behave in irrational ways, particularly when faced with novel situations. For this reason, the approximation used by the Tricogramma wasp, is only accurate over the natural range of egg sizes she encounters; if presented with an abnormally small or large host egg, her rule of thumb fails and she would lay too many or too few eggs (Wehner 1987). In humans, the approximations for relatedness we use for incest avoidance cause problems in

China and Taiwan where occasionally parents adopt a female infant and rear her alongside their son to whom she will eventually marry. Compared to marriages where spouses were raised apart, such marriages result in lower fertility and higher divorce rates (Wolf 1995). There is a vast and growing literature detailing how heuristic biases cause humans to make irrational decisions, particularly when faced with complex calculations or unfamiliar environments (Gilovich et al. 2002; Pohl 2004;

Haselton et al. 2005).

Third, the solution to fitness trade-offs can change across time and circumstances.

Behaviours may not appear to maximize anything because different fitness components are traded off against each other and, as our circumstances and life- stages change, these different components will change in importance (Stearns 1992).

26 For example, we may be risk-prone as adolescents, when our desire for social status outweighs our desire to avoid danger and become risk-averse once we have children when our desire to keep our children and ourselves from harm outweighs any desire to be ‘cool’ (Steinberg 2008). Similarly, an old lung or skin-cancer patient will probably wish they had not smoked or used sun beds when young. To an economist, such remorsefulness may appear irrational due to the inconsistency of preferences, but we are not designed to behave consistently, but to make the best of the situations we find ourselves in, continually altering our preferences throughout our lives in response to what makes us happy.

Fourth, novel environments produce non-optimal behaviour. Even behaviorally flexible organisms can only be expected to respond optimally within the range of natural variation they or their species have encountered before. Therefore in unfamiliar environments (such as a laboratory) organisms will make mistakes. In humans, this source of error is variously known as the ‘mismatch hypothesis’ (Hagen and Hammerstein 2006), artifact biases (Haselton et al. 2005) or, in cooperation research, the ‘big mistake hypothesis’ (Boyd and Richerson 2005b). As the physical and cultural environment humans live in is radically different from the one we evolved in we may exhibit traits that are not adapted for our current environment.

For example, in our ancestral environment, individuals with strong preferences for sugary, fatty, and salty foods were favoured, as they sought out higher quality diets.

Today in developed countries, these evolved preferences are contributing to obesity and heart disease (Birch 1999). Similarly, strong sexual desires drove males to find the best mates and more willing mates. Now they also maintain a multi-billion dollar-a-year porn industry. It is not immediately clear whether suffering from heart disease or spending income on pornography is maladaptive or not, but this is not the point, and attempts to find out are often misguided. The point is that if these

27 behaviours are not offset by other benefits then they will be selected against provided the environment remains stable over an evolutionary timescale. For example, it is possible that the factors contributing to heart disease may have inclusive fitness benefits in early life, and that an attraction towards pornography may be a relatively harmless side-effect of a beneficial sex-drive. As we discussed earlier, this idea is particularly relevant when studying cooperation because of the importance that demography and population structure play in the origin and maintenance of sociality.

Fifth, it is hard to measure fitness. Even when a behaviour is adaptive, the parameters being measured may be unable to detect the relationship. For example, behavioural ecologists use proxies of fitness, such as number of offspring, number of mates, longevity, food intake, body size, antler length etc. Better proxies (i.e. those more closely correlated with inclusive fitness) are expected to result in a better fit between the model prediction and the observed behaviours. Models that assume people are designed to maximize their income or social status are like models that assume a stag is designed to maximize its antler length or a bee its nectar intake; they will be accurate only insofar as income, social status, antler length or nectar intake are accurate proxies for inclusive fitness. Furthermore (due to the reasons discussed above) behaviours are more likely to appear mistaken (i.e. the adaptive relationship will be missed), if observed for only a short time, in isolation, in a single individual, or when the individual is in an unfamiliar environment.

Sixth, and finally, not all traits are the target of selection. This means that the existence of such traits, which may be genetically or culturally inherited, cannot be understood with economic tools. For example, a genetic trait of interest may be a by- product of selection for another gene; this applies to fleece colour of Soay Sheep (Ovis aries; Gratten et al. 2007) and red-ginger hair colour in humans (Valverde et al. 1995).

28 In Soay Sheep, the gene for dark fleeces is favoured as it is tightly linked (i.e. physically close on the chromosome) to a gene that is essential for cell protein function (Gratten et al. 2007). Similarly in humans, ginger hair has no adaptive value however the gene for ginger hair also causes pale skin, which is selected for when there is little sunshine as it increases vitamin D synthesis from UVB radiation

(Jablonski and Chaplin 2000). Adaptive explanations may help explain the fundamental desires that drive our culturally inherited behaviours. However, they say little about the cultural activities such as music, literature, sport, fashion and art that we undertake to fulfill those desires. In other words, much of what we value does not have an adaptive explanation so we do not believe that the minutiae of human behaviour can be studied using economic tools.

1.1.5 Is measurably optimal behavior ever expected in humans?

As optimal behaviour is not expected in the natural world, behavioural ecologists do not use economic tools to test whether animals behave optimally. Instead, they use them to make testable predictions that can help them understand why animals show the adaptations they do. Behavioural ecology is often a qualitative science, where evolutionary explanations are established by comparative studies across different conditions or between species (Frank 1998; Davies et al. 2011). Having said that, in the natural world, in vary rare cases optimality models fare much better, and can accurately predict quantitative differences. The best-known case of such ‘as-if’ optimization is that of sex ratio where very simple models can make accurate quantitative predictions (Charnov 1982; West 2009). For example, there is a tight fit between observed sex-ratios in the fig-pollinating wasps and those predicted by theory, which says that as the number of females laying eggs in a fruit increases, the sex ratios in the broods should become less female-biased (West et al. 2000).

29 Why do the models and data match so well in the case of sex-ratios? The first reason is that sex allocation represents a very basic tradeoff (male vs female), so it’s easy to construct models that accurately reflect the constraints and choices available to the organism. Second, the traits can be measured precisely as it is relatively easy to sex offspring and count the number of surviving grandchildren. Third, the sex ratio is very strongly tied to fitness so it is strongly selected for, meaning that natural selection will be the dominant evolutionary force, allowing the trait to get close to the goal of optimality. Finally, the sex ratio is a neat quantitative trait, with the expectation that mutations of small effect can readily arise that will nudge the sex ratio by a small amount, allowing a great precision of adaptation even in a very simple organism like the fig wasp. Cases like the sex ratio indicate when as-if optimality is most likely to be observed: when there is a simple set of choices, where decisions can be precisely quantified, when there are high-stakes and where the individual can control their choice. However it should be noted that even in the case of sex-ratios, deviations from predictions are observed, but these deviations are not random. In fact, as predicted, they negatively correlate with the strength of selection and with the reliability of environmental cues (West et al. 2002c). Other instances of

‘as-if’ optimization in animals are found in clutch sizes and various aspects of foraging behaviour (Davies et al. 2011).

The human capacity for learning, reasoning and behavioural flexibility means that, despite the mistakes we sometimes make, in a wide range of cases we do behave in ways that match the predictions of optimality models. Indeed, we are so good at fine- tuning our real time behaviour toward achieving specific goals that it is typically studied using rational choice models. The maximand for a rational choice model is an individual’s utility. Utility is maximized over short timescales and correlates to desires or wants (Marshall 1920) so can be related to inclusive fitness (Grafen 1998). It

30 is a numerical representation of a preference ranking over a set of alternatives. Given its properties as a universal exchange rate, it is no surprise that in today’s world so many people have decided that making money is an effective way to be happy. This justifies the oft-made simplifying assumption that utility represents a single preference for income maximization.

So long as we model the constraints and choices humans face carefully, rational choice models can accurately describe human behavior, i.e. we do achieve ‘as-if’ optimization of utility in real time in many situations (Tversky and Kahneman 1986).

Like ‘as-if’ optimality in nature, we predict that human behaviour is more likely to quantitatively match the predictions of rational choice models when: the decision can be precisely measured (as is the case for market behaviour); the choices are simple or routine (so there is a clear optimal decision and we do not need to rely on heuristics); the stakes are high (so the individual is motivated to ‘care’ about the outcome); and when the individual it in total control of their choice. It is often hard to give economic explanations for our actions, particularly when they affect others or have other far- reaching consequences. When attempting to model such behaviour, care is needed to avoid unrealistic assumptions, for example by only specifying cognitively feasible concerns. Furthermore, parsimonious explanations are often possible with simpler models that only focus on personal costs and benefits. For example, an experimental participant that acts in a way that benefits or harms others may not be motivated beyond personal concerns; such externalities may be a by-product of experimental design (Kümmerli et al. 2010).

Humans are so good at achieving ‘as-if’ optimization of utility in the way rational choice models predict, it is no surprise that most economic theories assume we behave in this way. However, we do not match the predictions of rational choice models all the time (Sen 1977). This is especially true for non-market behaviour. In

31 the previous section, we discussed why our behaviour may not appear to maximize anything; economists will be familiar with many of these reasons. In addition to these however, the predictions about human behaviour arising from the expectation that individuals will strive to maximize their inclusive fitness will not always match the predictions of models which assume we strive for real-time utility maximization.

Like body weight and antler length, income and happiness are proxies for inclusive fitness and this is the only maximand that the average lifetime consequences of a person’s actions might be consistent with. Furthermore humans are not expected to have consistent preferences during their lives (as rational choice theory assumes).

Instead, preferences vary in importance and will shift over time as an individual’s personal circumstances and knowledge changes. These preference changes may seem irrational from an economic perspective, but they can make sense from an inclusive fitness point of view. Most importantly, whether we appear to solve a particular problem in an economically rational way will depend on the time frame of the problem. Over short time frames, we do maximize our pleasure. However, we are not designed to maximize pleasure in the same way over our lifetime, as what gives us pleasure will continually change. Finally, optimization models, no matter how sophisticated or complex, cannot describe all our behaviour as we are a product of evolution, not just natural selection. In other words, we may be good, but we are not optimally designed.

1.1.6 What do Humans Maximise?

How would an alien biologist that was capable of observing our behaviours and reading our minds sum up what humans maximize? They would probably say that, intentionally, humans do not aim to maximize any one thing over their life times, but do generally try to maximize their pleasure and minimize their pain over short time

32 frames. The alien may note that we receive pleasure and pain from adaptively sensible sources, and that we are very good at learning or inventing new ways to increase our pleasure. They may also note that the happiest people are those whose lives have improved constantly, rather than those who started at the top and just had to stay there. Finally, whilst realizing that humans engage in many instances of cooperation, their energies appear to be mostly directed towards benefitting themselves, their reproductive opportunities, and the well being of their relatives.

Whilst fascinating for many reasons, the alien would conclude that humans, along with all other organisms, are best described as striving to maximize their inclusive fitness over their lifetimes, yet are imperfect and non-optimal, but often in predictable ways.

1.2 Thesis outline

In the first two chapters of this thesis, I develop general theory which provides theoretical overviews of mechanisms that favour the evolution of cooperation: chapter 2 explores how knowing whether one has dispersed or not could effect the evolution of social behaviours and chapter 3 considers the evolution of cooperation in groups via the repression of within-group competition. Given their generality, these results may be relevant to diverse species. In chapters 4 and 5, I focus specifically on extending and developing ideas originally presented in the human literature: chapter 4 examines how competition between groups can select for social behaviours, and chapter 5 extends the model presented in chapter 3 to consider the coevolution of reproductive levelling and competitivness. In the final two chapters, I present the result of collaborative work on specific non-human models. Chapter 6 considers the effect of promiscuity on the evolution of cooperative breeding and chapter 7 combines theoretical and experimental work to consider how partner

33 choice may help plants maintain cooperation in the plant-fungal mutualism. A more detailed summary of each chapter is given below.

Chapter 2: When might social behaviours, which depend on the dispersal status of the actor, evolve? There has been much interest in the evolution of social behaviour in viscous populations (Hamilton 1964). Whilst low dispersal increases the relatedness of neighbours, which tends to promote the evolution of indiscriminate helping behaviour, it can also increase competition between neighbours, which tends to inhibit the evolution of helping and may even favour harming behaviour (Queller

1992; West et al. 2002a). In the simplest scenario, these two effects exactly cancel, so that dispersal rate has no impact on the evolution of helping or harming (Taylor

1992a). In this chapter I show that the dispersal rate does matter when individuals can adjust their social behaviour conditional on whether they have dispersed or whether they have remained close to their place of origin. I find that non-dispersing individuals are weakly favoured to indiscriminately help their neighbours, whilst dispersing individuals are more readily favoured to indiscriminately harm their neighbours.

Chapter 3: Policing is regarded as an important mechanism for maintaining cooperation in human and animal social groups. A simple model providing a theoretical overview of the coevolution of policing and cooperation has been analysed by Frank (1995, 1996b, 2003, 2009), and this suggests that policing will evolve to fully suppress cheating within social groups when relatedness is low. Here,

I relax some of the assumptions made by Frank, and investigate the consequences for policing and cooperation. First, I address the assumption that the individual cost of investment into policing is reduced when selfishness dominates. I find that relaxing this assumption leads to policing being favoured only at intermediate relatedness.

Second, I address the assumption that policing fully recovers the loss of fitness

34 incurred by the group owing to selfishness. I find that relaxing this assumption prohibits the evolution of full policing. Finally, I consider the impact of demography on the coevolution of policing and cooperation, in particular the role for kin competition to disfavour the evolution of policing, using both a heuristic ‘open’ model and a ‘closed’ island model. I find that large groups and increased kin competition disfavour policing, and that policing is maintained more readily than it invades. My results suggest that policing may be harder to evolve than previously thought.

Chapter 4: Conflict, or warfare between groups has been suggested as a potentially important mechanism favouring the evolution of altruism in humans as the cost of being an altruist may be outweighed by the benefits of succeeding in between-group competition (Choi and Bowles 2007; Bowles 2006; 2009). Here, I build upon models by Bowles (2006; 2009) to examine in more detail what forms of sociality are favoured when groups compete. I show that: (1) group competition can favour the evolution of pro-social or anti-social behaviours, what matters is that they improve the group’s ability to compete; (2) even when they are anti-social and reduce the fitness of group mates, the behaviour is still mutually beneficial or altruistic because for group members, the fitness costs of the anti-social behaviour are outweighed by the benefits of increased group survival. In contrast to earlier models, my results suggest that competition between groups alone cannot explain the evolution of pro-social behaviour in humans.

Chapter 5: Reproductive levelling refers to mechanisms where conflict between group members is limited such that resources are equally distributed.

Anthropological evidence suggests that egalitarianism is common in small-scale societies (Fried 1967; Boehm 1982; 1997; 2001, Knauft 1991), which suggests that mechanisms such as reproductive levelling may have been important during human

35 evolution. Previous models of reproductive levelling have assumed its existence, and considered its impact upon the evolution of cooperation, without considering the evolution of reproductive levelling itself (Bowles 2006). Here, I extend the model presented in chapter 3 to examine the co-evolution of traits for costly reproductive levelling and competitiveness. I consider both the genetical and cultural evolution of reproductive levelling and competitiveness and find that: (1) costly culturally- inherited reproductive levelling can be selectively favoured, in such a way that it selects for genetically-inherited self-restraint; (2) reproductive levelling is most strongly favoured when group members are genetically unrelated but share the same culture. Assuming that cultures can evolve in the way my model assumes, my results suggest that reproductive levelling could be a potentially important mechanism to explain how gene-culture coevolution may lead to the evolution of pro-sociality in large, genetically-heterogeneous groups.

Chapter 6: Empirical data suggest that monogamy or low levels of promiscuity have played a key role in the evolution of cooperative breeding and eusociality (Boomsma

2007, 2009; Hughes et al. 2008; Cornwallis et al. 2010). However, from a theoretical perspective, low levels of promiscuity, can favour dispersal away from the natal patch, and have been argued to select against cooperation, in a way that cannot be explained by inclusive fitness theory (Nonacs 2011). Here, I use an inclusive fitness approach to model selection for juveniles to stay and help parents breed in a simple patch-structured population. The model predicts that the level of promiscuity has either no, or a slightly negative influence on selection for helping. This prediction is driven by the fact that staying to help leads to increased competition between relatives for the breeding position - when promiscuity is low (and relatedness is high), the best way to aid relatives is by dispersing to avoid competing with them.

Furthermore, my coauthors found the same results with an individual based

36 simulation, showing that this is not an area where inclusive fitness theory ‘gets it wrong’ (contra Nonacs 2011). My results suggest that whether promiscuity has a negative or positive effect on selection for staying to help will depend upon a large number of biological assumptions. The aim of this chapter is not to argue that low promiscuity disfavours cooperation – there is plenty of empirical evidence that it does. Rather, this model highlights that in nature the benefits of helping must be decoupled from the costs of competition.

Chapter 7: Explaining mutualistic cooperation between species remains a major challenge for evolutionary biology (Herre et al. 1999; Sachs et al. 2004; West et al.

2007b). Why cooperate if defection potentially reaps greater benefits? It is commonly assumed that spatial structure (limited dispersal) aligns the interests of mutualistic partners (Frank 1994; Doebeli and Knowlton 1998; Bever and Simms 2000; Hoeksema and Kummel 2003; Lion and van Baalen 2008; Platt and Bever 2009; Hodge et al.

2010). But is spatial structure consistently positive for cooperation? Here, I examine the role of spatial structure in the maintenance of mutualisms and show that when hosts interact with multiple partners and enforce cooperation at a fine scale, then spatial structure (limited dispersal) can actually disfavour cooperation by limiting the suite of potential partners. I find that the effect of spatial structuring depends on the scale (fine or coarse level) at which hosts reward their partners. My coauthors tested these model predictions using novel molecular methods to track the abundance of competing, closely related, cooperative and less-cooperative arbuscular mycorrhizal

(AM) fungal symbionts on host roots over multiple generations. The results concur with my model predictions as when the soil is mixed to reduce spatial structure, the relative success of the more cooperative AM fungal species increases. This challenges previous suggestions that high spatial structuring is critical in stabilizing cooperation in the mycorrhizal mutualism (Bever et al. 2009).

37 Chapter 2. Nice natives and mean migrants: the evolution of dispersal-dependent social behaviour in viscous populations.

This chapter appears as the following publication: El Mouden, C and Gardner, A.

2008. ‘Nice natives and mean migrants: the evolution of dispersal-dependent social behaviour in viscous populations’, Journal of Evolutionary Biology 21, 1480-1491.

2.1 Introduction

Cooperation abounds in nature, and poses one of the greatest problems for evolutionary biology. If natural selection is the survival of the fittest, then how can individuals be favoured to promote the reproductive success of others? (Hamilton

1996; West et al. 2007a). Despite this apparent problem, cooperation occurs at all levels of biological organisation from within eukaryotic cells, to multi-cellular organisms to complex human and animal societies (Leigh 1977; Buss 1987; Maynard

Smith and Szathmáry 1995).

One explanation for cooperation is provided by kin selection (Hamilton 1963, 1964,

1970; Maynard Smith 1964). Because relatives share genes in common, an individual who improves the reproductive success of her relatives may increase her overall genetic contribution to future generations, and genes that promote cooperation may increase in frequency. Hamilton (1964) suggested two major mechanisms for ensuring that cooperative social behaviours be directed primarily towards high relatedness social partners. Firstly, he suggested that individuals might be able to

38 recognise and act preferentially towards their genealogical kin (kin discrimination).

However, this mechanism may require complicated cognitive faculties, and may be inherently unstable (Crozier 1986; Rousset and Roze 2007), so it appears to be rather limited in its application. Secondly, he proposed that if individuals do not disperse far from their place of origin (population viscosity), then neighbours would tend to be genealogical relatives, and hence even indiscriminate cooperation may be favoured (see also Wright 1945). This requires no cognitive function, nor indeed any perception of the external world, and hence viscosity appears to provide a very general explanation for cooperation at all levels of biological organisation (West, et al.

2007b).

However, later work showed that cooperation is not so readily favoured in viscous populations. Although reduced dispersal increases the relatedness of social partners, which acts to favour the evolution of cooperation, it also increases competition between social partners for space and resources, which acts to disfavour the evolution of cooperation (reviewed by Queller 1992 and West et al. 2002a). In the simplest scenario these two effects exactly cancel, so that there is no net impact of the dispersal rate on the evolution of cooperation (Taylor 1992a, b; see also Wilson et al.

1992). Experimental and observational data support these predictions showing that local competition can reduce or completely negate any effect of limited dispersal on selection for cooperation (West et al. 2001; Griffin et al. 2004; Kümmerli et al. 2009).

Indiscriminate cooperation is favoured in a purely viscous population only as readily as it is favoured in a fully dispersing population.

In addition to disfavouring cooperation, localised competition may also favour harming behaviour (reviewed by Gardner and West 2004b). However, Taylor’s

(1992a) model, which implicitly describes the evolution of indiscriminate harming as

39 well as helping behaviour in viscous populations, again predicts that there is no net impact of dispersal rate. Whilst limited dispersal increases competition between neighbours, promoting the evolution of harming, it also increases the relatedness of neighbours, which inhibits the evolution of harming, and these two effects exactly cancel. Neither indiscriminate helping nor indiscriminate harming are affected by the dispersal rate in this simplest scenario.

Taylor’s (1992a) simple model provides a useful benchmark upon which to rest and conceptualise more elaborate (and potentially more realistic) models in which the relationships between dispersal, relatedness and competition are altered so that limited dispersal can make an impact on helping and harming behaviour. Analytical extensions of the basic model that have been shown to promote the evolution of indiscriminate helping behaviour include the inclusion of overlapping generations

(Taylor and Irwin 2000; Irwin and Taylor 2001; Grafen 2007; Lehmann et al. 2007a;

Taylor, et al. 2007), trans-generational social behaviour and niche construction

(Lehmann 2007), elastic population structure (Taylor 1992b; van Baalen and Rand

1998; Mitteldorf and Wilson 2000), group (or ‘budding’) dispersal (Gardner and West

2006; Lehmann et al. 2006), sex-biased dispersal (Johnstone and Cant 2008), and kin discrimination (Perrin and Lehmann 2001). The evolution of harming has received much less attention (e.g. Gardner et al. 2004a, 2007b; Lehmann et al. 2007a; Johnstone and Cant 2008). A major challenge is to find minimal additions to the basic model that allow social behaviour to be promoted by population viscosity, without invoking unrealistic cognitive and other faculties that will reduce the generality of the mechanism.

In this article, we consider that individuals can assess their dispersal status, i.e. whether they have dispersed or whether they have remained close to their place of origin, and we ask whether they will be favoured to adjust their social behaviour

40 conditional upon this information. Whilst dispersers and non-dispersers do not differ in the resource competition they experience with neighbours, they can differ with respect to their relatedness to their neighbours. Specifically, non-dispersers are generally predicted to be more highly related to their neighbours and dispersers will be less related to their neighbours, relative to Taylor’s (1992a) model for the same degree of local competition. Thus, we predict that helping will more readily be favoured among non-dispersers and harming more readily favoured among dispersers, relative to the case where social behaviour is not adjusted according to dispersal status. This mechanism does not require that individuals assess any aspect of their social partners directly, nor indeed necessarily any aspect of their external environment. Thus, it appears to be rather general in its application.

2.2 Models and Analyses

2.2.1 Basic model

Our analysis is based on the infinite island model analysed by Taylor (1992a; see also

Wright 1931). A summary of the model notation used in this article is provided in

Table 2.1. We consider an infinite population of asexual haploid individuals organised into patches of n individuals, and initially we assume that no social interaction takes place. Individuals produce a large number K >> 1 of offspring, then die, and each offspring disperses to another, randomly chosen, patch with probability m or else, with probability 1-m, remains in its patch of origin. After dispersal, n offspring are chosen on each patch to mature to adulthood, while the rest die, and this returns the population to its original size for the beginning of the next lifecycle.

41 Table 2.1. A summary of symbols used in the analysis

Symbol Definition Symbol Definition

a Fecundity cost of harming for harmer M Mutant

A Personal-cost function for harming n Number of breeding spaces per patch

b Fecundity benefit of helping for N non-disperser recipients

B Recipient-benefit function for helping qX Abundance of class X (Appendix 2.4 only) ˜ Fitness effect of social behaviour on r Coefficient of relatedness between b recipient patch mates (i.e. not including self) c Fecundity cost of helping for helper R Coefficient of relatedness within a patch (i.e. including self) C Personal-cost function for helping t Generation number € c˜ Fitness effect of social behaviour on u Generic non-disperser class (Appendix actor 2.4 only)

cX Reproductive value of class X U Set of all non-disperser classes (Appendix 2.4 only) (Appendix 2.4 only)

€ d Fecundity cost of harming for victims v Generic disperser class (Appendix 2.4 only)

D Victim-cost function for harming V Set of all disperser classes (Appendix 2.4 only)

D disperser w Darwinian fitness

F Fecundity of a focal individual x Helping strategy

Average fecundity of focal patch Mutant helping strategy F xˆ

g Genetic ‘breeding’ value for social Average helping strategy behaviour (Appendix 2.4 only) x € i Number of mutants among a focal y Harming strategy € individual's patch mates K Average number of offspring per yˆ Mutant harming strategy individual in the population € m Dispersal rate y Average harming strategy

€

€

Assuming that all genetic variation in the population is neutral with respect to

reproductive success, we can use the recursion approach of Taylor (1992a) to

42 determine the relatedness structure of the population at equilibrium. We note that

the expected relatedness of any individual in generation t to a randomly chosen

member of her own patch (including herself) is Rt = 1/n + ((n-1)/n)rt, where 1/n is

the probability of choosing herself, in which case the relatedness is 1, and (n-1)/n is

the probability of choosing a different individual, in which case the relatedness is rt,

the relatedness of two different individuals in the same patch. The only way for two

different individuals from the same patch to be related is if neither dispersed, which

occurs with probability (1-m)2. In this case, the two individuals will have an expected

relatedness equal to the relatedness of two individuals chosen at random from that

same patch in the previous generation, Rt-1. This allows us to write down a recursion:

1 n −1 2 Rt = + (1− m) Rt−1, (1) n n

which can be solved for equilibrium (Rt-1 = Rt = R) to yield an equilibrium relatedness € of an individual to her group (including herself) of:

1 R = 2 , (2) n − (n −1)(1− m)

as determined by Taylor (1992a). Substituting this into the expression R = 1/n +((n- € 1)/n)r and solving for r obtains the equilibrium relatedness between different

individuals within the same patch:

2 (1− m) r = 2 , (3) n − (n −1)(1− m)

i.e. the FST of Wright (1943). Note that both these ‘whole-patch’ and ‘others-only’ € relatedness coefficients (R and r; Pepper 2000) are expressed as averages over all

individuals, with dispersers and non-dispersers taken together. However, dispersing

43 individuals always find themselves in a patch containing no relatives other than themselves, and so restricting attention to dispersers only, the equilibrium relatedness is RD = 1/n to the whole patch and rD = 0 to the other individuals in the patch. Noting that R = m RD + (1-m)RN and r = m rD + (1-m)rN, where RN and rN are the whole-patch and others-only relatedness for non-dispersing individuals, we can solve to find that:

n + (n − 1)m(1− m) RN = 2 , (4) n(n − (n − 1)(1− m) ) and:

1− m rN = 2 , (5) n − (n − 1)(1− m)

for non-dispersers. Note that for 0 < m < 1, we have RD < R < RN and rD < r < rN, i.e. dispersers experience a lower relatedness to their social partners, and non-dispersers a higher relatedness to their social partners, than average (Figure 2.1). Taylor (1992a) found that unconditional helping behaviour cannot be favoured if it incurs a net fecundity cost for the actor; the result also extends to unconditional harming behaviour, revealing that this is also disfavoured if the actor incurs a net fecundity cost. In the next two sections, we explore whether and when the increased relatedness experienced by non-dispersers favours helping behaviour that is expressed only by non-dispersers, and whether and when the decreased relatedness experienced by dispersers favours harming behaviour that is expressed only by dispersers.

44

Figure 2.1. Relatedness coefficients in viscous populations. Increased dispersal (m) separates kin and hence reduces the relatedness between patchmates (r) and the relatedness of an individual to the patch as a whole (i.e. including itself; R). These relatedness coefficients correspond to Taylor's (1992a) model of unconditional social behaviour, in which neither costly helping nor costly harming are favoured by selection. Our model considers that nondispersing individuals will experience higher than average relatedness to their patch (rN > r and RN > R) whilst dispersing individuals will experience lower than average relatedness to their patch (rD < r and RD < R), and thus nondispersers may be favoured to exhibit indiscriminate helping behaviour, and dispersers may be favoured to exhibit indiscriminate harming behaviour. Numerical solutions are calculated assuming a patch size of two individuals (n = 2).

2.2.2 Evolution of helping in non-dispersers

We now consider a vanishingly rare genetic mutation that causes the bearer to express indiscriminate, others-only helping behaviour that incurs a relative personal fecundity cost c and gives a relative fecundity benefit b that is shared equally by her n-1 patch mates; these fecundity effects are expressed relative to the baseline fecundity, which is set equal to 1, and we assume b, c << 1. Crucially, we assume that this helping behaviour is expressed conditionally with respect to the individual’s own dispersal status; in particular, we assume that only non-dispersers, who have remained in their natal patch, exhibit helping behaviour.

45 In order for the mutation to be favoured by selection it must, on average, increase the expected fitness of its bearers, i.e. the number of surviving offspring that the individual produces over her lifetime (Price 1970). There are two routes by which a gene may impact on the fitness of its bearer: firstly, by having a direct impact on that individual’s personal reproductive success, and secondly, by having an indirect impact on the individual’s personal reproductive success due to its presence among relatives who socially interact with her (neighbour-modulated fitness; Hamilton 1963,

1964). There are neither direct nor indirect effects of the gene on carriers that have dispersed away from their patch of origin, since they do not express the helping behaviour and they do not encounter any other individuals carrying the mutant gene. So, the condition for the gene to be favoured is if it leads, through direct and/or indirect effects, to an increase in the expected number of surviving offspring (i.e.

Darwinian fitness) of non-dispersers (see Appendix 2.4 for mathematical details). The average fitness in the population is 1, because the population size remains constant from generation to generation, and dispersers and non-dispersers have the same expected fitness, so a condition for the mutant gene to be favoured by natural selection is that the expected fitness of a disperser carrying the gene exceeds 1.

We now proceed to calculate the expected fitness of a non-disperser who is carrying the mutant gene, and this is a function of her fecundity and the intensity of competition experienced by her offspring for one of the n breeding spaces within whichever patch they find themselves in after the dispersal event. The relative (to the population average) fecundity of a mutant non-disperser that shares her patch with i other mutant non-dispersers is:

i F = 1− c + b . (6) MN,i n − 1

46 Thus, the focal individual produces a total of KFMN,i offspring, of which mKFMN,i

disperse while (1-m)KFMN,i remain in the natal patch.

We consider first the fate of the mKFMN,i dispersing offspring of the focal mutant

individual. Each arrives in a separate patch, where they compete with (1-m)nK locals

and mnK other dispersers. As there are n breeding spaces within the patch, the

probability that a particular dispersing offspring of the focal mutant survives

competition to find a breeding space is n/(1+(1-m)nK + mnK) ≈1/K. Thus, the

expected number of survivors among the dispersing offspring of the focal mutant is

mKFMN,i/K = mFMN,i,.

We now consider the fate of the (1-m)KFMN,i non-dispersing offspring of the focal

mutant individual, who compete a breeding space on their natal patch. The

competitors include (1− m)K F MN ,i non-dispersing locals plus mKnF incoming

migrants, where F MN ,i is the average relative fecundity of parents in the focal patch, € which is:

i + 1 F MN,i = 1+ (b − c) . (7) n

Again, there are n breeding spaces available, and so the expected number of

successful non-dispersing offspring due to the focal mutant individual is

(1− m)KFMN,i /((1− m)KF MN,i + mK) which, to first order in b and c, is:

# i # i +1& & € (1− m)%1 − c + b − (b − c)% (( 1− m)( . (8) $ n −1 $ n ' '

By adding the expected number of surviving dispersing and non-dispersing € offspring we can find the expected fitness of the focal mutant non-disperser, which is:

47 i 2 i + 1 wMN,i = 1− c + b − (1− m) (b − c) . (9) n − 1 n

This expected fitness is conditional on there being i other (non-dispersing) mutants in the focal individual’s patch. As outlined above, the condition for invasion of the mutant gene is wMN > 1, where wMN is the average fitness of all non-dispersing mutant individuals, i.e. wMN = EMN(wMN,i). This average fitness obtains:

" i % 2 " i + 1% wMN = 1− c + bEMN $ ' − (1− m) (b − c)EMN $ ' . (10) # n − 1& # n &

Note that i/(n-1) is the proportion of (non-disperser) mutants among the social partners of the focal individual, and so the average of this quantity, taken over all non-dispersing mutants in the population, is equal to rN, the others-only relatedness experienced by non-dispersers. This makes use of the assumption that the mutant is vanishingly rare in the population. Similarly, (i+1)/n is the proportion of (non- disperser) mutants among all the individuals in the focal individual’s patch, including herself; so, averaging this quantity over all non-disperser mutants in the population obtains RN, the whole-group relatedness experienced by non-dispersers.

Making this substitution, we obtain a condition, in the form of Hamilton’s rule, for when the helping mutant will invade from rarity:

2 −c + brN − (1− m) (b − c)RN > 0 . (11)

Because we have assumed weak selection, this condition for increase is frequency independent (Rousset 2004, p80), and therefore it also describes the progress of the mutant gene when it is no longer rare. Although condition (11) has been derived using a neighbour-modulated (or ‘direct’) fitness approach to kin selection, it easily yields an inclusive fitness interpretation. Helping by non-dispersers is favoured

48 when the sum of three components, equal to the inclusive fitness effect of their

helping, is positive. Firstly, their help leads to a personal loss of c offspring. Secondly,

their help leads to b extra offspring for the other individuals on the patch, and these

are valued at rN, the relatedness to patch mates. Thirdly, the net increase of b-c

offspring on the patch leads to enhanced competition for locals, to the extent that

these extra offspring do not disperse and nor do the other locals (i.e. (1-m)2), and this

leads to the competitive exclusion of local offspring (including own offspring) which

are valued by RN, the whole-group relatedness appropriate for non-dispersers.

Condition (11) is expressed in terms of fecundity effects, but can be re-written as ˜ ˜ −c˜ + b rN > 0, in terms of the direct ( −c˜ ) and indirect ( b ) fitness effects of helping. The

2 direct cost to the actor is c˜ = c + (1− m) (b − c)/n and the benefit to the recipient is

2 € € b˜ = b(1− m) (b − c)(n −1)/n ; both fitness effects€ are positive under the assumption b > € c, and so the helping behaviour is altruistic ( c˜ & b˜ > 0; Hamilton 1964). € Note that the relatedness coefficients can be written in the form of equations (4) and

(5); making this substitution into the€ inequality (11) obtains an invasion condition

that is expressed wholly in terms of model parameters (n, m, b and c):

3 n 2 (2 − m) − (n −1)(1− m) b/c > . (12) 1+ m(2n − (n −1)(3− m)m − 3)

The above analysis has determined when a mutation that incrementally increases (or € decreases, if we change the sign of b and c) the level of helping indiscriminate

exhibited by non-dispersers will invade from rarity under the action of natural

selection. The RHS of inequality (12), which we may denote f(n,m), presents a

quantity that must be exceeded by the benefit/cost ratio of helping in order for non-

dispersing individuals to be favoured to indiscriminately help their patch mates. We

49 see that as n increases, the RHS of the inequality becomes larger (∂f/∂n > 0) i.e.

helping is less favoured as the number of individuals per patch increases (Figure

2.2a). We see that helping is most readily favoured at intermediate migration rates

(∂f/∂m < 0 for m < m* there is some threshold 0 0 for m > m*;

Figure 2.2a). At the extreme of a very viscous population, m → 0, the condition for

helping to be favoured in non-dispersers is b/c > 1+(2n-1)n. However, the selection

pressure here is vanishingly small (LHS (11) → 0 as m → 0), and so helping is a

nearly-neutral character in extremely viscous populations and is not expected to be

elaborated by natural selection. At the other extreme of a fully mixing population,

costly helping is never favoured (f → ∞ as m → 1). Numerical investigation reveals

that the optimum migration rate for helping (i.e. m = m*) is less than 0.5, but

converges upon 0.5 as n → ∞, and the condition for helping to evolve at this optimum

is approximately b/c > 4n-2.

We have determined when a small increase or decrease in helping by non-dispersers

will be favoured by natural selection, and we can use this to determine the absolute

level of helping that is expected to evolve, i.e. the evolutionarily stable strategy (ESS;

Maynard Smith and Price 1973). If we assume that expressing a level x of helping

incurs a personal fecundity cost C(x) and an others-only fecundity benefit B(x) (where

B, C << 1; see Appendix 2.4), then the incremental increases in cost and benefit that

result from adopting a mutant strategy xˆ , rather than the resident strategy x are

C(xˆ ) − C(x ) = C #( x )(xˆ − x ) = c and respectively. Substituting these into the invasion € condition (12), we find that an incremental increase in the helping strategy of non-

€ dispersers ( xˆ >x) is favoured when the condition is satisfied, an incremental

decrease in helping ( xˆ <x) is favoured when the reversed condition is satisfied, and € a candidate (non-zero) evolutionarily stable strategy (x*; i.e. a resident strategy that is €

50 uninvadable by any mutant strategy xˆ ) is implicitly defined by changing the

invasion condition to an equation: € 3 B "( x *) n 2(2 − m) − (n −1)(1− m) = . (13) C "( x *) 1+ m(2n − (n −1)(3− m)m − 3)

Strictly, this condition applies only if the ESS takes a positive value; and x* = 0 if € Bʹ(0)/Cʹ(0) < 0). An explicit solution for the ESS level of helping x* requires explicit

functional forms for C(x) and B(x). Numerical solutions for an illustrative example

are given in Figure 2.2b.

2.2a 2.2b

50

40

30 c / b

Ratio 20

10

0 0 0.2 0.4 0.6 0.8 1 Dispersal rate (m) Figure 2.2. Indiscriminate helping by non-dispersers. (a) The benefit/cost ratio (b/c) below which a mutant gene increasing the level of helping is favoured by selection, for n = 2, 5 and 10 individuals per patch, and a range of dispersal rates 0 < m < 1. (b) The ESS level of helping,

assuming C(x) = kx and B(x) = kαxβ and α = 4, β = 0.5, k → 0, n = 2, 5 and 10, and 0 < m < 1.

2.2.3 The evolution of harming in dispersers

We now investigate the evolutionary success of a vanishingly rare mutant gene that

causes bearers to express harming behaviour upon dispersal to a new patch, and

which confers a relative personal fecundity cost a and a further relative fecundity cost

d that is shared equally among the n-1 other individuals on the patch. Again we

51 express a and d relative to baseline fecundity, and we assume that a, d << 1. This mutant gene will spread if, on average, it increases fitness of its bearers (whether they are dispersers or non-dispersers), either due to direct or indirect fitness effects. There are no direct or indirect effects of the mutant gene on non-dispersing mutants as these do not express the gene nor do they encounter dispersing mutants who do express the harming behaviour. Therefore, the mutant gene can only invade from rarity if it increases the average fitness of mutant dispersers (see Appendix 2.4 for mathematical details). The relative fecundity of a mutant disperser is simply:

FMD = 1− a . (14)

This focal individual produces a total of KFMD offspring, of which mKFMD disperse to a new patch and (1-m)KFMD remain in the natal patch. Using the same approach as in the previous section, the probability that any one of the dispersing offspring of the focal mutant will survive the competition and find a breeding space is 1/K. This means that the expected number of dispersing offspring due to the focal individual is mFMD = m(1-a).

We now consider the fate of the (1-m)KFMD non-dispersing offspring of the focal mutant disperser, which remain on the natal patch to compete for breeding spaces.

The competitors consist of (1− m)K F MD non-dispersing locals, plus mKn incoming

migrants, where F MD is the average relative fecundity of parents in the focal patch € and is given by:

1 F MD = 1− (a + d) . (15) n

52 Again, there are n breeding spaces available, and so the expected number of

successful non-dispersing offspring due to the focal mutant individual is

n 1 m KF / 1 m K F MDn mKn ( − ) MD (( − ) + ) , which is, to first order in a and d, equal to:

2 1 (1− m)(1− a) − (1− m) (a + d) . (16) n

By adding the expected numbers of dispersing and non-dispersing offspring together € this we can find the average fitness of a mutant disperser, and this is:

2 1 wMD = 1− a + (a + d)(1− m) . (17) n

Note that 1/n is equal to RD, the relatedness coefficient for the whole patch the

mutant is found in (for dispersing mutants, rD is 0). In order for the mutant gene to

invade from rarity it must provide a fitness advantage i.e. wMD > 1. This gives an

inequality in the form of Hamilton’s rule:

2 1 −a + (a + d)(1− m) > 0 (18) n

Again, we can derive an inclusive fitness interpretation. Harming by dispersers is € favoured when the sum of three components, equal to the inclusive fitness effect of

harming, is positive. Firstly, the harming act leads to a personal loss of a offspring.

Secondly, the act will inflict a cost d on fellow patch mates; but, since these are related

by rD = 0, this component of inclusive fitness is zero. Thirdly, the net decrease of a+d

offspring on the patch leads to decreased competition for locals and hence an

effective increase in local offspring (including the mutants own offspring) that would

otherwise have been competitively excluded. These are valued by RD= 1/n the

whole-group relatedness for dispersing mutants.

53 Note that condition (18) can be recovered from condition (11) by substituting a = c

and d = -b and replacing the relatedness coefficients for non-dispersers (rN and RN)

with those for dispersers (rD and RD). As we did for condition (11), condition (18) can

be re-expressed in terms of fitness effects, i.e. c br , where c˜ and ˜ are the direct − + D b

2 and indirect fitness consequences of harming. We find that c˜ = a − (1− m) (a + d)/n € 2 and b˜ = −d + (1− m) (a + d)(n −1)/n ; both fitness effects are€ negative when condition € (18) is satisfied so harming, whenever it is favoured by selection, is selfish ( c˜ & b˜ < 0;

€ Hamilton 1964).

We now rearrange inequality (18) to give a condition: €

n d /a > 2 −1, (19) (1− m)

which can be used to determine when a mutation that incrementally increases (or € decreases) the level of indiscriminate harming exhibited by dispersers will invade

from rarity under the action of natural selection. It is helpful to denote the RHS of

condition (19) by g(n,m). We find that, as with the helping analysis considered in the

previous section, harming by dispersers is progressively more difficult to evolve as

the number of individuals per patch increases (i.e. ∂g/∂n > 0; Fig. 2.3a). Also,

harming is less easily favoured with increasing dispersal rates (i.e. ∂g/∂m > 0; Fig.

2.3a): at the extreme of a very viscous population (m → 0) harming is favoured

whenever the ratio of the harm inflicted on patch mates to harm inflicted upon self is

greater than the number of individuals the disperser shares a patch with (i.e. d/a > n-

1; Hamilton 1971), and at the opposite extreme of a fully mixing population harming

will not evolve under any circumstances (g → ∞ as m→1). We note that, if the

dispersal rate is tuned so as to be most favourable to harming (i.e. m → 0) then

54 harming by dispersers evolves more easily than does helping by nondispersers in the

corresponding best-case scenario for the latter to evolve (i.e. m = m*; and n-1 < 4n-2).

Again, we can locate a candidate ESS to examine the absolute level of harming that

we expect to evolve among dispersing individuals. We assume that expressing a level

y of harming incurs a personal fecundity cost A(y) and an others-only fecundity cost

of D(y) (where A, D << 1; see Appendix 2.4), hence the incremental increases in

personal cost and damage to victims that results from adopting the mutant strategy

yˆ rather than the resident strategy y are given by D(yˆ ) − D(y ) = D# ( y )(yˆ − y ) = d and

A(yˆ ) − A(y ) = A #( y )(yˆ − y ) = a respectively. Substituting these into invasion condition € (18), we find that an incremental increase in€ the harming strategy of non-dispersers (

€ yˆ >y) is favoured when the condition is satisfied, an incremental decrease in

harming ( yˆ

€ i.e. a resident strategy that is uninvadable by any mutant strategy y) is implicitly

defined€ by changing the invasion condition to an equation:

D" ( y *) n = 2 −1. (20) A "( y *) (1− m)

Again, this applies strictly to ESSs that take intermediate values, and y* = 0 if € Dʹ(0)/Aʹ(0) < 0. An explicit solution for the ESS level of helping y* requires explicit

functional forms for D(y) and A(y). A numerical solution for an illustrative example is

given in Figure 2.3b. Note that, whilst we have modelled helping and harming

behaviours as separate traits, an alternative (but equivalent) analysis could treat this

social behaviour as a single trait varying from extreme conflict to extreme

cooperation.

55 2.3a 2.3b

Figure 2.3. Indiscriminate harming by dispersers. (a) The harm/cost ratio (d/a) below which a mutant gene increasing the level of harming is favoured by selection, for n = 2, 5 and 10 individuals per patch, and a range of dispersal rates 0 < m < 1. (b) The ESS level of harming,

assuming A(y) = ky and D(y) = kαyβ and α = 4, β = 0.5, k → 0, n = 2, 5 and 10, and 0 < m < 1.

2.3 Discussion

Hamilton (1964, 1971) suggested that indiscriminate helping behaviour should be commonest in viscous populations, because these were associated with greater relatedness between neighbouring individuals. However, more recent work has emphasised the potential for kin competition to inhibit helping behaviour in viscous populations (reviewed by Queller 1992 and West et al. 2002a), and in the simplest scenario of an inelastic island model there is no net impact of dispersal rate upon the evolution of helping (Taylor 1992a). There is much interest in determining which minimal modifications to the basic model can be made in order to allow social behaviour to flourish. We have examined the possibility that social behaviour is adjusted conditionally upon the dispersal status of the actor, and have shown that this allows both helping and harming behaviour to evolve in viscous populations.

Because no external information need be assessed by the individual – for example, its

56 dispersal status may be inferred according to whether it is a winged or wingless morph or whether it experienced a period of cooling or desiccation associated with aerial dispersion – and hence there is no reliance on complex cognitive function, this appears to be a rather general mechanism for the evolution of social behaviours in viscous populations.

In the simplest case of an inelastic island model with dispersal-independent social behaviour (Taylor 1992a) the effect of increased relatedness associated with population viscosity is exactly cancelled by the effect of increased competition between social partners. Thus, both helping and harming behaviours evolve no more readily in viscous populations than they do in fully mixing populations. In contrast, dispersal-dependent social behaviour decouples relatedness and competition so that, while dispersers and non-dispersers experience the same level of competition with their social partners, they differ in their genetic relatedness. In particular, non- dispersers are more related than average to their social partners and dispersers are less related than average to their social partners, and this can allow for the evolution of indiscriminate helping by non-dispersers and indiscriminate harming by dispersers. In the special cases of a completely viscous or completely mixing population, individuals are either all non-dispersers or all dispersers, and so at these extremes our analysis recovers Taylor’s (1992a) results and here we predict that neither costly helping nor costly harming will evolve (for example, whilst dispersers are expected to invest more into harming as the population becomes more viscous, the frequency of dispersers is decreasing towards zero). Thus, we predict that costly helping and harming behaviours are most abundant at intermediate population viscosity.

We have found a basic asymmetry between helping and harming evolution such that, over all, helping is relatively weakly favoured and harming is relatively strongly

57 favoured in viscous populations. This is because zero relatedness (on average) is guaranteed by one’s own dispersal, irrespective of the population level of dispersal, whereas the relatedness achieved by not dispersing is nonzero only insofar as one’s social partners are also non-dispersers. In a highly viscous population conditions are favourable to the evolution of harming by dispersers, because they experience strong local competition and zero relatedness to their social partners. Conversely, while the low competition experienced in highly dispersive populations is conducive to helping, even those individuals not dispersing will have low relatedness to their social partners, who will often be immigrants. So whilst the level of harming exhibited by a disperser is predicted to increase with increasing population viscosity, the level of helping by a non-disperser is predicted to maximise at intermediate population viscosity. Thus, if a sample of species were collected that had relatively low viscosity then we would support Hamilton’s (1964) prediction that, within this sample, there will be a positive correlation between viscosity and the level of helping.

However, a sample of species with relatively high population viscosities might show the reverse trend.

The implications of this simple mechanism for the evolution of indiscriminate helping by non-dispersers were also considered by Perrin and Lehmann (2001).

However, in that previous study, the authors evaluated fitness by considering the fecundity of actors averaged over the disperser and non-disperser classes (their equation 5a), instead of averaging the fitness component obtained by disperser and non-disperser, with the result that the fitness consequences for disperser and non- disperser where not properly separated. This led to an underestimation of the intensity of kin competition (Lehmann 2003, pp14-16; L. Lehmann, pers. comm.). Thus, the results of the present analysis on the evolution of helping are quantitatively and also qualitatively different from those obtained previously. For example, Perrin and

58 Lehmann’s (2001) analysis predicted that the evolutionarily stable level of helping should increase monotonically with population viscosity, whereas we have described a dome-shaped relationship, with maximal helping at intermediate viscosity. In addition, Perrin and Lehmann (2001) did not consider the evolution of indiscriminate harming by dispersing individuals.

Our model of dispersal-dependent social behaviour may provide insight into the evolution of social parasitism in aphids, whereby individuals disperse to new colonies and, upon arrival, downregulate their usual altruistic colony-defence function and upregulate selfish, personal reproduction (Abbot et al. 2001; Foster

2002). Although social parasitism is known in several taxa, this might usually be attributed to direct discrimination of own versus alien colonies. However, aphids are presently understood to lack the faculty for kin recognition that would permit them to directly assess genetic relatedness of their neighbours, and so it may be the simple act of dispersal that provides the cue for this facultative change in their social behaviour (Foster 2002). The present analysis is also conceptually related to a study of dispersal-dependent sex allocation by Taylor and Crespi (1994), which predicted

(and empirically tested and confirmed in a species of thrips) that dispersing and non- dispersing individuals would allocate differentially to sons versus daughters due to the correlation between dispersal status and relatedness to neighbours. In particular,

Taylor and Crespi’s (1994) diploidy model predicted that non-dispersers should exhibit a slight bias towards daughters, and dispersers a strong bias towards sons.

This is analogous to our result that non-dispersers are weakly favoured to help and dispersers are strongly favoured to harm, and indeed in both cases the pattern is due to the different relatedness experienced by dispersers and non-dispersers.

A correlation between dispersal status and social behaviour can have important consequences for population demography, suggesting avenues for future research

59 directions. By modifying the fitness differences between dispersers and non- dispersers, this association can drive selection for changes in the rate and pattern of dispersal. Perrin and Lehmann (2001) considered the role for indiscriminate helping by non-dispersers to mediate the evolution of dispersal, and the subsequent feedback on the helping trait itself. As with our analysis, Perrin and Lehmann (2001) found that helping was only weakly favoured in viscous populations, and hence the co- evolution of helping and dispersal did not have a dramatic impact on either trait.

However, the role for harming by dispersers to mediate the evolution of population viscosity remains to be considered, and our analysis suggests potential for rather stronger effects. An association between dispersal status and social behaviour might also impact upon inter-specific interactions, influencing the distribution and abundance of competing species. For example, Duckworth (2007) has described a positive correlation between dispersal and aggression in blue birds that appears to mediate species range expansions.

More generally, we have highlighted that the kin selection coefficient of relatedness describes the conditional information that an individual has with regard to the genetic similarity of its social partners, rather than their actual genetic similarity (e.g.

Frank 1998, ch 6). Typical extensions of Taylor’s (1992a) simple island model that promote the evolution of social behaviour in viscous populations do so by decoupling the genetic structure of the population from the intensity of local competition; for example, by introducing ‘elasticity’ so that cooperative patches can grow in size, localised competition is reduced while genetic structuring remains high.

In contrast, the degree of genetic structure and the intensity of localised competition described in our model is exactly as occurs in Taylor’s (1992a) model, and the decoupling of relatedness and localised competition occurs through use of conditional information: whether or not the individual dispersed.

60 2.4 Appendix

2.4.1 Natural selection and social behaviour

We consider a locus G controlling social behaviour of one type of individual (i.e. non-

disperser or disperser), and we denote an individual’s genetic ‘breeding’ value (Price

1970; Falconer 1981; Grafen 1985) for the behaviour as g. The change in the

population average breeding value is given by:

Δg = ∑cu covu(w /w u,g) + ∑cv covv (w /w v,g), (A1) u∈U v∈V

where: U is the set of all non-disperser classes (defined according to the number of € non-dispersers among their social partners); V is the set of all disperser classes (also

defined according to the number of non-dispersers among their social partners); cu

and cv are the class reproductive values of non-disperser class u and disperser class v

respectively; wu and wv are the average fitnesses for individuals of class u and v

respectively; and cov is the statistical covariance taken over individuals of the

indicated set (Price 1970; Taylor 1996b).

2.4.2 Natural selection and social behaviour in non-dispersers

Consider first that the gene affects social behaviour in non-dispersers only. We can

expand the selection component due to dispersers as:

∑cv covv (w /w v,g) = ∑cvβv (w /w v,g)varv (g), (A2) v∈V v∈V

€

61 where β is a least-squares linear regression, and var is the statistical variance, both

taken over the indicated class of individuals. Expanding the regression into partial

regressions, we have:

βv (w /w v,g) = βv (w /w v,g | gN# & gD# ) + βv (w /w v,gN# | g & gD# )βv (gN# ,g) (A3) + βv (w /w v,gD# | gN# & g)βv (gD# ,g),

where g´N and g´D are the average breeding values of the individual’s non-disperser € and disperser social partners, respectively. Note that, because the gene is not

expressed in dispersers, βv(w/wv,g|g´N and g´D) = βv(w/wv, g´D |g andg´N) = 0. Also,

because the focal individual is a disperser it is unrelated to the other individuals on

its patch, and so βv(g´N,g) = βv(g´D,g) = 0. Hence, βv(w/wv,g) = 0 and so, from

equations (A1) and (A2), selection on the gene is given by:

Δg = ∑cu covu(w /w u,g). (A4) u∈U

Thus, if the gene is expressed only in non-dispersers, selection for or against the gene € is determined solely by the fitness consequences for non-dispersers.

2.4.3 Natural selection and social behaviour in dispersers

Alternatively, consider that the gene affects social behaviour in dispersers only. We

can expand the selection component due to non-dispersers as:

∑cu covu (w /w u,g) = ∑cuβu(w /w u,g)varu (g), (A5) u∈U u∈U

and, expanding the regression into partial regressions, we have: €

βu (w /w u,g) = βu (w /w u,g | gN# & gD# ) + βu(w /w u,gN# | g & gD# )βu (gN# ,g) (A6) + βu (w /w u,gD# | gN# & g)βu(gD# ,g),

€ 62 where g´N and g´D are once again the average breeding values of the individual’s non-

disperser and disperser social partners, respectively. Note that, because the gene is

not expressed in non-dispersers, βu(w/wu,g|g´N and g´D) = βu(w/wu, g´N |g and g´D) =

0. Also, because the focal individual is unrelated to its disperser social partners,

βu(g´D,g) = 0. Hence, βu(w/wu,g) = 0 and so, from equations (A1) and (A5), selection

on the gene is given by:

Δg = ∑cv covv (w /w v,g). (A7) v∈V

Thus, if the gene is expressed only by dispersers, selection for or against the gene is € determined solely by the fitness consequences for dispersers.

2.4.4 Invasion condition for non-disperser social behaviour

We now determine the invasion condition for a gene affecting social behaviour of

non-dispersers, and we do so under the assumptions that: (1) the gene has

vanishingly small impact on the phenotype; and (2) social behaviours in general have

vanishingly small impact upon fecundity. Because the population is not growing in

size, we can write the fitness of a focal non-disperser individual j ∈ Ju in class u as wuj

= 1 + δwuj, the average fitness of individuals of this class as wu = 1 + δwu, and the

reproductive value of this class as cu = qu + δcu, where qu is the proportion of all

individuals in the population that belong to class u. These three deviation terms are

all vanishingly small quantities (δwuj, δwu, δcu << 1).

Taking equation (A4), and making the individual notation more explicit, the action of

selection is given by:

63 Δg = ∑(qu + δcu)covu(wuj /w u,guj ) u∈U (A8) = ∑qu covu (wuj /w u,guj ) + ∑δcu covu(wuj /w u,guj ). u∈U u∈U

Due to the weak-selection assumption, covu(wuj/wu,guj) << 1, and so the second term €

of equation (A8) is negligible. Using wuj = 1 + δwuj and wu = 1 + δwu, and discarding

higher order terms of δ, we can rewrite equation (A8) as:

Δg = ∑qu covu(wuj −δw u,guj ) u∈U

= ∑qu covu (wuj ,guj ) (A9) u∈U q ' q * = q u ) uj w g − w g ,, N ∑ q ) ∑ q uj uj u u, u∈U N ( j ∈Ju u +

where qN = ΣU qu = 1-m is the total fraction of individuals in the population that are € non-dispersers. Making the standard assumption that gene frequencies are equal in

all classes, i.e. gu = g for all u ∈ U, this obtains:

& q ) Δg = (1− m)( uj w g − w g + ( ∑ ∑ q uj uj N + ' u∈U j ∈Ju N * (A10)

= (1− m)covN (wuj ,guj ),

where covN is a covariance taken over the set of all non-disperser individuals in the € population. Natural selection acts to increase the average genetic breeding value of

the population when Δg > 0. In order to track the invasion success of a rare variant

gene, we may assign it a value of g = 1 and all its alleles a value of g = 0, in which case

we can write Δg = (1-m)g(wMN-wN), where wMN is the average fitness of all the non-

disperser carriers of the variant gene. Hence, the variant gene invades if, on average,

64 it increases the fitness of its non-disperser carriers above that of the average non-

disperser individual (wMN > wN).

2.4.5 Invasion condition for disperser social behaviour

The action of selection upon genes for disperser social behaviour (equation (A7)) has the same form as that for non-disperser social behaviour (equation (A4)), and an invasion condition can be determined by following the procedure of the last section and simply switching the non-disperser notation to the corresponding disperser notation (i.e. u to v, U to V, and N to D). This reveals that a rare genetic variant that alters the social behaviour of disperser individuals is favoured if the average fitness of its disperser carriers is greater than the average fitness of all dispersers in the

population (wMD > wD).

65 Chapter 3. The enforcement of cooperation by policing.

This chapter appears as the following publication: El Mouden, C., West, S. A. and

Gardner, A 2010. ‘The enforcement of cooperation by policing’, Evolution. 64, 2139-

2152.

3.1 Introduction

Explaining cooperation is a major challenge for evolutionary biology. Natural selection favours those individuals who have the greatest relative fitness (Darwin

1859; Fisher 1930; Price 1970). However, all else being equal, cooperation reduces the relative fitness of the actor, as it involves providing a benefit to other individuals. The challenge is therefore to explain how cooperation can evolve by natural selection

(Hamilton 1966; Sachs et al. 2004; Lehmann and Keller 2006; Lehmann et al. 2007c;

West et al. 2007a). In the social sciences, the problem of cooperation is famously encapsulated by the “tragedy of the commons” (Hardin 1968) which considers a group of herdsmen deciding how many animals to graze on a shared pasture. The benefit of adding an extra animal accrues directly to the individual herdsman whilst the cost of overgrazing is shared by all. Hence, each herdsman is favoured to add further animals to the commons, resulting in severe overgrazing and the destruction of the shared resource. Finding solutions to avert this tragedy is of interest to evolutionary biologists, economists and moral and political philosophers (Ostrom

1990; Binmore 1994, 1998; Frank 2003, 2009).

66 Two major ways for the tragedy of the commons to be averted are through kin- selected self-restraint or repression of competition (Frank 2003). The former mechanism works when relatives interact, as individuals that improve the reproductive success of those with which they share genes are more successful in transmitting copies of these genes to future generations (Hamilton 1963, 1964a;

Maynard Smith 1964; Hamilton 1970). However, theory predicts that even groups of close relatives can foster significant internal conflict, and so there is still often scope for very costly competition within groups (Frank 1995, 1996b, 2003). Repression of competition is when there is a mechanism that reduces competition among the members of a group (Frank 2003). In contrast to kin-selected self-restraint, complete repression of competition is predicted to lead to harmonious groups because the individual can pursue the maximization of its inclusive fitness only by promoting the success of its group (Leigh 1971, 1977; Alexander 1987; Frank 2003; Gardner and

Grafen 2009).

A potentially important means of repressing competition is by policing. The theory of policing has developed in two directions: (1) general theory intended to capture how mechanisms that enforce an equitable distribution of shared group resources may evolve (e.g. Frank 1995, 1996b, 2003, 2009); and (2) theory developed specifically for eusocial insects (Ratnieks et al. 2006; Ratnieks and Helanträ, 2009) such as honeybees where egg-laying workers are policed (their eggs are eaten) (Ratnieks 1988; Ratnieks and Visscher 1989). This work considers how mechanisms of kin-discrimination can lead to policing in haplodiploid societies. Here, we are concerned with the former approach, which aims for an illustrative overview of how the shared benefits of cooperation can drive the evolution of policing mechanisms to limit within-group conflict. Such policing is found at all scales of biological complexity: for example, intragenomic conflict is limited by fair meiosis (Leigh 1977); uncooperative symbionts

67 can face sanctions from their host, such as in the legume-rhizobia mutualism (West et al. 2002b; Kiers et al. 2003; 2006b) and ‘reproductive leveling’ in human tribal groups is enforced via strict food sharing norms (Bowles, 2006). Such general models highlight how comparable evolutionary mechanisms may result in cooperation evolving in very different contexts.

We develop the general theory of ‘group-benefits’ driven policing (Frank 1995,

1996b), in three ways. First, we relax an implicit assumption in the original model that the individual cost of investment into policing is lower in more selfish groups.

Second, we relax the assumption that the group cost of selfishness is totally recovered by policing. Third, we consider policing in a demographic context: the population processes that allow relatedness to build up (e.g. low dispersal) can often lead to intensified competition between relatives (Taylor, 1992a; Queller 1992; West et al.

2002a), but this has been neglected in previous models of policing. We consider the general impact of increased kin competition on the evolution of group-benefits driven policing, and provide an illustration using Wright’s (1931) infinite island model.

3.2 Models and Analyses

3.2.1 Frank’s model of selfishness and policing

In this section, we introduce and re-derive the results of the policing model developed by Frank (1995, 1996b, 2003, 2009), which forms the basis of our subsequent analyses. Frank (1995, 1996b) considered the evolution of a policing behaviour that represses competition and limits the detrimental impact of within- group competition on the level of group resources, assuming that the fitness w of an individual is given by:

68 % z ( w ∝' a# +(1− a# ) − ca* 1− (1− a #) z # , (1) & z # )( )

where: a is the individual’s investment into group policing; a' is the average level of € policing by members of the group; z is the individual’s investment into selfishness; z'

is the group average level of selfishness; and c scales the individual cost of

investment into policing (a summary of all notation is given in Table 3.1). Fitness is

given by the individual’s relative share (first multiplicative term) of total group

resources (second multiplicative term). The individual’s relative share of group

resources is given by 1 in all policed encounters, and by her relative selfishness z/zʹ

in all unpoliced encounters. The total group resources are reduced as unpoliced

encounters and selfishness increase within the group. (A variant of the model

considered by Frank (1996b), implements a personal cost of investment into policing

alternatively as a third multiplicative fitness component, which does not qualitatively

affect results.) Importantly, Frank’s (1995) model assumes that investment into

policing incurs a reduction in the individual’s share of the group resources, and hence

the absolute cost of policing is lower when the group has fewer resources (i.e. it is

cheaper to police a more selfish group). We relax this assumption in the next section.

In the absence of policing (a=a'=0), the model describes a basic tragedy of the

commons scenario (Frank 1996a, 2009, 2010), where fitness is given by:

z w ∝ (1− z #) , (2) z #

€ where an individual’s share of group resources increases with its relative investment

in selfishness (z/z'), which favours the evolution of increased selfishness, and the

tragedy occurs because a more selfish group has reduced resources (1-z'). The model

shows that one way the tragedy can be averted is when there is within-group

69 relatedness, which favours some degree of self-restraint. If within-group relatedness is r, then the evolutionarily stable (Maynard Smith and Price 1973) level of selfishness is z*=1-r (Frank 1995, 1996b).

Table 3.1. A summary of symbols used in the analysis

Symbol Definition Symbol Definition

fitness cost to actor of a social a policing behaviour C behaviour

number of breeding spaces per z selfishness / competitiveness n group (group size) c cost of policing F fecundity of an individual

efficiency of policing at restoring group b mean fecundity of group resources Fʹ

coefficient of relatedness (“whole- r mean fecundity of population group” relatedness) F w Darwinian fitness s scale of competition

R “others-only” relatedness m dispersal rate

fitness benefit to the recipient of a social B K number of offspring behaviour

Frank (1995) found that relatively high relatedness disfavours policing (a* = 0 when r>1-c; Figure 3.1a). This is because kin-selected self-restraint can maintain relatively high cooperation in high-relatedness groups, so that the benefits of policing do not outweigh its costs. Note, however, that even in groups with considerable relatedness, there is scope for significant selfishness. For example, in a large group of full siblings, relatedness is approximately one half, so the ESS level of selfishness is also equal to one half, and hence the group’s fitness is only half as great as it could be if selfishness were completely abolished. Conversely, a relatively low within-group relatedness favours full investment into policing (a* = 1 when r<1-c; Figure 3.1a). A further result, not emphasized by Frank (1995), is that policing results in increased selfishness z* > 1-r; Figure 3.1b). Hence, Frank’s model of repression of competition

70 by policing works by recovering the group cost of selfishness and not by reducing selfishness itself.

Frank (1995, 1996b) considered the evolutionary dynamics and statics of policing and selfishness, but did not show how policing fits into the standard classification of social behaviours provided by Hamilton (1964b; Table 1.1). The classification depends on the others-only relatedness of group mates (R; the expected relatedness of two individuals sampled at random from the same group without replacement;

Pepper 2000). Details of how the classification is determined are given in Appendix

3.4. We find that:

Result 1 – In Frank’s (1995) model, policing may be altruistic or mutually beneficial.

Full policing is altruistic when others-only relatedness is high (R > [n(1-c)-1]/[(n-

1)(1+cn)]) and mutually beneficial when others-only relatedness is low (R < [n(1-c)-

1]/[(n-1)(1+cn)]).

3.1a 3.1b

Figure 3.1 Frank’s model of policing and selfishness In Frank’s policing model (Frank 1995; Figure 3.1a) full policing (a*=1), which results in a complete repression of within-group competition and a fair distribution of group resources, is predicted to invade whenever relatedness is low (r<1-c). When relatedness is high (r>1-c), no policing occurs (a*= 0) and the model reduces to the tragedy of the commons and selfishness (Figure 3.1b) is limited only by kin-selected self-restraint (z*=1-r). Policing always favours increased selfishness (dz*/da>0, z*=(1-r)/r(1-c)).

71 3.2.2 The individual cost of policing

Owing to multiplicativity of terms, Frank’s (1995, 1996b) model of policing implicitly

assumes that the magnitude of the individual cost of investment into policing

increases with group resources (i.e. ∂w/∂a = -c(1-(1-aʹ)zʹ), where 1-(1-a´)z´ is the total

group resources). In other words, policing gets cheaper to maintain as groups

become more selfish. However, it is difficult to think of a biological scenario where

this would be the case. In this section, we assume instead that policing requires a

fixed investment of time or energy. The relationship between the cost of policing and

selfishness will depend on the biological details; our aim here is to illustrate the

importance of such details and provide a base upon which more complex models,

that consider specific biological cases, may be developed.

We make the marginal cost of policing a fixed proportion of group fitness (∂w/∂a = -

c), which we implement using a modified fitness function:

% z ( w ∝' a# +(1− a# ) * 1− (1− a# ) z # −ca. (3) & z # )( )

In Appendix 3.4, we show that this leads to the following predictions: €

Result 2 – Full policing (a*=1) is only favoured at intermediate relatedness

(specifically, when r(1-r)>c; Figure 3.2a). At the extremes of relatedness (r(1-r)

policing cannot evolve and the model recovers the tragedy of the commons result

(a*=0 and z*=1-r). This is in contrast to Frank’s (1995, 1996b) model, in which policing

is favoured at arbitrarily low relatedness.

Result 3 – When there is full policing, selfishness increases (z*>1-r), but tends to a

lower level than in Frank’s (1995, 1996b) model (z* = (1-r)/r; Figure 3.1b).

72 3.2.3 The group cost of selfishness

Frank’s (1995, 1996b, 2003) models assumed that policing fully recovers the loss of

group fitness due to selfishness. This seems appropriate if policing removes the

opportunity for individuals to behave selfishly (as in fair meiosis; (Leigh 1977).

However, if policing simply works to remove the fitness benefits of selfishness,

without directly suppressing selfishness itself (as in the eating of worker-laid eggs in

honeybees, which does not directly prevent worker egg-laying but removes the

benefits of doing this; Ratnieks et al. 2006), then we would expect selfishness to

continue to incur some group-level cost even in the presence of full policing. We now

develop the model described in the previous section to allow for a less than perfect

efficiency of policing, with only a proportion b of the group fitness lost to selfishness

being recovered by policing. In particular, we write:

% z ( w ∝' a# +(1− a# ) * 1− (1− ba# ) z # −ca . (4) & z # )( )

We find that this adjustment impacts upon the model behaviour in several ways (see € Appendix 3.4 for details):

Result 4 – Full policing is never stable (a* < 1; a full expression for a* is given in

Appendix 3.4). This is in contrast to Frank’s (1995, 1996b) model, in which full

policing may be stable (a* = 1), and an intermediate level of policing is never stable.

Result 5 – Reduced efficiency of policing (lower b) reduces the scope for policing to

invade (a* > 0 only if r(1-r) > c/b, see Appendix 3.4 for full expression of a*) and leads

to a lower ESS level of policing when it does invade (da*/db > 0; Figure 3.2a).

73 Result 6 – Policing is more easily maintained by selection than it initially evolves.

The condition for invasion of policing (instability of a = 0) is more stringent that the condition for an internal stable equilibrium to exist (0 < a* < 1; Figure 3.4).

Result 7 – Policing promotes self-restraint (z*<1-r; Figure 3.2b). When there is policing, selfishness is z* = c/rb which, in the region where policing can invade from rarity (r(1-r) > c/b, is lower than that which is obtained in the absence of policing (z*

= 1-r)). This is in contrast to Frank’s (1995, 1996b) model, where policing always favours increased selfishness (z*>1-r; Figure 3.1b)

Result 8 – Policing is an altruistic behaviour. This is in contrast to Frank’s (1995,

1996b) model, in which policing could be either altruistic or mutually beneficial (see

Result 1).

3.2a 3.2b

Figure 3.2 Modified individual and group cost models. In Frank’s model, the cost of policing was proportional to the value of group resources so high levels of selfishness were favoured as they allowed vanishingly cheap policing. When the cost is adjusted to be proportional only to the level of policing an individual performs (i.e. ∂w/∂a = -c, we assume here c=0.1; Figure 3.2a) policing is favoured to invade only at intermediate relatedness (when r(1-r)>c/b). If the efficiency with which a policing behaviour restores group resources is imperfect (b<1) intermediate levels of policing are favoured (Figure 3.2a) and when it invades, it always promotes self-restraint (dz*/da<0, z*=c/br; Figure 3.2b). This is in contrast to perfect policing (b=1, a=1) which favours increased selfishness (z*→(1-r)/r; Figure 3.2b). When policing is not stable (a*=0 is stable when r(1-r)

74 r). As the efficiency of a policing behaviour at restoring group resources b is reduced, the ESS level of policing falls (da*/db<0; see Appendix 3.4).

3.2.4 Demography and policing – illustrative overview

Frank’s (1995, 1996b) models consider the impact of whole-group relatedness (r)

upon the coevolution of policing and cooperation, and our developments of this basic

model reveal that an intermediate level of relatedness is crucial for policing to evolve.

However, the ecological and demographic processes that lead to relatedness within

groups can often also lead to intensified competition between relatives (reviewed by

Queller 1992; West et al. 2002a), which has not been explicitly considered in Frank’s

models. For example if individuals do not disperse far from where they were born

(population viscosity, Hamilton 1964b), then relatives must compete more intensely

for space and resources.

We extend the model developed in the previous section, by making explicit the

impact of policing and selfishness upon fecundity, and by making explicit the impact

of personal fecundity and the fecundity of competitors upon Darwinian fitness. We

employ a “scale of competition parameter” (s; Frank, 1998), which measures the

extent to which an increase in an individual’s fecundity leads to intensified

competition within its social group. This approach has been useful for both

theoretical (West et al. 2002a; West and Buckling 2003; Gardner and West 2004b, 2006;

Gardner et al. 2009; Lively 2009; Gardner, 2010) and empirical (West et al. 2001;

Griffin et al. 2004; West et al. 2006; Kümmerli et al. 2009) studies of social evolution.

Following the approach taken by those earlier studies, we define the fitness of an

individual as:

F w = , (5) sF " + (1− s)F

€ 75 where: F is the relative fecundity of the individual, and is given by

$ z ' F = & a" +(1− a" ) ) 1− (1− ba" ) z " −ca ; (6) % z " (( )

F´ is the average fecundity of the individuals in its social group, and is given by €

F " =1− (1− ba" ) z "− ca" ; (7)

and F is the average fecundity of individuals over the whole population, and is € given by

F =1− (1− ba )z −ca . (8)

Thus the fitness of an individual is determined by their fecundity (F) relative to the € fecundity of their competitors, a proportion s of which are locals (group mates, who

have average fecundity F') and a proportion 1-s of which are random individuals

drawn from the population as a whole (having average fecundity F). In Appendix

3.4, we derive the following results:

Result 9 – Local competition reduces the scope for policing to invade (a*>0 when r(1-

r)[(1-s)/(1-rs)2]>c/b, which is increasingly stringent as s increases; for a*, see Appendix

3.4) . Increased local competition both narrows the range of relatedness values that

favour the invasion of policing (d[rmax-rmin]/ds<0 where rmax and rmin are the maximum

and minimum relatedness permitting policing to invade from rarity) and shifts this

range to higher levels of relatedness (drmax/ds>0, drmin/ds>0; Figure 3.3a). When

competition is entirely local (s=1) policing is never favoured. Policing is still

maintained more readily than it invades (Figure 3.4).

Result 10 – Increased local competition leads to higher selfishness (dz*/ds > 0). In the

absence of policing z*=(1-r)/(1-rs) which rises from z*=1-r when s=0 to z*=1 when

76 s=1. As before, policing promotes self-restraint (z*=c(1-rs)/br(1-s)) in the region where it can invade from rarity (when r(1-r)[(1-s)/(1-rs)2]>c/b).

Queller (1994) suggested that the effects of local competition can be easily accounted for by redefining relatedness so that it measures genetic similarity of two individuals with respect to their “economic neighbourhood” (the arena in which they compete) rather than the with respect to the population as a whole. The condition for policing r(1-r)[(1-s)/(1-rs)2]>c/b may be re-written as ρ(1-ρ) >c/b by substituting whole-group relatedness, r with Queller’s relatedness, ρ where ρ=(r-sr)/(1-sr) (see also Gardner and West 2006). Queller’s relatedness captures both the genetic structure of a population (relatedness, r) and the effect of changing the scale of competition (s), with Queller’s relatedness being an increasing function of the former (∂ρ/∂r > 0) and a decreasing function of the latter (∂ρ/∂s < 0). Results 9 and 10 may be derived from the basic model results (Results 2, 4-7) by substituting “r” with “ρ”. If competition is entirely local, regardless of the genetic relatedness, there is no incentive to perform cooperative behaviours as all group members are valued equally (i.e. when s=1, ρ=0).

This is why the effect of kin-selected self-restraint on selfishness disappears as competition becomes localised which then confines policing to higher levels of relatedness.

77 3.3a 3.3b

Figure 3.3 Policing and Demography. Increased local competition reduces the scope for policing (a*>0 when r(1-r)[(1-s)/(1-rs)2]>c/b). In the open model (Figure 3.3a, where we assume c=0.1, b=0.7), as the scale of competition (s) increases, policing is restricted to higher, narrower levels of relatedness. When s=0, policing is equal to the line for b=0.7 on Figure 3.2a. In the Island model (Figure 3.3b, where we assume c=0.1), which considers a population subdivided into groups of size n with migration occurring between them, we find that policing invades less readily in larger groups (a*>0 when (n-1)/n2>c/b).

3.2.5 Demography and policing - island model

In the previous section we provided an illustrative overview of how intensified kin competition impacts upon the coevolution of policing and cooperation. This revealed that, all else being equal, increased competition reduces the scope for policing and self-restraint to evolve. In particular, we examined the impact of increased kin competition (s) under the assumption of fixed relatedness (r). However, the same demographic processes may determine the degree of relatedness and kin competition, so it can be misleading to consider the effects of changing the value of either of these quantities in isolation. Here, we explicitly consider the co-dependency of relatedness and kin competition in a specific infinite island model setting (Wright

1931). The model assumes that an infinite population of asexual haploid individuals compete for resources in groups of size n. The relative fecundity of an individual is determined by its success in social interactions within the group (as in Equations 6-8).

Individuals then have a very large number of offspring (K) proportional to their

78 fecundity, of which a proportion m migrate to randomly chosen groups and a proportion 1-m remain in the natal group. We assume that parents die at each generation and that dispersing individuals will not find relatives in their new group.

In the next generation, native and immigrant individuals compete in every group for the n breeding spots. The fitness of an individual is determined by its success relative to others in its group. In Appendix 3.4, we show that:

Result 11 –The migration rate m has no net effect on the evolution of policing

(da*/dm = dz*/dm = 0). Decreasing the migration rate both increases relatedness

(r=1/(n-(n-1)(1-m)2, so dr/dm<0) and also increases local competition (s=(1-m)2, ds/dm<0). However these effects have no net impact on Queller’s relatedness ρ, which is the relatedness of an individual to its social partners relative to its competitors, (substituting the island model values for r and s into ρ gives an effective relatedness of ρ=1/n which is not a function of m).

Result 12 – Increasing group size disfavours the invasion of policing (policing can increase from rarity when (n-1)/n2>c/b, in large groups this is approximately when b/c>n, for a* see Appendix 3.4; Figure 3.3b). The larger the group, the lower the whole-group relatedness (which falls from r=1 when the group size n=1 to r→0 as n→∞) and the higher the selfishness. Policing is maintained in larger groups more readily than it invades in them (Figure 3.4). In the absence of policing, selfishness is stable at z*=(n-1)/n and as before, in the region where policing can invade, it always promotes self-restraint, given by z*=cn/b.

79

Figure 3.4 Policing and Demography - Stability Analysis. A numerical investigation (following the methodology of Otto and Day (2007, pp 237-242)) shows policing is always stable when it invades and that policing is maintained outside the region defined by the invasion condition. If the ESS for policing is stable, (the real part of its associated eigenvalue is negative), we plot a white or grey point; white denotes when only policing is stable (no policing is unstable if the marginal fitness for policing, dw/da is positive when a*=0, which is when r(1-r)[(1-s)/(1-rs)2]>c/b)) and grey denotes when both policing and no policing are stable i.e. policing cannot invade but it would be stable if already present. If only no policing is stable (the real part of the ESS’s associated eigenvalue is positive and r(1-r)[(1-s)/(1-rs)2]

3.3 Discussion

Group-benefits driven policing – enforcement of fairness in the distribution of resources among group mates – has been regarded as a potentially important mechanism for the evolution of cooperation and the major transitions in evolution

80 (Leigh 1977; Alexander 1987; Frank 2003; Gardner and Grafen 2009). In this paper we have examined Frank’s (1995, 1996b) model of policing, which was developed to provide a general overview of group benefits driven policing theory. The basic predictions of his model was that policing evolves most readily when within-group relatedness is low, that this leads to an increase in selfishness and, when it does evolve, it will completely abolish competition among group mates (Figure 3.1).

However, his model assumes that the individual cost of investment into policing vanishes as group selfishness increases, and also that policing totally recovers any loss of group fitness due to selfishness – even though individuals continue to invest in selfishness. We have examined the consequences of relaxing these assumptions and find these basic results are changed substantially. Specifically we find that policing does not evolve when relatedness is low (Figure 3.2a), favours a reduction in individual selfishness (Figure 3.2b), cannot repress all within-group conflict (i.e. a*<1;

Figure 3.2a), is easier to maintain at lower relatedness than to initially evolve (Figure

3.4), and is altruistic.

We also examined the consequences of placing group-benefits policing in its proper demographic context. We have shown that within-group relatedness is crucial for policing to be favoured, and yet the demographic processes that can generate within- group relatedness may also lead to intensified kin competition, which has not been considered explicitly in previous models. We found that, in general, an increased competition for resources within groups leads to a reduction in policing. Further, in the context of Wright’s (1931) infinite island model, we found that the relatedness- inflating effects of low migration exactly cancel the associated increase in local competition, so the evolution of policing is solely dependent upon group size, with larger groups disfavouring policing (Figure 3.3b). In short, all of our model developments reduce the scope for policing to invade from rarity. However our

81 results also show that policing is easier to maintain than it is to evolve, specifically that it is stable at lower relatedness (Figure 3.4). This result mirrors previous work on the evolution of punishment and cooperation which showed that punishment was maintained more readily than it evolved (Gardner and West 2004a).

We found that group-benefits policing requires intermediate relatedness within groups in order to be favoured by natural selection. We emphasize that this refers to the ‘whole-group’ relatedness, which includes the relatedness of the individual to itself, and not the ‘others-only’ relatedness, which is the relatedness between social partners (Pepper 2000). Thus, the model could be applied to study policing in interspecific mutualisms, where a host individual is not genetically related to its symbiont(s), but nevertheless could reap the benefits of policing its social partners

(mediated through non-zero whole-group relatedness). Our model predicts policing evolves most readily when whole-group relatedness is one half, which is obtained when unrelated partners are in a 1:1 relationship or when one individual forms a symbiosis with a clonal group. However, our model is not intended to fully capture the biology of mutualisms. Importantly, it treats all individuals as equivalent, incurring the same costs and benefits for the same investments into policing and selfishness, whereas real-world mutualisms bring non-equivalent individuals from different species together in a complementary relationship. Development of this basic policing model to explicitly incorporate the biology particular to mutualisms represents an important direction for future research (Murray 1985; West et al

2002a,b; Yu et al 2004; Foster and Wenseleers 2006).

Group-benefits driven policing may have been important in human evolution

(Bowles 2006, 2009). As policing is maintained more readily than it invades, it could have evolved at a time when humans lived in small tightly-knit groups and then been retained as much larger societies developed, particularly if social norms

82 allowed the cost of policing to be reduced. To investigate the evolution of social behaviours in humans, the effect of both population structure and cultural transmission (if they are culturally inherited) on their evolution must be understood.

Whilst the importance of cultural transmission in enhancing/limiting selection for cooperation has received considerable attention (Boyd and Richerson 1985; Henrich and Boyd 1998; Henrich and Gil-White 2001; Boyd and Richerson 2005a; Lehmann and Feldman 2008a; Lehmann et al. 2008), the importance of population structure has not. For example, budding dispersal, where groups migrate to new territories (the

‘tribe-splitting’ model of Haldane 1932) may be a better description of human population movements than the basic island model used here (which assumes individuals disperse independently). Both theoretical and empirical work (Gardner and West 2006; Lehmann et al. 2006; Kümmerli et al. 2009) show that budding dispersal favours the evolution of cooperation. This illustrates the importance of considering whether models capture the details of human population structure when investigating the evolution of behaviours such as policing. Extending the island model (Wright 1931; Taylor 1992a) to study policing in a demographic context which more accurately reflects the structure of human Palaeolithic society is an important task for the future.

In this paper, we have focused upon models of policing that work owing to group benefits (following Leigh 1977; Alexander 1987; Frank 2003). However, group benefits are not necessary for policing to be favoured and for policing to favour cooperation. Perhaps the best understood policing system, from a theoretical and empirical standpoint, is that of honeybees and other hymenoptera. Here, workers prevent the rearing of eggs laid by other workers by eating them upon discovery, and this has been shown to reduce the extent to which workers do lay eggs (Wenseleers et al. 2004a, b). The selective force favouring policing in such cases is the nepotistic

83 interests of the egg-eating workers, who value queen-laid unfertilized eggs above worker-laid unfertilized eggs, because they are more genetically related to the former than to the latter (Ratnieks 1988; Ratnieks and Visscher 1989; Wenseleers et al 2005;

Wenseleers and Ratnieks 2006; Ratnieks and Helanterä 2009). Consequently, this egg- eating behaviour can evolve even when it does not provide a colony-level benefit

(Wenseleers et al. 2004a; Wenseleers et al. 2004b). However, such policing can also provide colony level benefits, which can maintain policing even in cases where there are not relatedness asymmetries (Martin et al 2002).

How do our results relate to our understanding of the major evolutionary transitions? A major transition can occur when individuality shifts from a lower to higher level (Buss 1987; Maynard Smith and Szathmary 1995; Queller 2000). For such transitions to be possible, internal conflict must be resolved (Frank 2003). This can occur owing to clonality of individuals within social groups, or if there is a total repression of within-group competition (Buss 1987; Maynard Smith and Szathmary

1995; Gardner and Grafen 2009). Policing has been suggested as a potentially important factor in driving major transitions (Frank 1995, 1996b, 2003; Ratnieks and

Helanterä 2009). We suggest caution, as this may depend upon whether policing evolves via group-benefits or kin discrimination, and as the extent to which individuals can discover other ways to pursue their selfish interests. This suggestion is consistent with recent work on social insects which suggests lifetime monogamy

(which leads to high within-colony relatedness) was an essential step in the evolution of eusociality (Boomsma 2007, 2009; Hughes et al. 2008). It appears that worker policing was only important after eusociality became irreversible as a means to limit within-colony reproductive conflicts that emerged as queens began multiply mating

(Ratnieks et al. 2006; Ratnieks and Wenseleers 2008). More generally, this emphasises

84 how the origin versus the maintenance evolution of cooperation behaviours can involve different selective forces (West et al. 2007a).

To conclude, we have shown that policing mechanisms that enforce the fair sharing of group resources are harder to initially evolve than previously thought. We show that contrary to previous suggestions, policing is unlikely to be favoured at low relatedness and that, when it evolves, it often favours self-restraint. Furthermore policing is unlikely to result in a complete repression of within-group competition.

This suggests group-benefits driven policing will be of limited importance in major transitions. Major tasks for the future include the study of policing in different demographies, particularly with reference to humans and understanding the role of policing in intra-specific mutualisms.

3.4 Appendix

3.4.1 Frank’s model of selfishness and policing

Here, we analyse the dynamics and statics of Frank’s (1995) model of the coevolution of selfishness and policing, deriving the results stated in section (a) of the main text.

In the context of this model, the average fitness of an individual is given by equation

(1) of the main text. The direction of selection acting upon any trait is given by the marginal fitness with respect to that trait, i.e. dw/dz and dw/da for selfishness and policing, respectively. These derivatives may be separated into partial derivatives, describing the effect of an individual’s trait value, and the effect of the average trait value of its group, on this individual’s fitness, using the chain rule methodology of

Taylor and Frank (1996; see also Frank 1995, 1996b, 2003; Taylor 1996a; Taylor et al.

2007). This gives dw/dz = ∂w/∂z + r ∂w/∂z´ and dw/da = ∂w/∂a + r ∂w/∂a´, where r

85 = dz´/dz = da´/da is the kin-selection coefficient of (whole-group) relatedness

(Taylor and Frank 1996; Pepper 2000), or:

dw # (1− r)(1− (1− a)z) & = (1− a)% − r(1− ca)( , (A1) dz $ z '

and: €

dw = rz(1− ca) − c 1− (1− a)z . (A2) da ( )

First, we consider the evolution of selfishness in the absence of policing (a = 0). € Evaluating the RHS of equation (A1) at a = 0 and z → +0 gives +∞, which is positive

and hence selfishness is always favoured in the absence of policing. The ESS level of

selfishness z* is determined by setting the RHS of equation (A1) to zero and solving

for z; in the absence of policing (a = 0), this gives z* = 1-r. More generally, holding the

level of policing fixed at a, the ESS level of selfishness is given by z*=(1-r)/[1-a(1-r(1-

c))]. When there is full policing (a*=1) selfishness is neutral (dw/dz=0), but in the

limit of full policing (a → 1), the ESS level of selfishness is z*→(1-r)/(1-c)r. This is

stable as any reduction in selfishness (z=[(1-r)/(1-c)r]-δ, where δ is a vanishingly

small quantity), favours an increase in policing (dw/da>0).

Next, we determine when policing can evolve, by evaluating the RHS of equation

(A2) at a = 0 and z = 1-r. This gives dw/da=r(1-r)-cr, and hence the condition for

policing to invade is r<1-c. Assuming that policing does evolve, we determine the

evolutionary stability of full policing (a = 1) by evaluating the RHS of equation (A2)

at a = 1 and z =(1-r)/(1-c)r, which obtains dw/da=1-c-r. Hence, full policing is

evolutionarily stable (a* = 1) if r < 1-c. Setting the RHS of both equations (A1) and

(A2) to zero, and simultaneously solving for a and z, yields no biologically feasible

86 solutions; hence, an intermediate level of policing (0 < a* < 1) is never evolutionarily stable. This means that when r>1-c, policing cannot evolve (a* = 0), whereas when r<1-c, then policing can evolve, and the level of policing increases until full policing is established (a* = 1). Policing favours an increase in selfishness: differentiating the

ESS level of selfishness z*=(1-r)/[1-a(1-r(1-c))] with respect to a yields a positive slope

(dz*/da=[(1-r)(1-(1-c)r)]/[1-a(1-(1-c)r)]2>0) when r<1/(1-c) which is always met as

1/(1-c)>1.

To investigate whether policing evolves due to direct or indirect fitness benefits, we must rearrange the marginal fitness for policing to be in the form of Hamilton’s rule.

Above, we expressed the marginal fitness for policing as dw/da= dw/da+r(dw/daʹ), where r is the whole-group relatedness (relatedness to average group member, including onself; Pepper 2000). We now re-write in terms of others-only relatedness R

(relatedness to average group member, not including self; Pepper 2000), which is the relatedness term that appears in a formal statement of Hamilton’s rule. Substituting in the expression r=1/n + [(n-1)/n]R, we may rearrange equation (A2) into the form

RB-C, where -C is the direct fitness effect and RB is the indirect fitness effect and their sum is the inclusive fitness effect (Hamilton 1963, 1964b). Hence, we have B=[(n-

1)/n](dw/daʹ) and –C=dw/da+[(1/n)(dw/daʹ)]. For Frank’s model, this obtains

B=[(n-1)/n](z(1-ac)) and C=c(1-(1-a)z)-[(1/n)(z(1-ac))]. We evaluate B and C at the ESS a=a*=1 and z=z*=(1-r)/(1-c)r, which gives B = [(n-1)(n(1-c)-1)(1-R)]/[(1-c)n(1+R(n-1))] and C =c+1/n-1/(1+R(n-1)). Social behaviours are classified according to the sign of B and C (Hamilton 1964; Table 1.1): we find that policing is altruistic (B > 0 and C > 0) if

R > [n(1-c)-1]/[(n-1)(1+cn)], and otherwise it is mutually beneficial (B > 0 and C < 0) – and this is Result 1 of the main text.

87 3.4.2 The individual and group cost of selfishness

Here, we determine the evolutionarily stable levels of policing and selfishness for our

modified model of policing (sections (b) and (c) of the main text). Note that the model

presented in section (c) of the main text is a generalization of the model presented in

section (b); thus, the results of both sections can be derived from the model of section

(c), with those pertaining to section (b) being given in the special case of b = 1. The

average fitness of an individual is given by equation (4) of the main text. Marginal

fitness for selfishness (z) and policing (a) are given by:

dw (1− r)(1− a) = − (1− ab) 1− a(1− r) , (A3) dz z ( )

and: €

dw = brz − c. (A4) da

We first consider selfishness. Holding the level of policing fixed at a, and setting the € RHS of equation (A3) equal to zero and solving for z, we obtain an ESS level of

selfishness of z*=[(1-a)(1-r)]/[(1-ab)(1-a(1-r))]. If we assume there is no policing (set

a=0), as in Frank’s model, this gives selfishness as z*=1-r. The marginal fitness for

policing at this point is given by evaluating equation (A4) at z = 1-r, obtaining dw/da

=br(1-r)-c, and hence policing can invade only when r(1-r)>c/b (Results 2 and 6 of the

main text). For full policing (a=1) when b=1 we find that selfishness is neutral

(dw/dz=0) at a=1. In the limit of full policing (a → 1) selfishness tends to z* →(1-r)/r

(Result 2 and 3, main text). This ESS is stable as any reduction in selfishness (z=(1-

r)/r-δ where δ is a vanishingly small quantity) favours an increase in policing

(dw/da>0). For full policing (a=1) when b<1 we find that dw/dz=-(1-b)r, so

selfishness is disfavoured and will decline to zero and this obtains dw/da=–c which

88 is always negative. Hence for b<1, full policing is never stable (Result 4 of the main

text). This means that when policing can invade from rarity (r(1-r)>c/b), the ESS

level of policing must be intermediate (0 < a* < 1). To find this explicitly, we set the

RHS of equations (A3) and (A4) to zero and simultaneously solve a=a* and z=z*. This

yields two results, only one of which is a biologically feasible ESS at z*=c/rb and

2 bc + (c − br)(1− r) + 4bc(1− r)(br(1− r) − c) + (bc + (c − br)(1− r)) a* = (A5), 2bc(1− r)

(Result 4 and 8 of the main text). The maintenance and stability of this ESS (Results 6 € of the main text) is discussed in the next section. Using the approach outlined in the

previous section, we may rearrange the RHS of equation (A4) into the form RB-C,

where B=bz(n-1)/n and C=c-bz/n. Evaluating this at the above ESS values of z* and

a*, we find that B > 0 and C > 0, and hence policing is always altruistic in this variant

model (Result 8 of the main text). Finally, we examine the impact of b – the efficiency

with which a policing behaviour restores the group resources – upon the ESS level of

policing.

Differentiating the RHS of equation (A5) with respect to b obtains:

2 da∗ bc − (c − br)(1− r) − 4bc(1− r)(br(1− r) − c) + (bc + (c − br)(1− r)) = (A6), 2 db 2b2 4bc(1− r)(br(1− r) − c) + (bc + (c − br)(1− r))

which is always positive, hence the ESS level of policing is stable at higher values as € the efficiency of policing improves (see Figure 3.2b).

3.4.3 Demography and policing - Illustrative overview

Here, we consider the evolution of policing and selfishness in a demographic context,

using the illustrative overview presented in section (d) of the main text where the

89 average fitness of an individual is given by equation (5). The marginal fitness for

selfishness (z) and policing (a) are given by:

dw (1− a)(1− r) − (1− ab)(1− a(1− r) − rs)z = , (A7) dz z(1− ac − z(1− ab))

and: €

dw (1− s)brz − c(1− rs) = . (A8) da 1+ a(bz − c) − z

First we consider the evolution of selfishness. The ESS level of selfishness (z*) is € found by setting the RHS of equation (A7) to zero and solving for z; in the absence of

policing (a = 0), this gives z* = (1-r)/(1-rs). More generally, if the level of policing is

held at a, the ESS level of selfishness is given by:

(1− r)(1− a) z* = (A9) (1− ab)((1− a)(1− r) + (1− s)r)

We determine the evolutionary stability of policing by evaluating the RHS of € equation (A8) at a = 0 and z = (1-r)/(1-rs), which obtains an invasion condition for

policing of r(1-r)[(1-s)/(1-rs)2]>c/b. For full policing (a=1) we find that dw/dz=-[(1-

b)r(1-s)]/[1-c-(1-b)z] which is zero for b=1 and negative for all b<1, so selfishness is

neutral in the special case of fully efficient policing and is otherwise disfavoured and

is expected to decline to zero. Evaluating equation (A8) at z = 0 and a = 1 yields –[c(1-

rs)]/(1-c), which is always negative and hence full policing (a=1) is never stable. This

means that when policing can invade from rarity (r(1-r)[(1-s)/(1-rs)2]>c/b), the ESS

level of policing must be intermediate (0 < a* < 1). To find this ESS, we set both

equations (A7) and (A8) equal to zero, and solve simultaneously for a and z at a=a*

and z= z*. This yields a single biologically feasible result of z*=c(1-rs)/br(1-s)) and a

90 rather complicated a* (which can be recovered by replacing Queller’s relatedness (r-

sr)/(1-sr) for r in equation (A5)).

To show that policing is maintained more readily that it invades we evaluate the

marginal fitness for policing (A8) at the ESS level of selfishness (A9). This yields:

dw br(1− a)(1− r)(1− s) − c(1− ab)(1− a(1− r) − rs)(1− rs) = . (A10) da (1− ab)(ca2(1− r) + r(1− s) − a(c(1− rs)))

Evaluating equation (A10) at a = 0 and c=br(1-r)[(1-s)/(1-rs)2] (the maximum cost of € policing at which policing can invade), we find some policing is favoured (dw/da>0)

when b>ρ (where ρ is Queller’s relatedness (r-rs)/(1-rs)). As full policing a = 1 is

never favoured (dw/da<0), this means an ESS must exist between 0

dw/da=0 outside the region where policing can invade (Result 6 of the main text).

We conduct a numerical investigation of the stability of this ESS using the

methodology of Otto and Day (2007, pp 237-242; results in Figure 3.4) which confirms

that when policing can invade, the ESS is always stable and that policing is

maintained in regions where it cannot invade (particularly at lower relatedness and

higher c/b ratios). When the ESS is stable, the eigenvalues are negative complex

numbers indicating that policing and selfishness spiral in towards their ESS values.

To examine the effect of local competition s on the coevolution of selfishness and

policing, the invasion condition r(1-r)[(1-s)/(1-rs)2]>c/b is solved for r, to yield the

maximum and minimum relatedness values that permit policing to invade from

rarity. These bounds take the form rmax=X+Y and rmin=X-Y where:

1− s(1− 2c b) X = , (A11) 2(1− s(1− sc b))

and: €

91 (1− s) 1− 4c b Y = . (A12) 2(1− s(1− sc b))

To know whether local competition favours or disfavours the evolution of policing, € we first examine how an increase in local competition affects the range of relatedness

values over which policing may invade. This range evolve is given by rmax-rmin = 2Y.

Differentiating Y with respect to s yields

1 dY s(c b)(1− 2 s)(1− s) 1− 4c b = − 2 ds (1− s)(1− s(1− s(c b)))

(A13),

€ which is negative when c/b < ¼, which is a necessary (but not sufficient) condition

for the evolution of policing, hence the range of relatedness values over which

policing invades decreases with increasingly local competition (d(rmax-rmin)/ds < 0;

Result 9 of the main text).

If increasing local competition leads to policing invading at higher levels of

relatedness, then drmax/ds and drmin/ds are positive. For these to be positive, both

dX/ds must be positive and dX/ds must be greater than dY/ds. Differentiating X

with respect to s yields:

2 1 dX c b(1− s + s ( 2 − c b)) = 2 (A14), ds (1− s(1− sc b))

which is positive when c/b>0. Setting the RHS of equation (A13) equal to the RHS of € (A14) gives the condition that dX/ds> dY/ds when c/b>0. Hence as local competition

increases, policing is favoured to invade at higher levels of relatedness (Result 8 of

the main text).

92 To investigate the effect of increased local competition on selfishness we differentiate

z* (A9) with respect to s, which yields

dz∗ r(1− r)(1− a) = 2 (A15). ds (1− ab)(1− a(1− r) − rs)

This is positive for all 0 < r < 1, which is a necessary (but not sufficient) condition for € the evolution of policing which means selfishness increases as competition becomes

more localised (Result 10 of the main text).

3.4.4 Demography and policing - Island Model

We now consider the evolution of policing in an island model setting (Wright 1931;

Taylor 1992a), as described in the main text of section (e). We assume that the

average fecundity F of a focal individual is given by equation (6), the average relative

fecundity of all individuals in a group F' is given by equation (7) and the average

relative fecundity of all individuals in the population F is given by equation (8) in

the main text. The focal individual has KF offspring of which mKF disperse and (1-

m)KF remain in the natal group. Each dispersing offspring arrives in a new group

with (1-m)FnK natives and mFnK other immigrants. Therefore the chance that a

particular dispersing offspring succeeds in obtaining one of the n breeding spaces is

n/[1+ (1-m)FnK +mFnK] ≈ 1/FK assuming K is very large. Therefore, the expected

fitness of the focal individual that is obtained through its dispersing offspring is

mFK/FK = mF/F. Turning now to the (1-m)FK non-dispersing offspring of the

focal individual, these find themselves competing for the n breeding spaces in the

natal patch with (1-m)F'K natives (including each other) and mFnK immigrants,

therefore the total number of offspring is (1-m)F'nK+mFnK. Hence the expected

fitness of the focal individual that is obtained through its non-dispersing offspring is

93 [(1-m)FnK]/[(1-m)F'nK+mFnK] =[(1-m)F]/[(1-m)F'+mF]. Thus the total number of

dispersing and non-dispersing offspring of a focal individual (which is equal to an

individual’s fitness) is given by:

F F(1− m) w = m + (A16) F (1− m)F # + mF

Using equation (A16), we substitute the fecundity values given in section (d) to € calculate the marginal fitnesses for selfishness (z) and policing (a) where:

2 2 dw brz 1− (1− m) − c 1− r(1− m) = ( ) ( ) (A17), da 1+ a(bz − c) − z

and €

2 dw (1− a)(1− r) − (1− ab) 1− a(1− r) − r(1− m) z = ( ) (A18). dz z(1+ a(bz − c) − z)

Note that equations (A7) and (A8) recover equations (A17) and (A18), respectively, if € we make the substitution s = (1-m)2, i.e. in the context of the inelastic island model the

scale of competition is equal to the probability that two randomly chosen group

mates both fail to disperse (Gardner and West 2006). Relatedness r may also be

expressed as a function of model parameters, by deriving a recursion for relatedness

over successive generations and calculating its equilibrium value (Taylor 1992a). In

generation t, the expected relatedness of an individual to a randomly chosen member

of its group (including itself), is rt = 1/n + ((n-1)/n)Rt where 1/n is the probability of

choosing itself (in which case relatedness is 1) and (n-1)/n is the probability of

choosing another individual (in which case relatedness is Rt, the “others-only”

relatedness). The only way for two different individuals from the same group to be

related is if neither of them dispersed (which occurs with a probability of (1-m)2), in

94 which case their relatedness is equal to the whole-group relatedness of the previous

generation (rt-1). Hence, we may write the following recursion:

1 n −1 2 rt = + (1− m) rt−1 (A19), n n

which can be solved at equilibrium (rt-1=rt=r) to give an equilibrium relatedness of: € 1 r = 2 (A20), n − (n −1)(1− m)

as determined by Taylor (1992a). Substituting equation (A20) into equations (A17) € and (A18) obtains:

2 dw 1− (1− m) ((1− a)(n −1) + (1− ab)(a(n −1) − n)z) = ( ) (A21), 2 dz z(1+ a(bz − c) − z)(1+ (n −1)(1− (1− m) ))

and: €

2 dw 1− (1− m) (bz − cn) = ( ) (A22). 2 da (1+ a(bz − c) − z)(1+ (n −1)(1− (1− m) ))

We now determine the evolutionarily stable levels of policing and selfishness. € Considering selfishness first, we find the ESS level of selfishness (z*) by setting the

RHS of equation (A21) to zero and solving for z whilst holding the level of policing at

a; this gives:

(1− a)(n −1) z* = (A23). (1− ab)(a + n(1− a))

In the absence of policing (a = 0), this is z* = (n-1)/n. Evaluating equation (A22) at z* € = (n-1)/n, gives an invasion condition for the evolution of policing of (n-1)/n2 >c/b

95 (Result 11 of the main text). To find the stable level of policing, we set both equations

(A21) and (A22) equal to zero, and solve simultaneously for a and z at a=a* and z=z*.

This yields one biologically feasible solution of z*=(c/b)n and a complicated a*, which

can be recovered by substituting ρ=1/n = r in equation (A5) (Result 12 of the main

text). To investigate whether policing may be maintained in larger groups than it can

invade, we evaluate the marginal fitness for policing (A22) at the ESS level of

selfishness (A23), which yields:

2 2 dw m(m − 2) cn(a + n(1− a)) + b(1− n + cna (n −1) + a(n −1− cn )) = ( ) (A24) da (1− ab)(1+ 2m(n −1) − m2 (n −1))(1+ ca2(n −1) − acn)

Evaluating equation (A24) at a = 0 and c=(b(n-1))/n2 (the maximum cost of policing at € which policing can invade), we find policing is still favoured (dw/da>0) when b>1/n

(which is equivalent to the open model result of b>ρ as for the Island model,

Queller’s relatedness ρ = 1/n). Full policing a = 1 is not stable (dw/da<0), so an ESS

must occur between 0

We conduct a stability analysis for the Island model ESS using the methodology of

Otto and Day (2007, pp 237-242). The results can be obtained from Figure 3.4 by

substituting s for (1-m)2 and r for 1/[n-(n-1)s].

To investigate how changing migration rates affect both relatedness and the intensity

of local competition, we first differentiate relatedness (r=1/(n-(n-1)(1-m)2) with

respect to m, which yields dr/dm=-[2(1-m)(n-1)]/[1+(1-(1-m)2)(n-1)]2. This is always

negative, so increasing the migration rate decreases relatedness. Differentiating the

scale of competition term, s=(1-m)2, in terms of m gives ds/dm=-2(1-m) which is also

always negative meaning that increasing migration decreases local competition. Both

these effects must cancel resulting in the migration rate m having no net effect on the

96 evolution of policing as a* and z* and ρ are not functions of m, so da*/dm = dz*/dm = dρ/dm=0 (Result 11, main text).

97 Chapter 4. Group competition and the evolution of cooperation in humans.

4.1 Introduction

Darwin (1871) was the first to suggest that hostility between groups could have played a key role in the evolution of cooperative social behaviours in humans. The presence of cooperation in human society poses a problem from an evolutionary perspective as, all else being equal, helping another individual reduces the relative fitness of the helper so natural selection should favour non-helpers over helpers

(Hamilton 1964). Theoretical and empirical work suggests that conflict or war between groups may favour the evolution of cooperative behaviours, because the benefit of being in a winning cooperative group may outweigh the cost of cooperation (Choi and Bowles 2007, Bowles 2006; 2009; Lehmann and Feldman 2008).

Theoretical model of group conflict have been parameterized with archaeological and ethnographic data, to estimate empirically the influence of conflict (Bowles 2006;

2009). However, it is not always clear exactly why cooperation is favoured in these models.

In this paper we ask whether conflict only favours pro-social traits (e.g. teamwork) or whether it can also favour anti-social traits (e.g. uncontrolled aggression). This latter case, that conflict between-groups may favour harmful traits, has not been examined.

Furthermore, we examine to what extent predictions from existing models are applicable to all forms of social behaviour, as currently it is unclear whether conflict

98 favours any sort of social trait (e.g. food sharing), or just those directly related to conflict (e.g. helping to make spears for the group) (Lehmann and Feldmann 2008).

Our investigation into the influence of group competition on the evolution of social behaviours uses Bowles’s (2006; 2009) theoretical framework as a starting point to illuminate the selective forces involved.

4.2 Model and Analysis

The model we develop is only intended to provide a heuristic overview and for that reason, we consider the simplest possible case of a social behaviour p, which influences both the relative reproductive success of the individual within the group, and the ability of her group to compete with other groups. We assume the social behaviour is directly costly to the individual performing it so that if she invests in this trait, she incurs a fitness cost (scaled by c, such that for an individual i, in group j, her relative fitness within the group is reduced by cpij). We further assume that the social behaviour has a fitness effect on the pj group members (we do not specify the way in which this fitness effect is conferred). This effect is scaled by b, such that the individual’s relative fitness within her group is changed by bpj. For generality, we allow b to be positive or negative, and so consider social behaviours that either help or harm the other group members. For example, a helping trait may be collective childcare and a harming trait might be fighting other group members.

In addition to affecting within-group relative fitness via helping or harming, we assume that investing in the social behaviour, p helps the group succeed if it comes into conflict with other groups. Every generation we assume, that with probability κ, the group will compete against another randomly chosen group in the population and that the group will win that contest with probability λ. Bowles (2006, 2009) used

99 different functions for λ and we examine both here. In Bowles (2006), λ was related to the average (absolute) level of the social behaviour expressed by members of group j

(so λ=f(pj)) and in Bowles (2009), λ was related to the average level of the social behaviour expressed by the group pj, relative to the population average, p (so

λ=f(pj/p)). Following Bowles (2006; 2009), we assume competitive encounters are ‘all or nothing’ events, so if the group loses, it goes extinct and if it wins it momentarily doubles in size and then splits into two daughter groups (of equal size). We assume that only between-group competition determines the group’s fitness, so if groups do not compete (which occurs with probability 1-κ), group fitness is 1. Summing these

probabilities gives the average fitness of a group j as wj = 0[κ(1-λ)]+2[κλ]+1[(1-κ)] = 1-

κ+2κλ.

Table 4.1. A summary of symbols used in the analysis

Symbol Definition Symbol Definition p social behaviour r within-group relatedness c personal cost w Darwinian fitness b helping or harming n group size

fitness benefit to the recipient of a social frequency of group contest B κ behaviour

probability of winning a C fitness cost to actor of a social behaviour λ contest

The overall fitness of an individual i in group j, wij, is found by multiplying her success within the group by her group’s resources, so the more resources she has, the higher her fitness. More formally, if the group’s resources are given by FG, and her

100 share is determined by her within-group success FI, relative to her group mates’

successFI, then her fitness wij = FG(FI/FI). Here, we have

1+ bp j − cpij wij = (1−κ + 2κλ) (1) 1 bp cp + j − j ,

where pij is the individual’s investment into the social behaviour and pj is the average € level of the social behaviour performed by members of the group. Model notation is

summarised in Table 4.1. We use the neighbour modulated fitness modelling

approach (Taylor and Frank, 1996) to assess the evolutionary stability of the social

behaviour p.

4.2.1. Probability of winning a group contest, λ is related to the average (absolute) level of the social behaviour in the group.

Here, we follow Bowles (2006) and assume that the probability of winning a contest

λ, is related by a scaling factor z to the average level of the social behaviour pj, so that

the probability of winning a conflict is given by: λ=z +(1-2z)pj, where z may vary

between 0 and 0.5. When z=0, the average level of the social behaviour, pj is equal to

the probability of winning and as z increases, the social behaviour becomes less

important in conflict such that when z=0.5, the probability of winning a conflict is

independent of pj. In Appendix 4.4, we show three main results.

First, group conflict favours the evolution of both helping (+b) and harming (-b)

behaviours (Figure 4.1A). All that matters is that the within-group social behaviour

(b) is correlated with winning contests (λ), and that the inclusive fitness benefit the

individual gains from winning contests outweighs the costs (-c and –b) incurred by

performing the behaviour. The within-group effect (b) of the social behaviour (p) does

affect the evolutionary stable level of the social behaviour (ESS; Maynard Smith and

101 Price, 1973), but it does not affect when the social behaviour can invade from rarity

(Figure 4.1A). The reason why helping and harming are both favoured may be because we followed Bowles (2006, 2009) and assumed that within-group social

behaviour (b) does not affect the group’s fitness (wj=1-κ+2κλ). Here group fitness is solely determined by the effect between-group competition has on the group’s resources (FG). For this reason, in the absence of group contest, the social behaviour is never favoured (p*=0; dw/dp<0 if κ=0) as if an individual incurred the cost c of performing the social behaviour, it would gain less of the group’s resources relative

to any non-social group mates (FI <FI), resulting in a lower relative fitness.

Figure 4.1 Helping and harming behaviours may be favoured when groups compete. The within-group social behaviour, b may be helpful (+b) or harmful (-b), however this has no effect on the invasion condition for the social behaviour (p>0; solid lines are b=0, right/lower dotted lines are b=-0.3, left/higher dotted lines are b=0.3). Whether the social behaviour is favoured or not depends on its ability to improve the group’s chance at surviving contests, λ. If group contests are more frequent (higher κ), this favours higher levels of the social behaviour because the benefits from increased group survival will outweigh the costs of performing it (determined by b and c). If we follow Bowles (2006) and assume the group’s chance of

surviving is related to the average level of the social behaviour in the group (λ=f(pj); Panel A), we find when contests are infrequent (low κ), the social behaviour is not favoured. By contrast, if we follow Bowles (2009) and assume the group’s chance of surviving is related to the average level of social behaviour in the group relative to the population average level of the social

102 behaviour (λ=f(pj/p); Panel B), then, if groups compete, the social behaviour is always favoured whatever the cost (p*>0 when κ>0). Plotted for n=32, R=0.076 (where r=(1/n)+((n-1)/n)R), Panel A: z=0; Panel B: µ=2.

Panel A. Absolute: λ=f(pj) Panel B. Relative: λ=f(pj/p)

Figure 4.2 The social behaviour is either altruistic or mutually beneficial. When compared to the fitness of a rare defector who does not perform the social behaviour, the social behaviour may be altruistic or mutually beneficial. It may sound nonsensical to say a harming behaviour is also be altruistic or mutually beneficial, but this is because harming represents only a partial fitness effect, whilst the ‘B’ and ‘C’ terms in Hamilton’s rule captures the overall fitness effect, which includes the group-level benefit the social behaviour yields due to increased group survival. Hamilton’s rule states that a behaviour or trait will be favoured by selection, when RB-C>0; where -C is the fitness cost to the actor, B is the fitness benefit to the recipient weighted by R, their others-only relatedness. A behaviour is termed altruistic if B>0 and C>0, mutually beneficial if B > 0 and C < 0 (Hamilton, 1964; West et al., 2007b). Panel A plotted for n=10, r=0.5, κ=0.3 and z=0. Panel B plotted for n=10, r=0.3, κ=0.6 and µ=10.

Second, when compared to the fitness of a rare defector who does not perform the social behaviour (p), the social behaviour may be altruistic or mutually beneficial

(Figure 4.2A; Table 1.1). According to Hamilton’s rule, social behaviours are favoured when RB-C>0, where R is the average relatedness between actors and recipients and

B and C are the fitness effects on the recipient and actor respectively (Hamilton,

1964). The social behaviour always yields indirect fitness benefits, even if the within- group term b is negative, because the benefit group mates receive from increased group success always outweighs any fitness effect of the within-group trait, b. Note, the ‘B’ and ‘C’ terms in Hamilton’s rule differ from ‘b’ and ‘c’ model terms because

103 they capture the overall fitness effect, not the within-group component of fitness.

Altruism occurs when the actor incurs a direct fitness cost (-C>0) whilst group mates enjoy a fitness benefit (B>0) and mutually beneficial behaviours occur when both the actor and recipients benefit (-C<0, B>0). In this model intermediate levels of the social behaviour, p*, are always altruistic. When p*=1, the behaviour is mutually beneficial when the cost, c is very low (the value is given in equation A11 of Appendix 4.4), otherwise the behaviour is altruistic (Figure 4.2A).

Third, the optimal level of the social behaviour (p*) is correlated with the likelihood of group contests (κ), within-group relatedness (r) and the cost of performing the social behaviour (c). The optimal level of the social behaviour is positively correlated with the likelihood of group contests, κ (dp*/dκ>0), as if conflicts becomes more likely, the benefit gained by winning more of them becomes greater relative to any within group costs. It is positively correlated with within group relatedness

(dp*/dr>0, where r, is Wright’s (1922) coefficient of relatedness, which measures how much of the genetic variance in the social behaviour occurs between-groups) as higher within-group relatedness increases the indirect fitness benefits the individual receives. Finally, the optimal level of the social behaviour is negatively correlated with the individual cost of performing the behaviour, c as higher costs reduce the overall fitness benefit an individual receives. If the social behaviour is very costly, it cannot evolve (p*=0 when c>[2r (1-2z)κ]/[(1-r)(1-(1-2z)κ)]) and conversely, when the social behaviour is very cheap, maximum expression occurs (p*=1 when c<[2(1+b)r(1-

2z)κ]/[1-r+(1-r)(1-2z)κ]). For costs between these values, intermediate levels of the social behaviour are favoured (p*=[2κr(1-2z)- c(1-r)(1-(1-2z)κ]/[2κ(1-2z)(c-br)]).

104 4.2.2 Probability of winning a contest is related to the relative level of the

social behaviour expressed by the group (pj/p).

Bowles (2006) assumes that the likelihood of the winning a between-group contest, λ

is determined by the average (absolute) investment in social behaviour by group

members. However, the likelihood of winning is more likely to depend upon how

much the group invests in social behaviour relative to the groups it competes against

(i.e. whoever invests more wins; Bowles 2009). To allow for this, we follow Bowles

(2009) and define the conflict success term, λ, in (1) as:

1 λ = , (2) $ p j ' 2µ&1 − p ) 1+ e % (

where e is the natural logarithm and µ is a scaling function. With this adjustment, if €

the average investment by group members in the social behaviour (pj) is the same as

the population average (p), then on average it is evenly matched against other groups

so will win contests with a probability ½ (i.e. when pj=p, λ=0.5). If the group invests

below the population average in the social behaviour (pj

half the time (λ<0.5), and likewise, if it invests more than the population average

(pj>p), it will win more than half the time (λ>0.5). The scaling function, µ determines

the degree to which having higher/lower levels of social behaviour than the

population average alters the probability of winning (becoming more pronounced as

µ increases). In Appendix 4.4 we show that this modification qualitatively changes

the previous results in two ways.

First, if there is any competition between groups (κ>0), then the social behaviour is

always favoured (p*>0), irrespective of how costly it is (Figure 4.1B). This is because if

groups did not perform the social behaviour, any group investing in a non-zero level

105 of the social behaviour would win every contest, and all other groups would go extinct. Consequently, no matter how costly, a non-zero optimal level of the social behaviour (p*) is always favoured (p*=1 when c< [rκµ(1+b)]/[1- r(1-κµ)] and when the cost is higher than this p*=(κµr)/[c(1-r(1-κµ))-brκµ)]). With the previous assumption, a fractional increase in the level of the social behaviour led to only a fractional increase in a group’s success so the cost of investing in the trait could be greater than the benefit. Second, whilst the Hamiltonian classification still shows that the behaviour is always mutually beneficial or altruistic, now intermediate levels of the social behaviour, p*, may be mutually beneficial or altruistic (Figure 4.2B). As before, the social behaviour is mutually beneficial when the cost is low, and altruistic when the cost is high (the condition is given in A13 of Appendix 4.4).

4.3 Discussion

We investigated the effect of competition between groups on the evolution of social behaviours. For completeness, we calculated the likelihood a group will succeed in conflict success in two different ways: linked to (1) the average expression of the social behaviour in the group (absolute; figures 4.1A and 4.2A) and (2) this average expression relative to the population average expression (relative; figures 4.1B and

4.2B). Our results show that whichever way we calculate the probability of success, between-group competition can favour any social behaviour that helps the group win contests, be it pro-social (associated with helping group-mates) or anti-social

(associated with harming group mates). Furthermore, we show that when it evolves, this behaviour is always altruistic or mutually beneficial because, even if it causes harm to group mates, the costs are more than compensated by the benefits of greater success in conflicts. As we show that helpful or harmful social behaviours are

106 favoured so long as they are associated with improved competitiveness, our results complicate the widely held view that the presence of competition between groups may be a mechanism driving the evolution of general pro-social preferences in humans.

Our results suggest that group competition only favours within-group social behaviours that are directly associated with the ability to win conflicts. We follow earlier models (Bowles 2006; 2009) and assume that the evolution of any within- group helping or harming behaviour, b, is perfectly linked to the group’s ability to win contests, λ (they are both properties of a single trait p). This assumption means that the within-group social behaviour, b in our model is not an adaptation i.e. it is not selected due to its fitness-maximising properties; it occurs because we assume it is inseparable from a behaviour that is adaptive (the ability to win conflicts). Whilst natural selection will always minimize the deleterious consequences of an adaptive behaviour, the adaptive benefits of some behaviours (e.g. aggression) may be inseparable from their harmful side effects.

Our model assume that the average fitness of a group member (wj) is entirely determined by competition between groups. This is unrealistic; if group members harm each other we would expect their group to do worse than a group where individuals help each other. However we favour this assumption for two reasons.

First, we are interested in the simplest case, and if we added the complication that within-group social behaviours could affect group fitness, it would only make a quantitative difference to our results. There would still be parameter values where helping and harming are possible. Second, relaxing this assumption would not aid our understanding of how group competition influences the evolution of social behaviours. It would merely capture the well-known fact that helping behaviours

107 may be favoured if they increase group fitness and harming behaviours may be disfavoured if they decrease group fitness. In reality, competition between groups will not be the only factor affecting the average fitness of a group member. For example fitness will also be affected by how group members share food, care for children and sanction bad behaviour. Given that many factors contribute to an individual’s inclusive fitness, there are always evolutionary trade-offs and so even if harming behaviour helped groups succeed in conflicts, it may not be favoured by natural selection if, for example, people who harmed made bad parents.

Our results do not support the view that between-group competition alone can favour generalized individually costly pro- or anti-social tendencies. Instead, we find it only favours behaviours directly associated with increased competitiveness.

However, our model considers group competition in isolation; a fruitful avenue for future research will be to re-examine the established evolutionary explanations for cooperation in the presence of group competition (West et al. 2007b). For example, between-group competition may favour cooperation between non-kin when combined with the mechanism of group augmentation (Kokko et al. 2001). Models of group augmentation assume that group size correlates with individual survival, and show that costly cooperation between non-relatives can evolve if it helps groups to grow in size (e.g. by helping to rear unrelated offspring; Kokko et al. 2001). As group size is probably a good predictor of success in contests, the presence of between- group competition may favour the evolution of cooperation in large non-kin groups.

It will be interesting to see whether group competition also increases selection for punishment mechanisms (Clutton Brock 1995).

Our results show the social behaviour always provides indirect fitness benefits, no matter how harmful it is (Figure 4.2). This result highlights the important difference between specific fitness effects (such helping or harming other group members) and

108 the overall fitness consequences of a behaviour (in this case altruistic or mutually beneficial). Hamilton’s rule measures the overall lifetime fitness consequences of a behaviour, not partial fitness effects such as the fitness effects within a group or effects measured over a short timescale (Hamilton 1964). It is a complete and general description of the action of natural selection (Gardner et al. 2011). Behaviours are often altruistic, even if they have short-term costs. For example, if bees die stinging a honey badger, colony productivity will drop which is costly to the remaining bees, however stinging remains altruistic because the benefit of having the hive saved far outweighs any costs the remaining bees suffer from reduced productivity. Similarly, in this model, the harmful social behaviour is indirectly beneficial as the costs group mates incur from being harmed are outweighed by the benefits they gain if their group wins contests. It is worth noting that models sometimes adopt other definitions for altruism by defining the relatedness, benefits and costs differently from the very specific definitions used in Hamilton’s rule (Table 1.1; West et al. 2011).

For example, previous models of group competition (Bowles 2006, 2009) defined altruism as individually costly but group beneficial by likening the model’s ‘b’ and ‘c’ terms to the indirect and direct fitness terms in Hamilton’s rule. If we used that definition here, we would have classified the harming behaviour as spiteful (West et al. 2011), allowing us to conclude that between-group competition favours the evolution of spite in humans. This potential for confusion, highlights why it is important to use the Hamiltonian definitions when discussing the evolution of social behaviours.

How could our models be tested with empirical data? One possibility, used by

Bowles (2006, 2009) is to attempt to parameterise our model with empirical data, and examine if that leads to conditions where certain forms of social behaviour would be favoured. However, Frank (1998) has argued that this is not how such theory should

109 be used. The problem is that countless other models could be developed which also predict the occurrence of social behaviours, in response to a number of other factors, and so it would be impossible to distinguish a role of between group competition per se. Put another way, if alternative hypotheses are allowed for, it is hard to distinguish between them. This is the case even with much simpler traits, where the relevant traits can be measured much more easily, such as population sex ratios (West 2009), let alone with the archaeological data used by Bowles (2006, 2009), which will be subject to huge confidence intervals. The alternative and much more powerful approach is to use our model to make comparative predictions for how the levels of social behaviours should vary across groups, populations, cultures or species (Frank

1998; West 2009). For example, one could test our model predictions by seeing whether social behaviours are correlated with the frequency of between group conflicts or success rates in conflicts.

To conclude, our aim here was not to argue that group competition could not be a factor in the evolution of helping. One can readily imagine why traits for coalitional, cohesive groups may be favoured. However, one can also imagine why traits that reduce cooperation within the group, such as aggression and violence, could also be favoured when groups compete. We suggest that the presence of group competition alone is insufficient to explain the existence of pro-social, cooperative groups and that more theoretical and empirical work is needed to understand the role of group competition on the evolution of cooperation.

4.4 Appendix

Here, we determine the evolutionarily stable levels of the social behaviour for our group competition model. We calculate the direction of selection acting upon the

110 social behaviour p from the individual fitness in equation (1) of the main text, by

calculating the marginal fitness with respect to p, i.e. dw/dp. To do this we use the

neighbourhood modulated fitness (or kin selection) approach. This uses the chain

rule methodology of Taylor and Frank (1996) to give dw/dp = ∂wij/∂pij + r ∂wj/∂pj

where r = ∂pj/∂pij which is the kin-selection coefficient of (whole-group) relatedness

(Taylor and Frank, 1996; Pepper 2000). This derivative may be separated into partial

derivatives, which describes the effect of an individual’s trait value (∂wij /∂pij), and

the effect of the average trait value of its group (∂wj/∂pj), on a focal individual’s

fitness.

4.4.1 Contest success is determined by absolute (group average) level of the social behaviour.

We now derive the results presented in section 4.2.1 of the main text by conducting

an invasion and stability analysis of (A1). Assuming that λ=z +(1-2z)pj, in equation (1)

in the main text, at the population average (pij= pj=p), this obtains:

dw c = r(2κ(1− 2z)) − (1− r) 1−κ + 2κ(z + (1− 2z)p) (A1) dp 1+ p(b − c) ( )

First, we evaluate equation (A1) assuming there is no social behaviour (i.e. set p=0), € this obtains dw/dp =2κr(1-2z)+c(1-r)κ(1-2z), and hence the social behaviour cannot

invade (dw/dp <0) when:

c>[2κr(1-2z)]/[(1-r)(1-(1-2z)κ)], (A2)

note this is not a function of the within-group social behaviour term b. When the cost

is below this, some level of the social behaviour is favoured (dw/dp <0). For the

maximum expression of the social behaviour (p*=1) we find that dw/dp=2κr(1-2z)-

111 [c(1-r)(1+κ(1-2z)]/[1+b-c] and hence the social behaviour will reach its maximum

expression when:

c<[2(1+b)r(1-2z)κ]/[1-r+(1-r)(1-2z)κ]. (A3)

For intermediate costs, an intermediate ESS level of the social behaviour is favoured.

To find this, we set the RHS of equation (A1) equal to zero and solve for p, which

obtains:

2κr(1− 2z) − c(1− r)(1− (1− 2z)κ) p* = (A4) 2κ(1− 2z)(c − br)

We now assess how different model parameters affect levels of the social behaviour. € In the absence of group contest (κ=0), equation (A1) becomes dw/dp=-(1-r)[c/(1+p(b-

c))] which is always negative, so selection on the social behaviour, p is neutral when

groups do not compete. When groups do compete, to examine how increasing the

frequency of group contests (κ) affects the ESS level of the social behaviour p we

differentiate the RHS of (A4) with respect to κ. This obtains:

dp* c(1− r) = (A5) dκ 2(c − br)(1− 2z)κ 2

which is positive when 0

relatedness between group members, r affects the ESS level of the social behaviour p,

we differentiate the RHS of (A4) with respect to r, which obtains:

dp* c(c + (2 + b − c)(1− 2z)κ − b) = 2 . (A6) dr 2(c − br) (1− 2z)κ

€

112 This is positive when c>[b-(2+b)(1-2z)κ]/(1-(1-2z)κ). If we compare this inequality to

the boundary conditions for 0

levels of p* are favoured. To examine how increasing the cost, -c of performing the

social behaviour affects its evolution, we differentiate the RHS of (A4) with respect to

-c, which obtains:

dp* 2(1− 2z)κ − b(1− r)(1− (1− 2z)κ) = r 2 (A7) d(-c) 2(c − br) (1− 2z)κ

which is negative when b>2(1-2z)κ/[(1-r)(1-(1-2z)κ)]. For intermediate levels of the € social behaviour, p*, this inequality is always satisfied so increasing the cost will

decreases the stable level of the social behaviour. If the within-group

helping/harming behaviour b, is below this value, intermediate levels of the social

behaviour are not favoured. This can be shown as the condition for p*>0 (A2) is

higher than the condition for p*=1 (A3) when b>2(1-2z)κ/[(1-r)(1-(1-2z)κ)]. Finally, to

investigate how changing levels of within-group helping/harming affect levels of the

social behaviour, we differentiate the RHS of (A4) with respect to b, which obtains:

dp* c(1− r + 2r)(1− 2z)κ − c(1− r) = r 2 (A8) db 2(c − br) (1− 2z)κ

which is positive when c<[2κr(1-2z)]/[(1-r)(1-(1-2z)κ)], which is always satisfied as € this is the condition for the social behaviour to be favoured (see A2).

We investigate whether the social behaviour is maintained by direct or indirect

fitness benefits by calculating the fitness components of Hamilton’s rule (Hamilton

1964) and establishing when they are positive or negative. The change in relative

fitness is ΔW=Wij-W, whereW=1 and Wij=wij/w. If fitness, wij is given by equation

(1) in the main text, then assuming λ=z +(1-2z)pj, Wij is:

113 $ 1+ bp − cp ' 1−κ + 2κ z + 1− 2z p & j ij ) ( ( ( ) j ))& ) % 1+ bp j − cp j ( W = ; (A9) ij 1−κ + 2κ(z + (1− 2z)p)

We consider the change in relative fitness, ΔW, of a rare defector who does not € perform the social behaviour. Assuming weak penetration (the ‘tracer allele’

approach; Grafen 2006b), we have pj =p(n-1)/n to account for the group member

(representing a fraction 1/n of group j) who defects. To find the direct fitness term in

Hamilton’s rule –C, we assume the focal individual is a defector, so we set pij=0 and

pj=(n-1)p in (A9) which yields:

2κp(1− 2z)(n + b(n −1)p) − cpn(n −1)(1−κ + 2κ(z + (1− 2z)p)) −C = − (A10) n(n + p(b − c)(n −1))(1−κ + 2κ(z + (1− 2z)p))

and this is always positive in the region where 00 with the

boundary conditions for p*). To examine the region where p*=1 is favoured, we

substitute p=1 for p in A10 and examine when it is negative (-C<0), which is when:

2(b(n −1) + n)(1− 2z)κ c < (A11) (n −1)n(1+ (1− 2z)κ)

thus –C>0 for costs, c above this. If not performing the social behaviour is directly € costly, this implies performing it directly beneficial. Therefore, at intermediate levels

and for costs above the value given in the RHS of equation (A11), we have a directly

beneficial behaviour, otherwise, the social behaviour is directly costly.

To calculate the indirect benefits term B, we assume the focal individual performs the

social behaviour, but that one member of her group is a defector, so we set pij=p and

pj=(n-1)p in (A9). This yields:

114 cpn(1−κ + 2κ(1− 2z)) + 2pκ(1− 2z)(n + bp(n −1)) B = − . (A12) n(n + p(b − c)(n −1))(1−κ + 2κ(z + (1− 2z)p))

which is always negative in the region where p*>0 is favoured. If not performing the € social behaviour always has indirect costs, this implies performing it indirectly

beneficial, hence our result that the social behaviour is altruistic except for at low

costs when p=1, where it is mutually beneficial (A11; Figure 4.2A).

4.4.2 Contest success is determined by the average level of the social behaviour in the group relative to the population average.

We now derive the results presented in section 4.2.2 of the main text by assuming

2µ(1-(pj/p)) that λ=1/[1+e ] in equation (1) in the main text. At the population average (pij=

pj=p), this obtains:

dw κµr(1+ bp) − cp 2(1− r) +κµr = ( ) (A13) dp 2p(1+ bp − cp)

In the absence of group competition (κ = 0), the social behaviour cannot spread (as

dw/dp<0 when κ = 0). If κ>0, some positive level of the social behaviour (p>0) is

always favoured as p*=0 is never stable (when p=0, dw/dp→+∞). Maximum

expression of the social behaviour (p*=1) occurs when dw/dp > 0 at p=1 which is

when:

κµr 1+ b c < ( ) (A14) κµr + 2(1− r)

When the cost of performing the social behaviour is above this threshold, an

intermediate level of the social behaviour is stable. To find the ESS level of the social

behaviour, we set the RHS of equation (A13) to zero and solve for p which gives:

115 κµr p* = . (A15) c(2(1− r) +κµr) − bκµr

Following the methodology described in section 4.4.1, we now calculate the fitness components of Hamilton’s rule and determine when they are positive or negative. If fitness, wij is given by equation (1) in the main text, then as w=1, Wij =wij. The direct fitness term is:

" % n + bp n −1 κ tanh µ − cp n −1 ( ( )) #$ n &' ( ) −C = − (A12) n + p(b − c)(n −1) and this is positive (-C>0), when:

" % n + bp n −1 κ tanh µ ( ( )) #$ n &' c > (A13) p(n −1) and negative (-C<0) for values of c below this. Both intermediate (0

(p*=1) expressions of p are favoured for costs above and below the value given on the

RHS of A13, therefore the social behaviour may be directly costly (-C>0) or directly beneficial (-C<0) for p ≤1.

Following the methodology described in section 4.4.1, the indirect benefits term B is found to be:

" % n 1+ p b − c − bp κ tanh µ + cp ( ( ( ) )) #$ n &' B = − (A14) n + p(b − c)(n −1) which is always negative (B<0) for all the model parameter values. Thus, if not performing the social behaviour always incurs an indirect cost, this implies that performing it is always indirectly beneficial, hence our result that the social

116 behaviour is either altruistic or mutually beneficial, depending on whether the cost, c lies above or below the cost threshold given in A13 (Figure 4.2B).

117 Chapter 5. Reproductive levelling and the evolution of cooperation between non-kin.

This chapter is submitted for publication in Evolution and Human Behaviour: El

Mouden, C., West, S. A. and Gardner, A. (2011) Reproductive levelling and the evolution of cooperation between non-kin.

5.1 Introduction

Cooperation can be found at all levels of biological complexity, between genes within a genome, between cells within multicellular organisms, and between individuals within societies (Maynard Smith and Szathmáry, 1995; Bourke 2011). However, cooperative behaviours which benefit others whilst reducing the actor’s relative fitness pose a challenge for the biological and social sciences. This is because natural selection favours individuals with a higher relative fitness so, all else being equal, cooperation should be selected against (Darwin 1859; Fisher 1930; Price 1970). A large theoretical literature has arisen to address this problem, showing that cooperation can be favoured between genetic relatives (kin selection) or when cooperation leads to a direct benefit for the actor (reviewed by Sachs et al. 2004; Lehman and Keller

2006; West et al. 2007a).

One potentially important way cooperation in humans may have evolved is via norms or social institutions which promote reproductive levelling (Bowles 2006).

Reproductive levelling is when fitness differences between group members are reduced via the equal distribution of resources, or the limitation of competition

118 between group members, and hence leads to more egalitarian groups. A possible role of reproductive levelling in our evolution is supported by: archaeological and anthropological evidence that ancestral humans lived in broadly egalitarian societies

(Knauft 1991; Fried 1967; Boehm 1982, 1993, 1997, 2001); economic experiments showing that human possess preferences for equality (Dawes et al. 2007); and empirical studies suggesting that these preferences limit inequalities in wealth, status or power (Woodburn 1982; Boehm 1982, 1993, 1997, 2001).

However, from a theoretical perspective, it is not clear how and why reproductive levelling itself should be favoured (Bowles 2006; Boyd 2006; Bell 2009). Previous theoretical work assumes that cultural norms for reproductive levelling are present, and consider its impact upon the evolution of cooperation, without considering the evolution of reproductive levelling itself (Bowles 2006). However, because reproductive levelling can be a form of cooperation, this approach examines the consequences of cooperation (reproductive levelling) for the evolution of cooperation

(other cooperative traits), and hence avoids the ultimate problem of whether reproductive levelling itself would be favoured (West et al. 2011). It has been suggested that natural selection on genes alone cannot explain the extent and complexity of social behaviours found in humans (Bell et al., 2009, Henrich and

Henrich 2007; Richerson and Boyd 2004) These authors suggest that culturally inherited cooperation created a novel selective environment which led to uniquely human forms of genetically inherited cooperation evolving. The results of earlier work on reproductive levelling (Bowles 2006), in particular, have been cited in support of the idea that pre-existing culturally inherited cooperation leads to natural selection favouring genes for sociality that would otherwise be maladaptive (Bell et al. 2009).

119 Our aim here is to examine if and how reproductive levelling can be favoured and whether it can lead to more cooperative groups. The idea of reproductive levelling is captured by models in the biological literature which consider how cooperation can be favoured by what is sometimes termed policing or repression of competition

(Leigh 1971, 1977; Ratnieks 1989; Frank 1995, 2003). In such models, a policing trait is costly to the individual performing it, but leads to a more equitable sharing of resources within groups. We allow such a trait to coevolve with a competitiveness trait that determines the extent to which individuals exploit resources cooperatively or selfishly. Although our interest is the scenario when reproductive levelling is a culturally inherited trait and competitiveness is a genetic trait, we consider, for completeness, all possible combinations of cultural or genetic inheritance.

5.2 Model and Analyses

In the absence of reproductive levelling, fitness is determined by the share of group resources that an individual obtains, which we capture with Frank’s (1995, 1996a,b,

2003, 2009, 2010) tragedy of the commons model. We assume an individual i in group j displays a level of competitiveness sij and that her share of the group resources is determined by how competitive she is, relative to her group mates (sij/sj). If an individual invests in an above-average level of competitiveness (sij/sj>1), she outcompetes her group mates and receives a greater share of the group’s resources.

We also assume that this competition between group members reduces the group’s resources, such that the total resources (we scale resources such that a fully cooperative group has one unit) available for group members is given by 1-sj.

Consequently, in the absence of reproductive levelling, an individual’s fitness is given by wij=(sij/sj)(1-sj). This leads to the classic ‘tragedy of the commons’ scenario,

120 where the evolution of increased competitiveness is favoured, even though this leads to a reduction in the available resources (Hardin 1968; Frank 1995, 2010).

We assume that an individual can invest in a reproductive levelling behaviour at level τ, (0<τ<1), which has three fitness consequences. First, it reduces fitness differences between group members by ensuring resources are shared more evenly; a

proportion τj of resources are distributed equally between group members (where τj is the average group investment in reproductive levelling). The individual’s share of

the remaining portion 1-τj of group resources is still given by her relative

competitiveness, sij/sj. Second, reproductive levelling stops the group’s resources

being lost due to selfish exploitation; the result is to increase group fitness by bτj (so

group fitness is wj=1-(1-bτj)sj), where b (0

Third, an individual pay a fitness cost, -c to perform her personal level of the

behaviour τij. Model notation is summarized in Table 5.1. If wij denotes the relative fitness of an individual i in group j, then:

. (1)

This is the same fitness function used in chapter 3, which builds upon previous policing models by Frank (1995, 1996b, 2003). The individual faces a trade off between the cost of reproductive levelling, which she pays for personally, and the benefit she receives, which is determined by the average investment in reproductive levelling by all group members. The individual benefits from increased group

121 resources, however in addition to the personal cost, increased reproductive levelling

also limits the individual’s ability to compete and selfishly exploit the group’s

resources.

Table 5.1. A summary of symbols used in the analysis

Symbol Definition Symbol Definition

coefficient of (whole-group) reproductive levelling r τ s relatedness for the selfishness trait

s Competitiveness / selfishness w Darwinian fitness

others-only relatedness (the ‘R’ c cost of reproductive levelling R term in Hamilton’s rule) efficiency reproductive levelling restores fitness benefit to the recipient of a b B group resources social behaviour coefficient of (whole-group) relatedness fitness cost to actor of a social C rτ for the reproductive levelling trait behaviour

5.2.1 The tragedy of the commons and reproductive levelling € We are primarily concerned with how culturally inherited reproductive levelling

coevolves with genetically inherited cooperation. However, before considering that

scenario, it is useful to consider two other special cases of our model. In Appendix 5.4

we show that:

Result 1. In the absence of reproductive levelling, (τij =τj =0), we recover the classic

‘tragedy of the commons’ model of Frank (1995, 1996a,b, 1998, 2003, 2009, 2010). In

this case, if within-group relatedness for the competitiveness trait is rs, then the

evolutionarily stable (ESS; Maynard Smith and Price 1973) level of selfish

competition is s*=1-rs (Frank 1995, 1996a; Taylor and Frank 1996; Figure 5.1b).

Competitiveness is negatively correlated with relatedness because higher relatedness

favours increased kin-selected self-restraint.

122 Result 2. If both traits have the same mode of inheritance (either both genetic, or both cultural), such that both traits have the same within-group relatedness, r, then high levels of costly reproductive levelling can be favoured at intermediate relatedness

(when r(1-r)>c/b) and this leads to a reduction in competitiveness (Figure 5.1). The results of this special case are identical to that presented in an earlier paper on the genetic evolution of policing (see chapter 3).

5.2.2 Gene-culture coevolution

We now consider the scenario where the relatedness terms may differ. We follow

Bowles (2006) to examine the case when the competitiveness trait is genetically inherited (rs measures average genetic similarity) and the reproductive levelling trait is culturally inherited ( rτ measures average cultural similarity). These relatedness terms, rs and rτ , are statistical measures of variation, describing how closely the € individual’s trait values (sij and τij) match her group’s average trait value (sj and τj); the tighter€ the correlation, the higher the relatedness. Previous analyses of this model assumed the selfishness and reproductive levelling were inherited in the same way i.e. that r=rs= rτ (see Chapter 3). However, the two relatedness terms could differ if the traits have different mechanisms of inheritance (i.e. one is cultural, one is genetic).

We€ make no assumptions about the transmission or inheritance structure of the trait such as whether it is horizontal vs vertical (Lehmann et al. 2008) or blended vs particulate (Gardner 2011). Consequently, these relatedness terms could measure the cultural or genetic similarity between the focal individual and their group; we return to this topic in the discussion.

In Appendix 5.4, we show that this leads to the following predictions:

123 Result 3. Costly reproductive levelling can evolve, and it is most strongly favoured

when genetic relatedness is very low (rs→ 0) and cultural relatedness very high ( rτ →

1; τ*>0 when rτ (1-rs)>c/b; Figure 5.2). This is because low genetic relatedness leads to € high competitiveness, which increases the benefits of a mechanism such as reproductive€ levelling that can prevent the tragedy of the commons. High genetic

relatedness disfavours reproductive levelling (τ* = 0 when rτ (1-rs)

Fig. 5.1a Fig. 5.1b

Figure 5.1. The coevolution of reproductive levelling and competitiveness when both traits are inherited in the same way (i.e. both genetically or both culturally). Reproductive levelling is favoured only at intermediate relatedness (when r(1-r)>c/b; Figure 5.1a) and when it invades, it always promotes self-restraint (ds*/dτ<0, s*=c/br; Figure 5.1b). If reproductive levelling is not perfectly efficient at restoring group resources (b<1), intermediate levels of reproductive levelling, and smaller reductions in competitiveness are favoured. When reproductive levelling is not favoured (τ*=0 is stable when r(1-r)

124 Result 5. When it evolves, intermediate levels of reproductive levelling (0<τ<1) are

favoured. Full reproductive levelling is not favoured (τ*<1) so the ESS level must be

an intermediate value (τ* is given in Appendix 5.4). For the model parameter values,

τ* is bounded between 0<τ*<1, and so when it evolves, reproductive levelling

enforces the fair distribution of some, but not all of the group’s resources (τ*<1 so

wj<1). In other words, competitiveness is reduced, but some within group conflict

remains, and so individuals are not acting so as to maximise group fitness.

Result 6. When reproductive levelling does evolve, we find a lower level of

competitiveness is favoured (s*=c/b rτ ) than when it is absent (s*=1-rs). This

repression of competition occurs because reproductive levelling reduces the relative

benefit of competing over resources.€

Fig. 5.2a Fig. 5.2b

Figure 5.2. The coevolution reproductive levelling and competitiveness when reproductive levelling is inherited culturally, and competitiveness is inherited genetically. Reproductive levelling is favoured and enforces the fair distribution of the group’s resources when genetic

relatedness (rs) is low and cultural relatedness ( rτ ) is high (Figure 5.2a; τ*>0 when rτ (1-rs)>c/b). Here, relatedness is a statistical measure of the average cultural or genetic similarity between group members compared to the population average; it makes no assumptions about the

nature of inheritance. The greatest reduction€ in genetically inherited competitiveness€ (s*=c/b rτ ; Figure 5.2b) occurs when everyone shares the same cultural norms for reproductive levelling (

rτ →1). When reproductive levelling is not favoured (τ**=0 is stable when rτ (1-rs)

125 € € model reduces to the tragedy of the commons (s*=1-rs) and competitiveness is only limited by kin-selected self-restraint (Figure 5.1b, 5.2b). Plotted where the individual cost of performing reproductive levelling, c=0.05, and the efficiency of reproductive levelling, b=0.5 (i.e. at best, reproductive levelling can restore the group’s resources to half of what could be possible if competitiveness were absent).

To investigate whether reproductive levelling is maintained by direct or indirect fitness benefits, in the Appendix, we calculate the fitness components of Hamilton’s rule (Hamilton 1964). We examine the change in relative fitness ΔW, of a rare defector who performs no reproductive levelling. We find that:

Result 7. When it is at its ESS level, reproductive levelling is a mutually beneficial trait, by which we mean that it provides a benefit to both the actor and the other individuals of the group (West et al 2007b; Table 1.1). By performing the trait, an individual gains both direct and (if R>0) indirect fitness benefits, when compared to the fitness of a rare defector who does not perform the trait.

5.3 Discussion

We have examined the evolution of a reproductive levelling trait, which reduces fitness differences between group members by enforcing the fair distribution of the group’s resources. We have shown that reproductive levelling: (1) can be favoured by either genetic and/or cultural selection (Figure 5.1a, 5.2a); (2) is mutually beneficial;

(3) favours a reduction in competitiveness (Figure 5.1b, 5.2b); (4) is most strongly favoured when group members share the same culture but are genetically unrelated

(Figure 5.2). We chose to model a relatively simple scenario, in order to elucidate the underlying selective forces, and also to clarify links with the biological literature on how policing or repression of competition can favour cooperation. These results confirm that reproductive levelling mechanisms could play a role in the evolution of

126 cooperation, and counters the suggestion that such mechanisms cannot evolve genetically (Bell et al. 2009; Svensson 2009; Bowles 2006; Boyd 2006).

Why is reproductive levelling favoured? Reproductive levelling is a mutually beneficial trait that is favoured because both the actor and any related group-mates benefit from improved group resources (Hamilton 1964). If we assume that both the competitiveness and reproductive levelling traits share the same mode of inheritance

(i.e. both genetical or both cultural), reproductive levelling is most strongly favoured at intermediate relatedness. This is because of a trade off: relatedness must be high enough for the indirect benefits of reproductive levelling to be significant, but not too high, otherwise kin-selected self-restraint reduces competitiveness to a point where reproductive levelling is no longer cost-effective (Figure 5.1). The potential benefit of gene-culture coevolution is that if the traits are inherited differently, the relatedness coefficicents are different, so these opposing selective constraints are decoupled.

Reproductive levelling is most strongly favoured when genetic relatedness is low, as this increases the benefits of a mechanism to limit the harmful effects of competitiveness, and when cultural relatedness is very high as this maximizes the indirect benefits the actor gains from performing the trait. Estimates of genetic and cultural relatedness among human groups indicate this situation, of low genetic and high cultural relatedness, was common during human evolution (Bell et al. 2009).

If we assume that both reproductive levelling and competitiveness traits are inherited genetically, our results are the same as those predicted by policing models (see chapter 3 and Frank 1995, 1996b, 2003, 2009; Leigh, 1971, 1977). Policing is a well- known and important mechanism by which within-group competition is repressed, and evidence for it is found at all scales of biological complexity (Leigh 1977; West et al. 2002b; Kiers et al. 2003; Ratneiks 1988; Ratnieks and Visscher 1989; Ratneiks and

Wenseleers 2008). It is possible that genetically controlled policing mechanisms are

127 important in humans too as it can enforce cooperation in large groups where relatedness is low (see chapter 3). Assuming genetic inheritance does not imply behaviours are hard-wired or that culture does not matter. For example, natural selection may favour a genetically-inherited preference for egalitarianism between group members, however, as discussed in chapter 1, the way in which an individual can fulfil this preference would vary greatly, depending upon the learning opportunities, social structure, environment and culture they experience. Given the experimental and ethnographic evidence that suggests a preference for egalitarianism may be universal (Dawes et al. 2007; Boehm 1982, 1997, 2001), this seems a fruitful avenue for further research and it highlights the benefits that increased interaction with the biological literature may bring.

To conclude, our results present a possible mechanism whereby gene-culture co- evolution could favour cooperation in large groups of genetically unrelated individuals. However, there are two caveats here, which could apply to many models of gene-culture coevolution. First, we examine the coevolution between genetically inherited competiveness and culturally inherited reproductive levelling, which implicitly assumes the cultural environment is stable enough to alter selection on genes in a consistent way. Given that almost all the complex traits, like competitiveness, that have been studied involve many genes with small effects, any selective change would take hundreds of generations (Lynch and Walsh 1998). Thus for our model predictions to apply, it would require long periods of cultural stability

(generating a high cultural relatedness long enough to favour genes for reduced competitiveness). Assuming both traits are inherited genetically avoids this problem as the population dynamics for both traits operate on the same timescale.

Second, we assume that natural selection can lead to fitness maximizing adaptations via culturally inherited traits. A special feature of Mendelian inheritance is that

128 genetic relatedness will be approximately equal across most of the genome, in such a way that all genes will tend to be striving towards the same fitness maximization

(Grafen 1985, 2006, 2007). It is precisely this which allows the construction of complex adaptations which are based upon multiple genes, and provides justification for the maximization or optimizing approach (Grafen 2007; Gardner 2009). In contrast, the same unity of interest cannot be assumed with cultural selection, because the pattern of transmission can vary across space, time or traits. Therefore it is unclear whether a trait such as reproductive levelling could evolve via cultural selection. It would be extremely useful to apply Grafen’s (1999, 2007) formal Darwinism methods to cultural selection, and determine what, if any, maximizing principle would emerge from the cultural dynamics (Gardner 2009).

5.4 Appendix

Here we perform an invasion analysis and assess equilibrium conditions, deriving the results stated in the main text. To understand when the traits of competitiveness and reproductive levelling are favoured by selection, we examine how changing the levels of these traits affect an individual’s fitness. The direction of selection is found using the neighbour-modulated fitness methodology of Taylor and Frank (1996; see also Frank 1995, 1996b, 2003; Taylor 1996a; Taylor et al. 2007). Assuming vanishing variation in the trait about the population average, the marginal fitness for competitiveness is:

dwij/dsij = ∂wij/∂sij + rs ∂wij/∂sj, (A1) and for reproductive levelling is:

dwij/dτij = ∂wij/∂τij + rτ ∂wij/∂τj, (A2)

€ 129 where rs = dsj/dsij and rτ = dτj/dτij represent the kin-selection coefficients of whole-

group relatedness for the competitiveness and reproductive levelling traits

respectively (relatedness€ to average group member, including oneself; Taylor and

Frank 1996; Pepper 2000). The average fitness of an individual is give by equation (1)

in the main text. For the population average, (sij=sj=s and τij=τj=τ), this gives the

marginal fitnesses for competitiveness (s) and reproductive levelling (τ) as:

dw (1− rs)(1−τ) = − (1−τb) 1−τ 1− r , (A3) ds s ( ( s))

and: €

dw = br s − c . (A4) dτ τ

We first consider the evolution of competitiveness in the absence of reproductive € levelling. To do this we set τ =0 in equation (A3), set the RHS to zero, and solve for s.

This gives competitiveness as s*=1-rs, (result 1 of the main text). The marginal fitness

for reproductive levelling at this point is found by evaluating equation (A4) at s*=1-rs,

which obtains dw/dτ =b rτ (1-rs)-c; hence for reproductive levelling to invade (which

occurs when dw/dτ >0), we need rτ (1-rs)>c/b (result 3 of the main text). If both traits € are inherited in the same way (rs= rτ ), then this becomes r(1-r)>c/b (result 2 of the € main text). If τ=1, we have dw/ds=-(1-b)rs, and, as we assume the efficiency of

reproductive levelling b<1,€ competitiveness, s, is disfavoured and will decline to zero

and this obtains dw/dτ=–c which is always negative. This means reproductive

levelling cannot reach fixation (result 5 of the main text). To find the ESS levels of

reproductive levelling and competitiveness, we set the RHS of equations (A3) and

130 (A4) to zero and simultaneously solve for τ=τ* and s=s*. This yields s*=c/b rτ (result 6

of the main text) and two results for τ*, only one of which is a biologically feasible: €

2 bc c br 1 r 4bc 1 r br 1 r c bc c br 1 r + ( − τ )( − s) + ( − s)( τ ( − s) − ) + ( + ( − τ )( − s)) τ* = 2bc 1 r ( − s) , (A5)

€ which is result 5 of the main text.

To investigate whether reproductive levelling is maintained by direct or indirect

fitness benefits, we must calculate the fitness components of Hamilton’s rule

(Hamilton, 1964). The change in relative fitness is ΔW=Wij-W, whereW=1 and

Wij=wij/w. If fitness, wij is given by equation (1) in the main text, Wij is:

τ j + (1−τ j )(sij s j ) 1− (1− bτ j )s j − cτij W = ( )( ) (A6) ij 1− s(1− bτ) − cτ

We consider the change in relative fitness, ΔW, of a rare defector who performs no € reproductive levelling. Assuming weak penetration (the ‘tracer allele’ approach;

Grafen, 2006b), we have τj =τ(n-1)/n as one group member (representing a fraction

1/n of group j) performs no reproductive levelling. To find the direct fitness term in

Hamilton’s rule –C, we assume the focal individual is a defector, so we set τij=0 and

τj=(n-1)τ. This yields:

$ $ 1'' −C = −s&1 − bτ&1 − )) (A7) % % n((

which is always negative (-C<0) for our model parameters. If not performing € reproductive levelling is directly costly, this implies performing it is directly

beneficial. To find the indirect benefits term B, we assume the focal individual

131 performs reproductive levelling, but that one member of their group is a defector, so

we set τij=τ and τj=(n-1)τ. This yields:

$ $ 1'' cτ B = −s&1 − bτ&1 − )) − (A8) % % n(( 1− cτ − s(1− bτ)

this is always negative (B<0) for our model parameters. If not performing € reproductive levelling has indirect fitness costs, this implies performing it yields

indirect fitness benefits, thus, when compared to the fitness of the defector,

reproductive levelling is mutually beneficial (result 7 of the main text).

132 Chapter 6. Promiscuity and the evolution of cooperative breeding.

This chapter has been submitted for publication in Proceedings of the Royal Society,

B: Leggett, H., El Mouden, C., Wild, G., and West, S. A. (2011) Promiscuity and the evolution of cooperative breeding.

6.1 Introduction

It has been argued that monogamy or low levels of promiscuity have played a key role in favouring the evolution of cooperative breeding and eusociality. Hamilton’s inclusive fitness theory (Hamilton 1964, 1970; Grafen 2006a) explains how cooperation can be favoured when it is directed towards relatives who also carry the gene for cooperation. Higher levels of promiscuity would decrease relatedness within family groups (Figure 6.1a), which could therefore reduce selection for staying and helping in the natal family group (Charnov 1981). Empirical support for this idea has come from the observations that: (a) eusociality has only evolved in species with strict lifetime monogamy, and (b) both the occurrence and amount of cooperative breeding in birds is negatively correlated with the level of promiscuity (Boomsma

2007; Hughes et al. 2008; Boomsma 2009; Cornwallis et al. 2010).

However, it is unclear whether greater promiscuity should always select against staying to help in natal groups. From an empirical perspective, some of the highest promiscuity rates ever recorded are cooperative breeders, such as the superb fairy wren, and the Australian magpie (Cornwallis et al. 2010). From a theoretical

133 perspective, previous predictions have been based on verbal or heuristic models.

More explicit models, which allow the action of selection to emerge as a consequence of demographic assumptions, have shown that an increased relatedness will not always favour cooperation because it can be negated by increased local competition between relatives (reviewed by West et al. 2002a and Lehmann and Rousset 2010).

Furthermore, a class of theoretical models make the opposite prediction, showing that because higher relatedness within groups leads to increased competition between relatives, selection can favour dispersal away from the natal group to reduce competition between relatives (Figure 6.1b; e.g. Hamilton and May 1977; Taylor 1988;

Frank 1998; Gandon and Rousset 1999; Ronce et al. 2000; Rousset and Gandon 2002).

Finally, a recent attempt to explicitly model the effect of promiscuity on cooperation by Nonacs (2011), with population genetic simulations, found that the level of promiscuity had little, or even a positive effect on selection for cooperation, and that this might be an area where inclusive fitness theory “gets it wrong”.

Figure 6.1. Promiscuity, relatedness and dispersal rate within cooperatively breeding groups. (A) Promiscuity and relatedness. Promiscuity (number of mates) plotted against the mean genetic relatedness between potential helpers and both their siblings and offspring. An individual is always related to its offspring with r= 0.5. However, as promiscuity increases, the relatedness to siblings decreases from r=0.5 to r=0.25 (full-siblings to half-siblings) (Charnov 1981); redrawn from (Cornwallis et al. 2010). (B) Higher relatedness within groups can lead to increased competition between relatives, which selects for dispersal away from natal groups to reduce competition between relatives (Hamilton and May 1977).

134 Here, we use an inclusive fitness approach to model how promiscuity influences selection to stay and help in natal groups. Our aim is to consider the simplest possible scenario to explicitly examine how the level of promiscuity, through its effect on within group relatedness, has multiple consequences. Our approach is therefore analogous to that taken by Taylor (1992a), when considering how limited dispersal influences cooperation. In particular, we examine how lower levels of promiscuity lead to higher within-group relatedness, which can, in turn, increase the indirect benefits of both staying to help relatives and dispersing to reduce competition among relatives (Wild 2006). In addition, we check the validity of our inclusive fitness model by comparing its predictions with that made by an individual based model.

6.2 The Model

We consider a population of cooperatively breeding, diploid individuals. To address the effect of promiscuity on cooperative breeding behaviour, we must assume the individuals are sexual. In order to avoid complications of modelling different sexes, we assume individuals are sexual hermaphrodites.

We assume the population is subdivided into a very large, but fixed, number of breeding patches (say, P patches). Each patch is of equal quality and supports one adult breeder at a time, plus a large number of juvenile offspring. In nature, the cooperative behavior of juvenile helpers can increase the breeder’s survival and/or fecundity (Heinsohn and Legge 1999; Griffin and West 2003; Rousset and Ronce 2004;

Russell et al. 2007). We focus on the former case, where the survival of the breeder is boosted by the presence of helpers. We avoid the latter case in order to avoid complications associated with the cascading effects that fecundity benefits exert on

135 breeder fitness (more helpers result in more offspring, producing greater numbers of helpers, which in turn increases fecundity further, etc).

Since juvenile helpers often direct their help toward the individual(s) occupying the same patch on which helpers, themselves, were born (e.g. helpers help their own parents (Clutton-Brock 2009; Hatchwell 2009), we assume that an offspring’s tendency to help is linked to its decision to disperse. In other words, an offspring that remains on its natal patch does so in order to help the incumbent breeder, while an offspring that disperses from its natal patch will compete (without helping) on a new patch for the opportunity to breed, which occurs upon the (possible) death of a breeder. We explain our assumptions about the basis for an individual’s “tendency” to help (or “decision” to not disperse) in greater detail below. For now, simply note that all helpers exert the same level of effort, and it is just this “tendency” to help that varies among individuals in our model population.

Now suppose we observe the population described above at discrete, evenly spaced points in time. We assume that, between successive observations, the following series of five events occurs:

1. Mating. Each breeder chooses exactly M mates from the global population of

breeders. Mate choice occurs uniformly at random, and with replacement. Note

that larger (resp. smaller) M indicates that breeders are more (resp. less)

promiscuous.

2. Birth. Each breeder produces a very large number (K) of offspring, and each

offspring carries one chromosome pair. An offspring’s “maternally” inherited

chromosome is one of the two chromosomes (chosen uniformly, at random)

belonging to the local breeder, while the offspring’s “paternally” inherited

136 chromosome is one of two chromosomes (chosen uniformly, at random)

belonging to one of the breeder’s M mates (chosen uniformly, at random with

replacement). The local breeder, then, will be considered “mother” to offspring

produced on her patch, while the local breeder’s mates will be considered

potential “fathers”. Of course, the local breeder will also be potential “father” to

offspring produced off-site, while the local breeder’s mates will also be “mother”

to offspring produced on the patches they, themselves, occupy.

3. Dispersal. Genes found on the chromosomes carried by an individual offspring

determine d, the probability with which that individual disperses from its natal

patch. We assume that maternally inherited and paternally inherited genetic

information contribute equally and additively to the offspring’s dispersal

phenotype. Thus, we treat d as the mean of two phenotypes: one that would be

produced by two identical copies of the maternally inherited gene, and a second

that would be produced by two identical copies of the paternally inherited gene.

We assume that offspring that disperse successfully find their new patch by

choosing from among the P patches in the population uniformly at random. Note

there is a possibility that an offspring actually disperses to its natal patch, and in

this case becomes a non-helping, individual-in-wait. Since offspring that do not

disperse stay to help the incumbent breeder, the probability 1 – d quantifies an

individual’s aforementioned “tendency” toward helping.

4. Helping. Once the decision to help/disperse has been made, the direct benefit to

the breeder can accrue. We assume that breeder mortality decreases at a rate

proportional to the total number of helpers found on its patch immediately

following dispersal. We assume that the constant of proportionality is small (say,

of order 1/K), and model the probability of breeder mortality with the function,

137

µ(d) = exp{−k (1 −d)}, (1)

where k is a positive constant that controls how quickly breeder mortality decays

with increasing numbers of helpers (larger k means helping is more effective at

reducing mortality), and whered is the mean d among a breeder’s current brood

of offspring. Given this model for breeder mortality, we model the probability of

breeder survival as

Sb(d) = 1 − µ(d). (2)

5. Competition for vacant patches. If a breeder survives, then we assume it retains its

patch, and breeds there again in the next time step. However, if a breeder does

not survive, then we assume that competition for the vacant patch occurs among

helpers (those native offspring that did not disperse) and non-helping

individuals-in-wait (those non-native offspring that dispersed to the contested

patch). Because a helper may compete for a breeding site more/less successfully

than a non-helper due to, for example, different territory inheritance patterns

(Griffin and West 2003), costs of dispersal (Ronce 2007), costs of helping itself

(Heinsohn and Legge 1999), we introduce a positive constant c that allows the

competitive ability of a non-helper to vary relative to that of a helper. When c > 1

a non-helper has a competitive advantage over a helper, when c < 1 a non-helper

has a competitive disadvantage, and when c = 1 helpers and non-helpers are

competitively equivalent. Mathematically, c acts as a competitive weight given to

non-helpers, so that the terms and

express the probability that a given helper and given non-helper, respectively,

secure the breeding site on which they compete.

138 When the competition phase (event 5, above) is complete, all unsuccessful competing offspring die (this includes offspring competing on patches where the incumbent breeder survived), and the cycle repeats.

6.3 Methods of Analysis and Results

We analyzed the model using the neighbour-modulated (direct) fitness approach of

Taylor and Frank (1996; see also Frank 1998; Taylor, et al. 2007). For convenience, our inclusive fitness treatment assumes that the number of patches (P) and brood size (K) are both very large (ideally infinite). We make the standard suite of genetic assumptions (weak selection, additive gene action, etc.), discussed in detail elsewhere

(Taylor 1989; Taylor and Frank 1996; Rousset and Billiard 2000; Rousset 2004;

Gardner et al. 2011).

We show in Appendix 6.5 that, given the life history assumptions described above, the marginal inclusive fitness of a focal juvenile with a reduced tendency to help (i.e. increased tendency to disperse) can be described by:

ΔW = (1 − Sb) [(c − 1)/(1 − d + cd)] I

+R Sbʹ II

+ r (1 − Sb) [1/(1 − d + cd)] III

− r Sbʹ IV

(3)

where Sbʹ is the marginal survival of the breeder (this is negative, since increased dispersal leads to reduced breeder survival), R = 0.5 is the average relatedness

139 between a juvenile and its parent, and r is the average relatedness between a focal juvenile and the average juvenile that competes on the focal juvenile’s natal patch.

The coefficient r can be written as r = [(1− d )/(1 − d + cd)] [(1/M)+1]/4, where M is the number of different mates a breeder has (i.e. M is the degree of promiscuity).

Figure 6.2 The costs and benefits of helping at the natal patch. Numerals relate to the lines of equation 3. – is a cost, + is a benefit: (I)-/+ is the direct cost/benefit associated with a decreased/increased chance of becoming a breeder; (II)+ is the benefit of increasing parent survival; (III)- is the cost of increasing parent survival, leading to decreased chances of the helper or siblings becoming a breeder; (IV)- is the cost of not dispersing, leading to increasing competition among siblings.

Equation 3 shows that the inclusive fitness effects of reduced helping (i.e. increased dispersal) are given by four components (Figure 6.2).

I. The direct fitness effect of reduced helping. That is, the difference between

the focal individual’s own probability of securing a breeding site in the

case that it disperses (i.e. does not help), and the focal individual’s own

probability of securing a breeding site in the case that it does not disperse

(i.e. does help). When c > 1 this term counts as a benefit, and when c < 1

this term counts as a cost.

140 II. The change in adult breeder survival that is associated with reduced help.

This term counts as a cost, since Sbʹ = −kµ < 0.

III. Reduced local competition due to the focal individual’s increased

tendency to disperse (i.e. decreased tendency to help). This is a benefit

given to the average juvenile that competes on the focal juvenile’s natal

patch.

IV. The benefit of increased opportunity to succeed the current breeder. When

there is less help available to the breeder, adult survivorship is decreased

which, in turn, increases the probability that the local breeding site will

become available to a juvenile competing on the patch. Again, this benefit

is given to the average juvenile that competes on the focal juvenile’s natal

patch.

The sign of ΔW determines when reduced helping confers a selective advantage.

Reduced helping is favoured when ΔW > 0 and is disfavoured when ΔW < 0. When

ΔW = 0, the population-wide level of helping is at an equilibrium. Simple algebra shows that the sign of ΔW is determined by a quadratic function of d, so it is possible to provide an exact expression for the equilibrium value of d, when it exists. That said, the expression for such a d is complicated, and its existence requires that a second complicated expression (in mathematical terms, the second expression is called the discriminant) be positive. To avoid complicated mathematical expressions that convey no biological insight, and to explore the possibility that long-term action of selection leads to non-equilibrium levels of behaviour we turned to numerical simulation.

141 Our numerical procedure began by assuming a set of initial conditions that covered the entire range of phenotype space evenly. In this way, the numerical procedure simulated several evolutionary trajectories simultaneously. Any given trajectory was constructed by using the most recently observed population average d to calculate

ΔW. When ΔW was positive (resp. negative) the population average d was increased

(resp. decreased) by an amount that was proportional to ΔW itself. These two basic steps were repeated until all trajectories of a given simulation run were sufficiently close to one another. A subset of the numerical results was confirmed with individual-based simulation of the model system (Figure 6.4; details of the assumptions and a version of the script used to generate simulation results are available on request).

Figure 6.3 Helping and promiscuity. The predicted rate at which individuals stay and help (1- d) is plotted against the level of promiscuity (M), for different rates of the competitive ability of non-helpers (c) and the survival benefit conferred upon breeders (k). The proportion of individuals selected to stay and help decreases with increased competitive abilities of non- helpers, c, and with decreased survival benefit conferred upon breeders, k.

142 We found that the predicted level of helping increased with increasing promiscuity

(Figure 6.3). Hence, the tendency to help juveniles born on the same patch increased with decreasing relatedness (from increasing promiscuity). The reason for this is that the rate of promiscuity only has consequences for lines III and IV in equation 3.

Promiscuity does not influence either direct fitness consequences (Line I) or relatedness to parent (Line II). In contrast, lower levels of promiscuity lead to a higher relatedness to other helpers on a patch, which increases the indirect benefit of dispersing to reduce competition for relatives.

Figure 6.4 Inclusive fitness versus simulations. The predicted rate at which individuals stay and help (1-d) is plotted against the level of promiscuity (M), for different rates of the competitive ability of non-helpers (c) and the survival benefit conferred upon breeders (k). The results obtained using inclusive fitness theory (solid lines) and individual based simulations (dots) are in close agreement.

143 6.4 Discussion

Our model predicts that selection for staying to help: (1) either shows no relationship or increases with higher levels of promiscuity; (2) is decreased when non-helpers are more likely to inherit a territory (higher c); (3) increases when helping provides a greater survival benefit to the breeder (Figure 6.3). In addition, we found close agreement between the predictions of our inclusive fitness analyses and an individual based simulation (Figure 6.4).

Our model shows that, even in a very simple scenario, the consequences of promiscuity for the level of cooperation can be far more complicated than expected from heuristic and verbal arguments (see equation 3 and Figure 6.2). Previous theoretical arguments have focused on how promiscuity reduces relatedness to siblings, and hence reduces the indirect benefit of staying to help, leading to the prediction that promiscuity selects against helping (Figure 6.1). However, when all the population level consequences of staying to help are explicitly modelled, there can also be indirect costs of staying to help. Specifically, staying to help increases the chance that both the breeder survives, and that the helper inherits the patch if the breeder dies (Lines IV and III of equation 3 respectively). Consequently, staying to help decreases the chance that another sibling who has stayed to help will inherit the patch.

Why does our model predict that promiscuity has either no or a positive effect on selection to stay and help (Figure 6.3)? Equation 3 shows that in our model, staying to help has four fitness consequences: (I) direct cost of not dispersing to breed elsewhere; (II) indirect benefit of increased survival of parent; (III and IV) indirect cost of creating greater competition for non-dispersing siblings by reducing the likelihood that the breeder will die, and the helper staying to compete for the

144 breeding spot if the breeder dies. The indirect benefit of staying to help (II) is not altered by promiscuity, because the helper is always related to their parent by R=0.5.

Consequently, the effect of promiscuity is driven by its indirect cost of making greater competition for siblings (III and IV). Lower levels of promiscuity lead to a higher relatedness between siblings, which favours dispersal to reduce the competition that these siblings face. Another way of looking at this is that low promiscuity does favour helping, but that the way to help siblings is to disperse away (to reduce competition), rather than to stay and help. This illustrates the general point that verbal arguments can be misleading because they focus on conspicuous traits and ignore less conspicuous consequences.

From a theoretical perspective, one of the most important points that arises from our model is that, even in a simple situation, things can be much more complicated than expected. Specifically, competition between relatives can either negate or reverse predictions that follow from simple intuition. In this respect, our paper is analogous to when Taylor (1992a) showed that limited dispersal didn’t necessarily favour cooperation, because it could also lead to increased competition between relatives. A major theme since Taylor’s pivotal paper has been examining exactly what kind of demographic factors lead to limited dispersal favouring cooperation (reviewed by

West et al. 2002a; Lehmann and Rousset 2010). Analogously, we are not saying that the idea that monogamy and low levels of promiscuity will favour cooperation is wrong, but rather saying that this is only likely to hold given certain biological assumptions.

A problem with our model is that by concentrating on survival benefits, our model neglected one of the main aspects of helping that we had hoped to capture - the benefit to siblings from staying to help. Extending our model to include survival benefits to siblings did not change things (results not shown), because the strict

145 density dependence means that this is cancelled out by the increased competition that it also creates. We suggest that the best avenue for future research is to take an entirely different approach, and examine fecundity benefits in a population where there is not strict density dependence (Rousset and Ronce 2004; Grafen 2007; Alizon and Taylor 2008; Wild et al. 2009; Taylor and Grafen 2010; Wild 2011). This task will be non-trivial, as even the simple scenario we considered here, with only survival effects, and strict density dependence, led to a model that we had to solve numerically, rather than analytically. We suspect that this alternative approach would lead to a negative correlation between selection to stay to help and promiscuity, as has been found in the empirical literature (Boomsma 2007, 2009;

Hughes et al. 2008; Cornwallis et al. 2010). Noteworthy here is that the species where the benefits of helping are primarily on the survival of breeders, the superb fairy- wren, also has one of the highest levels of promiscuity (Russell et al. 2007).

Finally, our results contradict the possibility that the influence of promiscuity on the evolution of cooperation could be an area where inclusive fitness theory makes

“spurious predictions” (Nonacs 2011). Nonacs (2011) found that a population genetic simulation predicted a flat or slightly positive relationship between promiscuity and helping, and contrasted this with arguments based on kin selection theory that promiscuity should select against cooperation (e.g. Charnov 1981; Boomsma 2007,

2009; Cornwallis et al. 2010). However, he did not also construct an inclusive fitness model, and so he was comparing scenarios with very different assumptions, rather than comparing different theoretical approaches (for a related example on the evolution of dispersal, see pp. 117-120 of Frank 1998). We have found the same pattern as Nonacs with an inclusive fitness analysis, and also shown that when population genetic simulations are produced for the same scenario, they give the same results. This is not surprising, given the large literature showing the

146 equivalence of different approaches (Queller 1992; Taylor 1996b; Rousset 2004;

Gardner et al. 2007a; Gardner et al. 2011). More generally, the comparison of our

simulation and inclusive fitness analyses support the previous suggestion that the

inclusive fitness approach will often be computationally simpler, and hence facilitate

both the deriving of predictions and our conceptual understandings of why those

predictions arise (p. 671 of Taylor 1996b; p. 120 of Frank 1998; p. 119 of Pen and

Weissing 2000).

6.5 Appendix

In this Appendix, we develop a class-structured model (Taylor and Frank, 1996) to

calculate the marginal inclusive fitness of a focal juvenile that is given in equation (3)

of the main text.

If d is the probability of dispersal for a focal individual, d is the average level of

dispersal from the natal patch, and dʹ is the population average level of dispersal. To

calculate the probability that a juvenile becomes a breeder, we note that a helper will

share a patch with (1-d)K other helpers and dʹcK immigrants, thus, all else being

equal, if the parent dies, a focal helper will become a breeder with probability [(1-d)]/

[K((1-d)+dʹc)]. A dispersing juvenile will be one of dʹcK dispersers and will share the

patch with (1-dʹ)K helpers. Thus if the breeder dies, all else being equal, it will inherit

the breeding site with probability dc/[K(1-dʹ)+Kdʹc]. Therefore, the probability a focal

juvenile will inherit a breeding site is given by:

1− d cd w 1 S d 1 S d AJ = ( − b ( )) + ( − b ( #) ) (A1) K((1− d ) + cd# ) K((1− d# ) + cd# )

€

147 In this model, we consider the fitness effect the focal individual (actor) has on its

parent breeder and juvenile helper siblings (the recipients). Both these two classes of

recipients have adult and juvenile fitness components. For the adult class, the adult

breeder (wAA) gains fitness by surviving, which occurs with probability wAA =Sb(d)

and gains fitness by producing juveniles (wJA), where wJA=KSb(d). For the juvenile

class, a juvenile gains fitness by inheriting a breeding site (wAJ) , which is given by

equation (A1) and gains fitness by producing offspring if the juvenile becomes an

adult breeder (wJJ), where wJJ =K wAJ. These fitness components are combined using

the following reproductive values:

" % " wJJ wJA % 1− Sb (d () KSb (d( ) W = $ ' = $ ' (A2) w w S d K S d # AJ AA & d =d =d ( # b ( () b ( () &

To calculate the juvenile and adult components of fitness we weight the class €

fitnesses by the reproductive values. The fitness of a focal juvenile, WJ, is given by

$ ' 1− S (d# ) 1 1− d cd W b w w & 1 S d 1 S d ) J = JJ + JA = & ( − b ( )) + ( − b ( #) ) ) , (A3) KSb Sb (d# ) KSb (d# ) % (1− d ) + cd# (1− d# ) + cd# (

and the fitness of a focal adult, WA, is given by €

1− Sb (d# ) Sb (d ) WA = wJA + wAA = . (A4) KSb (d# ) Sb (d# )

The direction of selection acting upon the dispersal trait, d is given by the marginal € fitness with respect to that trait weighted by the class reproductive values. This is

given by:

dW dW dW = K J + A (A5) dd" dd" dd"

€ 148 These derivatives may be separated into partial derivatives, describing the effect of an individual’s trait value, and the effect of the average trait value of its group, on the juvenile and adult class fitnesses, using the chain rule methodology of Taylor and

Frank (1996; see also Taylor 1996; Taylor et al. 2007). This gives dWJ/ddʹ = ∂WJ/∂d +r

∂WJ/∂d and dWA/ddʹ = ∂WA/∂d + R ∂WA/∂d, where r = dd/dd is the kin- selection coefficient of (whole-group) relatedness between a focal individual and other members of the juvenile class, and R = dd/dd is the kin-selection coefficient of (whole-group) relatedness between a focal individual and members of the adult class (Taylor and Frank, 1996; Pepper 2000). This yields the marginal fitness given in the main text, which uses the notation dʹ=d and gives the relatedness between two juveniles as r=r[(1-d)/(1-d(1-c)].

149 Chapter 7. Spatial structure and inter-specific cooperation: Theory and an empirical test using the mycorrhizal mutualism

7.1 Introduction

Despite their ubiquity, explaining the evolutionary persistence of cooperative mutualisms between species remains a major challenge (Sachs et al. 2004; West et al.

2007).The problem is that mutualisms are vulnerable to ‘free-riders’ or ‘cheaters’, partners who do not cooperate, but may gain the benefit of others cooperating

(Axelrod and Hamilton 1981). These exploiters are common in mutualistic systems, ranging from rhizobia that fail to provide fixed N2 to their legume hosts (Denison and Kiers 2011) to pollinators that lay eggs in hosts but defect from pollinating duties

(Goto et al. 2010). Theory suggests that, without mechanisms to prevent exploitation, natural selection should favour these ‘free-riders’ because they pay reduced costs but still reap benefits of the partnership.

The spatial structuring or partitioning of symbiotic partners is often invoked as a mechanism for maintaining mutualisms (Frank 1994; Doebeli and Knowlton 1998;

Bever and Simms 2000; Hoeksema and Kummel 2003; Foster and Wenseleers 2006;

Gardner et al. 2007; Lion and van Baalen 2008; Platt and Bever 2009; Hodge et al.

2010). In particular, spatial structuring has been proposed to be important in stabilizing mutualisms in which hosts differentially allocate resources to partners varying in benefit. It has been argued that spatial structure can facilitate the evolution of cooperation by separating patches of high-quality mutualists from their

150 antagonistic counterparts on a single host (Bever et al. 2009). When mutualists of similar quality are clustered, a host may preferentially reward patches of cooperators and/or punish patches of cheaters. However, spatial structure could also limit the local diversity of partners, reducing the range of partners among which hosts might discriminate.

Here, we model the role of spatial structure in maintaining mutualism (interspecific cooperation), focusing on the symbiosis between the majority of land plants and arbuscular mycorrhizal (AM) fungi. This symbiosis, which primarily involves the exchange of carbohydrates from plants for mineral nutrients from the fungal partners, offers ideal experimental opportunities to study spatial structuring: a single host root can be colonized by several AM fungal species that range in benefit from mutualistic to antagonistic (Hoeksema et al. 2010). Although spatial structure has been suggested to play an important role in the maintenance of the mycorrhizal mutualism (Chanway et al. 1991; Wilkinson 1998; Bever et al. 2009; Hodge et al.

2010), observations from the field and lab experiments do not consistently reveal strong structuring of different fungal communities (van Tuinen et al. 1998; Jansa et al.

2003; Alkan et al. 2006) but rather distantly related AM fungi are found to intermingle on a small spatial scale in host roots (Jansa et al. 2003). Whether hosts discriminate among intermingled fungal partners, allocating resources preferentially to the best ones, has been the subject of an on-going debate (Fitter 2006; Kiers and van der Heijden 2006; Bever et al. 2009; Helgason and Fitter 2009; Smith et al. 2009).

Recent manipulative work supports the hypothesis that hosts reward fungal symbionts at a fine scale, preferentially allocating C to small patches of hyphae when they deliver more P resources to their host (Kiers et al. 2011). The reciprocal was also found to be true, with fungi preferentially allocating P to small patches of roots that delivered more C. This bi-directional control allows the mutualism to function like a

151 biological market (Noë and Hammerstein 1995), in which partners offering the best rate of exchange are rewarded. However, it is not yet known how fungal spatial structuring influences this partner control.

To generate predictions, we first determined theoretically how the relative fitness of

‘high-quality’ and ‘low-quality’ mutualists should depend on the extent to which they are spatially structured. If plants can only preferentially allocate resources to fungi at a relatively coarse level, such as sections of the root system, we predict that spatial structuring will increase the relative fitness and proliferation of the high- quality mutualist. In contrast, if plants preferentially allocate resources at a much finer level, such as groups of cortical cells or at the level of the arbuscule (the site where nutrient transfer occurs; Parniske 2008), then we predict that spatial structuring will decrease the relative fitness and proliferation of the high-quality mutualist. These opposite predictions arise because spatial structure can limit the local availability of partners, reducing the extent to which hosts can reward more cooperative partners on a fine scale. We then test these competing predictions by using a multi-generational selection experiment, manipulating spatial structuring to determine how this influences the relative fitness of the low-quality AM fungal species. In the past, it has been difficult to track AM fungal fitness in mixed cultures over multi-generations. We resolved those constraints by developing specific quantitative molecular markers for our focal AM fungal species, allowing us to follow their relative fitness - when in direct competition - over multiple generations.

We manipulated spatial structure by either mixing the soil or leaving it non-mixed after every successive generation.

152 7.2 Predicting the consequences of spatial structure

Our aim is to provide a general model, that makes qualitative predictions, and which could also be applied to other mutualisms, rather than a more specific model that makes quantitative predictions. We examine the situation where there is a host interacting with two mutualists (in our case mycorrhizal fungal species) that differ in mutualistic quality. Our model defines the high-quality fungal species as a mutualist that transfers sufficient resources to the plant to be considered beneficial, and the lower-quality species as one that transfers fewer resources to the plant, (less beneficial/less cooperative). We assume that both fungal mutualists gain some baseline fitness benefit (w0) by interacting with the plant, regardless of the degree of spatial structuring, but as the high-quality fungus transfers more resources, it experiences a higher energetic cost c. The fitness benefit experienced by the high- quality fungi depends upon the plant’s ability to differentiate between high- and low-quality fungi.

We assume, as has recently been shown empirically, that plants preferentially allocate resources to higher-quality mutualists (Kiers et al. 2011). We consider two evolutionary scenarios, which differ in the scale at which plants can preferentially allocate resources (coarse or fine-scale rewarding). The fungi occur in patches, where a patch is a section of the root network on a single plant. If plants can only adjust their level of resource allocation between different sections of the root network, then differential allocation is at a “coarse” level, between patches. In contrast, if plants are able adjust their level of resource allocation at a finer scale, for example towards specific groups of cortical cells or at the level of the arbuscule or interface with one fungus (Kiers et al. 2011), then differential allocation is at a “fine level”, within patches.

153 We examine the influence of population structuring, by considering the extent to which fungal strains are mixed between and within root patches. At one extreme, in highly structured populations, we would expect to see each patch (associated with a single host plant) dominated by one or a small number of fungal strains, but to find different strains in different patches. At the other extreme, in highly unstructured

(panmictic) populations, we would expect to see a mixture of strains in each patch, and a similar mix in all patches. This corresponds to Wright’s (1935) Fst, a measure of population structuring, going from high (1.0) to low (0). Mixing the soil would reduce the fungal population structure, resulting in a lower Fst. We denote the fitness effect of population structuring by f(Fst) where f increases monotonically with Fst and the sign of the function f differs for high and low quality fungal strains and is determined by the level of control the plant has over resource allocation. This function need not be the same for the two types of fungal species or for the different levels of plant control. To highlight this we use fHc(Fst), fHf(Fst), fLc(Fst) and fLf(Fst) for the high (H) and low (L) quality fungal strains under coarse (c) and fine (f) control respectively.

When there is coarse control, preferential allocation by hosts is to different parts of the root network. Under this scenario, plants will be best able to discriminate between high- and low-quality fungi, when each part of the root network (patch) tends to contain only one type of fungi, but different types occur in different patches.

Higher population structuring (higher Fst) will allow greater discrimination and hence favours the high-quality fungal strain. Conversely, low quality fungi do better if patches are more evenly mixed so the plants will be less able to discriminate against them. Therefore fitness of a high-quality fungus (wHc) is given by wHc = w0 - c + fHc[Fst] and the fitness of a low quality fungus (wLc) is given by wLc = w0 - fLc[Fst]. This leads to the prediction that, under coarse control, the relative fitness of low-quality

154 mutualists (vLc = wLc / wHc) decreases as populations becomes more structured (fig.

7.1). Conversely, we predict that spatial structuring (high Fst, low mixing) under coarse control favors high-quality mutualists and hence selects for cooperation.

Figure 7.1 The effect of spatial structure (Fst) on the relative fitness of low-quality (less

cooperative) fungal species (vL). Fst, is a measure of population structuring; high structuring

(Fst →1) implies no mixing, low structuring (Fst →0) implies mixing. The black line represents

coarse level control (vLc, which assumes a single plant interacts with patches of fungi on its root

system) and the grey line represents fine level control (vLf, which assumes plant interacts with individual strains of fungi at the level of the arbuscule). Fine control predicts that high spatial

structuring (no mixing, Fst →1) favours low-quality strains whereas course control predicts high

spatial structuring (no mixing, Fst →1) favours high-quality strains. Here, we illustrate the simplest case, where the function f is the same for high and low quality fungi and for coarse

and fine control, but this need not be the case. We assume the fitness function, f of Fst takes the y form f = x(±Fst) + z. Fig. 7.1 is plotted for w0=1, x = 0.5, y = 2, z = 0 and the cost to the high- quality fungus, c = 0.1. Qualitatively, the result that the low-quality mutualist is favoured at high Fst under fine control and disfavoured under coarse control, holds for all values of x, y, z and c > 0.

Under fine control, we assume that the plant can allocate resources at a finer scale, towards specific fungal strains within patches of a single root-system. Consequently, plants will be best able to discriminate between low- and high-quality fungi when

155 both types are evenly mixed (low Fst). Because the fitness of high-quality fungi will be positively correlated with the extent to which plants can discriminate, the fitness of a high-quality fungus when there is fine control (wHf) is given by wHf = w0 - c - fHf[Fst].

Conversely, if patches are composed solely of low-quality partners, the plant is less able to preferentially choose the high-quality fungi, hence the fitness of a low quality fungus (wLf) is given by wLf = w0 + fLf[Fst] as lower-quality mutualists will be more successful when there is low mutualist diversity (high Fst). These fitness equations lead to the opposite prediction from the coarse control scenario. When there is fine control, they predict that the relative fitness of a low-quality mutualist (vLf = wLf / wHf) increases as populations become more structured (fig. 7.1). Here, spatial structuring favours low-quality fungi and selects against cooperation. Although the extent to which this is the case, would be reduced if plants can also move resources between patches (i.e. a combination of fine and coarse control).

7.3 Material and methods

7.3.1 Culture of plants and fungi

For our experimental test, we used two closely related AM fungi of the highly cosmopolitan sub-genus Glomus Ab (Schwarzott et al. 2001), Glomus intraradices

Schenck & Smith (but see Stockinger et al. 2009 for discussion of G. intraradices re- classification) and Glomus custos Cano & Dalpé. The use of closely-related AMF allowed us to focus only on fungal cooperative strategy while excluding differences associated with radically contrasting life history traits (Denison and Kiers 2011). Both strains were isolated from Southwest Spain but represented two ends of the mutualistic continuum. At comparable colonization levels, inoculation with the

“high-quality mutualist” G. intraradices results in significantly higher host biomass

156 production and phosphorus (P) content compared to the “low-quality mutualist” G. custos in both monocot and dicot plant species (Kiers et al. 2011). We chose Plantago lanceolata L. as a host because it is a model plant species in AM fungal research and is readily colonized and highly responsive to a broad range of AM fungal taxa

(Maherali and Klironomos 2007).

7.3.2 Experimental setup

Our experiment consisted of three fungal treatments: G. intraradices alone, G. custos alone, and an equal mixture of both (“combined”). We planted seedlings in 2.6 L pots, grown from seeds that were surface-sterilized with 10% sodium hypochlorite.

We inoculated roots of seedlings with ~1500 spores and 1.0 g of in vitro root material of either G. intraradices or G. custos (provided by Mycovitro S.L. Biotechnología ecológica, Granada, Spain). These cultures were originally started from single spores, and thus considered to exhibit high genetic uniformity (but see Angelard et al. 2010).

In the “combined” treatment, we halved inoculum amounts so that seedlings received equal amounts of both AM fungal species. Plants grew in autoclaved, nutrient-poor dune sand (as in: Scheublin et al. 2007), with the following characteristics: pH: 7.2; OM: 0.2 %; P: 0.3 mg/kg (CaCl2-extracted); N-total 190 mg/kg. We maintained soil humidity at 70% water holding capacity and biweekly, we added nutrients, 8 ml of 1/2 P Hoagland solution (Arnon and Hoagland 1940) per pot.. We randomized pots biweekly on benches, and they grew for 12 weeks per generation under semi-controlled light intensity with a 16 h light and 8 h dark rhythm.

We grew plants for three consecutive seasons. After the first generation, we randomly assigned the soils to either a “mixed soil” treatment (reduced spatial structuring i.e. low population structure) or “non-mixed soil” treatment (control

157 spatial structuring i.e. high population structure). In both treatments, we removed the central root with a core (2 x14 cm), and replaced it with an equal core of sterile soil (± 50 g). For the mixed treatment, we removed all soil and roots from the pots and cut the roots into ~1 cm pieces. Contents were then mixed and transferred back into their respective pots separately. In the non-mixed treatment, we left the soil and roots outside the core intact. We treated all pots with an excess of demineralized water to standardize for soil compactness, and then planted a new seedling. We replicated each fungal treatment x soil treatment 12 times for a total of 72 pots. In the beginning of the third generation, one plant in the high-quality/non-mixed treatment died resulting in only 11 biological replicates.

7.3.3 Plant and fungal measurements

After each generation, we harvested the aboveground plant material, which was oven dried it at 70°C for 72 h, and weighed. After the first generation, we determined

P content by the molybdate blue ascorbic acid method following acid digestion

(Watanabe and Olsen 1965) and N and C contents by dry combustion on an

Elemental Analyzer (NC2500, ThermoQuest Italia, Rodana, Italy). For each pot, we removed four evenly spaced soil cores (same as used for removing central root system) and which were replaced by sterilized sand. Half the cores were freeze-dried overnight and the other half used to measure the mycorrhizal colonization using trypan blue staining and the magnified intersection method (McGonigle et al. 1990;

100 intersections per replicate).

To compare the relative abundance of competing fungal species in the “combined” fungal treatments, we performed DNA analyses after each generation and measured the fungal abundances in the single species treatments after generation three. We pooled roots from two cores per replicate, weighed and ground them in a

158 microcentrifuge tube using a 4 mm glass bead in a FP120 bead beater (ThermoSavant,

Holbrook, USA; 4.0 m s-1, 2 x 10 s). We spiked each sample with 2 x 1010 copies of a plasmid containing the cassava mosaic virus sequence motif. This served as an internal standard in order to exclude variations in isolation efficiency or pre-analysis

DNA degradation (see also Appendix 7.67.). We extracted DNA from the resulting mixture using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the standard protocol and eluted in 100 µl of the provided elution solution. We analyzed the abundance of each species by quantitative PCR primers targeting the mitochondrial large subunit (mtLSU) rRNA gene (see: Kiers et al. 2011). These primers were specifically designed to quantify the mtLSU copy numbers of each species, and have been found to be highly specific and accurate (for design, optimization and validation see Appendix 7.6).

7.3.4 Statistical analyses

We performed all statistical analyses using SPSS version 17.0. To establish the effects of the factor “species” (with levels: “high-quality”, “low-quality”, “combined”) on plant biomass, we tested N content, P content, and hyphal colonization of roots by

MANOVA (Hyphal colonization of roots was arcsin-transformed to meet assumptions). Then, in order to assess which of these response variables contributes most to differences between inoculation treatments, we performed a Discriminant

Analysis. To test whether response variables were significantly different between plants inoculated with different species we performed a one-way ANOVA for each response variable separately, followed by Tukey’s b post-hoc test and pair-wise

ANOVAs for each of three species treatments.

We performed separate ANOVAs for shoot biomass differences in mixed and non- mixed pots between first and final generation for the “combined” treatment. We

159 performed an ANOVA on colonization responses (i.e. percentage root colonized by hyphae, arbuscules or vesicles as indication of differences in the abundance of fungal nutrient transfer structures (e.g. arbuscules) versus storage structures (e.g. vesicles, see Johnson et al. 2010) in the final generation. For the single species treatments, we also compared relative abundance of arbuscules compared to vesicles for both mixed and non-mixed treatments, which were ln transformed to meet assumptions for

ANOVA. We analyzed qPCR-derived abundance data of each species for the “high- quality”, “low-quality”, and “combined” fungal treatment between mixed and non - mixed soil treatments using the non-parametric Mann-Whitney U test because the requirement for a normal data distribution of ANOVA was not met. Ranked abundance of each strain did not significantly deviate in their homogeneity of variance - which is a requirement of the Mann-Whitney U test (Ruxton and

Beauchamp 2008). In case of “no detection”, we replaced species abundance by the detection level, after confirming that these cases did not appear to represent statistical outliers.

7.4 Results

7.4.1 Benefits conferred by fungal species to host plants

Plant growth and nutritional status after the first generation confirmed that the two fungal species differed in mutualistic benefits conferred to their hosts. This is consistent with previous plant growth results that utilized these fungal species (Kiers et al. 2011). MANOVA analysis indicated a significant treatment effect on plant responses (Pillai’s trace, V = .98, F8,134 = 4.64, P < .001), and Discriminant Analysis suggested an especially high difference between the low-quality mutualist versus the high-quality mutualist and combined treatment (Appendix 7.6). Above ground

160 biomass contributed most strongly to this separation, followed by shoot P content, hyphal colonization and lastly shoot N content (Appendix 7.6).

Inoculation by the high-quality mutualist (G. intraradices) resulted in significantly higher aboveground biomass (F1,46 = 18.15, P =.001) and shoot P content ( F1,46 = 10.92,

P = .002) than the low-quality mutualist (G. custos; fig. 7.2). The plants that were inoculated with both species (“combined treatment”) did not differ from plants that were only inoculated with the “high-quality” species G. intraradices (Biomass: F1,46 =

0.40, P = .531; P content: F1,46 = 0.20, P = .661; fig. 7.2). We found no significant differences in plant N content (F2,69 = 0.34, P = .710) suggesting that biomass differences are predominantly the result of an improved P nutritional status and/or lower C costs. The hyphal colonization levels of the high-quality and low-quality species in the single-inoculation treatments were high (80 ± 1.8% SEM versus 71 ±

3.3%, respectively; all error estimates onwards represent 1 SEM unless stated otherwise), confirming that the host was able to form effective associations with both partners. Although these colonization levels differed significantly (F1,46 = 6.67, P =

.013), the high colonization (above 70%) in each treatment and its low contribution to treatment separation in Discriminant Analysis (see Appendix 7.6) suggest that differences in mycorrhizal colonization were not the primarily cause for differences in host benefit.

161 A a a

b Shoot biomass (g) biomass Shoot

High Low Combined quality quality B a a

b content (mg) content – Shoot P Shoot High Low Combined quality quality C a a a content (mg) – N Shoot High Low Combined quality quality

Figure 7.2. Effect of fungal treatments on shoot biomass and P content. Shoot biomass (A) and P content (B) of plants from generation one were significantly higher for plants inoculated with the high-quality species than with the low-quality species. For N content (C) there were no significant differences between treatments. When inoculated with both AM fungal species (“combined treatment”), the host plant biomass did not differ significantly from plants that were only inoculated with the high-quality species. Bars represent mean ± 1 standard error. Different letters represent significant differences according to Tukey’ b post-hoc test.

162 7.4.2 Mixed versus non-mixed soil, effects on AMF mutualists

We found that soil mixing significantly affected the success of competing high- and low-quality mutualists. In the combined treatment, the abundance of the low-quality mutualists was significantly reduced after two consecutive generations of soil mixing

(Z1,22 = -2.83, P = .004, fig. 7.3). This effect was not due to a higher tolerance to soil mixing by the high-quality species. In the single species treatments, mixing reduced the abundance of the high-quality species (from 8.2 (± 1.4) ⋅107 to 4.4 (± 0.8) ⋅107

-1 marker copies g root; Z1,21 = -2.28, P = .023), whereas there was no significant negative effect of mixing on the abundance of the low-quality species in the single species treatment. There was even a marginal positive effect of soil mixing on the low-quality species when grown alone (from 6.2 (± 0.4) ⋅106 to 10.5 (± 2.3) ⋅ 106 marker

-1 copies g root; Z1,22 = -1.67, P = .083).

7.4.3 Plant benefit under mixed and non - mixed conditions

When both high- and low-quality fungal mutualists were present, soil mixing had a marginally significant positive effect on host shoot biomass compared to non-mixed conditions (F1,22 = 3.40, P = .079; fig. 7.4A). A positive effect of soil mixing on the symbiosis is further supported by the finding that in host roots, under mixed conditions, more arbuscules and fewer vesicles were formed than under non-mixed conditions, whereas there was no difference in hyphal colonization (arbuscules: F1,22 =

8.77, P = .007; vesicles: F1,22 = 7.92, P = .010; hyphae: F1,22 = 0.03, P = .860; fig. 7.4B). A high relative abundance of arbuscules (structures important for nutrient transfer) compared to vesicles (structures for storage) has been shown to be a useful indicator for mutualistic quality of AM fungi (Johnson et al. 2003; Johnson et al. 2010). This higher arbusculae abundance could indicate increased colonization by the high- quality species because we found that high arbuscule abundance (compared to

163 vesicle abundance) is a consistent identifiable trait of the high-quality species under both mixed (F1,22 = 42.28, P = .001) and non-mixed conditions (F1,21 40.38, P = .001).

Non – mixed soil 2,02 High quality species A Mixed soil root) 1 - 1.51,5

1,01

0,5

marker copies g copies marker 0.5 8 AM fungal abundance

(x 10 (x 0,00 5,05 B Low quality species root) 1 - 4,04 ** 3,03

2,02 marker copies g copies marker 1,0 5 1 AM fungal abundance

(x 10 (x 0,00 1 2 3 Generation

Figure 7.3. Effect of soil treatments on abundance of the fungal species, (A) high-quality Glomus intraradices and (B) low-quality Glomus custos in the “combined” treatment for each. In generation three, abundance of the low-quality species after mixing was significantly reduced compared to the non - mixed control. Error bars represent 1 standard error. Significant differences between mixed versus non - mixed are indicated with asterisks (** = P ≤ .01). Note that mixing occurred twice (generation 1 to 2 and generation 2 to 3).

164 A (*) biomass (g)biomass Increase shoot Increase

Non - mixed Mixed

B ** **

= arbuscules = vesicles Colonization % Colonization

Non - mixed Mixed

Figure 7.4 Effect of soil mixing on plant biomass and mycorrhizal colonization in “combined” treatment. Host plant increase in biomass from first to third generation when both AMF species were present (A) was marginally significantly larger (i.e. (*), P = .079) when soil was mixed. This effect was associated (B) with a significant increase of arbuscular and decrease of vesicular colonization. Asterisks indicate significant difference (** = P ≤ 0.01). Hyphal colonization did not differ between treatments (not shown). Bars indicate mean ± 1 standard error.

7.5 Discussion

Spatial structure is frequently assumed to favour cooperation within mutualistic interactions (Frank 1994; Doebeli and Knowlton 1998; Bever and Simms 2000;

Hoeksema and Kummel 2003; Foster and Wenseleers 2006; Gardner et al. 2007; Lion and van Baalen 2008; Platt and Bever 2009; Hodge et al. 2010). Other models (e.g.

(Sherratt and Roberts 2002) have found that cooperation can be maintained even in

165 the absence of spatial structure depending on investment-response rules. Our model predicts that spatial structuring can indeed have contrasting effects on selection for cooperation, depending on the scale (fine or coarse) at which hosts enforce cooperation (fig. 7.1). We tested these predictions using the arbuscular mycorrhizal symbiosis as a model system. We found that when we mixed the soil (reduced spatial structuring) in the combined treatment, this decreased the success of the low-quality species (fig. 7.3B). In contrast, mixing soil had no significant effect on the success of the high-quality species (fig. 7.3A).

There is a large theoretical literature examining how spatial structure would influence the evolution of cooperation. These include models for cooperation both between (see above) and within (reviewed by Lehmann and Rousset 2010; Nowak et al. 2010) species. Within species, population structuring keeps relatives together, and hence can favour cooperation through kin selection (Hamilton 1964), as long as this is not negated by increased competition between relatives (West et al. 2002b; Lehmann and Rousset 2010). Between species, population structure can generate spatial correlations in the tendency to cooperate, which can favour cooperation (Frank 1994).

In all cases, a general feature is that population structuring usually favours cooperation, because of its tendency to create genetic correlations either within or between species (for an exception, see Hauert and Doebeli 2004). We gain the opposite prediction, because in our model, population structure influences the evolution of cooperation in a very different way. Specifically, it alters the extent to which one partner is able to preferentially reward more cooperative partners.

We used soil mixing as a mechanism to decrease the extent to which fungi populations are structured (i.e. to produce a lower Fst). Our model predicts that if plants can enforce cooperation at a fine-scale then lower spatial structuring will be relatively detrimental to the success of low-quality species (fig. 7.1). This effect arises

166 because it will lead to patches containing both high- and low-quality mutualists, which facilitates the extent to which plants can differentially allocate resources to high-quality mutualists. Our empirical results are consistent with this model prediction (fig. 7.3B), although there may be other explanations (see discussion below). We also found that soil mixing led to a small increase in benefit for the host itself (fig. 7.4A). This benefit likely arose because the host was associating predominately with the high-quality species.

Our work suggests that if hosts are able to enforce cooperation on a fine scale then high spatial structuring of the symbiont community may not be critical for stabilizing the partnership, at least in the model-system tested here. Recent work has highlighted potential mechanisms plants may employ to enforce cooperation at the fine-scale assumed in our model (e.g. level of arbuscule). For example, stunted arbuscule morphology has been described when arbuscular-specific plant phosphate transporters were knocked out (Javot et al. 2007), suggesting that when fungal symbionts fail to deliver P, plant hosts may reduce C allocation locally (Parniske

2008) or even digest individual arbuscles (Kobae and Hata 2010). Other work has pointed to the role of lysophosphatidylcholine (LPc), a compound that may help hosts to sense P concentrations, potentially allowing hosts to evaluate the amount of

P delivered via the mycorrhizal pathway (Bucher et al. 2009). Now more work is needed to establish how plants physically control C transport across individual interfaces.

Are there alternative explanations for how soil mixing could favour the spread of the high-quality symbiont? Other differences between the species, such as ability to tolerate disturbance, could explain why the high-quality symbiont responds better to mixing. However, we can tentatively reject this hypothesis because the low-quality species appears to have greater inherent tolerance of mixing. In the single species

167 treatments, mixing reduced the abundance of the high-quality species (decrease of

~50%, P = .023), whereas mixing had no significant effect (even a small mean increase of ~60%, P = .083) on the low-quality species.

A second way in which soil mixing could favour one symbiont over the other would be through changes in the competitive dynamics between the two species not related to host C-rewards. For instance, variation in life-history traits could mean that one species proliferates more, or earlier than the other, depending on our experimental conditions. However, by choosing closely-related species, we reduced the variation associated with diverse life-history strategies typically found across genera (Denison and Kiers 2011). Future work should aim to extend these types of multi-generational tests to more fungal species pairs so broader conclusions can be drawn.

One important assumption of our work was that the control treatment, without soil mixing, showed a higher level of spatial structuring than the mixed treatment. Spatial structuring of AM fungal communities can arise from differential soil colonization patterns of strains (Smith et al. 2000; Oehl et al. 2005) or clustered sporulation (Jansa et al. 2002). In line with previous recommendations (e.g. Bever et al. 2009), we examined fungal competition under relatively realistic spatial conditions, i.e. the structuring that originates during one or two generations of a successful symbiotic interaction (non-mixed) compared to a reduction of that structuring (mixed). We assume that even our non-mixed soil treatment was exposed to some mixing from root forging and other processes. No soil in nature would ever be free from mixing, especially at the fine-level of the arbuscule. Manipulative tests of spatial structure should be extended to the field where natural processes such as non random host-

AMF associations (Öpik et al. 2009), burrowing animals (Mangan and Adler 2002) and fungal aggregation in response to plant signaling cues (Yoneyama et al. 2008) can both increase and decrease spatial structure of symbiont communities.

168 Is there evidence for strong spatial structuring of symbiont communities in nature?

While studies have demonstrated spatial structuring at the field or plot level

(Mummey and Rillig 2008), the few studies that have examined spatial distribution of

AM fungal species within individual plant roots, find fungal strains to be relatively well-mixed (Alkan et al. 2006). This could be because relatively large root fragments

(± 5 cm) were examined (Jansa et al. 2003) or small fragments were pooled (Wolfe et al. 2007). The smallest root fragments studied thus far to our knowledge (~1 cm), did not reveal strong structuring (van Tuinen et al. 1998; Janouskova et al. 2009). A quantitative approach (e.g. Alkan et al. 2006) in which strain abundance (rather than presence-absence) is measured in small root fragments, will be the most successful at determining the degree of spatial structuring of AM fungal communities.

We found that the low-quality mutualist was not completely replaced by the high- quality mutualist. If the experiment had been continued for more generations, the abundance of the low-quality mutualist would have likely decreased further.

However, our results suggest that the high-quality mutualist does not completely replace the other. One explanation is that host plants are physically unable to exclude colonization of a particular AM fungal species. Another explanation is that some fungi may provide benefits not measured (e.g. defense against pathogens; Herre et al.

2007), or only provide benefits under different conditions, and thus even ‘putative’ low-quality mutualists are retained in low abundances (Smith et al. 2009; Palmer et al. 2010).

Can our model be extended to other systems? It would be interesting to test our model predictions in the legume-rhizobia mutualism, where fine-scale rewarding (at the level of the individual nodule) is known to operate (Kiers et al. 2006; Simms et al.

2006; Oono et al. 2011). Like in the fungal system, we would predict that soil mixing has the potential to be positive (or at least not negative) on the success of the high-

169 quality strain by exposing the host to a greater diversity of potential partners. The only complication is that ‘mixed nodules’ (containing more than one rhizobial strain) are known to occur with some frequency in the field (e.g. 12-32%; Moawad and

Schmidt 1987). There is currently no evidence that hosts can differentially control resource distribution to strains within a single nodule (Denison and Kiers 2011). So if soil mixing also increased the frequency of ‘mixed nodules’, this could undercut any positive benefit for the host plant (Kiers et al. 2002; Friesen and Mathias 2010); Till and no-till agricultural systems could be useful for field-level experimental tests of our predictions, especially in cases where high-quality fungal and rhizobial inoculum have been introduced. We predict that tillage could have a positive effect in promoting cooperation in the mycorrhizal symbiosis, but this effect is likely overridden by life history differences among fungi in respect to their ability to handle extreme disturbances (Verbruggen and Kiers 2010).

More generally, whether spatial structure will stimulate or decrease mutualistic cooperation depends on the biology of the interaction. If a mutualism depends upon partners cooperating to produce a beneficial common good, without a mechanism to reward more cooperative partners (or punish less cooperative partners), then increased spatial structure favors cooperation, because it increases relatedness amongst symbionts leading to a shared interest between symbiont and host (Frank

1994; Bever and Simms 2000; West et al. 2002a; Foster and Wenseleers 2006). This is analogous to how spatial structuring favours cooperation within species (e.g.: Griffin et al. 2004; Diggle et al. 2007; Cornwallis et al. 2009; Lehmann and Rousset 2010). In other symbiotic interactions, like the mycorrhizal symbiosis, mixing may increase the potential suite of partners available to host plants. This could increase host plant benefit as long as plants can discriminate among partners at a fine-scale. Research is now needed to (i) identify potential physiological mechanisms that would allow

170 plants to control allocation processes, and (ii) test these dynamics in the field to determine the actual effect of symbiont mixing under natural conditions.

7.6 Appendix

7.6.1 Quantitative real-time PCR (qPCR) markers design, validation, optimization, and Calibration

Design of qPCR markers (primers with hydrolysis probes) for specific detection and quantification of Glomus intraradices and Glomus custos (isolates obtained from

Mycovitro S.L. Biotechnología ecológica, Granada, Spain) targeting the mitochondrial large subunit RNA gene.

DNA was extracted from AMF spores produced monoxenically with carrot root organ cultures, or from the colonized carrot roots from the same cultures, using the

DNeasy Plant Mini kit (Qiagen, Hombrechtikon, Switzerland), following manufacturer recommendations. The final volume of the DNA preparations from spores was 20 µl (instead of 100 µl recommended by the manufacturer) to maximize

DNA concentrations before PCR. The roots samples were eluted with 100 µl of the elution buffer. Monoxenic fungal cultures were provided by A. Bago, H. Bucking and

M. Hart.

DNA was subjected to PCR amplification of the mitochondrial large ribosomal subunit (mtLSU) RNA gene with following primer pair combinations, according to

Börstler et al (2008): RNL11- RNL17, RNL1-RNL14, or RNL1-RNL15. The PCR was carried out using Taq PCR Core kit with CoralLoad reaction buffer (Qiagen), using 25

µl PCR reaction volume, 1µM concentration of each primer, and 38 cycles

(denaturation at 95°C for 10s, annealing at 50°C for 90s and amplification at 72°C for

171 90s). Amplified DNA fragments were cloned into a blue-script vector (pGEM-T Easy vector system; Promega, Dubendorf,̈ Switzerland) and sequenced at Microsynth

(Balgach, Switzerland). The sequences were individually edited and the clones re- sequenced if quality of the reads proved insufficient. The identity of the sequences was revealed by BLAST search (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to sort out contaminant sequences (e.g., bacteria, unspecific amplifications of other genome regions etc.). Several partial sequences of the mitochondrial large subunit of the different AMF species were obtained, which were deposited in the GenBank under the accession numbers HQ706096-HQ706103.

These sequences were aligned with other mtLSU sequences available in the GenBank

(e.g., Glomus intraradices, Glomus proliferum, Glomus clarum) and at least two combinations per AMF species of species-discriminating qPCR primers with associated hydrolysis probes were designed using the AlleleID version 6 software

(Premier Biosoft International, Palo Alto, California, USA). Care was taken to target mtLSU regions coding for the ribosomal RNA, i.e. avoid putative introns described recently by Thiery et al (2010). Specificity of the primers and fluorescent probes was confirmed with a BLAST search and the oligonucleotides (primers and dually labeled hydrolysis probes, labeled with fluorescein at 5`-end and BHQ-1 quencher at 3`- end) were then synthesized in Microsynth. The primers were purified by preparative

HPLC and the probes by preparative polyacrylamide gel electrophoresis before lyophilization. Both primers and probes were diluted with PCR-grade water to achieve 25 µM concentrations, aliquoted (20 µl each) and frozen at -20°C.

172

→ Sequences 5` 3` Denaturation Annealing Amplification Marker Target Nr cycles (forward primer, reverse primer, (°C / s) (°C / s) (°C / s) hydrolysis probe)

TTTTAGCGATAGCGTAACAGC, TACATCTAGGACAGGGTTTCG Glomus intra mt5 65 95 / 10 60 / 10 72 / 1 intraradices FAM-AAACTGCCACTCCCTCCAT ATCCAA-BHQ1

72 /1 TCTAACCCCAGAAATGTATAG,

AAGGACTGCCTTGTGTTC Glomus cust 65 95 / 10 62 / 15 custos FAM-ATACAATAATGGGCAATC

AGACATATCGT-BHQ1

CGAACCTGGACTGTTATGATG, AATAAACAATCCCCTGTATT Cassava TCAC CMV 45 95/10 50/30 72/1 Mosaic Virus FAM-CACCAGGCACCAACA ACGACCATT-BHQ1

FAM – floursecein; BHQ1 – floursescece quencher

Table 7.1: qPCR markers for specific quantification of Glomus intraradices and Glomus custos by means of measuring gene copies of the mitochondrial large ribosomal subunit of the respective AMF species. Also represented are the markers for partial Cassava Mosaic Virus DNA sequence (Genbank accession number: AJ427910; inserted in vector pUC19) quantification, used as a control in the experiment.

7.6.1 Marker selection, optimization of cycling conditions, cross-reactivity testing

First, the markers were tested for their specificity under low stringency cycling conditions (denaturation at 95°C for 10 s, annealing at 52°C for 30 s, and amplification at 72°C for 5 s). In this assay, DNA extracts from Medicago truncatula roots colonized by the different AMF (3 individual samples per each AMF species) were used as templates. Second, markers showing greatest specificity towards their target species (either no cross-amplification with other species or the greatest difference in Cq value between target and non-target species) were selected for further optimization (Table 7.1). Stringency of cycling conditions was then stepwise increased for each of the markers so as to achieve a complete avoidance of

173 amplification with non-target samples (see Table 7.1 for details of the optimized cycling conditions). The markers were confirmed to amplify their target AMF species and to avoid non-target AMF species as well as the plant DNA (tested on spores and colonized roots of Medicago truncatula and Daucus carota roots). All qPCR assays were carried out in 9 µl reaction format, using the LightCycler® 2.0 instrument,

LightCycler TaqMan chemistry (LightCycler® TaqMan® Master, cat. Nr

04735536001) and 20µl-Lightcycler® glass capillaries (cat. Nr 04929292001). Final concentration of the primers was 0.5 µM each and the final concentration of the hydrolysis probe was 0.11 µM. 2.25 µl DNA template (i.e, sample) was included in each reaction.

Marker system Detection limit (Cq) Threshold mtLSU gene copy concentration (copies µl-1) intra mt5 37.62 199 cust 35.6 10 Table 7.2: Detection limits and minimal detectable target gene concentrations of the two qPCR assays

7.6.3 qPCR calibration, detection limits

Two to four individual plasmid preparations per AMF species were used for calibration of the qPCR markers. These plasmids were carrying fragments of mtLSU of the respective AMF species with 100% sequence match to the region amplified by the markers (confirmed by sequencing). Plasmids were isolated from overnight cultures of the transformed E.coli JM109 (Promega), grown on LB medium supplemented with Ampicillin (100 µg ml-1), using the Miniprep procedure

(Sambrook et al. 1989). The plasmids were linearized using EcoRI+ (Fermentas, Le

Mont-sur-Lausanne, Switzerland) digestion, incubating the samples at 37°C for 2h and then at 65°C for 20 min. The concentration of the DNA was then measured by

PicoGreen fluorescence assay (Invitrogen, cat. Nr P7589), using LighCycler 2.0 at

174 45°C and measuring the emission at 530 nm. The concentration of plasmid copies per unit of sample volume was calculated according to Jansa et al (2008), by knowing the concentration of DNA in each sample, length of insert (176 bp for G. intraradices and

438 bp for G. custos) , length of vector (3015 bp), and assuming molecular weight per nucleotide double-stranded DNA to be 660 Da. Plasmid preparations were serially diluted (5fold and 10fold) to achieve a range of plasmid concentration from few billions through less than 1 per microliter. These were used for establishment of:

1. Calibration curves for conversion of the qPCR detection cycle (Cq) to gene copy concentrations

2. Assessment of detection limits of the qPCR markers.

Figure 7.4 Typical response of qPCR assay to DNA template dilution. Linear response region contains values used for calibration of the qPCR assay and the background region has been used for assessment of detection limit of the qPCR assay.

Namely, the Cq is typically negatively and linearly correlated to the log-transformed template concentrations (Fig. 7.4, linear response region), until the detection limit of

175 the assay is reached and the Cq becomes independent of the further dilution (Fig. 7.4,

background region), or there is no response at all.

Linear response region of each of the calibration assay has been used to derive

equations allowing conversion of Cq values to mtLSU gene copies per unit volume of

the template (Fig. 7.5)

Detection limits were derived from the background region of the qPCR response

curve as follows:

DL AV 3 SD AV , = Cq(back) − × ( Cq(back) )

Where DL stays for detection limit of the assay (Cq value), AVCq(back) stays for mean of €

the Cq values in the background region and SD(AVCq(back)) stays for standard

deviation of this mean. The detection limits of the three marker systems and

corresponding threshold mtLSU concentrations are given in Table 7.2.

These assays were then used for assessment of mtLSU gene copy concentrations in

DNA samples taking into account any dilutions of the template during the sample

processing.

176

Figure 7.5 Calibration curves for the qPCR assays targeting mtLSU of Glomus intraradices and G. custos. Equations for conversion of the qPCR signal to mtLSU copy concentrations in template are given for each assay. CP stands for number of target gene copies per microliter template. (Glomus intradadices: R2 = 0.998, ;

Glomus custos: R2 = 0.999, .

177 Chapter 8. Discussion.

Each thesis chapter contained its own detailed discussion. The aim of this chapter is to summarise the main conclusions, highlight avenues for future research and consider some of the general issues my work raises about social evolution research, specifically when applied to the evolution of cooperation in humans.

8.1 Chapter summaries

In Chapter 2, I examined whether knowing ones dispersal status could affect the evolution of social behaviours. This is because, in general, non-dispersers are more highly related to their neighbours whilst dispersers are less related. I extended

Taylor’s infinite island model (Taylor 1992a), to compare the case when individuals cannot adjust their behaviour depending on their dispersal status, with the case when they can. I found that helping was more readily favoured among non- dispersers and harming more readily favoured among dispersers. I also showed that harming by dispersers evolved more readily than does helping by non-dispersers.

This is because zero relatedness (on average) is guaranteed by one’s own dispersal, irrespective of the population level of dispersal, whereas the relatedness achieved by not dispersing is nonzero only insofar as one’s social partners are also non- dispersers. In the future, I would like to build upon this work to examine the coevolution of dispersal and social behaviours in more detail, in particular the coevolution of dispersal and harming.

In chapter 3, I extended earlier work by Frank (1995, 1996b, 2003, 2009) to examine the evolution of cooperation by policing. The original aim was to investigate how

178 different population structures alter selection for enforcement mechanisms within groups. However, the focus became more general when I examined Frank’s model and found qualitatively different results were obtained by relaxing some potentially unrealistic implicit assumptions. One of these implicit assumptions was that the individual cost of investment into policing fell as selfishness increased, and it was this which generated the model’s result that policing favours high levels of selfishness. By relaxing this assumption, I found that policing was favoured only at intermediate levels of relatedness and that when it evolved it led to a reduction in selfishness. I also relaxed the assumption that policing perfectly recovers the group resources that are lost due to selfishness and found this modification meant full policing was disfavoured. I considered the impact of demography on the coevolution of policing and cooperation and showed that policing is maintained more readily than it invades (Figure 3.4). A future goal is to model the evolution of inequality by extending Frank’s (1996b) model of policing when resources vary, to consider the coevolution of selfishness and mechanisms that could enforce an inequitable distribution of resources in a group.

In chapter 4, I examined how competition between groups affected selection for social behaviours. Conflict, or warfare between groups is seen as a potentially important mechanism favouring the evolution of altruism in humans (Choi and

Bowles 2007; Bowles 2006; 2009), however the evolutionary logic underpinning this idea was not clear as I identified several inconsistencies in their model assumptions.

Using this previous work as a starting point, I investigated what forms of sociality are favoured when groups compete. I showed that in contrast with previous results, group competition could favour the evolution of any social behaviour within a group so long as it improves the group’s ability to compete. In the future, I would like to investigate the evolution of group competition as a means to maintain within-group

179 cooperation. This builds upon Hamilton’s (1996) idea that if relatives cannot avoid bunching in groups, then the group itself must expand at the expense of others because to be effective, altruism must put offspring in competition with non-altruists, not bunch them into wasteful competition with their own kind.

Chapter 5 extends the model developed in chapter 3 to investigate the idea of reproductive levelling. Reproductive levelling is the term used to refer to policing mechanisms in the human literature and has been suggested as an important factor in the emergence of egalitarian societies (Bowles 2006; Boehm 1997, 2001). Previous models of reproductive levelling examined how culturally inherited traits (e.g. norms for sharing resources) could aid the evolution of genetic cooperation (Bowles 2006).

However this work assumed a culturally inherited norm for reproductive levelling was already present in the population at no cost; it did not model how it might evolve in the first place. I considered the coevolution of a genetically inherited trait for competitiveness and a culturally inherited trait for reproductive levelling and demonstrated that costly, culturally-inherited traits for reproductive levelling could be favoured in such a way that they select for genetically-inherited self-restraint. I showed that reproductive levelling is most strongly favoured when group members are genetically unrelated but share the same culture. Therefore, the gene-culture coevolution of reduced competitiveness and reproductive levelling may explain why the simplest societies tend to be both cooperative and egalitarian. However, this result comes with an important caveat: it is unclear if cultures evolve in the way this model assumes. More work is needed to model cultural change realistically and for this reason, one of my research priorities is to use the formal Darwinism approach

(Grafen 1999, 2007) to investigate whether or when the dynamics of cultural transmission can to lead to a maximisation principle.

180 In chapter 6, I developed a class-structured model using an inclusive fitness approach to consider how parental promiscuity affected selection on juveniles to stay and help their parents breed. Previous authors (Hughes et al. 2008; Boomsma 2009) have suggested that low promiscuity or monogamy may favour the evolution of cooperative breeding as the relatedness between parent and offspring and the relatedness between siblings will be similar, meaning that the fitness benefits of raising siblings is potentially the same as the fitness benefits of reproducing independently. By contrast, I found that promiscuity had very little effect on selection for helping. This was because of a trade-off: staying to help benefitted the parent by increasing their survival, but it led to increased competition with siblings for the breeding position. This competition meant that when there was monogamy (high relatedness), cooperative breeding was disfavoured as the best way to aid relatives was by dispersing to avoid competing with them. In nature, the cost of competition between siblings must be decoupled from the benefits of cooperative breeding, for example via delayed dispersal, as there is ample evidence to suggest that monogamy does favour the evolution of cooperation (Hughes et al. 2008; Boomsma et al. 2009;

Cornwallis et al. 2010). This work illustrates the value of formalising verbal arguments as even straightforward ideas, like the monogamy hypothesis, can yield surprisingly complex results.

In chapter 7, I presented the results of some collaborative work I undertook with plant science researchers from the Netherlands. They wanted to know whether the soil-tilling practices used in industrial agriculture might affect the stability of the mutualism between plants and their arbuscular mycorrhizal (AM) fungal symbionts.

I developed a simple model to consider how changing the soil’s spatial structure (i.e. how mixed the soil is) interacts with the scale at which a plant has evolved to reward its fungal partners. The model showed that the effect of tilling on plant fitness

181 depended on the scale at which the plant rewarded its mutualistic partners.

This work led to an idea for some future work concerning mutualisms where a single host has a very large number of symbionts, such as the squid-bioluminescent bacteria mutualism. In these types of mutualism, if an empiricist wishes to estimate the inclusive fitness of the symbionts with the host, they must take great care as the relevant relatedness they need to measure will depend upon the scale at which the hosts meters out rewards or sanctions (Foster and Wenseleers 2006). At the extremes, if the host rewards or sanctions at the level of an individual symbiont cell, relatedness r=1 whereas if they reward or sanction at the symbiont population level then, in the limit, relatedness r→0). As high symbiont relatedness limits selection for free-riding, this suggests that mutualisms will be more stable if hosts possess adaptations that allow them to use targeted rewards or sanctions.

8.2 General themes

8.2.1 Implicit assumptions matter

Three of my chapters resulted from relaxing potentially unrealistic implicit assumptions in existing models. As many people who read theory papers are not mathematically inclined, they must trust that the theoretician’s written description reflects their maths accurately. If a model’s results rely upon unexplained assumptions, it is bound to cause problems. First, if the maths and written arguments are inconsistent, it obscures why cooperation occurs in the model and second, it makes testing the model’s predictions extremely difficult. If empirical researchers try to test such as model, they are likely to measure unsuitable parameters and may draw incorrect conclusions from their observations (for a related discussion from an

182 experimental perspective, see Appendix section 2). Implicit assumptions have led to widespread confusion in the human evolution literature where there are many instances of verbal arguments not describing properly how a model works (reviewed by West et al. 2011). For example Lehmann et al. (2007c) evaluated the implicit assumptions in two models that claimed to examine the coevolution of cooperation and punishment (Bowles and Gintis 2004; Gintis 2000). Contrary to the papers’ verbal arguments, Lehmann et al. (2007c) showed that the cooperative trait in these models merely acted as a tag for who carried the punishment trait; cooperation in itself did not generate the models’ results. Thus these models actually examined the evolution of spiteful greenbeards, which is a biologically unrealistic explanation for the evolution of cooperation in humans. I plan to continue to highlight the problem of implicit assumptions in models that examine the evolution of cooperation in humans.

I also plan to establish collaborations with empiricists who study human cooperation so as to formalize their research questions using simple, testable models, which may be presented alongside their quantitative results (akin to the project which resulted in chapter 7).

8.2.2 It is often easier to be nasty than nice

My work illustrates that punishment and enforcement mechanisms often evolve more readily than unenforced cooperation. In chapter 2, I found that harming by dispersers evolves more readily than helping by non-dispersers, as it is easier to guarantee the low relatedness needed for harming than it is the high relatedness needed for helping. Unenforced cooperation evolves when relatedness is high, but as

I illustrated in chapter 3, even with high relatedness there is significant scope for selfishness and the presence of enforcement mechanisms such as policing can still yield additional benefits. Furthermore, as Bourke (2011 and see my review in

183 Appendix section 1) emphasizes, once present, cooperation is not guaranteed to persist; if selfish individuals are not checked, they can make groups collapse or even cause species to go extinct (Rankin and López-Sepulcre 2005; Rankin et al. 2007).

My work also highlights why it is important to distinguish between evolutionary classifications for cooperation and the proximate details of how fitness effects are achieved. For example, in chapter 3, I showed that policing is altruistic in an evolutionary sense, however the proximate details of how groups enforce the distribution of resources need not pleasant. Similarly, in chapter 4, whilst social behaviours associated with group competition are always altruistic or mutually beneficial, I showed they could be anti-social and reduce the partial fitness of fellow group mates. For this reason, care is needed when engaging with research on human cooperation where the aim is typically to explain more proximate definitions of altruism, such as the altruistic act of helping another when there is no prospect of a reward. In Chapter 1 and in West et al (2011, reprinted in Appendix section 3), I explain that as we cannot know precisely what selective pressures favoured the behaviours we see today, it is impossible to say whether the adaptations which underpin specific behaviours, such as blood donation, voting, or recycling, which are described as altruistic, were favoured due to direct or indirect benefits.

8.2.3 Understanding how group and population structure affects the evolution of cooperation in humans is an important area for future research

My work highlights that details of a population’s structure, such as dispersal patterns

(chapter 2), group size and dispersal rates (chapter 3), the intensity of competition between groups (chapter 4) and the degree of mixing (chapter 7) are important determinants of whether or when social behaviours evolve. The importance of population structure for the evolution of cooperation is well known (Hamilton 1964,

184 Taylor 1992a) and there is an extensive body of work in the biological literature which examines how different facets of a population’s structure affects the evolution of cooperation, such as: overlapping generations (Taylor and Irwin 2000; Irwin and

Taylor 2001; Grafen 2007; Lehmann et al. 2007a; Taylor, et al. 2007), trans-generational social behaviours (Lehmann 2007), grouped (or ‘budded’) dispersal (chapter 1;

Gardner and West 2006; Lehmann et al. 2006), sex-biased dispersal (Johnstone and

Cant 2008) competition between groups (chapter 4; Hamilton 1996; Bowles 2006;

2009; Lehmann and Feldman 2008b) and changes in group size (Kokko et al. 2011).

Unfortunately, most of this research is unknown to researchers concerned with the evolution of cooperation in humans. Therefore, an important task for the future is to take this body of work and make it relevant to human cooperation researchers by developing specific models reflecting human demography that generate testable predictions.

8.2.4 The evolutionary importance of culture is unclear.

It is frequently assumed that cultural evolution is able to explain cooperation in cases where genetic selection cannot (Richerson and Boyd 2004; Henrich 2005, 2008).

Whilst humans do appear to be unique in the extent to which social learning allows them to adapt to local conditions, it is unclear what role culture plays in the evolution of cooperation. First, contrary to verbal claims (Henrich and Boyd 1998; Henrich and

Gil-White 2001; Henrich and McElreath 2003), whether or not cultural evolution promotes cooperation depends upon the exact mechanism of cultural transmission

(Feldman et al. 1985; Lehmann et al. 2007c, 2008; Lehmann and Feldman 2008). For example, prestige-biased transmission, where individuals copy a single successful individual in their group, may encourage the spread of harming behaviours if, in the short term, defectors appear more successful (Lehmann et al. 2008). Second, as

185 discussed in chapter 5, it is unclear whether culture itself possesses the inheritance properties necessary for natural selection to operate and lead to any maximising dynamics. It may be more accurate to view culture as a phenotypic tool where change is primarily guided by clever minds, which themselves are the product of natural selection.

One reason why cultural evolution receives so much attention appears to stem from misconceptions about the scope of genetic natural selection to favour cooperation

(see chapter 1 and Appendix section 3 for more discussion). It is widely assumed that evolutionary explanations for genetically inherited cooperation are limited to kin- selection and reciprocity. This means the role of genetic selection in the evolution of behaviours that stabilize cooperation in large groups or between non-kin is often neglected. Furthermore, the idea that culturally inherited institutions can enhance or encourage cooperative behaviours is receiving ever-more attention from think tanks and politicians (Bowles 2008; Willetts 2010). There is a real danger that in the near future, the results of evolutionary models that rely on untested or unrealistic assumptions will be used to inform public policy.

8.3 Conclusion

To conclude, the rapidly expanding field of human cooperation research offers great opportunities for synergistic collaborations between biologists, psychologists, economists, anthropologists, archaeologists and social scientists. One particularly challenging, yet rewarding, feature of human cooperation research is that because it spans so many disciplines, successful research is only possible when people from different traditions themselves learn how to cooperate. In order to facilitate future collaboration and the progression of the field’s research, there are several major

186 outstanding tasks. These include addressing the misconceptions about evolutionary theory before they become accepted wisdom, determining whether or when cultures evolve and how culture influences levels of cooperation and considering whether it is the insights from these diverse disciplines may be drawn together in a way that is appropriate for use by policy makers.

187 Literature Cited

Abbot, P., Withgott, J.H. and Moran, N.A. 2001. Bergmüller, R., Bshary, R., Johnstone, R. A., and Genetic conflict and conditional altruism in social Russell, A. F. 2007. ‘Integrating cooperative aphid colonies. Proc. Nat. Acad. Sci. USA 98: breeding and cooperation theory’, Behavioural 12068-12071. Processes 76: 61–72.

Alexander, R. D. 1987. The biology of moral Bever, J. D., and E. L. Simms. 2000. Evolution of systems. Aldine de Gruyter, New York. nitrogen fixation in spatially structured populations of Rhizobium. Heredity 85:366-372. Alizon S. and Taylor P. 2008 Empty sites can promote altruistic behavior. Evolution 62(6), 1335- Bever, J. D., S. C. Richardson, B. M. Lawrence, J. 1344. Holmes, and M. Watson. 2009. Preferential allocation to beneficial symbiont with spatial Alkan, N., V. Gadkar, O. Yarden, and Y. structure maintains mycorrhizal mutualism. Kapulnik. 2006. Analysis of quantitative Ecology Letters 12:13-21. interactions between two species of arbuscular mycorrhizal fungi, Glomus mosseae and G. Binmore, K. 1994. Game Theory And The Social intraradices, by real-time PCR. Applied and Contract Volume 1: Playing Fair. MIT Press, Environmental Microbiology 72:4192-4199. Cambridge, Massachusetts.

Arnon, D. I., and D. R. Hoagland. 1940. ‘Crop Binmore, K. 1998. Game Theory And The Social production in artificial culture solutions and in Contract Volume 2: Just Playing. MIT Press, soils with special reference to factors influencing Cambridge, Massachusetts. yields and absorption of inorganic nutrients’, Soil Birch, L. 1999. ‘Development of food preferences’, Science 50:463-485. Annual Review of Nutrition 19: 41-62. Axelrod, R. and Hamilton, W. D. 1981. ‘The Boehm, C. 1982. The evolutionary development evolution of cooperation’, Science 211: 1390–96 of morality as an effect of dominance behavior Bateson M, Nettle D, Roberts G. 2006. ‘Cues of and conflict interference. Journal of Social and being watched enhance cooperation in a real- Biological Sciences 5:413-22. world setting’, Biology Letters 2:412-414 Boehm, C. 1993. Egalitarian behaviour and Baum, W. B. 1994. Understanding behaviourism: reverse dominance heirarchy. Current Science, behaviour and culture. New York. Harper Anthropology 43(3): 227-254. Collins. Boehm, C. 2001. Hierarchy in the Forest: The Bell, A. V., Richerson, P. J. and McElreath, R. Evolution of Egalitarian Behavior. Harvard 2009. Culture rather than genes provides greater University Press. scope for the evolution of large-scale human Boehm, C. 1997. Impact of the Human Egalitarian prosociality. Proceedings of the National Academy of Syndrome on Darwinian Selection Mechanics. Sciences of the United States of America 106(42) The American Naturalist 150:S100-S121. 17671-17674

188 Boomsma, J.J. 2007. ‘Kin selection versus sexual Boyd, R. 2006. The Puzzle of Human selection: Why the ends do not meet’, Current Sociality, Science, 314: 1553 Biology 17(16), R673-R683. Brown, P. J., and Konner, M. 1987. ‘An Boomsma, J.J. 2009 Lifetime monogamy and the anthropological perspective on obesity’, Annals of evolution of eusociality. Philosopical Transactions the New York Academy of Sciences 499: 29–46. of the Royal Society B-Biological Science 364(1533), Bucher, M., Wegmuller, S. and Drissner, D. 2009. 3191-3207 ‘Chasing the structures of small molecules in Börstler B, Raab PA, Thiery O, Morton JB and arbuscular mycorrhizal signaling’, Current Redecker D. 2008. Genetic diversity of the Opinion in Plant Biology 12:500-507. arbuscular mycorrhizal fungus Glomus Burke, T., Davies, N.B., Bruford, M.W. and intraradices as determined by mitochondrial large Hatchwell, B.J. 1989. ‘Parental care and mating subunit rRNA gene sequences is considerably behaviour of polyandrous dunnocks Prunella higher than previously expected. New Phytologist modularis related to paternity by DNA 180:452-465 fingerprinting’, Nature 338: 249-251. Bourke, A.F.G. 2011. Principles of Social Evolution. Burnham, T. C. and Hare, B. 2007. ‘Engineering Oxford University Press. human cooperation. Does involuntary neural Bowles, S. 2006. ‘Group competition, activation increase Public Goods contributions?’, reproductive leveling and the evolution of Human Nature 18: 88-108. human altruism’, Science 314:1569-1572. Buss, L. 1987. The Origin of Individuality. Princeton Bowles, S. 2008. ‘Policies designed for self- University Press. interested citizens may undermine “the moral Charnov E.L. 1981 Kin selection and helpers at sentiments”: evidence from economic the nest - effects of paternity and biparental care experiments’, Science 320: 1605-1609 Animal Behaviour 29, 631-632. Bowles, S. 2009. ‘Did Warfare among ancestral Charnov, E.L. 1982. The theory of sex allocation. Hunter-Gatherer groups affect the evolution of Princeton University Press, Princeton, NJ human social behaviors’, Science 324:1293-98. Choi, J-K. and Bowles, S. 2007. The coevolution of Bowles, S. and Gintis, H. 2004. ‘The evolution of parochial altruism and war. Science 318:636-640. strong reciprocity: cooperation in heterogeneous populations’, Theoretical Population Biology, 65, Clutton-Brock T. H. and Parker G. A. 1995. 17–28. ‘Punishment in animal societies’, Nature 373: 209– 216 Boyd, R. and Richerson, P. J. 1985. Culture and the evolutionary process. Chicago University Press. Clutton-Brock T. 2009 Structure and function in mammalian societies. Proceedings of the Royal Boyd, R. and Richerson, P. J. 2005a. The origin and Society of London Series B-Biological Sciences evolution of cultures. Oxford University Press. 364(1533), 3229-3242. Boyd, R. and Richerson, P. J. 2005b. ‘Solving the Cohen, D., Nisbett, R. E., Bowdle, B. F. and puzzle of human cooperation’, in: Levinson, S. Schwarz, N. 1996. ‘Insult, aggression, and the (ed.), Evolution and culture, MIT Press, Cambridge southern culture of honor: An “experimental MA.

189 ethnography”’, Journal of Personality and Social Dawkins, R. 1976. The selfish gene. New York Psychology 70(5): 945-960. City: Oxford University Press.

Cornwallis, C. K., S. A. West, and A. S. Griffin. Dawson, A. 2008. ‘Control of the annual cycle in 2009. Routes to indirect fitness in cooperatively birds: endocrine constraints and plasticity in breeding vertebrates: kin discrimination and response to ecological variability’, Proceedings of limited dispersal. Journal of Evolutionary biology the Royal Society of London Series B-Biological 22:2445-2457. Sciences 363: 1621-1633

Cornwallis C.K., West S.A., Davis K.E., Griffin DeBruine, L., Jones, B. C., Little, A. and Perrett, A.S. 2010. ‘Promiscuity and the evolutionary D., I. 2008. ‘Social perception of facial transition to complex societies’, Nature 466(7309), resemblance in humans’, Archives of Sexual 969-U991. Behaviour. 37: 64-77.

Corriveau, K.H. and Harris, P.L. 2009. ‘Choosing Diggle, S. P., A. S. Griffin, G. S. Campbell, and S. your informant: Weighing familiarity and past A. West. 2007. Cooperation and conflict in accuracy’, Developmental Science 12: 426-437. quorum-sensing bacterial populations. Nature 450:411-U417. Cowden, C. C., and C. J. Peterson. 2009. ‘A multi- mutualist simulation: Applying biological market Doebeli, M., and N. Knowlton. 1998. The models to diverse mycorrhizal communities’, evolution of interspecific mutualisms. Proceedings Ecological Modelling 220:1522-1533. of the National Academy of Sciences of the United States of America 95:8676-8680. Crozier, R. H. 1986. ‘Genetic clonal recognition abilities in marine invertebrates must be Duckworth, R. A. and Badyaev, A.V. 2007. maintained by selection for something else’, Coupling of dispersal and aggression facilitates Evolution 40: 1100-1101. the rapid range expansion of a passerine bird. Proceedings of the National Academy of Sciences of Darwin, C. R. 1859. On the origin of species by the United States of America 104: 15017-15022. means of natural selection, or, the preservation of favoured races in the struggle for life. John Murray, Ehlinger, T. J. 1989. ‘Learning and individual London, UK. variation in bluegill foraging: Habitat specific technique’, Animal Behaviour 38: 643-658. Darwin, C. R. 1871. The descent of man and selection in relation to sex. London, John Murray. El Mouden, C, Burton-Chellew M., Gardner A,. West, S., A. 2011. ‘What do Humans Maximise?’ Davies, N.B., Hatchwell, B.J., Robson, T. and in, Okasha, S. and Binmore, K. eds., Evolution and Burke, T. 1992. ‘Paternity and parental effort in Rationality: Decisions, Cooperation and Strategic dunnocks Prunella modularis: how good are male Behaviour, Oxford University Press. In press. chick-feeding rules?’, Animal Behaviour 43: 729- 745. El Mouden, C., West, S.A. and Gardner, A. 2010. ‘The enforcement of cooperation by policing’, Davies, N.B., Krebs, J.R. and West, S.A. 2011. An Evolution 64: 2139-2152. Introduction to Behavioural Ecology, Fourth Edition. Wiley-Blackwell. Ernest-Jones, M., Nettle, D. and Bateson, M. 2011. ‘Effects of eye images on everyday cooperative Dawes CT, Fowler JH, Johnson T, McElreath R, behaviour: a field experiment’, Evolution and Smirnov O. 2007. ‘Egalitarian Motives in Human Behaviour 32: 172-178. Humans’, Nature 446: 794-796

190 Falconer, D.S. 1981. Introduction to quantitative the Royal Society of London Series B-Biological genetics (2nd ed.). Longman, New York. Sciences 263:339-344.

Fehr, E and Gachter, S. 2002. ‘Altruistic Frank S. 1998 Foundations of Social Evolution, punishment in humans’, Nature 415: 137-140. Princeton University Press.

Fehr, E. and Fischbacher, U. 2003. ‘The nature of Frank, S. A. 2003. ‘Repression of competition and human altruism’, Nature 425: 785-791. the evolution of cooperation’, Evolution 57, 693– 705. Feldman, M. W., Cavalli-Sforza, L. L., and Peck, J. R. 1985. Gene-culture coevolution: models for the Frank, S. A. 2009. Evolutionary foundations of evolution of altruism with cultural transmission. cooperation and group cohesion. In Games, Groups, Proceedings of the National Academy of Sciences of and the Global Good S. A. Levin, ed. Springer- the United States of America, 82, 5814–5818. Verlag.

Fisher, R. A. 1930 The genetical theory of natural Frank, S. A. 2010. Demography and the tragedy selection. Oxford, UK: Clarendon. of the commons. Journal of Evolutionary Biology 23:32-39. Fiske, A. P. 1999. ‘Learning culture the way informants do: Observation, imitation, and Fried, M. H. 1967. ‘The evolution of political society: participation’ Unpublished manuscript An essay in political anthropology’. New York: Random House. Fitter, A. H. 2006. ‘What is the link between carbon and phosphorus fluxes in arbuscular Friesen, M. L., and Mathias. A. 2010. ‘Mixed mycorrhizas? A null hypothesis for symbiotic infections may promote diversification of function’, New Phytologist 172:3-6. mutualistic symbionts: why are there ineffective rhizobia?’, Journal of Evolutionary biology 23:323- Foster, W.A. 2002. ‘Soldier aphids go cuckoo’, 334. Trends in Ecology and Evolution 17: 199-200. Funder, D. C. 1987. ‘Errors and mistakes. Foster, K. R., and T. Wenseleers. 2006. ‘A general Evaluating the accuracy of social judgement’, model for the evolution of mutualisms’, Journal of Psychological Bulletin 101: 75-90. Evolutionary biology 19:1283-1293. Gandon S., Rousset F. 1999. ‘Evolution of Frank, S. A. 1994. ‘Genetics of mutualism - the stepping-stone dispersal rates’, Proceedings of the evolution of altruism between species’, Journal of Royal Society of London Series B-Biological Sciences Theoretical Biology 170:393-400. 266(1437), 2507-2513. Frank, S. A. 1995. ‘Mutual policing and Gangestad, S.W., Thornhill, R., Garver-Apgar, repression of competition in the evolution of C.E., 2005. ‘Adaptations to ovulation’, in: Buss, cooperative groups’, Nature 377:520–522. D.M. (ed.), The handbook of evolutionary psychology. Frank, S. A. 1996a. Models of parasite virulence. John Wiley and Sons, Hoboken, NJ. Quarterly Review of Biology. 71:37-78. Gangestad, S.W., Thornhill, R., Garver, C., 2002. Frank, S. A. 1996b. Policing and group cohesion ‘Changes in women's sexual interests and their when resources vary. Animal Behaviour 52:1163- partners' mate retention tactics across the 1169. menstrual cycle: evidence for shifting conflicts of Frank, S. A. 1996c. Host-symbiont conflict over interest’, Proceedings of the Royal Society of London the mixing of symbiotic lineages. Proceedings of Series B-Biological Sciences 269: 975–982.

191 Gardner A 2009. ‘Adaptation as organism design’ Gardner A., West S.A., Wild G. 2011. ‘The Biology Letters 5: 861-864 genetical theory of kin selection’, Journal of Evolutionary Biology 24(5), 1020-1043. Gardner, A. 2010. ‘Sex-biased dispersal of adults mediates the evolution of altruism among Gigerenzer, G and Gaissmaier, W. 2011. juveniles’, Journal of Theoretical Biology 262: 339- ‘Heuristic decision making’, Annual Review of 345. Psychology 62: 451–8

Gardner A. 2011. Kin selection under blending Gilovich, T., Griffin, D. and Kahneman, D. 2002. inheritance. Journal of Theoretical Biology 284, 125- Heuristics and biases: The psychology of intuitive 129 judgment. Cambridge University Press.

Gardner A., West S. A. and Buckling A. 2004. Gintis, H. 2000. Strong reciprocity and human Bacteriocins, spite and virulence. Proceedings of sociality. Journal of Theoretical Biology, 206, 169– the Royal Society of London Series B-Biological 179. Sciences 271: 1529-1535. Gintis, H., Bowles, S., Boyd, R., and Fehr, E. Gardner, A., and S. A. West. 2004a. ‘Cooperation 2005. Moral sentiments and material interests – the and punishment, especially in humans’, American foundations of cooperation in economic life. MIT Naturalist 164:753-764. Press.

Gardner, A., and S. A. West. 2004b. ‘Spite and the Goto, R., T. Okamoto, E. T. Kiers, A. Kawakita, scale of competition’, Journal of Evolutionary and M. Kato. 2010. ‘Selective flower abortion Biology 17:1195-1203. maintains moth cooperation in a newly discovered pollination mutualism’, Ecology Letters Gardner, A and West, S. A. 2006. ‘Demography, 13:321-329. altruism, and the benefits of budding’, Journal of Evolutionary Biology 19: 1707-1716. Grafen, A. 1985. ‘A geometric view of relatedness’. Oxford Surveys in Evolutionary Gardner A., West S., Barton N. 2007a. ‘The Biology, 2, 28-90. relation between multilocus population genetics and social evolution theory’, The American Grafen, A. 1998. ‘Fertility and labour supply Naturalist 169(2), 207-226. in Femina economica’, Journal of Theoretical Biology, 194, 429-455. Gardner, A., Hardy I. C. W., Taylor, P. D. and West, S. A. 2007b. ‘Spiteful soldiers and sex ratio Grafen, A. 1999. ‘Formal Darwinism, the conflict in polyembryonic parasitoid wasps’,. The individual-as-maximising-agent analogy, and American Naturalist 169: 519-533. bet-hedging’. Proceedings of the Royal Society of London Series B-Biological Sciences, 266, 799-803. Gardner, A., A. Arce and J. Alpedrinha. 2009. ‘Budding dispersal and the sex ratio’, Journal of Grafen, A. 2002. ‘A first formal link between the Evolutionary Biology 22:1036-1045. Price Equation and an Optimization Program’, Journal of Theoretical Biology 217: 75-91 Gardner, A., and A. Grafen. 2009. ‘Capturing the superorganism: a formal theory of group Grafen A, 2006a. ‘Optimization of inclusive adaptation’, Journal of Evolutionary Biology 22:659- fitness’, Journal of Theoretical Biology 238, 541-563 671.

192 Grafen, A. 2006b. ‘A theory of Fisher’s Orzack, S. H. and Sober, E. (eds.) Adaptationism reproductive value’. Journal of Mathematical and Optimality. Cambridge University Press. Biology 53, 15-60 Haldane, J. B. S. 1932. The causes of evolution. Grafen A. 2007 ‘Detecting kin selection at work Longmans. New York. using inclusive fitness’. Proceedings of the Royal Haley, K. J. and Fessler, D. M. T. 2005. ‘Nobody's Society of London Series B-Biological Sciences watching? Subtle cues affect generosity in an 274(1610), 713-719. anonymous economic game’, Evolution and Grafen, A. 2007. An Inclusive Fitness analysis of Human Behavior 26: 245–256 altruism on a cyclical network. Journal of Hamilton, W.D. 1963. ‘The evolution of altruistic Evolutionary Biology. 20: 2278-2283. behaviour’. The American Naturalist 97: 354–356. Grafen, A. 2007. ‘The formal Darwinism project: a Hamilton, W. D. 1964. The genetical evolution of mid-term report’, Journal of Evolutionary Biology social behaviour, I and II. Journal of theoretical 20: 1243–1254. Biology 7:1-52. Grafen, A, and M. Archetti. 2008. ‘Natural Hamilton, W. D. 1966. The moulding of selection of altruism in inelastic viscous senescence by natural selection. Journal of homogeneous population’, Journal of Theoretical theoretical Biology 12:12-45. Biology, 252, 694-710. Hamilton, W. D. 1970. ‘Selfish and spiteful Gratten, J., Beraldi, D., Lowder, B. V., McRae, A. behaviour in an evolutionary model’, Nature 228: F., Visscher, P. M., Pemberton, J. M. and Slate, J. 1218–1220. 2007. ‘Compelling evidence that a single nucleotide substitution in TYRP1 is responsible Hamilton, W. D. 1971. ‘Selection of selfish and for coat-colour polymorphism in a free-living altruistic behaviour in some extreme models’, in population of Soay sheep’, Proceedings of the Royal Eisenberg, J. F. and Dillon, W. S. (eds.) Man and Society of London Series B-Biological Sciences Beast: Comparative Social Behavior Washington, 274:619-626 DC, Smithsonian Press.

Griffin A.S., West S.A. 2003 Kin discrimination Hamilton, W. D. 1996. Narrow roads of gene land: I and the benefit of helping in cooperatively Evolution of social behaviour. Oxford, W.H. breeding vertebrates. Science 302(5645), 634-636. Freeman.

Griffin, A.S., West, S.A., and Buckling, A. 2004. Hamilton W.D., May R.M. 1977. ‘Dispersal in ‘Cooperation and competition in pathogenic stable habitats’, Nature 269(5629), 578-581. bacteria’, Nature 430: 1024-1027. Harbaugh, W. T., Mayr, U. and Burghart, D. R.

Hagen, E. H. and Hammerstein, P. 2006. ‘Game 2007. ‘Neural responses to taxation and voluntary theory and human evolution: A critique of some giving reveal motives for charitable donations’, recent interpretations of experimental games’, Science 316: 1622-1625 Theoretical Population Biology 63: 339-348 Hardin, G. 1968. ‘The tragedy of the commons’,

Halama, K. J. and Reznick, D. N. 2001. Science 162: 1243-1248. ‘Adaptation, optimality, and the meaning of Harrison, M. J., G. R. Dewbre, and Liu, J. Y. 2002. phenotypic variation in natural populations’, in, ‘A phosphate transporter from Medicago truncatula involved in the acquisiton of

193 phosphate released by arbuscular mycorrhizal Henrich, J. 2008. ‘A cultural species’ in, Brown, fungi’, Plant Cell 14:2413-2429. M., Explaining culture scientifically. University of Washington Press Hart, M. M., and Reader, R. J. 2005. The role of the external mycelium in early colonization for Henrich, J., and R. Boyd. 1998. The evolution of three arbuscular mycorrhizal fungal species with conformist transmission and the emergence of different colonization strategies. Pedobiologia between-group differences. Evolution and Human 49:269-279. behavior 19:215-241.

Haselton, M. G., and Buss, D. 2000. ‘Error Henrich, J., and F. J. Gil-White. 2001. ‘The management theory: a new perspective on Biases evolution of prestige freely conferred deference in cross-sex mind reading’, Journal of Personality as a mechanism for enhancing the benefits of and Social Psychology. 78: 81-91 cultural transmission’, Evolution and Human behavior 22:165-196. Haselton, M. G., Mortezaie, M., Pillsworth, E. G., Bleske-Recheck, A. E., and Frederick, D. A. 2007. Henrich, J. and McElreath, R. 2003. ‘The ‘Ovulation and human female ornamentation: evolution of cultural evolution’, Evolutionary Near ovulation, women dress to impress’, Anthropology 12(3): 123-135. Hormones and Behavior 51: 40-45 Henrich, J. and Henrich, N. 2007. Why Humans Haselton, M.G., Gangestad, S.W., 2006. Cooperate. Oxford University Press. ‘Conditional expression of women's desires and Herre, E. A. 1985. ‘Sex ratio adjustment in fig men's mate guarding across the ovulatory cycle’, wasps’, Science 228: 896-898. Hormones and Behaviour 49: 509–51 Herre, E. A. 1987. ‘Optimality, plasticity and Haselton, M.G., Nettle, D. and Andrews, P.W. selective regime in fig wasp sex ratios’, Nature 2005. ‘The evolution of cognitive bias’, in, D.M. 329: 627–629. Buss (ed.), Handbook of evolutionary psychology. Hoboken: Wiley. Herre, E. A., L. C. Mejia, D. A. Kyllo, E. Rojas, Z. Maynard, A. Butler, and S. A. Van Bael. 2007. Hatchwell B.J. 2009. ‘The evolution of cooperative ‘Ecological implications of anti-pathogen effects breeding in birds: kinship, dispersal and life of tropical fungal endophytes and mycorrhizae’, history’. Philosopical Transactions of the Royal Ecology 88:550-558. Socoety B-Biological Sciences 364(1533), 3217-3227. Herre, E. A., N. Knowlton, U. G. Mueller, and S. Heinsohn R., Legge S. 1999. ‘The cost of helping’, A. Rehner. 1999. The evolution of mutualisms: Trends in Ecology and Evolution 14(2), 53-57. exploring the paths between conflict and Helgason, T., and A. H. Fitter. 2009. Natural cooperation. Trends in Ecology and Evolution 14:49- selection and the evolutionary ecology of the 53. arbuscular mycorrhizal fungi (Phylum Heyes, C. M. and Galef, B. G. 1996. Social learning Glomeromycota). Journal of Experimental Botany in animals: the roots of culture. San Diego 60:2465-2480. Academic Press. Henrich, J. 2005. ‘Cultural evolution of human Hodge, A., T. Helgason, and A. H. Fitter. 2010. cooperation’ in, Boyd, R and Richerson, P. (eds). Nutritional ecology of arbuscular mycorrhizal The Origin and Evolution of Cultures. Oxford fungi. Fungal Ecology 3:267-273. University Press.

194 Hoeksema, J. D., and M. Kummel. 2003. Jansa, J., A. Mozafar, T. Anken, R. Ruh, I. R. Ecological persistence of the plant-mycorrhizal Sanders, and Frossard, E. 2002. ‘Diversity and mutualism: A hypothesis from species structure of AMF communities as affected by coexistence theory. American Naturalist 162:S40- tillage in a temperate soil’, Mycorrhiza 12:225-234. S50. Jansa, J., A. Mozafar, G. Kuhn, T. Anken, R. Ruh, Hoeksema, J. D., V. B. Chaudhary, C. A. Gehring, I. R. Sanders, and Frossard, E. 2003. ‘Soil tillage N. C. Johnson, J. Karst, R. T. Koide, A. Pringle et affects the community structure of mycorrhizal al. 2010. A meta-analysis of context-dependency fungi in maize roots’, Ecological Applications in plant response to inoculation with mycorrhizal 13:1164-1176. fungi. Ecology Letters 13:394-407. Javot, H., R. V. Penmetsa, N. Terzaghi, D. R. Hollis K. L., Pharr V. L., Dumas M. J., Britton G. Cook, and M. J. Harrison. 2007. ‘A Medicago B., Field J. 1997. ‘Classical conditioning provides truncatula phosphate transporter indispensable paternity advantage for territorial male blue for the arbuscular mycorrhizal symbiosis’, gouramis (Trichogaster trichopterus)’, Journal of Proceedings of the National Academy of Sciences of Comparative Psychology 111: 219–225. the United States of America 104:1720-1725.

Hughes W.O.H., Oldroyd B.P., Beekman M., Jansa J., Smith, F. A. and Smith, S. E. 2008. ‘Are Ratnieks F.L.W. 2008. ‘Ancestral monogamy there benefits of simultaneous root colonization shows kin selection is key to the evolution of by different arbuscular mycorrhizal fungi?’, New eusociality’, Science 320(5880), 1213-1216. Phytologist 177:779-789

Irwin, A. J. and Taylor, P. D. 2001. Evolution of Johnson, N. C., D. L. Rowland, L. Corkidi, L. M. altruism in a stepping-stone population with Egerton-Warburton, and E. B. Allen. 2003. overlapping generations. Theoretical Population ‘Nitrogen enrichment alters mycorrhizal Biology 60: 315-325. allocation at five mesic to semiarid grasslands’, Ecology 84:1895-1908. Jablonski, N. G. and Chaplin, C. 2000. ‘The evolution of human skin coloration’, Journal of Johnson, N. C., G. W. T. Wilson, M. A. Bowker, J. Human Evolution 39: 57–106. A. Wilson, and R. M. Miller. 2010. ‘Resource limitation is a driver of local adaptation in Jander, K. C., and E. A. Herre. 2010. ‘Host mycorrhizal symbioses’, Proceedings of the sanctions and pollinator cheating in the fig tree- National Academy of Sciences of the United States of fig wasp mutualism’, Proceedings of the Royal America 107:2093-2098. Society of London B-Biological Sciences 277:1481- 1488. Johnstone, R.A. and Cant, M.A. 2008. ‘Sex- differences in dispersal and the evolution of Janouskova, M., P. Seddas, L. Mrnka, D. van helping and harming’, American Naturalist. Tuinen, A. Dvorackova, M. Tollot, V. Gianinazzi- 172(3):318-30 Pearson, M. Vosatka, and Gollotte, A. 2009. ‘Development and activity of Glomus intraradices Kahneman, D.; Tversky, A. 1972. ‘Subjective as affected by co-existence with Glomus probability: A judgment of claroideum in one root system’, Mycorrhiza 19:393- representativeness’, Cognitive Psychology 3: 430– 402. 454

195 Kiers, E. T., S. A. West, and R. F. Denison. 2002. breeding through group augmentation’, ‘Mediating mutualisms: farm management Proceedings of the Royal Society of London B- practices and evolutionary changes in symbiont Biological Sciences. 268: 187-196. co-operation’, Journal of Applied Ecology 39:745- Kummel, M., and S. W. Salant. 2006. ‘The 754. economics of mutualisms: optimal utilization of Kiers, E. T., R. A. Rousseau, S. A. West, and R. F. mycorrhizal mutualistic partners by plants’, Denison. 2003. ‘Host sanctions and the legume- Ecology 87:892-902. rhizobium mutualism’. Nature 425:78-81. Kümmerli, R., A. Gardner, S. A. West, and A. S. Kiers, E. T. and M. G. A. van der Heijden. 2006a. Griffin. 2009. ‘Limited dispersal, budding ‘Mutualistic stability in the arbuscular dispersal and cooperation: an experimental mycorrhizal symbiosis: exploring hypotheses of study’, Evolution 63:939-949. evolutionary cooperation’. Ecology 87:1627-1636. Kümmerli, R., Burton-Chellew, M.N., Ross- Kiers, E. T., R. A. Rousseau, and R. F. Denison. Gillespie, A. and West, S.A. 2010. ‘Resistance to 2006b. ‘Measured sanctions: legume hosts detect extreme strategies, rather than prosocial quantitative variation in rhizobium cooperation preferences, can explain human cooperation in and punish accordingly’, Evolutionary Ecology public goods games’, Proceedings of the National Research 8:1077-1086. Academy of Sciences of the United States of America 107: 10125-10130 Kiers, E. T., and R. F. Denison. 2008. ‘Sanctions, cooperation, and the stability of plant- Lasseka, W. D. and Gaulin, S. J. C. 2008. ‘Waist- rhizosphere mutualisms’, Annual Review of hip ratio and cognitive ability: is gluteofemoral Ecology Evolution and Systematics 39:215-236. fat a privileged store of neurodevelopmental resources?’, Evolution and Human Behavior 29: 26– Kiers, E. T., M. Duhamel, Y. Beesetty, J. A. 34. Mensah, O. Franken, E. Verbruggen, C. R. Felbaum et al. 2011. ‘Reciprocal rewards stabilize Law, R. 1979. ‘Optimal Life Histories under age- cooperation in the mycorrhizal symbiosis’, specific predation’, American Naturalist 114: 399- Science. 333(6044): 880-882 417.

Knauft, B. 1991. Violence and sociality in human Ledyard, John O., 1995. ‘Public Goods: A Survey evolution. Current Anthropology 32:391-428. of Experimental Research’, in: Kagel, John H., Roth, Alvin E. (eds.), The Handbook of Experimental Kobae, Y., and S. Hata. 2010. ‘Dynamics of Economics, Princeton University Press. periarbuscular membranes visualized with a fluorescent phosphate transporter in arbuscular Lehmann, L. 2003. The good, the bad and the choosy. mycorrhizal roots of rice’, Plant and Cell Thèse de doctorat, Faculté de Sciences, de Physiology 51:341-353. l’Université de Lausanne.

Koch, A. M., D. Croll, and I. R. Sanders. 2006. Lehmann, L. 2007. The evolution of trans- ‘Genetic variability in a population of arbuscular generational altruism: kin selection meets niche mycorrhizal fungi causes variation in plant construction. Journal of Evolutionary Biology 20: growth’, Ecology Letters 9:103-110. 181-189.

Kokko, H., Johnstone, R. A., and Clutton-Brock, T. H. 2001. ‘The evolution of cooperative

196 Lehmann, L., N. Perrin, and F. Rousset. 2006. Leigh, E. G. 1971. Adaptation and diversity. Population demography and the evolution of Cooper, San Francisco. helping behaviours. Evolution 60:1137-1151. Leigh, E G. 1977. How does selection reconcile Lehmann, L, and L Keller. 2006. The evolution of individual advantage with the good of the cooperation and altruism. A general framework group? Proceedings of the National Academy of and classification of models. Journal of Sciences of the United States of America 74:4542- Evolutionary Biology 19:1365-1378. 4546.

Lehmann, L., Keller, L., and Sumpter, D. 2007a. Lieberman, D., J. Tooby, and L. Cosmides. 2003. The evolution of helping and harming on graphs: ‘Does morality have a biological basis? An the return of the inclusive fitness effect. Journal of empirical test of the factors governing moral Evolutionary Biology 20: 2284-2295. sentiments relating to incest’, Proceedings of the Royal Society B Biological Sciences 270: 819-826. Lehmann, L., L. Keller, S. A. West, and D. Roze. 2007b. ‘Group selection and kin selection. Two Lion, S. and van Baalen, M. 2008. ‘Self-structuring concepts but one process’, Proceedings of the in spatial evolutionary ecology’, Ecology Letters National Academy of Sciences of the United States of 11:277-295. America 104:6736-6739. Lively, C. 2009. ‘Local host competition in the Lehmann, L., Rousset, F., Roze, D., and Keller, L. evolution of virulence’, Journal of Evolutionary 2007c. ‘Strong-reciprocity or strong ferocity? A Biology 22(6):1268-74 population genetic view of the evolution of Lynch, M., and J. B. Walsh. 1998. Genetics and altruistic punishment’, American Naturalist, 170: Analysis of Quantitative Traits. 980 pp. Sinauer 21-36. Assocs., Inc., Sunderland, MA. Lehmann L., Feldman M.W., Foster K.R., Maherali, H., and J. N. Klironomos. 2007. 2008. ‘Cultural transmission can inhibit the ‘Influence of phylogeny on fungal community evolution of altruistic helping’, American assembly and ecosystem functioning’, Science Naturalist 172(1), 12-24. 316:1746-1748. Lehmann, L., and M. W. Feldman. 2008a. The co- Mahometa M. J., Domjan M. 2005. ‘Classical evolution of culturally inherited altruistic helping conditioning increases reproductive success in and cultural transmission under random group Japanese quail Coturnix japonica’, Animal formation. Theoretical Population Biology 73:506- Behaviour. 69: 983–989 516. Mangan, S. A., and G. H. Adler. 2002. ‘Seasonal Lehmann L., Feldman M.W., 2008b. ‘War and the dispersal of arbuscular mycorrhizal fungi by evolution of belligerence and bravery’, spiny rats in a neotropical forest’, Oecologia Proceedings of the Royal Society B Biological 131:587-597. Sciences 275(1653), 2877-2885. Marshall, A. 1920. Principles of Economics. An Lehmann, L., and F. Rousset. 2010. ‘How life Introductory Volume. 8th edition. London: history and demography promote or inhibit the Macmillan. evolution of helping behaviours’, Philosophical Transactions of the Royal Society B-Biological Martin, S. J., Beekman, M., Wossler, T. C. and Sciences 365:2599-2617. Ratnieks, F. L. W. 2002. ‘Parasitic worker honey

197 bees, Apis mellifera capensis, evade policing’, National Academy of Science, USA 103: 15623- Nature 415, 163–165. 15628.

Maynard Smith, J. 1964. ‘Group selection and kin Møller, A. P. and Jennions, M. D. 2002. ‘How selection’, Nature 201:1145-1147. much variance can be explained by ecologists and evolutionary biologists?’, Oecologia 132: 492- Maynard Smith, J. 1982. Evolution and the Theory 500. of Games. Cambridge University Press. Mummey, D. L., and M. C. Rillig. 2008. Spatial Maynard Smith, J. and Price, G.R. 1973. ‘The logic characterization of arbuscular mycorrhizal fungal of animal conflict’, Nature 246: 15-18. molecular diversity at the submetre scale in a Maynard Smith, J. and Szathmary, E. 1995. The temperate grassland. FEMS Microbiology Ecology Major Transitions in Evolution. Freeman, Oxford. 64:260-270.

McGonigle, T. P., M. H. Miller, D. G. Evans, G. L. Murray, M.G. 1985. ‘Figs (Ficus spp.) and fig Fairchild, and J. A. Swan. 1990. ‘A new method wasps (Chalcidoidea, Agaonidae): hypotheses for an which gives an objective measure of colonization ancient symbiosis’, Biological Journal of the Linnean of roots by vesicular arbuscular mycorrhizal Society 26, 69-81. fungi’, New Phytologist 115:495-501. Nonacs P. 2011 ‘Monogamy and high relatedness McGrew, W.C. 2004. The cultured Chimpanzee: do not preferentially favor the evolution of reflections on cultural primatology. Cambridge cooperation’, BMC Evolutionary Biology 11, 9. University Press. Oehl, F., E. Sieverding, K. Ineichen, E. A. Ris, T. McKay, R. and Efferson, C. 2010. ‘The subtleties Boller, and A. Wiemken. 2005. ‘Community of error management’, Evolution and Human structure of arbuscular mycorrhizal fungi at Behavior 31: 309-319. different soil depths in extensively and Miller, G. F., Tybur, J., and Jordan, B. 2007. intensively managed agroecosystems’, New ‘Ovulatory cycle effects on tip earnings by lap- Phytologist 165:273-283. dancers: Economic evidence for human Olsson, S. B., Barnard, J. and Turri, L. 2006. estrus?’, Evolution and Human Behavior 28: 375- ‘Olfaction and identification of unrelated 381. individuals: examination of the mysteries of Mitteldorf, J. and Wilson, D. S. 2000. ‘Population human odor recognition’, Journal of Chemical viscosity and the evolution of altruism’, Journal of Ecology. 32: 1635-1645.

Theoretical Biology 204:481-496. Oono, R., C. G. Anderson, and R. F. Denison. Moawad, M., and E. L. Schmidt. 1987. Occurrence 2011. ‘Failure to fix nitrogen by non-reproductive and nature of mixed infections in nodules of symbiotic rhizobia triggers host sanctions that field-grown soybeans (Glycine max). Biology and reduce fitness of their reproductive clonemates’, Fertility of Soils 5:112-114. Proceedings of the Royal Society of London B: Biological Sciences 278(1718):2698-2703 Moll, J., Krueger, F., Zahn, R., Pardini, M., de Oliviera-Souza, R. and Grafman, J. 2006 ‘Human Öpik, M., M. Metsis, T. J. Daniell, M. Zobel, and fronto–mesolimbic networks guide decisions M. Moora. 2009. ‘Large-scale parallel 454 about charitable donation’, Proceedings of the sequencing reveals host ecological group specificity of arbuscular mycorrhizal fungi in a

198 boreonemoral forest’, New Phytologist 184:424- Perrin, N and Lehmann, L. 2001. ‘Is sociality 437. driven by the costs of dispersal or the benefits of philopatry? A role for kin-discrimination Ostrom, E. 1990. ‘Governing the Commons: The mechanisms’, The American Naturalist 158: 471-48. Evolution of Institutions for Collective Action.’ New York: Cambridge University Press. Perrins, C. M. 1979. British Tits. HarperCollins

Otto, S. P. and T. Day. 2007. ‘A Biologist's Guide to Pettit, P. 2005. ‘The rise of modern humans’ in Mathematical Modeling in Ecology and Evolution.’ Scarre, C. (ed). The Human Past. Thames and Princeton University Press Hudson.

Paley, W. 1802 Natural theology. London, UK: Platt, T. G., and Bever, J. D. 2009. ‘Kin Wilks and Taylor. competition and the evolution of cooperation’, Trends in Ecology and Evolution 24:370-377. Palmer, T. M., D. F. Doak, M. L. Stanton, J. L. Bronstein, E. T. Kiers, T. P. Young, J. R. Goheen, Pohl, R. 2004. Cognitive Illusions: A Handbook on and R. M. Pringle. 2010. ‘Synergy of multiple Fallacies and Biases in Thinking, Judgement and partners, including freeloaders, increases host Memory, Hove, New York. fitness in a multispecies mutualism’, Proceedings Porter, R.H., and Cernoch, J.M. 1983. ‘Maternal of the National Academy of Sciences of the United recognition of neonates through olfactory cues’, States of America 107:17234-17239. Physiology and Behavior, 30: 151-154. Parker, G.A. and Maynard Smith, J. 1990. Powell, J. R., J. L. Parrent, M. M. Hart, J. N. ‘Optimality theory in evolutionary biology’, Klironomos, M. C. Rillig, and Maherali, H. 2009. Nature 348: 27-33. ‘Phylogenetic trait conservatism and the Parniske, M. 2008. ‘Arbuscular mycorrhiza: the evolution of functional trade-offs in arbuscular mother of plant root endosymbioses’, Nature mycorrhizal fungi’, Proceedings of the Royal Society Reviews Microbiology 6:763-775. B-Biological Sciences 276:4237-4245.

Pearce, D. 1997. Animal learning and cognition: an Price, G.R. 1970. Selection and covariance. Nature introduction. 2nd Edition. Psychology Press. 227: 520-521.

Pen I., Weissing F.J. 2000. ‘Sexual Selection and Quartz, S. and Sejnowski, T. 2000. ‘Constraining the Sex Ratio: An ESS Analysis’, Selection 1, 59-69. constructivism: cortical and sub-cortical constraints on learning in development’, Penton-Voak, I. S. and Perrett, D. I. 2000. ‘Female Behavioral and Brain Sciences 23(5): 785-792 preference for male faces changes cyclically: further evidence’, Evolution and Human Behavior Quartz, S. R. 2002. Liars, lovers and heroes. New 21: 39-48 York, William Morrow.

Penton-Voak, I. S., Perrett, D. I., Castles D. L., Queller D.C. 1992. ‘Quantitative Genetics, Kobayashi, T., Burt, D. M., Murray, L. K. and Inclusive Fitness, and Group Selection’, The Minamisawa, R. 1999. ‘Menstrual cycle alters face American Naturalist 139(3), 540-558. preference’, Nature 399: 741-742 Queller, D. C. 1992. Does population viscosity Pepper, J. W. 2000. ‘Relatedness in Trait Group promote kin selection? Trends in Ecology and Models of Social Evolution’, Journal of Theoretical Evolution 7:322-324. Biology 206(3):355-368

199 Queller, D. C. 1994. Genetic relatedness in Rilling, J. K., Gutman, D. A., Zeh, T. viscous populations. Evolutionary Ecology 8:70-73. R., Pagnoni, G., Berns, G. S. and Kilts, D. S. 2002. ‘A neural basis for social cooperation’, Queller, D. C. 2000. Relatedness and the fraternal Neuron 35: 395-402. major transitions. Philosophical Transactions of the Royal Society B-Biological Sciences 355:1647-1655. Roberts, S. D., Havlicek, J., Flegr, J., Hruskova, M., Little, A. C., Jones, B. C., Perrett, D. I. and Quervain, D. J. F. de, Fischbacher, U., Treyer, V., Petrie, M. 2004. ‘Female facial attractiveness Schellhammer, M., Schnyder, U., Buck, A and increases during the fertile phase of the Fehr, E. 2004. ‘The neural basis of altruistic menstrual cycle’, Proceedings of the Royal Society of punishment’, Science 305: 1254–1258. London B-Biological Sciences. 271: S270-S272 Rankin, D.J. and López-Sepulcre, A. 2005. ‘Can Ronce O. 2007. ‘How does it feel to be like a adaptation lead to extinction?’, Oikos 111: 616- rolling stone? Ten questions about dispersal 619. evolution’, Annual Review of Ecology, Evolution and Rankin, D.J., Bargum, K. & Kokko, H. 2007. ‘The Systematics 38, 231-253. tragedy of the commons in evolutionary Ronce O., Gandon S., Rousset F. 2000. ‘Kin biology’. Trends in Ecology and Evolution 22: 643- selection and natal dispersal in an age-structured 651. population’, Theoretical Population Biology 58(2), Ratnieks, F. L. W. 1988. ‘Reproductive harmony 143-159. via mutual policing by workers in eusocial Rousset F. 2004 Genetic Structure and Selection in Hymenoptera’, American Naturalist 132:217-236. Subdivided Populations. Princeton, NJ, Princeton Ratnieks, F. L. W. H. and H. Helanterä. 2009. ‘The University Press. evolution of extreme altruism and inequality in Rousset F., Billiard S. 2000. ‘A theoretical basis for insect societies’, Proceedings of the Royal Society of measures of kin selection in subdivided London B-Biological Sciences. 364(1533):3169-79. populations: finite populations and localized Ratnieks, F. L. W., and P. K. Visscher. 1989. dispersal’, Journal of Evolutionary Biology 13(5), ‘Worker policing in the honeybee’, Nature 814-825. 342:796-797. Rousset F. and Gandon S. 2002. ‘Evolution of the Ratnieks, F. L. W., K. R. Foster, and T. distribution of dispersal distance under distance- Wenseleers. 2006. ‘Conflict resolution in insect dependent cost of dispersal’. Journal of societies’, Annual Review of Entomology 51:581-608. Evolutionary Biology 15(4), 515-523.

Ratnieks, F. L. W., and T. Wenseleers. 2008. Rousset F., Ronce O. 2004. ‘Inclusive fitness for ‘Altruism in Insect societies and beyond: traits affecting metapopulation demography’, voluntary or enforced?’, Trends in Ecology and Theoretical Population Biology 65(2), 127-141. Evolution 23:45-52. Rousset, F. and Roze, D. 2007. ‘Constraints on the Richerson, P. J. and Boyd R. 2004. Not by Genes origin and maintenance of genetic kin Alone. University of Chicago Press. recognition’, Evolution 61(10): 2320-2330.

Rigdon, M., Ishii, K., Watabe, M. and Kitayama, Russel, M. J. 1976. ‘Human olfactory S. 2009. ‘Minimal social cues in the dictator communication’, Nature 260: 520-522. game’, Journal of Economic Psychology. 30: 358-367.

200 Russel, M. J., Mendelson, T., and Peeke, H.V.S. Simms, E. L., D. L. Taylor, J. Povich, R. P. 1983. ‘Mothers’ identification of their infant's Shefferson, J. L. Sachs, M. Urbina, and Y. odor’, Ethology and Sociobiology 4: 29-31. Tausczik. 2006. An empirical test of partner choice mechanisms in a wild legume-rhizobium Russell A.F., Langmore N.E., Cockburn A., interaction. Proceedings of the Royal Society B- Astheimer L.B., Kilner R.M. 2007. ‘Reduced egg Biological Sciences 273:77-81. investment can conceal helper effects in cooperatively breeding birds’, Science 317(5840), Simões-Lopes, P. C., Fabián, M. E. and 941-944. Menegheti, J. O. 1998. ‘Dolphin interactions with the mullet artisanal fishing on Southern Brazil: a Sachs, J. L., Mueller, U. G., Wilcox, T. P., and Bull, qualitative and quantitative approach’, Revista J. J. 2004. ‘The evolution of cooperation’, The Brasileira de Zoologia. 15(3): 709-726. Quarterly Review of Biology 79: 135–160. Simon, H. A. 1990. ‘A mechanism for social Sachs, J. L., M. O. Ehinger, and E. L. Simms. 2010. selection and successful altruism’, Science 250: ‘Origins of cheating and loss of symbiosis in wild 1665-1668. Bradyrhizobium’,. Journal of Evolutionary biology 23:1075-1089. Singh, D. 1993. ‘Adaptive significance of female physical attractiveness: Role of waist–hip ratio’, Sambrook J, Fritsch EF, Maniatis T. 1989. Journal of Personality and Social Psychology 65: 293- Molecular Cloning, A Laboratory Manual. 2nd 307. edition. Cold Spring Harbor, Laboratory Press, New York. Smith, F. A., I. Jakobsen, and S. E. Smith. 2000. ‘Spatial differences in acquisition of soil Scheublin, T. R., R. S. P. van Logtestijn, and M. G. phosphate between two arbuscular mycorrhizal A. van der Heijden. 2007. Presence and identity fungi in symbiosis with Medicago truncatula’, New of arbuscular mycorrhizal fungi influence Phytologist 147:357-366. competitive interactions between plant species. Journal of Ecology 95:631-638. Smith, F. A., E. J. Grace, and S. E. Smith. 2009. ‘More than a carbon economy: nutrient trade and Schwarzott, D., C. Walker, and A. Schüßler. 2001. ecological sustainability in facultative arbuscular Glomus, the largest genus of the arbuscular mycorrhizal symbioses’, New Phytologist 182:347- mycorrhizal fungi (Glomales), is 358. nonmonophyletic. Molecular Phylogenetics and Evolution 21:190-197. Smith, S. E., and Read, D. 2008. Mycorrhizal Symbiosis (Third Edition). London, Academic Segall, M., Campbell, D. and Herskovits, M. J. Press. 1966. The influence of culture on visual perception New York, The Bobbs-Merrill Company. Smith, T. B. and Skúlason, S. 1996. ‘Evolutionary significance of resource polymorphisms in fish, Sen, A. K. 1977. ‘Rational fools: a critique of the amphibians and birds’, Annual Review of Ecology behavioural foundations of economic theory’, and Systematics 27: 111-113. Philosophy and Public Affairs 6(4): 317-344 Stanfey, A., G. 2007. ‘Social decision-making: Shutts, K., Kinzler, K.D., McKee, C.B., and insights from Game Theory and Neuroscience’, Spelke, E.S. 2009. ‘Social information guides Science 318: 598-602. infants' selection of foods’, Journal of Cognition and Development 10: 1-17

201 Stearns, S. 1992. The Evolution of Life Histories. Royal Society of London B-Biological Sciences. Oxford University Press 249(1326): 299-302.

Steinberg, L. 2008. ‘A social neuroscience Taylor, P. D. 1996a. ‘The selection differential in perspective on adolescent risk-taking’, quantitative genetics and ESS models’, Evolution Development Review 28; 78-106. 50:2106-2110.

Stockinger, H., C. Walker, and A. Schüßler. 2009. Taylor P.D. 1996b. ‘Inclusive fitness arguments in 'Glomus intraradices DAOM197198', a model genetic models of behaviour’, Journal of fungus in arbuscular mycorrhiza research, is not Mathematical Biology 34(5-6), 654-674. Glomus intraradices. New Phytologist 183:1176- Taylor, P. D. and Crespi, B. J. 1994. ‘Evolutionary 1187. stable strategy sex ratios when correlates of Sugiyama, L. S. 2004. ‘Is beauty in the context- relatedness can be assessed’, The American sensitive adaptations of the beholder?’, Evolution Naturalist. 143: 297-316. and Human Behaviour 25: 51-62 Taylor P.D., Frank S.A. 1996 ‘How to make a kin Sullivan K. A. 1988. ‘Age-specific profitability selection model’, Journal of Theoretical Biology and prey choice’, Animal Behaviour 36: 613–615. 180(1), 27-37.

Svensson, E. 2009. Understanding the egalitarian Taylor, P.D. and Irwin, A. J. 2000. ‘Overlapping revolution in human social evolution. Trends in generations can promote altruistic behaviour’, Ecology and Evolution 24(5): 233-235. Evolution, 54: 1135-1141.

Swami, V., and Tovee, M. J. 2006. ‘Does hunger Taylor P.D., Day, T. and Wild, G. 2007. ‘Evolution influence judgments of female physical of cooperation in a finite homogeneous graph’, attractiveness?’, British Journal of Psychology 97: Nature 447: 469-472. 353-363. Taylor, P. D., G. Wild, and A. Gardner. 2007. Taborsky, M. 1994. ‘Sneakers, satellites and ‘Direct fitness or inclusive fitness: how shall we helpers: Parasitic and cooperative behavior in model kin selection’, Journal of Evolutionary fish reproduction’, Advances in the Study of Biology 20:301–309. Behaviour, 21: 1-100 Taylor, P. D. and Frank, S. A. 2007. ‘How to make Taylor P.D. 1988. ‘An inclusive fitness model for a kin selection model’, Journal of Theoretical dispersal of offspring’, Journal of Theoretical Biology 180:27-37. Biology 130(3), 363-378. Taylor P.D., Grafen A. 2010 ‘Relatedness with Taylor P.D. 1989. ‘Evolutionary stability in one- different interaction configurations’, Journal of parameter models under weak selection’, Theoretical Biology 262(3), 391-397. Theoretical Population Biology 36(2), 125-143. Thiery O, Borstler B, Ineichen K, Redecker D. Taylor P.D. 1992a ‘Altruism in viscous 2010. ‘Evolutionary dynamics of introns and populations - an inclusive fitness model’, homing endonuclease ORFs in a region of the Evolutionary Ecology 6(4), 352-356. large subunit of the mitochondrial rRNA in Glomus species (arbuscular mycorrhizal fungi, Taylor, P. D. 1992b. ‘Inclusive fitness in a Glomeromycota)’, Molecular Phylogenetics and homogeneous environment’, Proceedings of the Evolution 55:599-610

202 Tomasello, M. 1999. The cultural origins of human Wehner, R. 1987. ‘“Matched filters”- neural cognition. Cambridge MA: Harvard University models of the external world’, Journal of Press. Comparative Physiology A, 161: 511–531.

Tooby, J. and Cosmides, L. 1990. ‘The past Wenseleers, T and F.L.W. Ratnieks. 2006. explains the present: Emotional adaptations and ‘Comparative analysis of worker reproduction the structure of ancestral environments’, and policing in eusocial Hymenoptera supports Ethnology and Sociobiology 11(4-5): 375-424. relatedness theory’, The American Naturalist 168: E163-E179. Trivers, R. L. 1971. ‘The evolution of reciprocal altruism’, The Quarterly Review of Biology 46(1): Wenseleers, T., A. G. Hart, and F. L. W. Ratnieks. 35–57. 2004a. ‘When resistance is useless: policing and the evolution of reproductive acquiescence in Tversky, A. and Kahneman, D. 1974. ‘Judgment insect societies’, The American Naturalist 164:E154- under uncertainty: Heuristics and biases’, Science E167. 185: 1124–1131 Wenseleers, T., H. Helanterä, A. Hart, and F. L. Tversky, A., Kahneman, D., 1986. ‘Rational choice W. Ratnieks. 2004b. ‘Worker reproduction and and the framing of decisions’, Journal of Business policing in insect societies: an ESS analysis’, 59(4): S251–S278 Journal of Evolutionary Biology 17:1035-1047. Valverde P., Healy E., Jackson I., Rees J.L., Thody Wenseleers, T., N.S. Badcock, K. Erven, A. A.J. 1995. ‘Variants of the melanocyte-stimulating Tofilski, F.S. Nascimento, A.G. Hart, T.A. Burke, hormone receptor gene are associated with red M.E. Archer and F.L.W. Ratnieks. 2005. ‘A test of hair and fair skin’, Nature Genetics 11: 328–330 worker policing theory in an advanced eusocial van Baalen, M., and Rand, D. A. 1998. ‘The unit wasp, Vespula rufa’, Evolution 59: 1306-1314 of selection in viscous populations and the West, S.A., Herre, E.A. and Sheldon, B.C. 2000. evolution of altruism’, Science 193(4):631-648 ‘The benefits of allocating sex’, Science 290: 288- van Tuinen, D., E. Jacquot, B. Zhao, A. Gollotte, 290. and V. Gianinazzi-Pearson. 1998. West, S. A., Murray, M. G, Machado, C., Griffin, ‘Characterization of root colonization profiles by A. S. and Herre, E. A. 2001. Testing Hamilton’s a microcosm community of arbuscular rule with competition between relatives. Nature, mycorrhizal fungi using 25S rDNA-targeted 409, 510-513. nested PCR’, Molecular Ecology 7:879-887. West S.A., Pen I., Griffin A.S. 2002a. ‘Conflict and Warren, J. M. 1965. ‘Primate learning in cooperation - Cooperation and competition comparative perspective’, in Schrier, A. M., between relatives’, Science 296(5565), 72-75. Harlow, H. F. and Stollnitz, F. (eds.) Behavior of Nonhuman Primates. Modern Research Trends. New West, S. A., E. T. Kiers, E. L. Simms, and R. F. York: Academic Press. Denison. 2002b. ‘Sanctions and mutualism stability: why do rhizobia fix nitrogen?’, Watanabe, F. S., and S. R. Olsen. 1965. ‘Test of an Proceedings of the Royal Society of London Series B- ascorbic acid method for determining Biological Sciences 269:685-694. phosphorus water and NaHCO3 extracts from soil’, Soil Science Society of America Journal 29:677- West, S.A., Reece, S.E. and Sheldon, B.C. 2002c. 678. ‘Sex Ratios’, Heredity 88: 117-124.

203 West, S. A., and A. Buckling. 2003. ‘Cooperation, Willetts D. 2010. The Pinch: how the baby boomers virulence and siderophore production in bacterial took their children's future - and why they should give parasites’, Proceedings of the Royal Society of it back. Atlantic Books. London Series B-Biological Sciences 270:37-44. Williams, G. C. 1966. Adaptation and Natural West, S. A., A. Gardner, D. M. Shuker, T. Selection. Princeton University Press. Reynolds, M. Burton-Chellow, E. M. Sykes, M. A. Wilson, D.S., Pollock, G.B. and Dugatkin, L.A. Guinnee, and A. S. Griffin. 2006. ‘Cooperation 1992. ‘Can altruism evolve in purely viscous and the scale of competition in humans’, Current populations?’, Evolutionary Ecology 6: 331-341. Biology 16:1103-1106. Wolf, A. P. 1995. Sexual attraction and childhood West, S.A., Griffin, A. S. and Gardner, A. 2007a. association: a Chinese brief for Edward Westermarck. ‘Social semantics: altruism, cooperation, Stanford University Press mutualism, strong reciprocity and group selection’, Journal of Evolutionary Biology 20: 415- Wolfe, B. E., D. L. Mummey, M. C. Rillig, and J. 432. N. Klironomos. 2007. ‘Small-scale spatial heterogeneity of arbuscular mycorrhizal fungal West, S.A., Griffin, A. S. and Gardner, A. 2007b. abundance and community composition in a ‘Evolutionary Explanations for cooperation’, wetland plant community’, Mycorrhiza 17:175- Current Biology, 17: R661-R672. 183. West, S. A. 2009. Sex Allocation. Princeton Wright, S. 1931. ‘Evolution in mendelian University Press, Princeton, NJ. populations’, Genetics 16:97-159. West, S.A., El Mouden, C. and Gardner, A. Wright, S. 1943. ‘Isolation by distance’, Genetics 2011. ‘Sixteen common misconceptions about the 28: 114-138. evolution of cooperation in humans’, Evolution and Human Behavior 32: 231-262. Wright, S. 1945. ‘Tempo and mode in evolution: a critical review’, Ecology 26: 415-419. Westermarck, E. A. 1921. The history of human marriage, 5th edn. London: Macmillan Yoneyama, K., X. N. Xie, H. Sekimoto, Y. Takeuchi, S. Ogasawara, K. Akiyama, H. Whiten, A. and Ham, R. 1992. ‘On the nature and Hayashi, and K. Yoneyama. 2008. Strigolactones, evolution of imitation in the animal kingdom: A host recognition signals for root parasitic plants reappraisal of a centaury of research’, in Slater P. and arbuscular mycorrhizal fungi, from Fabaceae J. B, Rosenblatt, J. S., Beer, C. and Milkinski, M. plants. New Phytologist 179:484-494. (eds.), Advances in the study of Behaviour, Volume 21. Academic Press, NY. Yu D. W., J. Ridley, E. Jousselin, E. A. Herre, S. G. Compton, J. M. Cook, J. C. Moore, G. D. Weiblen Wild G. 2006 ‘Sex ratios when helpers stay at the 2004. Oviposition strategies, host coercion and nest’, Evolution 60(10), 2012-2022. the stable exploitation of figs by wasps. , Wild G., Gardner A., West S.A. 2009 ‘Adaptation Proceedings of the Royal Society of London Series B- and the evolution of parasite virulence in a Biological Sciences 271: 1185-1195. connected world’, Nature 459(7249), 983-986.

Wild G. 2011 ‘Direct fitness for dynamic kin selection’, Journal of Evolutionary Biology, 24(7), 1598-1610.

204 Appendix

This Appendix contains the other publications resulting from my DPhil:

Section 1. El Mouden, C. (2011) Life: social to its core. Evolution and Human Behaviour 33(1): 79-80.

This is a book review I wrote of Bourke, A. F. G. (2011). Principles of Social Evolution. OUP.

Section 2: Harrison, F. H. and El Mouden, C. (2011). ‘Money for nothing, cooperation for free: exploring the effects of working for endowments on behaviour in standard economic games’, PLoS ONE 6(11): e27623.

This is an experimental paper examining whether individuals behave differently in economic games if they must first earn their endowment by performing dull or difficult tasks.

Section 3: West, S.A., El Mouden, C. and Gardner, A. (2011). ‘Sixteen common misconceptions about the evolution of cooperation in humans’, Evolution and Human Behavior 32(4): 231-262.

This paper reviews recent work on the evolution of cooperation in humans and highlights common misunderstandings regarding how natural selection operates and the conditions under which cooperation can be favoured.

205 Evolution and Human Behavior xx (2011) xxx–xxx

Book Review

Life: social to its core There are three stages in a major transition: social group Principles of Social Evolution by Andrew F. G. Bourke formation, social group maintenance and social group Oxford: Oxford University Press; 2011. 279 pp. £65. ISBN transformation. Social group formation occurs when genes 9780199231157. Paper, D52.95, £29.95. ISBN for social living spread in a population. While the contexts 9780199231164. Oxford Series in Ecology and Evolution. vary, common themes emerge in the evolutionary paths Reviewed by Claire El Mouden. Department of Zoology, members of different taxa take towards sociality. Drawing University of Oxford, Oxford, UK from work on bacteria, yeasts, algae, birds, fish, insects and mammals, Bourke identifies the genetic and ecological factors that promote or prevent social groups forming. Before Principles of Social Evolution, when social Environmental factors include environmental variability, scientists or economists asked me to recommend a book on food shortages, habitat saturation and predation. The key the evolution of cooperation, I never quite knew what to say. genetic factor, unsurprisingly, is relatedness between group Davies et al.'s An Introduction to Behavioural Ecology members. High relatedness most commonly occurs when (2011), the stalwart of undergraduate Zoology courses, is an offspring do not disperse or when a new group is founded by excellent, example-laden introduction to social evolution and a single propagule or parent. However, genetic factors still the wider field, but is too basic for most. Maynard Smith and matter when social partners are nonrelatives (e.g., partners in Szathmary's The Major Transitions in Evolution (1995) an interspecific mutualism); here, cooperation is favored illustrates the importance of sociality but lacks the conceptual when partners share reproductive fates — most commonly framework that explains how it evolves. At the other extreme, achieved by the vertical transmission of the social interaction Frank's Foundations of Social Evolution (1998) explains the so that partnerships are continued into the next generation. theory but does not discuss any empirical work. I usually The second stage, social group maintenance, refers to the suggest this trilogy and some more recent review papers (e.g. evolutionary processes that maintain the stability and Lehmann and Keller, 2006; Sachs, Mueller, Wilcox, & Bull, integrity of a group once formed. Groups are subject to 2004; West, Griffin, & Gardner, 2007a). Now I will just exploitation, both from within (group members becoming direct people to the Principles of Social Evolution. This social cheats) and without (parasites and pathogens). Social highly readable account uses the conceptual framework of cheats are limited by various mechanisms that often act inclusive fitness theory, combined with a wealth of empirical together such as negative frequency dependent selection examples, to illustrate the common principles that underlie (i.e., they do less well as they become more common), self- the evolution of sociality at all levels of complexity. limitation (when the per capita costs of cheating become Bourke, an evolutionary biologist from the University of excessive) or by a host of coercion mechanisms that have East Anglia, explains how sociality evolves by introducing evolved to prevent, suppress, police or sanction defection. the major transitions view of life. This view sees the repeated Bourke rightly emphasizes that the evolutionary mechanisms grouping together of biological units to form new higher-level driving the initial spread of cooperation differ from those that units as the central theme in history of life. If the selective maintain it — for example, high relatedness (so often key for conditions are right, these higher-level units can consolidate the initial formation of cooperative groups) is often not such that a new level of individuality emerges. Unlike the needed to maintain the group's stability — indeed in some traditional view of the history of life as a grand branching tree cases, low relatedness may be beneficial. of taxa (with mammals and humans on a branch somewhere The final stage in a major transition is social group near the top), this view emphasizes the multilayered transformation. This occurs when a social group becomes so complexity of life: genes form genomes, prokaryotic cells stable that it may be viewed as an adaptive unit in its own and protobacteria form eukaryote cells, eukaryote cells form right. This final stage occurs infrequently and has received multicelled organisms and multicelled organisms form much less empirical and theoretical attention than the first eusocial societies or join with other organisms to form two stages. Indeed, it is unclear how to best define when this mutualisms. Major transitions occur when the selfish interests stage is reached. Formally speaking, a group is an adaptive of previously independent entities are controlled so that the unit (i.e., its survival and reproduction may be seen as the collective interest of the group prevails — in other words, target of natural selection) only when the fitness interests of major transitions are explained by social evolution theory. its members are perfectly aligned. This only occurs when

1090-5138/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.evolhumbehav.2011.05.001 2 Book Review / Evolution and Human Behavior xx (2011) xxx–xxx competition within the group is completely repressed or when Bourke excludes human social behaviors from his discus- the group consists of genetic clones (Gardner and Grafen, sion. He justifies this by saying that our capacity for culture and 2009). Most people assume that the cells in a multicelled reciprocation sets us apart. Unfortunately, as studies of human organism meet this criteria and view whole organisms as the sociality have involved limited interaction with biologists, target of selection, but a closer examination reveals that misconceptions about the evolution of cooperation in the social conflicts of interest between higher and lower-level units sciences are widespread (West, El Mouden, & Gardner, 2011). rumble at all levels of the biological hierarchy (Queller & Thus, it would have been beyond the scope of this book to Strassman, 2009). Bourke argues that understanding how investigate how the common principles of social evolution that social groups consolidate to become individuals is one of the Bourke discusses apply to humans. Understanding the degree major outstanding questions for the field. to which human social behaviors fit within the framework While permanent cooperation within groups is essential to Bourke sets out remains an important task for the future. achieve biological complexity, Bourke stresses that complex Principles of Social Evolution is set to become a classic sociality is neither inevitable nor guaranteed to remain introduction to social evolution theory suitable both for graduate evolutionarily stable. Typically, natural selection only favors students and active researchers. To make it more accessible, weakly social groups (as occurs in many vertebrates) or I wish that it contained more photographs and figures, and fleeting cooperation associated with specific behaviors (such I would have liked more examples outside the social insects as pollination) — complex sociality therefore is the (in particular, more discussion of the work on cooperatively exception, not the rule. Furthermore, established cooperative breeding vertebrates). This is a minor complaint though; as groups sometimes lose in the ongoing battle with cheats, Bourke states, at fewer than 300 pages, this book was never pathogens and parasites. If such selfish behavior is not intended to provide a comprehensive review of the field. This controlled, it destroys groups. The parallel between the important book achieves it aim; it teaches us that to understand destruction caused by cancers in multicelled organisms and how social groups are formed and maintained is to understand rogue reproductive workers in eusocial insect societies is something fundamental about the nature of life itself. striking. In both cases, cells and workers cease cooperation and reproduce rapidly, causing death or serious harm to the Claire El Mouden organism or colony. Extreme cases of unchecked defection, Department of Zoology, University of Oxford, Oxford, UK such as the transmissible facial cancers in the Tasmanian E-mail address: [email protected] devil, are so catastrophic they could even make the species go extinct. Cooperation is indeed a fragile thing. References The book uses inclusive fitness theory as a framework to think about the process of natural selection. An individual's Davies N. B., Krebs J. R., & West S. A. (2011). An introduction to inclusive fitness (assuming no conflicts persist at lower behavioural ecology. 4th Edition. Hoboken, NJ: Wiley-Blackwell. levels of selection) is the quantity that natural selection Frank S. A. (1998). Foundations of social evolution. Princeton, NJ: Princeton University Press. maximizes (Grafen, 2009). It has two components, a direct Gardner A., & Grafen A. (2009). Capturing the superorganism: A formal component (the individual's genetic contribution to future theory of group adaptation. Journal of Evolutionary Biology 22: generations from its own offspring) and an indirect 659–671. component (the effect the focal individual's actions have Gardner A., West S. A., & Wild G. (2011). The genetical theory of kin – on the genetic contribution to future generations of its social selection. Journal of Evolutionary Biology 24:1020 1043. Grafen A. (2009). Formalizing Darwinism and inclusive fitness theory. partners, valued by its genetic relatedness to them). This is a Philosophical Transactions of Royal Society, Series B 364:3135–3141. general description of the action of natural selection which Lehmann L., & Keller L. (2006). The evolution of cooperation and altruism: may be encapsulated using Hamilton's rule. A general framework and classification of models. Journal of In the past, and more recently, inclusive fitness theory and Evolutionary Biology 19:1365–1376. Hamilton's rule have been subject to criticism. Bourke Lion S., Jansen V., & Day T. (2011). Evolution in structured populations: Beyond the kin versus group debate. TREE 26(4):193–201. devotes 15 pages to examine these criticisms and to discuss Maynard Smith J., & Szathmary E. (1995). The major transitions in the assumptions of inclusive fitness theory. For students evolution. Oxford: W.H. Freeman. encountering the field for the first time, this primer is worth the Queller D. C., & Strassmann J. E. (2009). Beyond society: The evolution of price of the book alone. Much confusion stems from the fact organismality. Philosophical Transactions of Royal Society, Series B – that (confusingly), apart from being the definition of what 364:3143 3155. Sachs J. L., Mueller U. G., Wilcox T. P., & Bull J. J. (2004). The evolution natural selection maximizes, inclusive fitness is also the name of cooperation. The Quarterly Review of Biology 79:135–160. given to a specific modeling technique (multievel selection West S. A., Griffin A. S., & Gardner A. (2007a). Evolutionary explanations and neighbor-modulated fitness are other common ap- for cooperation. Current Biology 17:R661–R672. proaches). All techniques assume that natural selection West S. A., Griffin A. S., & Gardner A. (2007b). Social semantics: altruism, maximizes the same thing; they just differ in how they carve cooperation, mutualism, strong reciprocity and group selection. Journal of Evolutionary Biology 20:415–432. up the total effect of selection and what assumptions they use West S. A., El Mouden C., & Gardner A. (2011). Sixteen common (for further discussion, see Gardner, West, & Wild, 2011; misconceptions about the evolution of cooperation in humans. Evolution Lion, Jansen, & Day, 2011; West, Griffin, & Gardner, 2007b). Human Behaviordoi:10.1016/j.evolhumbehav. 2010.08.001. Exploring the Effects of Working for Endowments on Behaviour in Standard Economic Games

Freya Harrison1*, Claire El Mouden1,2 1 Department of Zoology, University of Oxford, Oxford, United Kingdom, 2 Nuffield College, University of Oxford, Oxford, United Kingdom

Abstract In recent years, significant advances have been made in understanding the adaptive (ultimate) and mechanistic (proximate) explanations for the evolution and maintenance of cooperation. Studies of cooperative behaviour in humans invariably use economic games. These games have provided important insights into the mechanisms that maintain economic and social cooperation in our species. However, they usually rely on the division of monetary tokens which are given to participants by the investigator. The extent to which behaviour in such games may reflect behaviour in the real world of biological markets – where money must be earned and behavioural strategies incur real costs and benefits – is unclear. To provide new data on the potential scale of this problem, we investigated whether people behaved differently in two standard economic games (public goods game and dictator game) when they had to earn their monetary endowments through the completion of dull or physically demanding tasks, as compared with simply being given the endowment. The requirement for endowments to be ‘earned’ through labour did not affect behaviour in the dictator game. However, the requirement to complete a dull task reduced cooperation in the public goods game among the subset of participants who were not familiar with game theory. There has been some effort to test whether the conclusions drawn from standard, token-based cooperation games adequately reflect cooperative behaviour ‘in the wild.’ However, given the almost total reliance on such games to study cooperation, more exploration of this issue would be welcome. Our data are not unduly worrying, but they do suggest that further exploration is needed if we are to make general inferences about human behaviour from the results of structured economic games.

Citation: Harrison F, El Mouden C (2011) Exploring the Effects of Working for Endowments on Behaviour in Standard Economic Games. PLoS ONE 6(11): e27623. doi:10.1371/journal.pone.0027623 Editor: Attila Szolnoki, Hungarian Academy of Sciences, Hungary Received August 19, 2011; Accepted October 20, 2011; Published November 16, 2011 Copyright: ß 2011 Harrison, El Mouden. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This study was funded by an award from Oxford University Press’s John Fell Foundation to FH, and also by the University of Oxford via an undergraduate project studentship. FH is funded by a Fellowship by Examination at Magdalen College, Oxford and CEM by a BBSRC Doctoral Training Grant and a Postdoctoral Prize Research Fellowship at Nuffield College, Oxford. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected]

Introduction participants are given monetary tokens to invest in public goods or to divide with a partner in ultimatum, trust or dictator games. Cooperation – where one individual’s actions increase the These tokens are later converted into real earnings. Such fitness of another – can be favoured by natural selection for two economic games provide important insights into the roles of reasons. First, cooperation can increase the actor’s reproductive reciprocity [9,10], reputation [10,25], punishment [12,13], success (i.e. cooperation confers direct fitness benefits); second, the between-group competition [26,27], negotiation [28] and fairness actor can direct cooperation at individuals who also carry the norms [29] in maintaining cooperation. Experiments have also cooperative gene (i.e. cooperation confers indirect fitness benefits) demonstrated individual [18,30], cultural [31–33] and sex-based [1,2]. Evolutionary explanations for cooperation, and the effects of [34] differences in cooperative strategies. environment and population structure on selection for coopera- However, the extent to which these token-based games reflect tion, are well understood (reviewed by [1,3–5]). Our own species behaviour outside the laboratory is not clear (for discussion of this also possesses a variety of behavioural adaptations that promote issue, see [24,35–38]). On a simple level, Benz & Meier [39] report apparently selfless behaviour. These ‘proximate’ explanations that individuals make similar decisions about charitable giving (sensu [6,7]; see also [8]) of sociality include a tendency toward both in the lab and in a natural setting. A larger study of Ethiopian direct (e.g. [9]) and indirect [10,11] reciprocity and the forest user groups by Rustagi et al. [40] reports a correlation punishment of defectors (e.g. [12,13]). The neurological basis for between the number of conditional cooperators as identified in a cooperation has also received attention (e.g. [14–18]). This work lab game and the success of that group in managing the real-world helps to explain why, when people play anonymous one-shot public good. Consistent with this, Fehr & Liebbrandt’s study of economic games, they cooperate more than would be expected if Brazilian fishermen [41] showed that individuals who contribute they were purely self-interested [19–21]. more to the public good in a lab game use nets with larger holes, The most common way to investigate our propensity to which presumably allow more immature fish to escape and breed, cooperate, reciprocate or punish is to use economic games maintaining the real-world public good on which the participants’ ([8,9,20–23] and discussed by [24]). In these experiments, livelihoods depend (see also [42]). In contrast, Lamba’s [43] study

PLoS ONE | www.plosone.org 1 November 2011 | Volume 6 | Issue 11 | e27623 Working for Endowments in Standard Economic Games of Indian villagers found that an individual’s behaviour in a public much or as little of the endowment as they wanted to a well-known goods game did not predict whether or not they would exploit a UK charity. In the PGG, participants played five rounds of the real, valuable public good. In this study, play in a standard, game anonymously in groups of three. Groups were randomly structured economic game did not predict how selfishly people assigned to the M, T1 or T2 conditions. In rounds 1, 2, 4 and 5 behaved in the ‘real world’. participants in all conditions were simply given the endowment. In the real world of economic and biological markets, The nature of round 3 differed between experimental conditions: behavioural strategies incur real costs and benefits, in terms of in the M condition the endowment was provided gratis as in the effort, money, status, time, personal risk and, ultimately, other rounds, but in the T1 and T2 conditions endowments in Darwinian fitness. Clearly one cannot ask experimental partici- round 3 had to be earned by completing the tasks described above. pants to take personal risks or to incur fitness losses, but it is Embedding the T1 and T2 conditions in a sequence of M rounds possible to ask participants to use their own money in games or to allowed us to a) test for any pre-treatment differences between make participants ‘work’ for their endowment. A recent participants in the three conditions and correct for them if experiment [44] concluded that mean contributions to a public necessary; b) test for between-subjects differences in behaviour goods game were not affected when participants had to provide across the three conditions; and c) test for any ‘carry-over’ effects their own endowments; however, a re-analysis [45] disputed this of working for endowments when participants returned to being conclusion and revealed that the proportion of ‘free riders’ – provided with endowments in rounds 4 and 5. Because we wanted participants who contributed nothing to the public good – was to compare levels of cooperation between our treatment groups, actually higher among participants who used their own money to we imposed a game structure that, on the whole, favours play the game. Various authors have studied how behaviour in cooperation. Otherwise, differences between conditions may be standard games changes when the endowment level is dependent obscured by a general decay of cooperation towards a selfish upon a person’s performance in a task (usually a quiz [46]) optimum. We achieved this by introducing an element of between- compared with when endowments are randomly allocated. The group competition, a structure known to favour cooperation results are varied. A recent meta-analysis concluded that, on [1,26,27,52] – participants were told that if they were in one of the average, participants in dictator games gave less when they had three groups that earned the highest total amount of money, they earned their endowment [21]; however, Vilares et al. [47] showed would each be awarded a voucher for a major online retailer. there was no difference in behaviour in trust games when endowments were either supplied gratis or earned via a Results physically-demanding task. Aside from quizzes, we are aware of only three experiments that imposed tangible costs whilst 1. Task validation and dictator game specifically looking at variables that are predicted to affect a 90 participants took part in the DG, 30 being assigned to each person’s willingness to invest effort to benefit others. Heyman and of the three treatments. We used questionnaires during the DG to Ariely [48] used effort in a computer-based task to study the effect ascertain that the tasks we used were perceived differently from of reward magnitude on investment and Madsen et al. [49] and one another, and from the M condition. Approximately one-third Harrison et al. [50] used a physically-demanding exercise to of participants felt that they owned the money they had been investigate the effect of genealogical relatedness and social given, and this did not vary across treatments (10/30 for M, 12/30 2 proximity, respectively. for T1 and T2: x 2 = 0.378, p = 0.828). None of the participants in Given that these results are so varied, more work to explicitly the M condition felt that they had earned their endowment, while test how earning endowments affects behaviour in standard 19/30 and 17/30 participants in the T1 and T2 conditions economic games is needed. Here, we test how working to earn respectively felt it had been earned. The proportion of participants endowments affects behaviour in a highly structured economic who felt they had earned their money was not significantly 2 game (the public goods game, PGG) and in a much simpler different between the two task treatments (x 1 = 0.28, p = 0.598). measure of pro-social tendency (donation to a charity in a dictator Participants were more likely to consider the pipette task dull, as game, DG). We compared two qualitatively different tasks in our compared with the squatting task (20/30 vs. 10/30 participants in 2 ‘earning’ condition: a time-consuming task, and a physically- the two conditions said they found the task dull: x 1 = 6.67, demanding task. In the DG we also allocated heterogeneous p = 0.010). The squatting task was more likely than the pipette tip 2 endowments according to success in the task to test for any effect of task to be considered difficult (26/30 vs. 8/30 participants: x 1 = endowment size on the amount donated: not only do people earn 21.99, p,0.001). their resources in real life, but their earnings depend on their We tested whether the proportion of endowment donated by success and/or how hard they work [46]. In addition, we each participant was affected by sex, age, knowledge of game examined whether prior knowledge of game theory might theory, treatment and endowment size using a logit transformed moderate the effect of earning endowments: other authors (e.g. GLM. The best model, as defined using Akaike’s Information [51]) have shown that knowledge affects behaviour in standard Criterion (AICc: [53]) included only the main effects of these games. terms. The proportion of endowment donated did not differ Participants in our experiment played either a DG or a PGG. In between treatments (F2,15 = 0.02, p = 0.977, Figure 1), but there both cases, they were assigned to one of three treatments: was a slight negative correlation between endowment size and 2 monetary endowments were either given to the participant (M proportion donated (F1,58 = 6.66; p = 0.012, partial g = 0.103). condition) or had to be earned via completion of a dull (T1) or This is likely due to the very small and potentially ‘throwaway’ physically demanding (T2) task. The dull task required partici- amounts of money used at the lower end of our scale. There was pants to put pipette tips into boxes and the physically demanding no effect of sex (F1,58 = 0.48; p = 0.490), age (F1,58 = 0.38; task required them to squat in an isometric ski training exercise p = 0.542) or knowledge of game theory (F1,58 = 0.20; p = 0.660). (after [49] and [50]). The DG was a one-shot game in which participants were randomly assigned to the M1, T1 and T2 2. Public goods game conditions and told they would be given or made to earn their 72 participants took part in the PGG, 24 being assigned to each endowment accordingly and then have the option of donating as of the three conditions. The raw data are plotted in Figure 2. We

PLoS ONE | www.plosone.org 2 November 2011 | Volume 6 | Issue 11 | e27623 Working for Endowments in Standard Economic Games

knowledge of game theory and interactions between these variables. The best model, as defined using Akaike’s Information Criterion (AICc: [53]), is shown in Table 1. There was no main effect of condition (F2,65 = 2.82, p = 0.067), which is not consistent with our main hypothesis that earning investments alters behaviour. However, there was a significant condition x game theory interaction (F2,65 = 3.83, p = 0.007). As illustrated in Figure 3a, this was driven by a difference between the two task conditions among participants who were not familiar with game theory: in this subset of participants, investment following the dull, pipette tip task (T1) was lower than investment following the physical task (T2; Tukey post-hoc comparison, p = 0.049). Further, investment following T1 was lower among participants who were unfamiliar with game theory as compared with those who were Figure 1. Proportion of endowment donated to Oxfam in the familiar with game theory (Tukey post-hoc comparison, p = 0.003). dictator game. Black bars show participants who were given money There was also a significant effect of sex, such that women gave (M), grey bars participants who performed the pipette tip task (T1) and slightly more to the public good than men (F2,65 = 8.44, p = 0.005). white bars participants who performed the squatting task (T2). Bars We then sought to determine whether condition influenced show mean percentage donation 6 one standard error. The proportion behaviour in the two rounds after the treatment round, where of endowment donated did not differ between treatments (F2,15 = 0.02, p = 0.977), however people receiving larger endowments donated a participants in all conditions once again played the M game. smaller proportion to charity (F1,58 = 6.66; p = 0.012). Again, there was no main effect of condition (F2,136 = 1.58, doi:10.1371/journal.pone.0027623.g001 p = 0.211), knowledge of game theory (F1,136 = 1.47, p = 0.228) or round (F = 2.59, p = 0.110) but there was a significant first verified that our M condition produced results comparable 1,136 condition x game theory interaction (F = 3.93, p = 0.022). As with the existing literature on PGGs. In round 1, participants 2,136 illustrated in Figure 3b this interaction was due to a difference invested an average of 62.562.24% of their endowment. GLMM between investment following T2 when comparing participants showed that the trend in investment over the five rounds was best described by a linear function, with investment showing a shallow who were familiar or unfamiliar with game theory (Tukey post-hoc comparison, p = 0.046); post-hoc comparisons between conditions decline over time (F1,72 = 5.82, p = 0.018; slope = -3.161.28; partial g2 = 0.075). This is broadly consistent with results reported within each subset of participants were not significant (p.0.3). by other authors [20,54–57]. The slope differed significantly Therefore the divergent reaction to T1 and T2 among participants unfamiliar with game theory disappeared in rounds 4 and 5. between individual participants (F16,72 = 3.05, p = 0.001). We calculated the fitted slopes for each individual participant from this model and used GLMM to test for significant effects of sex, Discussion age and knowledge of game theory on the slope. No such effects The tasks we used appear well suited for experimental use: they were found (p $ 0.105). were easy to implement and were viewed differently by We then verified that there was no effect of condition on participants. We found no effect of performing either task to earn investment in rounds 1 and 2, when participants in all three endowments on behaviour in the DG. However, we found that conditions played the same game. Neither condition nor round people who completed the dull task and were unfamiliar with were significant as main effects (F = 2.96, p = 0.245 and 2,21 game theory cooperated less than the control and difficult task F = 0.27, p = 0.602, respectively) and neither was their 1,69 groups in the public goods game. Our results do not undermine interaction (F = 1.71, p = 0.188). 2,69 the standard uses of cooperation games. We did not find any Having thus satisfied ourselves that there were no pre-treatment differences between our experimental groups, we analysed investment behaviour in round 3, when conditions actually varied. Table 1. Analysis of variance for individual investment in a) We were interested in potential effects of condition, sex, age, round 3 and b) rounds 4 and 5 of the public goods game.

a)

Source DF F p condition 2 2.81 0.067 sex 1 8.44 0.005 game theory 1 8.00 0.006 condition x game theory 2 3.83 0.027

b)

Source DF F p condition 2 1.58 0.211 sex 1 5.45 0.021 Figure 2. Investment in the public goods game. Raw data (means game theory 1 1.47 0.228 6 standard error) of investment by participants in the three treatment condition x game theory 2 3.93 0.022 conditions in rounds 1-5 of the game. doi:10.1371/journal.pone.0027623.g002 doi:10.1371/journal.pone.0027623.t001

PLoS ONE | www.plosone.org 3 November 2011 | Volume 6 | Issue 11 | e27623 Working for Endowments in Standard Economic Games

task. In future it would be useful to ask participants explicitly about their feelings towards the tasks and towards their team-mates while performing tasks [52,59]. Further, we did not test the extent of our participants’ knowledge of game theory. In any future work, a more thorough assessment of how deeply participants understand game theory would facilitate a better exploration of how knowledge of game theory influences behaviour. Other authors have reported [34] that different game structures can lead to different levels of cooperation and earning seems to have different effects in dictator and trust games [21,47]. It is unclear why a) our two different tasks had divergent effects on behaviour in our PGG and b) a comparable effect was not present in the DG. While the framing [57] of actions in games has received a lot of general attention (e.g. [60–65]), the differences in framing and interpretation of different standard games would benefit from empirical research (e.g. [66]). In our case, the explanation may simply be that the PGG by nature is more complex than the DG, requiring more organisation and a more structured environment for participants. A separate but related issue is the extent to which a participant’s ‘cultural baggage’ may affect their interpretation of games and their behaviour in the lab [58,67]. In general, it is essential to understand what people are thinking, what features of the experimental environment they find salient and how they perceive different games. We have presented initial findings that behaviour in a PGG can be affected by requiring participants to earn their endowment by means of a dull task, provided that those participants are unfamiliar with game theory. Further exploration of how the Figure 3. Investment in the public goods game. a) In round 3, model tasks we employed affect economic/cooperative behaviour investment levels were constant across the three conditions among could prove very useful for future experimental work. For participants who had heard of game theory, but varied across example, it would be interesting to test how people perceive co- conditions among participants who had not heard of game theory players in games with and without tasks (see [52]). More (F1,65 = 3.83, p = 0.027). All pairwise comparisons of conditions among participants who had heard of game theory were non-significant importantly, we recommend a more thorough exploration of the (p.0.6). b) Average investment in round 4 and 5. Investment did not effects of these tasks on behaviour in a larger study population – differ between M, T1 and T2 conditions among participants who were perhaps a population that is expected to be naı¨ve with regard to unfamilar with game theory, nor between participants who were game theory – and with different game structures. familiar with game theory (p.0.3). Bars show mean 6 standard error; black bars show M condition, grey bars T1 and white bars T2. p-values for pairwise comparisons are from Tukey post-hoc tests. Methods doi:10.1371/journal.pone.0027623.g003 1. Ethics statement strong evidence against such games having the ability to explain This study was designed in accordance with the ethical human behaviour more generally. However, our results do suggest guidelines provided by the University of Oxford and the British that we should be aware that there are likely to be behavioural Psychological Society and received ethical approval from the differences between participant pools drawn from people cogni- University of Oxford’s Social Sciences and Humanities Interde- sant of game theory and people unaware of this field (see [58]). partmental Research Ethics Committee (reference: SSD/ Participants who had heard of game theory behaved differently CUREC1/10-284). Participants were recruited from students in the PGG from those who had not: this is not entirely surprising and staff at the University of Oxford – mainly those working or [51], but its interaction with performance of a dull task is attending lectures in the Department of Zoology, though a interesting. Further experimental work could usefully test the minority of participants were recruited by snowball sampling. All robustness of this effect and, if it stands up to more explicit participants provided written informed consent. Participants were scrutiny, address it in more detail. We hypothesise that, among advised that part of the experiment (T2 condition, see below) was participants who are familiar with game theory, the framing effects not suitable for people with back or knee problems; any of presenting them with ‘‘an economic game’’ induced strategic participants who said this would not be suitable for them were behaviour. In our design, inter-group competition for an assigned to another condition. Information supplied to participants additional payoff (shopping vouchers) went some way to aligning and questionnaires used are provided as Supporting Information individual and group interests and it is possible that people who S1. are familiar with game theory were more likely to focus on this potential gain and play strategically, whereas people who were not 2. Task validation and dictator games behaved more ‘‘emotionally’’ after performing the dull task. It is 90 participants (44 female) took part in this experiment, which interesting that the physical task did not have this effect and it is was framed as ‘‘a study of people’s feelings about money and possible that this instead was interpreted as a ‘‘team-building’’ giving to charity.’’ Participant age ranged from 18 to 45 years exercise. We noticed that participants appeared to be focussed on (mean 21.960.54 years). Participants were randomly assigned into their own pipette tip boxes and not engaging with one another, groups of 5 and each group randomly assigned to one of three whereas they seemed much more communicative in the physical treatment conditions: Money (M); Task 1 (T1); or Task 2 (T2). 30

PLoS ONE | www.plosone.org 4 November 2011 | Volume 6 | Issue 11 | e27623 Working for Endowments in Standard Economic Games participants took part in each condition. In the M condition, each Participants were randomly assigned into 24 teams of three group member was randomly assigned £2, £4, £6, £8 or £10. In people and each team assigned to either the money (M) or task the T1 condition, each group member was given four empty (T1, T2) conditions. 2-3 teams played the game in each pipette tip boxes (each designed to hold 96 200ml pipette tips) and experimental session and individuals were randomly assigned to asked to fill them with tips as fast as possible; the fastest person was teams, i.e. they did not know which of the other people in the assigned £10, the next £8, and so on down to the slowest person room were in their team. Teams remained fixed over the course of who received £2. In the T2 condition, participants were asked to the game. In the M condition, participants played five rounds of a squat in an isometric ski training position for as long as they could. simple public goods game: in each round, each participant The group member who squatted the longest received £10, the received an endowment of £1, a fraction of which (0, 0.2, 0.4, 0.6, second longest £8 and so on. Each group of five performed their 0.8 or 1) they could contribute to a team investment. Participants task simultaneously, i.e. there was no privacy within groups. kept any money they did not invest. In the T1 condition, rounds 1, Groups were, however, separated from one another; groups tested 2, 4 and 5 of the game were identical to the M game, but in round at the same time were separated by screens. Participants received 3, participants were given four empty pipette tip boxes and told their endowment in a numbered money bag. The endowment was that they would earn that round’s endowment in return for filling supplied divided into tenths (achieved by using appropriate the boxes with tips. The T2 condition was identical to the T1 denominations and/or taping coins together). condition except that participants earned their endowment in Median age did not vary across treatments (Kruskal Wallis Test: round 3 by squatting in an isometric ski training exercise for 45 H = 4.27, D.F. = 2, p = 0.118, n = 89 as one participant did not give seconds. In T1 and T2, all participants completed the task. In each their age). The proportion of females did not vary across treatments round, the total invested by a team was multiplied by 1.5 and 2 (x 2 = 3.29, p = 0.193, n = 88). 64 participants said that they had divided equally between the team members. Participants were told previously heard of game theory: males were more likely to have that the number of rounds in the game, and the instance of the 2 heard of game theory than females (x 1 = 6.07, p = 0.014, n =81), labour round, would be determined randomly. This was to remove but overall the proportion of participants who had heard of game the risk of a ‘‘last round’’ effect and/or of participants using 2 theory did not vary across treatments (x 2 = 77, p = 0.682, n =83). backward induction to determine the best investment strategy After receiving their endowment, participants answered a short [54]. Participants were told their individual payoff after each questionnaire asking about demographic information and their round and their total payoff at the end of the game. feelings about the endowment. Specifically, they were asked Because we wanted to compare levels of cooperation between whether they agreed or disagreed with the following statements: ‘‘I our treatment groups, we chose to impose a game structure that, earned the money I was given’’; ‘‘the money I was given belonged on the whole, favoured cooperation. Otherwise, differences to me’’; ‘‘the task I did was dull’’ and ‘‘the task I did was difficult’’. between conditions may have been obscured by a general decay Participants were instructed to go behind a screen, remove as of cooperation towards a selfish optimum regardless of condition. much of the endowment (in tenths) as they wanted to keep, re-seal Inter-group competition goes some way to aligning individuals’ the bag and place it in a donation box. Amounts donated were selfish interests with those of the group, favouring cooperation on linked to unique participant numbers, maintaining anonymity. both evolutionary [1] and behavioural [26,27,52] timescales. We Participants were informed that all donations went to Oxfam therefore imposed inter-group competition by awarding vouchers (http://www.oxfam.org.uk; this charity has long been a household for a major online retailer for the three groups which gained the name in the United Kingdom and has no political or religious highest total earnings across all five rounds of the game and across affiliation). The charity’s logo was used on posters in the all conditions. experimental area to reassure participants that money would Data on per-round individual and team investment were really be donated. A total of £337.80 was donated to Oxfam by analysed using GLMM. Participant was declared as a random participants in this experiment. factor nested in team, which was itself a random factor nested in We used general linear mixed models (GLMM) to determine condition. Round was coded as categorical variable to allow for whether the proportion of the endowment donated was affected by post-hoc pairwise comparisons between rounds and/or groups. treatment, sex, age, knowledge of game theory or endowment size, When all rounds in the M condition were analysed together, including group as a random factor. The results given are for individual investment was square root transformed to meet model analyses that exclude nine participants who did not state their sex assumptions. and/or knowledge of game theory; including these participants did not change the results. Proportion of endowment donated was Supporting Information logit transformed to meet the assumptions of homoscedasticity and normality of error. Supporting Information S1 Information and instructions provided to participants in the DG and PGG. 3. Public goods games (DOC) 72 participants (37 female) took part in this experiment. Participant age ranged from 18 to 40 (mean 22.260.44 years). Acknowledgments 39 participants said that they had previously heard of game theory Thanks to Anna Howell and Kevin Hilbert for helping us to collect data; and the frequency of participants who had heard of game theory all of our participants for their time and generosity; the Department of 2 was not different for females and males (x 1 = 0.93, p = 0.33). Zoology for allowing us to conduct experiments in communal space and Because significant written information was provided on game the Lamb & Flag, Oxford, for allowing us to conduct pilot studies on levels structure and payoffs, we verified that all participants spoke fluent of boredom induced by pipette tip racking on their premises. We would English. Before commencing the experiment, participants were also like to thank Max Burton-Chellew for valuable discussion and advice. given a table showing examples of possible team investment patterns and payoffs and asked to verify that they understood the Author Contributions game; participants were given the opportunity to ask questions Conceived and designed the experiments: FH. Performed the experiments: before the game began to ensure their understanding. FH CEM. Analyzed the data: FH. Wrote the paper: FH CEM.

PLoS ONE | www.plosone.org 5 November 2011 | Volume 6 | Issue 11 | e27623 Working for Endowments in Standard Economic Games

References 1. Frank SA (1998) Foundations of Social Evolution. Princeton, NJ: Princeton 35. Ku¨mmerli R, Burton-Chellow M, Ross-Gillespie A, West SA (2010) Resistance University Press, ix, 268. to extreme strategies, rather than prosocial preferences, can explain human 2. Hamilton WD (1964) The genetical evolution of social behaviour I & II. J Theor cooperation in public goods games. PNAS 107: 10125–10130. Biol 7: 1–52. 36. Hagen EH, Hammerstein P (2006) Game theory and human evolution: A 3. Gardner A, Foster KR (2008) The evolution and ecology of cooperation - history critique of some recent interpretations of experimental games. Theoretical and concepts. In: Korb J, Heinze J, eds. Ecology of Social Evolution: Springer. Population Biology 69: 339–348. pp 1–36. 37. Trivers R (2006) Mutual benefits at all levels of life (review of Genetic and 4. Sachs JL, Mueller UG, Wilcox TP, Bull JJ (2004) The evolution of cooperation. Cultural Evolution of Cooperation, MIT Press 2003, P Hammerstein, ed). Quarterly Review of Biology 79: 135–160. Science 304: 964–965. 5. West SA, Griffin AS, Gardner A (2007) Evolutionary explanations for 38. Levitt SD, List JA (2007) What do laboratory experiments measuring social cooperation. Current Biology 17: R661–R672. preferences reveal about the real world? The Journal of Economic Perspectives 6. Mayr E (1961) Cause and effect in biology. Science 134: 1501–1150. 21: 153–174. 7. Tinbergen N (1963) On aims and methods of ethology. Zeitschrift fur 39. Benz M, Meier S (2008) Do people behave in experiments as in the field? Tierpsychologie 20: 410–433. Evidence from donations. Experimental Economics 11: 268–281. 8. West SA, El Mouden C, Gardner A (2011) Sixteen common misconceptions 40. Rustagi D, Engel S, Kosfeld M (2010) Conditional cooperation and costly about the evolution of cooperation in humans. Evolution and Human Behavior monitoring explain success in forest commons management. Science 330: 32231-262. 961–965. 9. Gachter S, Herrmann B (2009) Reciprocity, culture and human cooperation: 41. Fehr E, Leibbrandt A (2011) A field study on cooperativeness and impatience in previous insights and a new cross-cultural experiment. Philosophical Transac- the tragedy of the commons. Journal of Public Economics 95: 1144–1155. tions of the Royal Society B-Biological Sciences 364: 791–806. 42. Carpenter J, Seki E (2010) Do social preferences increase productivity? Field 10. Wedekind C, Milinski M (2000) Cooperation through image scoring in humans. experimental evidence from fishermen in Toyama Bay. Economic Inquiry 49: Science 288: 850–852. 612–630. 11. Milinski M, Semmann D, Krambeck H-J (2002) Reputation helps solve the 43. Lamba S, Mace R (2011) Demography and ecology drive variation in ‘tragedy of the commons’. Nature 415: 424–426. cooperation across human populations. PNAS 108(35): 14426–14430. 12. Fehr E, Fischbacher U (2004) Third-party punishment and social norms. Evolution and Human Behavior 25: 63–87. 44. Clark J (2002) house money effects in public good experiments. Experimental 13. Fehr E, Gachter S (2002) Altruistic punishment in humans. Nature 415: Economics 5: 223–231. 137–140. 45. Harrison G (2007) House money effects in public good experiments: comment. 14. Behrens TE, Hunt LT, Rushworth MF (2009) The computation of social Experimental Economics 10: 429–437. behavior. Science 324: 1160–1164. 46. Cherry TL, Kroll S, Shogren JF (2005) The impact of endowment heterogeneity 15. Stevens JR, Cushman FA, Hauser MD (2005) Evolving the psychological and origin on public good contributions: evidence from the lab. Journal of mechanisms for cooperation. Annual Review of Ecology, Evolution, and Economic Behavior & Organization 57: 357–365. Systematics 36: 499–518. 47. Vilares I, Dam G, Kording K (2011) Trust and reciprocity: are effort and money 16. Suzuki S, Niki K, Fujisaki S, Akiyama E (2011) Neural basis of conditional equivalent? PLoS ONE 6: e17113. cooperation. Social Cognitive and Affective Neuroscience 6: 338–347. 48. Heyman J, Ariely D (2004) Effort for payment: a tale of two markets. 17. Tankersley D, Stowe CJ, Huettel SA (2007) Altruism is associated with an Psychological Science 15: 787–793. increased neural response to agency. Nat Neurosci 10: 150–151. 49. Madsen EA, Tunney RJ, Fieldman G, Plotkin HC, Dunbar RI, et al. (2007) 18. Wischniewski J, Windmann S, Juckel G, Bru¨ne M (2009) Rules of social Kinship and altruism: a cross-cultural experimental study. British Journal of exchange: game theory, individual differences and psychopathology. Psychology 98: 339–359. Neurscience and Behavioral Reviews 33: 305–313. 50. Harrison F, Sciberras J, James R (2011) Strength of social tie predicts 19. Fehr E, Fischbacher U (2003) The nature of human altruism. Nature 425: cooperative investment in a human social network. PLoS ONE 6: e18338. 785–791. 51. Frank RH, Gilovich T, Regan DT (1993) Does studying economics inhibit 20. Ledyard JO (1995) Public Goods: A Survey of Experimental Research. In: cooperation? The Journal of Economic Perspectives 7: 159–171. Kagel JH, Roth AE, eds. The Handbook of Experimental Economics. Princeton: 52. Burton-Chellew M, Ross-Gillespie A, West SA (2010) Cooperation in humans: Princeton University Press. competition between groups and proximate emotions. Evolution and Human 21. Engel C (2010) Dictator games: a meta study. Experimental Economics in press. Behavior 31: 104–108. DOI 10.1007/s10683-10011-19283-10687. 53. Hurvich CM, Tsai C-L (1989) Regression and time series model selection in 22. Fehr E, Rockenbach B (2004) Human altruism: economic, neural, and small samples. Biometrika 76: 297–307. evolutionary perspectives. Current Opinion in Neurobiology 14: 784–790. 54. Camerer CF (2003) Behavioral Game Theory. New York/Princeton: Russell 23. Hammerstein P, Hagen EH (2005) The second wave of evolutionary economics Sage Foundation/Princeton University Press. in biology. Trends in Ecology & Evolution 20: 604–609. 55. Ostrom E (2000) Collective action and the evolution of social norms. The 24. Falk A, Heckman JJ (2009) Lab experiments are a major source of knowledge in Journal of Economic Perspectives 14: 137–158. the social sciences. Science 326: 535–538. 56. Zelmer J (2003) Linear Public Goods Experiments: A Meta-Analysis. 25. Sommerfeld RD, Krambeck H-J, Semmann D, Milinski M (2008) Gossip as an Experimental Economics 6: 299–310. alternative for direct observation in games of indirect reciprocity. PNAS 104: 57. Chaudhuri A (2011) Sustaining cooperation in laboratory public goods games: a 17435–17440. selective survey of the literature. Experimental Economics: in press. 26. Puurtinen M, Mappes T (2009) Between-group competition and human 58. Henrich J, Heine SJ, Norenzayan A (2010) Most people are not WEIRD. Nature cooperation. Proceedings of the Royal Society B: Biological Sciences. pp 466: 29–29. 355–360. 59. Hertel G, Neuhof J, Theuer T, Kerr NL (2000) Mood effects on cooperation in 27. West SA, Gardner A, Shuker DM, Reynolds T, Burton-Chellow M, et al. (2006) small groups: Does positive mood simply lead to more cooperation?. Cognition Cooperation and the scale of competition in humans. Current Biology 16: & Emotion 14: 441–472. 1103–1106. 60. Tversky A, Kahneman D (1981) The framing of decisions and the psychology of 28. Cardenas JC, Stranlund J, Willis C (2000) local environmental control and institutional crowding-out. World Development 28: 1719–1733. choice. Science 211: 453–458. 29. Mellers BA, Haselhuhn MP, Tetlock PE, Silva JC, Isen AM (2010) Predicting 61. Andreoni J (1995) Warm-Glow versus cold-prickle: the effects of positive and behavior in economic games by looking through the eyes of the players. Journal negative framing on cooperation in experiments. The Quarterly Journal of of Experimental Psychology: General 139: 743–755. Economics 110: 1–21. 30. Kurzban R, Houser D (2001) Individual differences in cooperation in a circular 62. Cookson R (2006) Framing effects in public goods experiments. Experimental public goods game. European Journal of Personality 15: S37–S52. Economics 3: 55–79. 31. Ga¨chter S, Herrmann B, Tho¨ni C (2010) Culture and cooperation. 63. Levin IP, Schneider SL, Gaeth GJ (1998) All frames are not created equal: a Philosophical Transactions of the Royal Society B: Biological Sciences 365: typology and critical analysis of framing effects. Organizational Behavior and 2651–2661. Human Decision Processes 76: 149–188. 32. Henrich J, Boyd R, Bowles S, Camerer C, Fehr E, et al. (2005) ‘Economic man’ 64. Willinger M, Ziegelmeyer A (1999) Framing and cooperation in public good in cross-cultural perspective: behavioural experiments in 15 small-scale societies. games: an experiment with an interior solution. Economics Letters 65: 323–328. Behavioural and Brain Sciences 28: 795–855. 65. Dufwenberg M, Gaechter S, Hennig-Schmidt H (2011) The framing of games 33. Tracer DP (2003) Selfishness and fairness in economic and evolutionary and the psychology of play. Games and Economic Behavior: 73: 459–478. perspective: an experimental economic study in Papua New Guinea. Current 66. Larrick RP, Blount S (1997) The claiming effect: Why players are more generous Anthropology 44: 432–438. in social dilemmas than in ultimatum games. Journal of Personality and Social 34. Ku¨mmerli R, Colliard C, Fiechter N, Petitpierre B, Russier F, et al. (2007) Psychology 72: 810–825. Human cooperation in social dilemmas: comparing the Snowdrift game with the 67. Henrich J, Boyd R, Bowles S, Gintis H, Fehr E, et al. (2004) Foundations of Prisoner’s Dilemma. Procedings of the Royal Society B: Biological Sciences 274: Human Sociality: Economic Experiments and Ethnographic Evidence in Fifteen 2965–2970. Small-Scale Societies. Oxford: Oxford University Press.

PLoS ONE | www.plosone.org 6 November 2011 | Volume 6 | Issue 11 | e27623 Evolution and Human Behavior 32 (2011) 231–262

Review Article Sixteen common misconceptions about the evolution of cooperation in humans ⁎ Stuart A. West , Claire El Mouden, Andy Gardner Department of Zoology, Oxford University, OX1 3PS Oxford, United Kingdom Initial receipt 7 September 2009; final revision received 2 August 2010

Abstract

The occurrence of cooperation poses a problem for the biological and social sciences. However, many aspects of the biological and social science literatures on this subject have developed relatively independently, with a lack of interaction. This has led to a number of misunderstandings with regard to how natural selection operates and the conditions under which cooperation can be favoured. Our aim here is to provide an accessible overview of social evolution theory and the evolutionary work on cooperation, emphasising common misconceptions. © 2011 Elsevier Inc. All rights reserved.

Keywords: Altruism; Fitness; Inclusive fitness; Reciprocity

1. Introduction sciences. Consequently, in many cases, the evolutionary theory being applied in the social sciences is based on One of the greatest problems for the biological and secondary sources that were aimed at non-specialists (e.g. social sciences is to explain social behaviours such as Dawkins, 1976; Wilson, 1975b), some of which contain cooperation (Darwin, 1871; Hamilton, 1996). In the fundamental errors (Grafen, 1982; Dawkins, 1979) and, do biological sciences, the problem ranges from explaining not reflect the current state of the field. At the same time, cooperative helping behaviours in organisms such as evolutionary biologists have generally remained unaware of bacteria or birds to the evolution of complex social insect many important developments in the social sciences, such as societies (Sachs et al., 2004; West et al., 2007a). In the the vast theoretical literature on reciprocity (Binmore, 1998). social sciences, the problem ranges from explaining human These issues have led to many sources of confusion, such as morality and aspects of our underlying psychology to the the reinvention of old problems, the continuation of long- emergence of our institutions and societies (Binmore, finished debates, and very different explanations being given 2005b; Gintis et al., 2005a; Nettle, 2009). In principle, to the same empirical observations or theoretical predictions. Darwin's theory of natural selection provides a general Our overall aim in this article is to provide an overview of framework that has the potential to unite aspects of research the evolutionary study of cooperation in a way that is across these very different areas (Darwin, 1871). accessible across disciplines, emphasising common mis- However, there is relatively poor agreement between the conceptions. In the first part of our study (Sections 2–5), we social and biological sciences over the underlying evolu- provide a brief summary of the relevant aspects of tionary theory. Our understanding of social evolution theory evolutionary theory. Specifically, we summarise the modern has advanced hugely over the last 45 years, providing a interpretation of Darwin's theory of natural selection unified framework that can be applied to all organisms, from (Section 2), the evolutionary classification of social traits microbes to vertebrates (see Section 2). Unfortunately, these such as altruism (Section 3), the problem of cooperation advances have been communicated poorly to the social (Section 4) and the different ways in which the problem of cooperation can be solved (Section 5). We include a number of biological examples in Section 5, as this helps in the ⁎ Corresponding author. elucidation of general theoretical principles. Sections 2–5 E-mail address: [email protected] (S.A. West). could be skipped by readers familiar with the evolutionary

1090-5138/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.evolhumbehav.2010.08.001 232 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 literature. In the second part of the article, we discuss a Table 1 number of common misconceptions and sources of confu- Glossary sion, concerning social theory and the problem of cooper- Term Definition ation (Section 6). We hope that our purpose in this section Actor The focal individual performing a behaviour. does not come across as negative, as our aim is to facilitate Adaptation A trait that enhances fitness and that arose progress at the interface of the biological and social sciences. historically as a result of natural selection for its Finally, in the third part of the article, we focus on humans, current role (Rose & Lauder, 1996). The problem of adaptation is the need to explain the empirical fact discussing why they cooperate and whether they are special that organisms looked designed (Gardner, 2009). (Section 7). Altruism A behaviour that is costly to the actor and beneficial to the recipient or recipients. Costs and benefits are defined on the basis of the lifetime direct fitness 2. Evolutionary theory consequences of a behaviour (Hamilton, 1964). Cooperation A behaviour that provides a benefit to another 2.1. Adaptation and natural selection individual (recipient), and the evolution of which has been dependent on its beneficial effect for the The cardinal problem for evolutionary biology is to recipient (West et al., 2007a). explain adaptation (Leigh, 1971; Maynard Smith, 1995). The Direct fitness The component of fitness gained through the problem of adaptation is the need to explain the empirical impact of an individual's behaviour on the production of its own offspring; the component of fact that organisms appear designed (Paley, 1802). Within personal fitness due to one's own behaviour. this problem, there are two distinct issues: process and Inclusive fitness The effect of one individual's actions on purpose (Gardner, 2009). First, by what process (dynamics) everybody's production of offspring, weighted by does biological adaptation arise? Second, what is the purpose the relatedness; the sum of direct and indirect of adaptation: what is it that organisms appear designed fitness; the quantity maximised by Darwinian individuals (Grafen, 2006a; Hamilton, 1964). to do? Indirect fitness The component of fitness gained from aiding Darwin's theory of natural selection explains both the related individuals. process and the purpose of adaptation (see glossary in Kin selection Process by which traits are favoured because of Table 1; Gardner, 2009). The process of adaptation occurs their effects on the fitness of related individuals; the via the action of natural selection, which is driven by the way in which natural selection may be separated into direct and indirect components. differential reproductive success of individual organisms. Neighbour-modulated The personal fitness of an individual, which may be Those heritable characters that are associated with greater fitness dependent upon the behaviours of social partners. reproductive success tend to accumulate in natural popula- Mutual benefit A behaviour that is beneficial to both the actor and tions. Thus, Darwin explained the purpose of adaptation: the recipient (West et al., 2007a). he argued that evolved characters will appear designed as Personal fitness An individual's number of offspring, surviving to adulthood. In a class-structured population, each if to maximize the individual's reproductive success. This is offspring is weighted by their reproductive value. analogous to the idea in economics that individuals should Recipient An individual who is affected by the behaviour of be self-regarding utility maximizers—in both cases, it is the focal actor. not required that individuals are consciously striving to Relatedness A measure of the genetic similarity of two maximize their fitness or utility, only that selection will individuals, relative to the average; the least- squares linear regression of the recipient's genetic have led to individuals that do so (Darwin, 1859; Friedman, breeding value for a trait on the breeding value of 1953). the actor (Box 1; Grafen, 1985; Hamilton, 1970). These ideas were later formalised in mathematical terms Reproductive value The expected, relative genetic contribution of an by Fisher (Fisher, 1930, 1941), who used population individual to generations in the distant future; the genetics theory to describe natural selection in terms of relative probability that a gene drawn at random from a generation in the distant future will trace changes in gene frequencies. Specifically, Fisher showed back to the focal individual in the present that genes that are associated with greater individual fitness generation (Fisher, 1930; Grafen, 2006b). are predicted to increase in frequency; hence, natural Selfishness A behaviour which is beneficial to the actor and selection acts to increase the mean fitness of individuals in costly to the recipient. a population. Fisher interpreted this result, which he termed Social behaviours Behaviours which have a fitness consequence for both the individual that performs the behaviour the fundamental theorem of natural selection (see Supple- (actor) and another individual (recipient). mentary material), as proof that organisms will appear Spite A behaviour that is costly to both the actor and the increasingly designed so as to maximize their Darwinian recipient (Hamilton, 1970). fitness. At the time, the work of Fisher and others (Haldane, 1932; Wright, 1931) was celebrated for uniting Darwinism with Mendelian genetics, showing that they were not process is that natural selection leads to an increase in the competing alternative explanations for evolution (Provine, frequency of genes associated with greater fitness. The 2001). However, Fisher's work also formalised both the purpose is that natural selection will lead to organisms which process and the purpose of adaptation (Grafen 2002). The appear designed so as to maximize their individual fitness. S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 233 Since Darwin, the only fundamental change in our Inclusive fitness is not just a special case for interactions understanding of adaptation has been Hamilton's develop- between relatives. It is our modern interpretation of ment of inclusive fitness theory (Hamilton, 1964). The Darwinian fitness in its most general form, explaining both traditional Darwinian view struggled to explain many the process and purpose of adaptation (Grafen, 2007b, 2009). cooperative social behaviours, with the most famous The process is that genes or traits which lead to an increase in example being the sterile worker caste in eusocial insect inclusive fitness will be favoured, and that this increase can species, the ants, bees, wasps, and termites. Fisher (1930) occur via direct or indirect routes. The purpose is that realised that genes can spread not only through their impact individuals should appear as if they have been designed to on their own direct transmission (direct fitness) but also maximize their inclusive fitness. Grafen (1999, 2002, 2006a, through their impact on the transmission of copies of the 2007b) has formalised this link between the process and same allele in other individuals (indirect fitness; see also purpose of adaptation, by showing the mathematical Darwin, 1859, pp. 257–259), but he explicitly chose to equivalence between the dynamics of gene frequency change neglect the latter effects in his derivation of the fundamental and the purpose represented by an optimisation program theorem. Hamilton (1964) incorporated indirect fitness which uses an “individual as maximising agent” (IMA) effects into a genetical theory of social evolution and analogy. This emphasises that inclusive fitness is not just an showed that the characters favoured by natural selection are accounting method, it is the component of reproductive those which improve the individual's “inclusive fitness,” success an organism can influence and what organisms which is the sum of its direct and indirect fitness. Another should appear to be maximising. way of thinking about this is that inclusive fitness represents the components of reproductive success of the actor and their 2.2. Uses and multiple methods social partners over which the actor has control (Fig. 1). The easiest and most common way in which indirect fitness The idea that organisms can be viewed as maximizing benefits can occur is through helping close relatives, in agents has proven incredibly useful. This is because which case genes are identical by descent (i.e., from a inclusive fitness theory takes the dynamics of gene common ancestor), and so, this process is often referred to as frequency change (the gold standard of evolutionary theory) “kin selection” (Maynard Smith, 1964). and turns them into predictions about how individuals should behave (which can be tested with relative ease). The use of this approach in explaining a vast number of traits across a range of organisms can be seen in any animal behaviour or evolutionary ecology textbook (Alcock, 2005; Krebs & Davies, 1993; Westneat & Fox, 2010). Some of the most successful areas include sex allocation (West, 2009), policing and conflict resolution (Ratnieks et al., 2006), cooperation (this paper), kin discrimination (Griffin & West, 2003; Rousset & Roze, 2007), parasite virulence (Frank, 1996b), parent-offspring conflict (Trivers, 1974), sibling conflict (Mock & Parker, 1997), selfish genetic elements (Burt & Trivers, 2006), cannibalism (Pfennig et al., 1999), dispersal (Hamilton & May, 1977), alarm calls (Sherman, 1977) and genomic imprinting (Haig, 2002). The success of Maynard Smith's (1982) evolutionarily stable strategy (ESS) approach is also because it makes an IMA analogy and, hence, predicts the behaviour of Fig. 1. Inclusive fitness is the sum of direct and indirect fitness (Hamilton, individuals. Most ESS models are not concerned with social 1964). Social behaviours affect the reproductive success of self and others. behaviours so can assume that indirect fitness effects are The impact of the actor's behaviour (yellow hands) on its reproductive unimportant meaning that individuals should behave so as to success (yellow offspring) is the direct fitness effect. The impact of the maximize their personal fitness (Maynard Smith & Price, actor's behaviour (yellow hands) on the reproductive success of social 1973). This is a special case of the more general inclusive partners (blue offspring), weighted by the relatedness of the actor to the recipient, is the indirect fitness effect. In particular, inclusive fitness does not fitness result, and has been formally justified with population include all of the reproductive success of relatives (blue offspring), only genetics (Fisher, 1930; Grafen, 1999, 2002). Empirical that which is due to the behaviour of the actor (yellow hands). Also, success stories in this area include research on foraging, inclusive fitness does not include all of the reproductive success of the competing for resources and the evolution of mating systems actor (yellow offspring), only that which is due to its own behaviour (Krebs & Davies, 1993). (yellow hands; adapted from West et al., 2007a). A key feature of inclusive fitness is that, as defined, it describes the components of reproductive Inclusive fitness theory has well-developed links with all success which an actor can influence, and therefore which they could be the other areas of evolutionary theory, especially quantitative appearing to maximise. and population genetics (Frank, 1998; Gardner et al., 2007a; 234 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 Grafen, 2006a; Queller, 1992a; Rousset, 2004; Taylor, 1990, Table 2 1996; Taylor & Frank, 1996; Wolf et al., 1999). As Hamilton Social behaviours (1964) originally showed, an advantage of inclusive fitness Effect on actor Effect on recipient theory is that it can be applied at the genetic or phenotypic + − level [contra O'Gorman et al., 2008; Sober & Wilson, 1998]. + Mutually beneficial Selfish Put another way, it is a genetic theory to explain individual − Altruistic Spiteful level adaptations. Modern techniques for the development of A Hamiltonian classification scheme for social behaviours that have been inclusive fitness theory, termed the direct or neighbour- selected for by natural selection (West et al., 2007b). These classifications modulated fitness method, provide very general, powerful are based on the average consequences of a behaviour, which is what matters for natural selection. and simple methods for analysing the evolution of all forms of social behaviour (Frank, 1997, 1998; Rousset, 2004; Taylor, 1996; Taylor & Frank, 1996; Taylor et al., 2007b). Importantly, these methods allow the biology to lead the (b) approaches such as multilevel selection, which focus on maths, rather than forcing the biology to fit the assumptions the process of adaptation, can lead to confusion over the of stylized games such as the Prisoner's dilemma (Brown, purpose of adaptation (Misconception 12). 2001; West et al., 2007a). An introduction to the Second, it is sometimes assumed that inclusive fitness mathematics and methods of kin selection theory is provided theory cannot be applied under certain conditions, such as elsewhere (Gardner et al., Submitted). when there is frequency dependence, strong selection While the theoretical overview that we have given above (mutations of large effect) or multiplicative fitness effects. is the framework within which the majority of evolutionary However, this is not the case, as such assumptions are not biologists work, it is not accepted by all researchers in the required by inclusive fitness theory (Hamilton, 1970; discipline. There are two issues here. First, of course we are Queller, 1992c; Gardner et al., Submitted). Instead, it is not suggesting that inclusive fitness is the only way to model that naive applications of inclusive fitness theory (especially social evolution. A variety of methods exist, which each Hamilton's rule) can lead to mistakes in such circumstances have pros and cons. Many researchers mix methods, by using (Frank, 1998; Gardner et al., Submitted, 2007a). the neighbour-modulated fitness method to construct models, which they then interpret with inclusive fitness theory (Taylor & Frank, 1996). In some cases, other methods 3. Social traits are more useful. For example, if the co-evolutionary dynamicsbetweentraitsiskey,asisthecasewith Within evolutionary biology, social behaviours are punishment of non-cooperators, then multi-locus population defined according to their personal fitness consequences genetic methods offer many advantages (Gardner et al., for the actor and recipient. An individual's personal fitness 2007a). Multi-level selection theory offers another method- is defined as the number of offspring that she produces ology, although this tends to be used relatively little, because that survive to adulthood (Dawkins, 1982; Grafen, 2007b; it: (a) can can be hard or impossible to incorporate many Hamilton, 1964; Maynard Smith, 1983; also termed neigh- important biological complexities (Queller, 2004), especially bour-modulated fitness). From an evolutionary point of those that arise when populations are structured into classes view, a behaviour (or action) is social if it has fitness (e.g., sexes, or age groups; Frank, 1998); (b) seems to lead to consequences for both the individual that performs that many sources of confusion (Misconceptions 9–13). behaviour (the actor) and another individual (the recipient). A general point with all these alternative methods is that Hamilton (1964) classified social behaviours according to they analyse the dynamics of natural selection differently whether the consequences they entail for the actor and and, so, do not constitute competing hypotheses as to how recipient are beneficial (increase personal fitness) or costly adaptation occurs or what it is for. Whatever way you do the (decrease personal fitness) (Table 2). A behaviour which is maths, this does not change that organisms are predicted to beneficial to the actor and costly to the recipient (+/−)is maximize inclusive fitness (Gardner & Grafen, 2009; selfish, a behaviour which is beneficial to both the actor and Grafen, 2006a, 2007b; Hamilton, 1975). The IMA analogy the recipient (+/+) is mutually beneficial, a behaviour which does not work with alternatives such as neighbour- is costly to the actor and beneficial to the recipient (−/+) is modulated fitness or group fitness, because an individual altruistic, and a behaviour which is costly to both the actor cannot completely control these measures of fitness, except and the recipient (−/−)isspiteful(Hamilton, 1964; in special cases (Hamilton, 1964). In contrast, inclusive Hamilton, 1970; West et al., 2007b). fitness looks at natural selection from the perspective of the Social behaviours are defined in this way for two reasons. elements of fitness over which the individual has control First, the adaptationist approach provides a formal justifica- (Fig. 1). Consequently, we favour the inclusive fitness tion for the use of intentional language (Grafen, 1999). As approach because (a) inclusive fitness provides a single described in Section 2, there is a mathematical correspon- theory that describes both the process and purpose of dence (isomorphism) between the dynamics of natural adaptation (no other theory has been shown to do this) and selection and the idea that the individual organism is striving S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 235 to maximize her lifetime reproductive success. Consequent- beneficial to a dung beetle that comes along and uses ly, whether a behaviour is beneficial or costly is defined on that dung, but it is not useful to call this cooperation. We the basis of: (i) the lifetime fitness consequences of the would only call this cooperation if the elephant were behaviour and (ii) the fitness of individuals relative to the selected to increase its rate of dung production because it whole population. Alternative evolutionary definitions of gained some benefit from the byproducts of the dung beetle terms such as altruism that rely upon only the short-term using their dung. More generally, we could refer to “social fitness consequences (e.g., “reciprocal altruism”), or relative adaptations” if we wanted to consider social behaviours to only a fraction of a population (e.g., the local group, as in (Table 2) whose selection has been influenced by the the group selection literature) lack formal justification, fitness consequences for the recipient. because there is no corresponding maximizing agent view that supports them. Second, these intentional terms do not provide a 4. The problem of cooperation superficial gloss, but are defined in ways that convey important information about gene frequency dynamics. In The problem of cooperation is to explain why an particular, altruistic and spiteful behaviours could not be individual should carry out a cooperative behaviour that explained by the Darwinian view, formalized by Fisher benefits other individuals (Hamilton, 1963, 1964). All else (1930), that individuals strive to maximize their personal being equal (i.e., in the absence of one of the mechanisms we fitness and, hence, required consideration of indirect fitness discuss below), cooperation would reduce the relative fitness consequences (Hamilton, 1964). It is for these two reasons of the performer of that behaviour and hence be selected that Hamilton's definitions have proven so useful in fields against. To illustrate this, consider a population of such as animal behaviour (Krebs & Davies, 1993). unconditional cooperators in which an uncooperative free Altruistic behaviour is favoured when it is directed rider (cheat) arises through mutation or migration. In the towards individuals who share a genetic predisposition for absence of any mechanism to punish noncooperators, the altruism (positive relatedness), such as when they share the free rider benefits from the cooperative behaviour of its same genes for altruism. In Misconceptions 1 and 2, we will social partners, without paying any cost. Consequently, discuss some of the confusion that has come about through genes for free riding have greater fitness than the genes for researchers redefining altruism (Hamilton, 1964; West et al., cooperation, and the former spread through the population, 2007b). Spiteful behaviour is favoured when it is directed despite the fact that this will lead to a decline in population towards individuals who are genetically less similar than fitness. The problem of cooperation is often illustrated within average (negative relatedness; Hamilton, 1970). One way of the fields of economics and human morality, as the “tragedy conceptualizing this is that the reduced fitness of the of the commons” (Hardin, 1968) or the prisoner's dilemma recipient reduces competition for other individuals who are (Luce & Raiffa, 1957; Rapoport & Chammah, 1965), but a more related to the actor than the recipient, i.e., spite is a variety of other games have also been used (Binmore, 1994, form of indirect altruism (Gardner et al., 2007b). This 1998, 2005b). Explaining the apparent paradox of cooper- requires very restrictive conditions, and there are only a ation is one of the central problems of biology because couple of clear examples in the natural world, such as almost all of the major evolutionary transitions from chemical warfare in bacteria and the sterile soldiers in replicating molecules to complex animal societies have polyembryonic wasps (Gardner et al., 2004; Gardner et al., relied upon solving this problem [see supplementary 2007b). It seems extremely unlikely that these conditions material; Leigh, 1991; Maynard Smith & Szathmary, 1995]. would be met in humans, where apparently spiteful behaviours are more likely to provide a direct benefit and hence be selfish (West & Gardner, 2010). 5. The solutions to the problem of cooperation Cooperation is defined as a behaviour which provides a benefit to another individual (recipient) and which is As cooperation is in evidence at all levels throughout selected for because of its beneficial effect on the recipient the natural world, there must be one or many solutions to the (West et al., 2007b). This definition of cooperation problem. In this section, we shall give a brief overview of therefore includes all altruistic (–/+) and some mutually the potential solutions. Further details can be found in the beneficial (+/+) behaviours. The latter clause in this supplementary material or elsewhere (Lehmann & Keller, definition relates to the standard text book definition of 2006; Sachs et al., 2004; West et al., 2007a). adaptation (Rose & Lauder, 1996), and focuses our Theoretical explanations for the evolution of cooperation attention upon behaviours that are selected for because of (or any behaviour) are broadly classified into two categories: their social consequences [see also Scott-Phillips, 2008]. direct fitness benefits or indirect fitness benefits (Fig. 2). A Therefore, we do not include any behaviours that only cooperative behaviour yields direct fitness benefits when the incidentally produce a one-way byproduct benefit to others. reproductive success of the actor, who performs the For example, when an elephant produces dung, this is cooperative behaviour, is increased. Cooperative behaviours beneficial to the elephant (emptying waste) and also that benefit both the actor and the recipient(s) of the 236 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

cooperation

direct benefits indirect benefits

limited kin non-enforced enforced greenbeard dispersal discrimination

facultative fixed environmental genetic enforcement enforcement cues cues

shared prior reward punishmentsanctions reciprocity policing environment assessment

indirect direct (reputation-based) reciprocity reciprocity

Fig. 2. A classification of the explanations for cooperation. Direct benefits explain mutually beneficial cooperation, whereas indirect benefits explain altruistic cooperation (Hamilton, 1964). Within these two fundamental categories, the different mechanisms can be classified in various ways (Bergmüller et al., 2007; Frank, 2003; Lehmann & Keller, 2006; Sachs et al., 2004; West et al., 2007a). These possibilities are not mutually exclusive—for example, a single act of cooperation could have both direct and indirect fitness benefits, or interactions with relatives could be maintained by both limited dispersal and kin discrimination. Our dividing up of conditional enforcement strategies is for illustration only, as a detailed discussion is beyond the scope of this paper, and provided elsewhere (Bergmüller et al., 2007) (adapted from West et al., 2007a). behaviour are termed mutually beneficial—although they exhausted swimmer, enslaved ants rear the brood of the may appear altruistic, they are not (West et al., 2007b; see slave-making species that captured them, or a reed warbler Misconceptions 1 and 2). These “self-interested” behaviours feeds a cuckoo chick that is bigger than itself. are readily studied using standard economics models. A However, these “irrational” or seemingly maladaptive cooperative behaviour can be explained by indirect fitness behaviours can be trivially explained by considering the benefits if it is directed towards other individuals who carry average fitness consequences of such an evolved response. genes for cooperation (Hamilton, 1964). As mentioned Specifically, the underlying mechanism that leads to such above, this is usually referred to as “kin selection” (Maynard behaviours will have only been selected for if they, on Smith, 1964) because the simplest and most common way average, provide a direct or indirect fitness benefit. For indirect benefits can occur is if cooperation is directed at example, the behaviour of dolphins may be a byproduct of genealogical relatives (kin), who share genes from a selection for helping within dolphin groups, the rearing common ancestor (Frank, 1998). By helping a close relative behaviour of the enslaved ants is favoured because it is reproduce, an individual is still passing copies of its genes on usually directed towards related brood, and the reed warbler to the next generation, albeit indirectly. Cooperative feeds the cuckoo because the chicks in its nest will usually behaviours that are costly to the actor and beneficial to the be its own offspring. The general point here, that we shall recipient are termed altruistic (Hamilton, 1964; West et al., return to in Misconceptions 4 and 14, is that maximisation 2007b; see Misconceptions 1 and 2). of fitness does not lead to an expectation for perfect fitness- Before describing the mechanisms that can explain maximising behaviour in every real-time situation. Behav- cooperation, a general point about the differences between iour should be studied within the context of the evolutionary mechanisms and rational choice theory is that environment in which it was selected for and is being evolutionary mechanisms only explain the average con- maintained (Herre, 1987). The possibility for such irrational sequences of a behaviour. Therefore, it is quite normal in mistakes arises even before we start considering the time nature to observe seemingly “irrational” behaviour where an that it takes for selection to “catch up” with environmental observed cooperative behaviour provides no direct or change (e.g., the time taken for bird species to evolve the indirect fitness benefit, such as when: dolphins help an ability to spot and avoid cuckoos). S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 237 5.1. Kin selection and indirect fitness benefits between interacting individuals. Hamilton (1964) suggested two possible mechanisms through which a high relatedness Hamilton's inclusive fitness (kin selection) theory could arise between social partners: kin discrimination and explains how altruistic cooperation can be favoured between limited dispersal. relatives. This is encapsulated in a pleasingly simple form by Hamilton's (1963, 1964, 1970) rule, which states that a 5.1.1. Kin discrimination behaviour or trait will be favoured by selection, when The first mechanism for generating sufficiently high rb-cN0, where c is the fitness cost to the actor, b is the relatedness to make indirect fitness benefits important is fitness benefit to the recipient, and r is their genetic kin discrimination, when an individual can distinguish relatedness. The coefficient of relatedness (r) is a statistical relatives from non-relatives and preferentially direct aid concept, describing the genetic similarity between two towards them (nepotism) (Hamilton, 1964). This has been individuals, relative to the average similarity of all demonstrated in a range of organisms, from fungi to birds individuals in the population (Grafen, 1985; Hamilton, to humans (see supplementary material). A clear example is 1970; Box 1). Putting this inequality into words, altruistic provided by Britain's only cooperative breeding bird, the cooperation can therefore be favoured if the benefits to the long-tailed tit, where individuals that fail to breed indepen- recipient (b), weighted by the genetic relatedness of the dently, preferentially help at the nest of close relatives recipient to the actor (r), outweigh the costs to the actor (c). (Russell & Hatchwell, 2001). All the terms (b, c and r) can be positive or negative, and so Kin discrimination can occur through the use of Hamilton's rule can be applied to all forms of social environmental or genetic cues (Grafen, 1990b). Environ- behaviour. The generality of Hamilton's rule as a complete mental cues, such as prior association or shared environment, description of the dynamics of natural selection is discussed appear to be the most common mechanism of kin elsewhere (Frank, 1998; Gardner et al., Submitted). discrimination and have been found in organisms ranging Explanations for cooperation based on indirect fitness fromantstohumans(Helanterä & Sundström, 2007; benefits require a sufficiently high genetic relatedness (r) Lieberman et al., 2003), for example, in long-tailed tits,

Box 1 What is relatedness?

An individual's phenotype can be separated into its genetic (i.e., heritable) component and its non-genetic (i.e., environmental) component (Fisher, 1918). The former component is termed the individual's genetic value for the phenotypic character of interest, and the action of natural selection is formally defined with respect to the change in the average of this quantity (Fisher, 1930). Fisher (1930) separated the action of natural selection into direct fitness effects (impact on personal reproductive success) and indirect fitness effects (impact on the reproductive success of relatives), and Hamilton (1963, 1964, 1970) showed how the latter are mediated by coefficients of relatedness between social partners. The coefficient of relatedness is defined statistically, as measure of the genetical similarity between social partners, relative to the rest of the population (Grafen, 1985; Hamilton, 1970). Specifically, it is given by:

r = covðÞg; g0 =covðÞg; g ;

where g is the genetic value of a focal individual for the phenotypic character of interest, g′ is the genetic value of the social partner of this individual, and cov denotes a statistical covariance taken over all individuals in the population (Frank, 1998; Gardner et al., Submitted; Orlove and Wood, 1978). This covariance formulation has a useful interpretation: if we make a scatter plot of the genetic values of social partners (g′; y-axis) against an individual's own genetic value (g; x-axis), then the coefficient of relatedness is equal to the slope of the straight line fitted through these data by means of least-squares regression (Gardner et al., Submitted; Grafen, 1985). The statistic r can be positive or negative like any statistical correlation but will have a mean value within a population of zero (i.e., if the y-axis of the scatter plot represents the average genetic value of all individuals in the population, this will be a constant and the corresponding regression line will have slope zero). Grafen (1991) defines “the relatedness of a potential actor A to the potential recipient R [as] the extent to which A helping R is like A helping itself.” In other words, the important measure of genetic similarity when considering the “r” in Hamilton's inequality, is the genetic similarity between two individuals relative to that between random individuals in the population as a whole. This stresses that it is genetic similarity not kinship per se which drives indirect fitness benefits—kinship just happens to be by far the most important reason by which genetic similarity arises (greenbeard genes being the other possibility). 238 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 where individuals distinguish between relatives and non- example, where the benefits can be in the future, rather than relatives on the basis of vocal contact cues, which are learned immediate, is if cooperation leads to an increase in group from related adults during the nesting period (associative size, which increases the fitness of everyone in the group, learning) (Sharp et al., 2005). Genetic cues include examples including the individual who performs the cooperative such as the odour produced by scent glands in a mammal behaviour (Kokko et al., 2001; Wiley & Rabenold, 1984; (Grafen, 1990b). This has been demonstrated in a range of Woolfenden, 1975). This process, termed group augmenta- organisms, including social amoebae, ants, and mammals tion, has been argued to be important in many coopera- (Boomsma et al., 2003; Mateo, 2002). There are a number of tively breeding vertebrates, such as meerkats, where a studies on potential mechanisms for kin discrimination in larger group size can provide a benefit to all the members humans (see supplementary material). of the group through an increase in survival, foraging success and the likelihood of winning conflicts with other 5.1.2. Limited dispersal groups (Clutton-Brock, 2002). Similar arguments can The second mechanism for generating sufficiently high explain cases of helping between unrelated individuals in relatedness to make indirect fitness benefits important is wasps, where high mortality rates mean that there is an limited dispersal (Hamilton, 1964). Limited dispersal appreciable chance that a subordinate individual can inherit (population viscosity) can generate high degrees of the dominant position and, hence, also inherit any workers relatedness between interacting individuals because it will that they helped produce (Queller et al., 2000). tend to keep relatives together (Hamilton, 1964). In this case, unconditional cooperation directed indiscriminately at 5.2.2. Enforcement other group members (neighbours) could be favoured, The second way in which cooperation can provide direct because group members (those neighbours) are more likely fitness benefits is if there is some mechanism for enforcing to be relatives (have a coefficient of relatedness above the cooperation by rewarding cooperators or punishing cheaters. population average). This mechanism has the potential to Trivers (1971) emphasised that cooperation could be be important in a wide range of cases, from the simplest favoured in reciprocal interactions with individuals prefer- replicating molecules to humans and other vertebrates, entially aiding those that have helped them in the past, as because it does not require the evolution of any potentially encapsulated by the well known phrase “you scratch my costly mechanism of kin discrimination to work (West back and I'll scratch yours.” This idea dates back to Hume et al., 2002a). Instead, all that is required is that the level (1739) and had already been analysed in detail in the of cooperation evolves in response to the average economics literature before Trivers rediscovered it (reviewed relatedness between individuals who tend to interact by by Aumann, 1981; Aumann & Maschler, 1995; Binmore, chance. Direct experimental evidence for a role of limited 1994, 1998, 2005b, 2007; Fudenberg & Maskin, 1986; dispersal has come from observational field data and Kandori, 1992; Luce & Raiffa, 1957; Mailah & Samuelson, laboratory experimental evolution on social amoebae and 2006). Reciprocal helping is sometimes referred to as direct bacteria (Brockhurst et al., 2007; Diggle et al., 2007; reciprocity (help those who help you), to distinguish it from Gilbert et al., 2007; Griffin et al., 2004; Kümmerli et al., indirect reciprocity, where cooperation is directed at those 2009) and field data on cooperative breeding vertebrates who are known to cooperate with others, via some method of (Cornwallis et al, 2009, 2010). “image scoring” (help those who help others; Alexander, 1987; Nowak & Sigmund, 1998). 5.2. Direct fitness benefits The possibility for cooperation via reciprocity has The evolution of cooperation does not only depend upon attracted much enthusiasm, with a huge theoretical literature kin selection and indirect fitness benefits—cooperation can investigating its possibility. In addition, both direct and also provide a direct fitness benefit to the cooperating indirect reciprocity appear to be important in the evolution individual (Trivers, 1971). In this case, cooperation is and maintenance of cooperation in humans (Alexander, mutually beneficial, not altruistic, and hence would be 1987; Binmore, 1994, 1998, 2005b; Gächter & Herrmann, favoured by “self interested” or “selfish” agents (West et al., 2009; Henrich & Henrich, 2007; Milinski & Wedekind, 2007b). We divide the direct fitness explanations for 1998; Milinski et al., 2002; Nowak & Sigmund, 2005; cooperation into two categories: byproduct benefits and Palameta & Brown, 1999; Seabright, 2004; Trivers, 1971; enforcement (Fig. 1). Wedekind & Milinski, 2000). However, reciprocity is thought to be generally unimportant in other organisms, 5.2.1. By-product benefits which lack the cognitive capacity of humans (Bergmüller First, the direct benefits of cooperating may flow et al., 2007; Clutton-Brock, 2002, 2009; Hammerstein, 2003; automatically (passively) as a by-product of helping another Russell & Wright, 2008; Stevens & Hauser, 2004; Whitlock individual (Darwin, 1871; Chapter 3). Coordinated foraging et al., 2007). Even classic text book examples such as blood in groups appears to be an example of this, where everyone sharing in vampire bates (Wilkinson, 1984) can be explained gains an immediate benefit from increased acquisition of more simply without the need for reciprocity by mechanisms food, such as in African wild dogs. A more complicated such as by-product benefit (Clutton-Brock, 2009). Overall, S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 239 after 40 years of enthusiasm, there is a lack of a clear et al., 2008). Enforcement could also be favoured if it example of reciprocity in a non-human species, and so, it is provides an indirect fitness benefit (El Mouden et al., clearly not a major force outside of humans. 2010; Frank, 1995a; Gardner & West, 2004; Lehmann In contrast, there is increasing empirical support for a et al., 2007c; Ratnieks, 1988). An example of this is range of other mechanisms that enforce cooperation (see provided by species of ant, bee, and wasp, where workers supplementary material). These other possibilities have been selectively cannibalize or “police” eggs laid by workers so termed punishment, policing, sanctions, partner switching that resources can instead be invested into the offspring and partner choice (Bergmüller et al., 2007; Frank, 2003; of the queen, to whom they are more related (Ratnieks Sachs et al., 2004; West et al., 2007a). Empirical examples et al., 2006). include dominant female meerkats evicting subordinates that try to breed (Young et al., 2006), Superb Fairy Wrens 5.3. Interactions and the origins of cooperation punishing subordinates that don't help (Mulder & Langmore, 1993), cleaner fish clients punishing and avoid cleaners who Although we have emphasised how the different take a bite of their tissue (Bshary, 2002; Bshary & Grutter, mechanisms favouring cooperation can be divided up, 2002; Bshary & Schäffer, 2002), soybeans cutting off the there is considerable scope for interactions between them. supply of oxygen to rhizobia bacteria that fail to supply them In particular, many of the direct fitness benefits can also with Nitrogen (Kiers et al., 2003), a range of pollinator provide an indirect benefit if directed at relatives. Bypro- mutualisms where the plants abort overexploited flowers duct mechanisms such as group augmentation involve (Goto et al., 2010; Jander & Herre, 2010; Pellmyr & Huth, individuals gaining a direct benefit from larger group size; 1994), and the policing of worker laid eggs in the social however, they will also gain an indirect benefit if their insects (Ratnieks et al., 2006). group includes relatives, as will often be the case. Enforce- ment mechanisms can be selected for on the basis of either 5.2.3. Why enforce? direct or indirect fitness benefits. Indeed, such mechanisms While it is clear that enforcing behaviours such as of enforcement cut across the direct/indirect fitness punishment or policing favour cooperation, it is some- distinction because they can alter the relative cost and — times less obvious why the actual punishment or policing will benefit of cooperating the b and c terms of Hamilton's be favoured by selection. If behaviours such as punishment rule (Lehmann & Keller, 2006). are costly, then they themselves represent a second-order Different selective forces may be involved in the origin public good, and so individuals could be selected to avoid the and then subsequent elaboration/maintenance of a trait. In cost of punishment. A possible solution to this is the many cases where there could eventually be a direct fitness punishment of individuals who refuse to punish cheats, but benefit to cooperation, it can be hard or impossible for this just moves the problem up another level because cooperation to spread initially, because to not cooperate punishment of nonpunishers represents a third-order public (defection) is also an ESS. This is for instance the case with good (Henrich & Boyd, 2001; Sober & Wilson, 1998). direct reciprocity (Axelrod & Hamilton, 1981), indirect This problem has been solved by a number of reciprocity (Panchanathan & Boyd, 2004), punishment theoretical and empirical studies showing how enforcing (Gardner & West, 2004; Henrich & Boyd, 2001), group behaviours can provide a direct or indirect benefit. The augmentation (Kokko et al., 2001), and costly signalling simplest way in which punishment could provide a direct (Gintis et al., 2001). In cases where these processes are fitness advantage is if it led to the termination of invoked, it is therefore likely that cooperation initially arose interactions with relatively uncooperative individuals due to factors such as indirect fitness benefits or shared (ostracism) and, hence, allowed interactions to be focused interests, and that only after this do mechanisms such as on more cooperative individuals (Frank, 2003; Murray, reciprocity or punishment select for higher levels of 1985; Schuessler, 1989; West et al., 2002b). This cooperation, even when relatedness falls to zero. So, for mechanism appears to be operating in cases discussed humans, it may be unnecessary to prove how cooperation above such as the cleaner fish, pollinator mutualisms and can arise de novo in unrelated populations if it originated in soybeans. In meerkats, pregnant subordinates will kill a hominid that lived in groups of relatives. other young, even those of the dominant, and so, the dominant increases the survival of her offspring by 6. Common Misconceptions harassing and evicting pregnant subordinates (Young & Clutton-Brock, 2006). A more complicated possibility is In this section, we briefly run through sixteen common that the punished individuals change their behaviour in misconceptions about social evolution theory, which are response to punishment and are more likely to cooperate summarised in Table 3. There is some overlap and repetition with the punisher in future interactions (Clutton-Brock & between sections, partly because multiple misconceptions Parker, 1995). This mechanism is at work in cleaner fish, are made in the same areas of research and partly because as described above, and could be important in species such we wish that each can be read relatively independently. as cooperative breeding vertebrates or humans (Gächter Further misconceptions on the issue of whether and why 240 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 humans are special are discussed in Section 7.1. The particularly large problem with the term altruism (West interested reader is also directed towards the “Twelve et al., 2007b;p.419–423), which has been redefined in misunderstandings of kin selection,” of Dawkins (1979), evolutionary models in many ways, including: (a) a many of which are still pertinent today. decrease in fitness over the short term, so that reciprocity is “reciprocal altruism” (Becker, 1974; Fehr & Fischbacher, 2003; Trivers, 1971); (b) a decrease in the fitness of the 6.1. Kin Selection, Reciprocity, and Altruism focal individual, relative to the other members of its group 6.1.1. Misconception 1: The various redefinitions of (relatively costly to individual, relatively beneficial to the altruism (Baschetti, 2007; Becker, 1974; Bergstrom, 1995, group; sometimes termed weak altruism)(Baschetti, 2007; 2002; Bowles, 2006, 2009; Bowles & Gintis, 2004, 2008; Bergstrom, 1995; Bowles, 2006; Bowles & Gintis, 2004; Boyd et al., 2003; Fehr & Fischbacher, 2003; Gintis, 2000; Boyd et al., 2003; Gintis, 2000; Sober & Wilson, 1998; Sober & Wilson, 1998; Trivers, 1971; Wilson, 1975a) Wilson, 1975a); (c) playing cooperate in a prisoners' In Section 3 we emphasized how terms such as altruism dilemma game (Bergstrom, 2002); (d) a failure to harm have very specific meanings, that have formal justification others (Field, 2001); (f) giving up resources in order and convey useful information. If these terms are misused, to benefit others (Pradel et al., 2009); (g) the mechanism by or redefined, the result is confusion. This has been a which one individual is motivated to help others (Axelrod,

Table 3 Sixteen common misconceptions about social evolution theory Misconception Reality 1. The various redefinitions of altruism. Many behaviours that have been described as altruism actually involve a net direct fitness benefit, and so are mutually beneficial, not altruistic. The jargon associated with redefining altruism often obscures the underlying selective forces. 2. Kin selection and reciprocity are the major In the context of reciprocity, cooperation is not altruistic, and there are many other mechanisms by which competing explanations for altruism in cooperation can be favoured due to direct fitness benefits (Fig. 2). biological theory. 3. Mutually beneficial cooperation is less Mechanisms to provide direct fitness benefits to (mutually beneficial) cooperation can often be much more interesting. complicated, from both a theoretical and empirical perspective, than indirect benefits, which can arise through relatively simple processes such as limited dispersal or kin discrimination. 4. Proximate and ultimate explanations. Proximate answers cannot provide a solution to ultimate problems. 5. Kin selection requires kin discrimination. A sufficiently high relatedness can also arise through limited dispersal. 6. Relatedness is only high between members If there is population structuring (viscous populations or limited dispersal), then relatedness can be relatively of the nuclear family. high between group members who are not close kin. 7. Kin selection only applies to interactions Indirect fitness benefits can accrue if cooperation is directed towards non-relatives who share the same between relatives and greenbeard genes cooperative gene. Such “greenbeard” mechanisms are unlikely to be important in humans. can explain cooperation in humans. 8. Greenbeards are a type of costly signaling. Greenbeards and costly signalling are two different things. 9. Group selection is a formal theory with Group selection is used to mean at least four different things. one meaning. 10. Group selection can apply in situations Group selection and kin selection are simply different approaches to describing the same biological when kin selection cannot explain process. cooperation. 11. Kin selection is a subset of group No group selection model has ever been constructed where the same result cannot be found with kin selection selection. theory. The reverse is not necessarily true. 12. Group selection leads to group Group selection will only lead to group adaptations in the special circumstances where either: (a) the group is adaptations. composed of genetically identical individuals (clonal groups, r=1), or (b) there is complete repression of competition between groups (i.e., no conflict within groups). 13. Most evolutionary biologists view group The reason that most evolutionary biologists, both theoretical and empirical, do not use the group selection selection as completely wrong, or that approach is simply that it is less useful, and if they express negative views, it is because it has generated more there is some ulterior motive for the lack of confusion than insight. attention given to it. 14. Human cooperation in economic games The simplest explanations for cooperating and punishing in one-shot encounters are individuals making requires the novel evolutionary force of mistakes and/or that it is a byproduct of selection for cooperation in other conditions. strong reciprocity. 15. The theoretical models on strong The theoretical models of strong reciprocity work upon standard direct and indirect fitness benefits. reciprocity provide a novel solution to the problem of cooperation, that are outside of the usual inclusive fitness explanations. 16. The claims made in the empirical and the The work on strong reciprocity can be divided into four areas – what the empirical data show, what it is theoretical strong reciprocity literature are argued the empirical data show, what the theoretical models show, and what it is argued the theoretical compatible. models show. All of these four areas are in disagreement with each other. S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 241 1984); (h) the willingness to take mortal risks as a fighter like to give the impression that an evolutionary definition (Bowles, 2009). is the only valid one. However, in all the cases discussed The first problem with these redefinitions is that they above, the authors are considering the evolution and lack a formal justification to use intentional language from maintenance of cooperation or altruism, with reference to an evolutionary or ultimate perspective. This is because they the evolutionary literature. An even greater problem is require the costs and benefits to be defined in different ways when papers mix up definitions, starting with a statement and not with respect to lifetime reproductive success. As of how altruism (or spite) poses a problem for evolutionary discussed in Sections 2 and 3, natural selection produces theory (which is true based an evolutionary definition) but organisms that behave intentionally, as maximizing agents, then actually focus on altruistic behaviours as defined by at the level of lifetime reproductive success. motivational or mechanistic definition and, so, where the The second problem is that these redefinitions include evolutionary problem doesn't necessarily apply (Miscon- scenarios where cooperation could provide a direct fitness ception 4; West & Gardner, 2010). benefit, and hence be either mutually beneficial (+/+) or Finally, some confusion over terminology may also altruistic (−/+). Considering a specific case, Gintis (2000) have arisen from the Dawkins title “The Selfish Gene” compared the relative fitness of two different strategies: (Dawkins, 1976) because he defined terms at a different “self-interested agents” who do not punish or cooperate, level to which had been done before (i.e., the gene rather and altruistic “strong reciprocators” who cooperate and than the individual). As discussed in Sections 2 and 3, punish noncooperators. He labels strong reciprocators as Hamilton's (1964) use of intentional language (Table 2) altruistic because they “increase the fitness of unrelated followed from the idea that individuals should appear as individuals at a cost to themselves.” However, in this and maximizing agents and, hence, defined behaviours such as related models, cooperation is individually costly within altruism and selfishness according to their direct con- the social group but provides a benefit to all the members sequences for individuals (Grafen, 1999, 2007a). If this of the group through mechanisms such as increased same logic is applied to genes, then selection could favour productivity or reducing the rate of group extinction genes that are selfish or altruistic or mutually beneficial or (Bowles & Gintis, 2004; Bowles et al., 2003; Boyd et al., spiteful. However, Dawkins defined genes as selfish not 2003; Gintis, 2000; Gintis, Bowles, Boyd, & Fehr, 2003; from the perspective of a single copy of a gene found in an Henrich & Boyd, 2001). Consequently, any individual that individual, but from the perspective of all copies of that behaves cooperatively also gains this (direct) benefit, gene. In this case, as selection only favours genes that which can outweigh the cost of performing the behaviour increase in frequency, it can only favour genes that are (Binmore, 2005b; Burnham & Johnson, 2005; Lehmann et selfish (at the level of every copy of that gene) (Burt & al., 2007c; West et al., 2007b). This leads to the confusing Trivers, 2006). This would be analogous to the situation situation where: (a) cooperation can be favoured because it that would arise had Hamilton defined terms such as provides a direct benefit to the cooperator because it altruism at the level of the inclusive fitness of the increases the chance they and the rest of their group individual, in which case, because natural selection favours survive, but this is defined as altruistic rather than in their traits that lead to an increase in inclusive fitness, these self interest (West et al., 2007b); (b) a “selfish agent” traits would always be defined as selfish (at the level of (Bowles & Gintis, 2004) can have a lower direct fitness inclusive fitness). For social scientists in the 1970s, a than an altruist. misconceived view that “selfish genes” referred to an A general issue here is that redefinitions of altruism individual's gene copies appeared to support the econo- obscure the fundamental distinction between when direct mists description of individuals as purely “self-interested.” or indirect fitness benefits are required to explain the As the selfishness axiom was effectively challenged in observed cooperation (Dawkins, 1979; Smuts, 1999; West economics, so it was assumed that evolutionary theory too et al., 2007b). This can lead to the situation where a was unable to explain human sociality. This was the origin behaviour is described as altruistic but can be explained by of many of the misconceptions and “new” evolutionary direct fitness benefits (i.e., by self-interested or self- explanations for human behaviour we discuss. regarding behaviours). This also clouds the relation to other research. For example, the models discussed in the 6.1.2. Misconception 2: Kin selection and reciprocity are the above paragraph are related to models of group augmen- major competing explanations for altruism in biological tation (Section 5.2.1), where cooperation has been argued theory (e.g., Boyd & Richerson, 2005; Boyd et al., 2003; to provide both direct and indirect benefits. An analogous de Waal, 2008; Fehr & Gächter, 2002; Fehr & Rockenbach, example from the economics literature is the confusion that 2003; Fehr & Fischbacher, 2003; Fehr & Rockenbach, has arisen from the multiple redefinitions of the term 2004; Gintis et al., 2005b; Henrich & Boyd, 2001; “social capital” (Binmore, 2005b; Manski, 2000). Richerson & Boyd, 1999; Schloss, 2002; Silk, 2002) We appreciate that terms can have different meanings in This is wrong on two counts. First, reciprocity is not different fields, such as the motivational definition of altruistic—it provides a direct fitness advantage to cooperat- altruism in the psychology literature, and we would not ing. If an individual does not pay the cost of cooperation in 242 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 the short term then it will not gain the benefit of cooperation emphasized in Section 4, it is actually the evolution of in the long term. Consequently, cooperation is only favoured cooperation that is the central problem of sociobiology (see (between nonrelatives) if it leads to an overall benefit, in supplementary material). which case it is mutually beneficial (+/+). Much of the “ ” confusion here is due to the term reciprocal altruism of 6.3. Proximate and ultimate explanations Trivers (1971, 1985) the introduction of which was accompanied by multiple redefinitions of altruism (West Misconception 4: Proximate explanations provide a et al., 2007b, p. 420). It was for these reasons that Hamilton solution to the ultimate problem of cooperation. (1996, p. 263) thought that reciprocal altruism was mis- It is useful to distinguish between ultimate and proximate named, and several authors have used less confusing explanations of traits or behaviours (Mayr, 1961; Tinbergen, alternatives such as “reciprocity” or “reciprocal cooperation” 1963). Proximate explanations are concerned with the causal (Alexander, 1974; Axelrod & Hamilton, 1981; Binmore, mechanisms underlying a behaviour (how questions). 1994, 1998; West et al., 2007b). Ultimate explanations are concerned with the fitness Second, when considering explanations for cooperation, consequences of a behaviour (why questions). Evolutionary the major competing hypotheses are not kin selection and biology attempts to explain features of an organism from an reciprocity. Reciprocity is only one of the many ways in ultimate perspective—why are organisms the way they are? which cooperation can lead to direct fitness benefits (Fig. 2), The key point is that these different methodologies are and while it appears to be important in humans, it is complementary and not competing alternatives. relatively unimportant in other species. In some cases, this The Nobel Prize winner Niko Tinbergen (1963) famously misconception appears to arise from only considering the clarified the distinction between ultimate and proximate evolutionary literature up to approximately the late 1970s explanations for animal behaviour, in the most influential and, hence, missing the huge advances that have been made paper of his career (Kruuk, 2003); less well known to many since then (sometimes referred to as the “disco problem”). As biologists is that Niko's brother Jan won the 1969 Nobel well as in the papers cited above from the primary literature, memorial prize in Economics. One of Tinbergen's classic Misconception 2 or a close approximation occurs in a scarily studies to illustrate this distinction was on the removal of large number of undergraduate textbooks. eggshells from their nests by black-headed gulls. The mechanistic (proximate) explanation for this is that indivi- 6.2. Mutually beneficial cooperation duals are more likely to remove objects from their nest if they are white- or egg-coloured, have frilly edges, and if Misconception 3: Mutually beneficial cooperation is they are feather-light. The evolutionary (ultimate) explana- less interesting tion for this is that it makes aerial predators such as herring Misconception 1 illustrated the point that altruism is often gulls less likely to find their brood. These explanations are redefined so that it will include a particular case of clearly not competing (each answer cannot provide a cooperation that is being examined. Furthermore, researchers solution to the other problem), and a fuller understanding are often disappointed to discover particular cases fit into the is gained by considering both. mutually beneficial category (+/+) and are not altruistic (−/ A clear example of the confusion that may be caused by +). Indeed, altruism may be redefined so frequently because conflating ultimate and proximate factors is provided by researchers prefer their research problem to be altruism. This work on “strong reciprocity,” which is defined proximately reflects the common feeling that mutually beneficial but then given as a solution to an ultimate problem (Bowles behaviours are somehow less interesting. We strongly & Gintis, 2004; Fehr & Gächter, 2002; Fehr & Rockenbach, disagree. Indeed, mechanisms to provide direct fitness 2003, 2004; Fehr & Fischbacher, 2003, 2004; Fehr et al., benefits to cooperation can often be much more complicated, 2002; Gintis et al., 2003). A strong reciprocator has been from both a theoretical and empirical perspective, than defined as a combination of “a predisposition to reward indirect benefits, which can arise through relatively simple others for cooperative, norm-abiding behaviours” and “a processes such as limited dispersal or kin discrimination. propensity to impose sanctions on others for norm Determining the relative importance of direct and indirect violations” (Fehr & Fischbacher, 2003). This is a description benefits remains a key problem and has long been a major of a proximate mechanism. However, it is then given as a topic of debate in areas such as the evolution of helping in solution to an ultimate problem—for example, “Strong cooperative breeding vertebrates (Clutton-Brock, 2002; reciprocity thus constitutes a powerful incentive for Cockburn, 1998; Griffin & West, 2002; Jennions & cooperation even in nonrepeated interactions when reputa- Macdonald, 1994). A contributing factor here may be the tion gains are absent” (Fehr & Fischbacher, 2003), or often quoted statement from the sociobiology book of “cooperation is maintained because many humans have a Wilson (1975b, p.31) that: “the central theoretical problem predisposition to punish those who violate group-beneficial of sociobiology [is]: how can altruism, which by definition norms” (Bowles & Gintis, 2004). reduces personal fitness, possibly evolve by natural This is illustrated even more clearly with a discussion of selection?” (Becker, 1974). This is misleading because, as neurological work, where it is suggested that an S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 243 explanation for the punishment of individuals who do not 6.4. Inclusive Fitness, Kin Selection, Relatedness cooperate is that such punishment leads to “satisfaction” and Greenbeards (Fehr & Rockenbach, 2004; Quervain et al., 2004). For example, in two adjoining sentences, Quervain et al. (2004, There are three related misconceptions about how a p. 1254) follow an ultimate question “Why do people significant relatedness and indirect fitness benefits (kin punish violators of widely approved norms although they selection) can occur. ” reap no offsetting material benefits themselves? with a 6.4.1. Misconception 5: Kin selection requires “ proximate answer We hypothesize that individuals derive kin discrimination ” satisfaction from the punishment of norm violators. This In his original papers on inclusive fitness theory, does not solve the ultimate problem because it does not Hamilton pointed out a sufficiently high relatedness to answer why evolution should have produced a psychology favour altruistic behaviours could accrue in two ways—kin or nervous system that mechanistically encourages discrimination or limited dispersal (Hamilton, 1964, 1971, (rewards) such punishment. 1972, 1975). There is a huge theoretical literature on the This approach mixes up two different questions (how possible role of limited dispersal reviewed by Platt & Bever and why, or process and product). Claiming that cooper- (2009) and West et al. (2002a), as well as experimental ation is favoured because individuals have a predisposition evolution tests of these models (Diggle et al., 2007; Griffin to cooperate, and punish those that do not, is circular, as it et al., 2004; Kümmerli et al., 2009). However, despite this, does not explain why individuals should have a predispo- it is still sometimes claimed that kin selection requires kin sition to cooperate and punish in the first place. The discrimination (Oates & Wilson, 2001; Silk, 2002). Further- proximate question is how is cooperation maintained? The more, a large number of authors appear to have implicitly or answer to this is a predisposition to cooperate and avoid explicitly assumed that kin discrimination is the only punishment, i.e., what has been termed strong reciprocity. mechanism by which altruistic behaviours can be directed The ultimate question is why is cooperation maintained, or towards relatives and have reinvented the role of limited more specifically, why are cooperation and punishment dispersal, usually calling it something else, and claiming that (strong reciprocity) maintained? The possible answer to this indirect fitness, kin selection or relatedness is not important is because it provides either a direct and/or an indirect (Table 4). fitness benefit (Gardner & West, 2004). We are not arguing that proximate questions are not interesting, and we 6.4.2. Misconception 6: Relatedness is only high between appreciate that they are, with good reason, the focus of close family members (Bowles & Gintis, 2004; Boyd & much human research. Instead, our point is that it is very Richerson, 2005; Gintis, 2000) misleading to mix and match by posing and justifying a It is sometimes implicitly assumed in the theoretical problem from an ultimate perspective and then providing a literature that relatedness can only be high between close proximate answer. family relatives. One example is the various strong Similar confusion over proximate and ultimate factors reciprocity theoretical models where it is argued that kin occurs in numerous other places. One example is “social selection is not important (e.g., Bowles & Gintis, 2004; institution” models, where selection for cooperation is Gintis, 2000), but then limited dispersal is assumed of a form increased by “the commonly observed human practices of that can lead to a substantial relatedness between interacting resource sharing among group members” (Bowles, 2006; individuals (Lehmann et al., 2007c; West et al., 2007b) (see Bowles et al., 2003). However, as an institution is a form of also Misconception 15). Another example, is provided by the cooperation itself, it just provides a proximate answer claim that group selection is an alternative mechanism that (cooperation is explained by cooperation) that avoids the explains cooperation between non-relatives but that it only ultimate problem of why would the social institution of works when “groups are small and migration infrequent” cooperative resource sharing ever evolve? This question can (Boyd et al., 2005, p. 215), without realising that this is when be addressed with models which assume that mechanisms for relatedness is high (see also Misconceptions 9–13). repressing competition within groups are potentially costly These conclusions appear to be based on the well-known traits under selection (El Mouden et al., 2010; Frank, 1995a, approximation that relatedness is approximately r=1/2 1996a, 2003; Leigh, 1971; Ratnieks, 1988). Another between full siblings, r=1/4 between half siblings, r=1/8 example is the suggestion that “adults may support their between cousins, etc. However, these are only approxima- parents in order to imprint a corresponding behavior pattern tions for large well-mixed populations, and the formal on their own children” (Bergstrom, 1996). This is a definition of relatedness is a statistical measure of genetic proximate answer, and does not answer why such imprinting similarity (Box 1). If there is population structuring with would be favoured. Similar mixing up of proximate and limited migration (viscous populations or limited dispersal), ultimate factors occur in the literature on the evolution of then relatedness between group members can be relatively language (Scott-Phillips, 2007), the group selection literature high because it will tend to increase the genetic similarity (Smuts, 1999) and at the interface of the primate and human between interacting individuals (Hamilton, 1964, 1970, literature (de Waal, 2008). 1971, 1972, 1975). To give a specific example, consider a 244 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

Table 4 Some examples of the reinvention of how limited dispersal increases relatedness between interacting individuals, and can hence provide an indirect fitness benefits (kin selection) for cooperation Suggested explanation for cooperation Authors suggesting explanation Authors showing that the suggestion is a reinvention of kin selection via limited dispersal Games in spatial settings Nowak et al., 2010 Lehmann & Keller, 2006 Spatial structure MacLean & Gudelj, 2006; Pfeiffer Frank, 1998; Frank, 2010 et al., 2001 Group or multilevel selection Nowak, 2006; Traulsen & Nowak, 2006; Frank, 1986; Grafen, 1984; Lehmann et al., 2007b; Queller, 1992a Wilson, 1975a Population structure Killingback et al., 2006 Grafen, 2007c Network reciprocity via games Lieberman et al., 2005; Nowak, 2006 Grafen, 2007a; Grafen & Archetti, 2008; Lehmann et al, 2007a; on graphs Taylor et al., 2007a Strong reciprocity Bowles & Gintis, 2004; Gintis, 2000 Lehmann & Keller, 2006; Lehmann et al., 2007c

population split into groups of size 100, and where 1% of using coancestry to measure relatedness in natural popula- individuals disperse from their natal patch before breeding. tions? No. The most common method by which empirical In this case, the increased genetic similarity that results biologists measure relatedness is to use molecular markers from population structuring will lead to the average such as microsatellites, and then plug the data from those relatedness of group mates being approximately one third into programmes such as Kinship, which estimates (see supplementary material). Hence, the relatedness relatedness with the statistical definition (Queller & between first cousins will be more than one third, and Goodnight, 1989). The extent to which the statistical not the commonly assumed one eighth. Clear quantitative measure of relatedness is used by empirical biologists is support for the effects of population structure on clear from the fact that the Queller and Goodnight (1989) relatedness have been provided by experimental evolution methods paper has been cited N1200 times (Web of Science studies with bacteria (Brockhurst et al., 2007; Griffin et al., search, September 2010). Second, it is true that introduc- 2004; Kümmerli et al., 2009). tory animal behaviour textbooks such as Krebs and Davies The above discussion for Misconceptions 5 and 6 rest (1993) and Alcock (2005) define relatedness through co- upon the understanding that relatedness is a statistical ancestry and not statistically. However, the coancestry measure of genetic similarity (Box 1). It is sometimes definition is a useful approximation for teaching certain argued that relatedness was originally a simple measure of age groups of undergraduates. The primary literature needs genealogical relationship and that evolutionary theoreticians to build upon and relate to the primary literature, not to later reinvented it as a more general measure of genetic introductory textbooks. similarity, either in the 1980s (e.g., by Grafen 1985) or later (e.g., by Lehmann & Keller, 2006). However, this is 6.4.3. Misconception 7: Inclusive fitness only applies to completely incorrect. In his original papers, Hamilton interactions between relatives, and greenbeard genes can made clear that what mattered was genetic similarity per explain cooperation in humans (Bergstrom, 1995, 1996, se, discussing relatedness in terms of a regression coefficient 2002; Bowles & Gintis, 2004, 2008; Boyd & Richerson, (Hamilton 1963, p. 355) and possible green beard effects 2005; Frank, 1987; Gintis, 2000; Robson, 1990) among genealogically unrelated individuals (Hamilton, This follows on from the previous two misconceptions, 1964,p.24–25). He then went on to formalise this in his and is wrong on three counts. First, as discussed in Section 2, 1970 Nature paper (Hamilton, 1970, 1975; Michod & inclusive fitness is a very general encapsulation of Hamilton, 1980), providing the regression definition of evolutionary theory, not a special case; it applies equally relatedness that is at the centre of modern social evolution well to social and nonsocial characters. Second, as discussed theory (Frank, 1998; Gardner et al., Submitted; Grafen, in Misconception 6, relatedness can be high between 1985, 2006a; Taylor & Frank, 1996). As well as the huge individuals who are not close family members. primary literature on this issue, the fact that it is genetic Third, as pointed out by Hamilton in his original similarity that matters was also made clear in the formulation of inclusive fitness, indirect fitness benefits popularisations of inclusive fitness theory of Dawkins can accrue if cooperation is directed towards non-relatives (1976, 1982). Relatedness and inclusive fitness theory who share the same cooperative gene (Hamilton, 1964; have not been reinvented—the modern interpretation is p. 24–25). Dawkins (1976, 1982) illustrated this with a that developed by Hamilton in the 1960's. hypothetical example of a gene that causes its bearer to Two other points are worth considering here. First, how grow a green beard and also to preferentially direct do empirical biologists approach the concept of related- cooperation towards other green-bearded individuals. This ness? Is the statistical (regression) definition of relatedness mechanism can also occur without a visible tag, for purely a theoretical concept, with empirical biologists example, if the cooperative gene also causes some effect S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 245 on habitat preference that leads individuals who carry the Second, some (but not all) models of “strong gene to settle close together (Hamilton, 1964, 1975). reciprocity” assume that helping and punishment are Consequently, although this mechanism is usually termed a completely linked traits (Bowles & Gintis, 2004; Gintis, greenbeard, it more generally represents an assortment 2000). In these strong reciprocity models, the benefit of mechanism, requiring a single gene—or a number of helping has no influence on selection for strong tightly linked genes (e.g., physically close on a chromo- reciprocity because it is cancelled out by the increased some and so not separated during sexual reproduction by kin competition that is generated by the act of helping recombination)—that encodes both the cooperative behav- (Lehmann et al., 2007c). Instead, strong reciprocity is iour and causes cooperators to associate (Gardner & West, selected for, because helping acts as a tag of who is 2010; Lehmann & Keller, 2006). One way of conceptua- carrying the punishment allele, and so, punishment can be lising greenbeards is that they are an extreme end point on directed at individuals who do not carry that allele, the genetic kin discrimination continuum, with no reducing competition for individuals who do carry this recombination between the tag and helping loci (Rousset allele. Consequently, in contrast to the verbal claim that & Roze, 2007). these models are examining the evolution of cooperation Greenbeard genes are likely to be extremely rare in the (Bowles & Gintis, 2004; Gintis, 2000), they are actually real world (Gardner & West, 2010; West & Gardner, 2010). examining the evolution of spiteful greenbeards (Lehmann The idea of greenbeards was initially developed as a thought et al., 2007c)! Furthermore, not only is selection driven experiment to illustrate that what matters for inclusive fitness by indirect fitness consequences, but the trait is costly to is genetic similarity at the locus (or loci) being considered, the group — this is the exact opposite of what is claimed rather than genealogical relationship per se (Hamilton, 1964, verbally in the original papers. The confusion that the can 1970, 1971, 1975). It was assumed that that greenbeards be caused by a such a mismatch between how a model would be unimportant in the real world because cheaters, works, and how it is claimed to work, is nicely illustrated which display the green beard or assorting behaviour without by the fact that Fehr & Fischbacher (2005a) cite Gintis also performing the cooperative behaviour, could invade and (2000) as showing how strong reciprocity can favour overrun the population (Dawkins, 1976, 1982). To date, only cooperation in humans in a paper where the main focus five examples of possible greenbeard genes have been found was to argue that greenbeards cannot explain cooperation in nature, three cooperative and two spiteful, four in in humans. microbes and one in an ant (Gardner & West, 2010). The feasibility of greenbeard genes is greatest in simpler 6.4.4. Misconception 9: Greenbeards are a type of costly organisms, such as bacteria, where there can be a relatively signaling (Henrich, 2004; Owren & Bachorowski, 2001) simple link between genotype and phenotype and, hence, Greenbeards and costly signalling are two different the possibility that a single gene could have the required things. As discussed above, the greenbeard mechanism multiple (pleiotropic) effects. involves a trait and a tag being encoded by the same gene, Models for the evolution of cooperation that rely upon or tightly linked genes (i.e., genetic linkage prevents lying). greenbeards are unlikely to be important in humans (Fehr & In contrast, costly (or honest) signalling is the idea that Fischbacher, 2005a, 2005b; Gardner & West, 2010; Henrich, signalling can be evolutionary stable if the signal is costly 2004). This is because the polygenic nature of behaviours and cannot be faked (i.e., lying is too costly; Grafen, 1990a; would readily allow the evolution of cheats who displayed a Spence, 1973). For example, if cooperative behaviours are tag or performed the assortative behaviour, but did not costly, then cooperation could function as a signal of quality cooperate. Despite this, two classes of models of cooperation because individuals in better condition would be able to in humans have been proposed which rely upon a greenbeard behave more cooperatively (even though, in principle, mechanism, and which are therefore based upon an unlikely anyone could perform cooperative behaviours; Gintis et al., and evolutionary unstable assumption. In both cases the 2001). This is further illustrated by considering the smiling assumption of a greenbeard mechanism was implicit and not andlaughingexamplediscussedabove(Owren & realised by the original authors. First, it has been suggested Bachorowski, 2001). In order for laughing and smiling to that individuals who cooperate differ from individuals who be favoured as a signal of cooperative behaviour via a cheat in “some observable characteristic” other than the greenbeard mechanism, we would require that laughing and cooperation phenotype itself (Amann & Yang, 1998; Frank, smiling be controlled by the same gene(s) (or tightly linked 1987; Robson, 1990). This represents the original green genes) as cooperative behaviours. In contrast, for smiling beard scenario, which is unlikely to work in humans, as and laughing to be favoured as a signal of cooperative described above. Owren & Bachorowski (2001) provide a behaviour via a costly signalling mechanism, it would more specific version of this scenario, where the observable require that laughing and smiling are too costly for characteristic is smiling and laughter. However, there is no individuals who have chosen not to cooperate. This also reason to expect genes for cooperative behaviours to be seems unlikely—given that laughing and smiling are likely tightly linked to, or the same as genes that control smiling to be relatively cost free, it seems more likely that laughing and laughter. and smiling act as a signal or bond between individuals 246 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 with a shared interest. A similar argument can be made this idea. Theory showed that this type of group selection about blushing, which is more easily described as mutually would only work under extremely restrictive conditions, beneficial signal of appeasement (Crozier, 2001). An and so, its importance would be rare or nonexistent (Leigh, analogous problem occurs in the evolution of language 1983; Maynard Smith, 1964, 1976; Williams, 1966). For literature when it is suggested that traits such as politeness example, selection will produce behaviours that maximize are costly honest signals, but where the costs and signal are group success if all the individuals within a group are dissociated, and arise from later behaviours such as genetically identical clones, or if there is complete reciprocity (Knight, 1998, 2008; van Rooy, 2003; see also repression of competition within groups, such that the Scott-Phillips, 2007; Scott-Phillips, 2008). reproductive success of members of the group cannot differ Howard's (1971) metagames with “transparent disposi- (Gardner & Grafen, 2009). These correspond to the tion” and Gauthier's (1986) theory of “constrained maximi- extreme cases where maximizing group success is the zation” are also relevant here. In these cases, it is assumed same as maximizing inclusive fitness (Fig. 4). Empirical that the second player in a one shot PD can choose a fixed work supported these theoretical conclusions by showing disposition (e.g., always defect, always cooperate, play tit- that individuals were reproducing at the rate that for-tat, etc.) that can be detected by the first player and that maximized their inclusive fitness and were not adapted to the first player can adjust their strategy accordingly. Given maximize group fitness (Krebs & Davies, 1987, 1993; that the second player can predict what the first player will do Lack, 1966; West et al., 2008). depending upon their chosen disposition, the second player It is this old form of group selection that leads people to can choose the disposition that will lead to the maximum the false conclusion that individuals behave for the good of payoff (backward induction). The assumption here is that the population or species or ecosystem, or that human disposition can be chosen facultatively, and so, in order for societies can be viewed as superorganisms in the same way this to work, disposition must be a costly honest signal, as certain social insect colonies (Kohn, 2008; Shennan, which seems very unlikely (at least to good politicians and 2002; Wilson et al., 2008; see also the review of the poker players; Binmore, 1994, pp. 174–186). A greenbeard anthropological literature by Soltis et al., 1995). For version of this hypothesis could also be constructed, but this example, as summed up by quotes such as “the concept would require that the outward appearance of disposition be of social groups as like single organisms” (Wilson & controlled by (or strongly linked to) the genes that control O'Brien 2009) and “Our species is the primate equivalent cooperation, which seems even more unlikely. of a beehive or a single organism” (Kohn, 2008). Similar confusion surrounds some discussions of punctuated 6.5. Group selection equilibrium, where it seems to be assumed that this In this section, we summarise the five misconceptions would lead to group-level or species-level adaptations generated by the group selection literature—the interested (Arnold, 1993; Shennan, 2002; Zeder, 2009). reader is directed towards more detailed reviews elsewhere (Gardner & Grafen, 2009; West et al., 2007b, 2008). 6.5.1.2. New group selection. In the 1970s and 1980s, a “new” form of group selection was championed by Wilson 6.5.1. Misconception 9. Group selection is a formal theory and others, which examined the consequences of interac- with one meaning tions in small structured populations (Colwell, 1981; A major part of the confusion surrounding group selection Hamilton, 1975; Wilson, 1975a, 1977). These models stems from the fact that the term has been used to mean at assumed that there are multiple levels of selection, which least three or four different things (Fig. 3; Okasha, 2004, can vary in their importance, and showed that cooperation 2006; West et al., 2007b, 2008). could be favoured if the benefits at the group level (between-group) outweighed the benefits at the individual 6.5.1.1. Old group selection and group adaptations. level (within-group). It was suggested that this new group During the 1960s, Wynne-Edwards (1962) argued for the selection approach provided an alternative explanation to importance of group selection in its original or “old” form. cooperation or altruism, in situations where kin selection or He argued that in groups consisting of selfish individuals, inclusive fitness could not. However, it has since been resources would be over exploited, and the group would go realized that group selection and kin selection were just extinct. In contrast, groups consisting of cooperative different ways of conceptualizing the same evolutionary individuals would not over exploit their resources and, process. For example, while the earliest group selection so, avoid extinction. Hence, by a process of differential models (e.g., Colwell, 1981; Traulsen & Nowak, 2006; survival of groups, behaviours evolved that were for the Wilson, 1975a, 1977) were reinventing how indirect fitness good of the group. Another way of looking at this is that benefits (kin selection) can work via limited dispersal, later selection would favour traits that maximize group success, models (e.g., Wilson & Dugatkin, 1997; Wilson & termed group adaptations. Hölldobler, 2005) were reinventions of the green beard During the 1960s and 1970s, a large amount of process (Dawkins, 1979; Foster et al., 2006; Frank, 1986; theoretical and empirical evidence was piled up against Grafen, 1984; Hamilton, 1975; Harvey et al., 1985; S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 247

A “old” group selection B “new” group selection

C “newer” group selection D cultural group selection

competition T T

Fig. 3. The different types of group selection. The white circles represent cooperators, whereas the grey circles represent relatively selfish individuals who do not cooperate. Panel A shows the “old” group selection, with well-defined groups with little gene flow between them (solid outline). Competition and reproduction is between groups. The groups with more cooperators do better, but selfish individuals can spread within groups. Panel B shows the “new” group selection, with arbitrarily defined groups (dashed lines), and the potential for more gene flow between them. The different groups make different contributions to the same reproductive pool (although there is also the possibility of factors such as limited dispersal leading to more structuring), from which new groups are formed. Panel C shows the “newer” group selection, which emphasises the more proximate mechanism of inter-group competition as a factor shaping the evolution of social behaviours. Panel D shows cultural group selection, in which social behaviours can be horizontally transmitted between group mates, for example with all individuals in the group imitating the behaviour of one “teacher” (T).

Lehmann & Keller, 2006; Lehmann et al., 2007b; Maynard response to within-group selection, but it is straightforward Smith, 1976). to recover Hamilton's rule from this. Both approaches tell The key point here is that this new group selection us that increasing the group benefits and reducing the (multilevel selection) is just a different way of looking at individual cost favours cooperation. Similarly, group the dynamics by which inclusive fitness is maximized. selection tells us that cooperation is favoured if we increase They are mathematically identical (Frank, 1986, 1995b; the proportion of genetic variance that is between-group as Gardner, 2008; Gardner & Grafen, 2009; Gardner et al., opposed to within-group, but that is exactly equivalent to 2007a; Grafen, 1984, 2006a; Hamilton, 1975; Lehmann saying that the kin selection coefficient of relatedness is et al., 2007b; Queller, 1992a; Wade, 1985). New group increased (Frank, 1995a). In all cases, where both methods selection models show that cooperation is favoured when have been used to look at the same problem, they give the response to between-group selection outweighs the identical results (Table 5). This is not surprising given how they are both formalized with the Price equation (Frank, 1986; Gardner, 2008; Gardner et al., 2007a, submitted). As we shall discuss in further detail in Misconception 13, the reason that most biologists focus on the inclusive fitness or kin selection approach is that it is much easier to develop models and apply them to real organisms (West et al., 2008).

Fig. 4. The scope of inclusive fitness theory and group adaptation. Irrespective 6.5.1.3. Newer group selection. More recently, over the of the extent to which selection is within or between groups, natural selection last decade, group selection has been used in a third will lead to organisms that appear to be maximising their inclusive fitness “ ” (Frank, 1986; Grafen, 2006a; Hamilton, 1975). In contrast, individuals will newer way. In these models, it is argued that a key factor only be selected to maximise group fitness in the extreme scenario where there favouring cooperation is direct competition between is negligible within group selection (Gardner & Grafen, 2009). groups, and this is referred to as group selection (Binmore, 248 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

Table 5 Kin selection and new group section Area Papers claiming that a result requires group selection Papers showing equivalent result can be obtained with kin selection / inclusive fitness Sex ratios with local Colwell, 1981; Sober & Wilson, 1998; Wilson & Frank, 1986; Grafen, 1984; Harvey et al., 1985 mate competition Colwell, 1981 Sex ratios with Avilés, 1993 Gardner et al., 2009 budding dispersal Eusociality Wilson & Wilson, 2007; Wilson & Hölldobler, 2005 Foster et al., 2006; Helanterä & Bargum, 2007; Hughes et al., 2008 Strong reciprocity Bowles & Gintis, 2004; Bowles et al., 2003; Boyd & Gardner & West, 2004; Lehmann et al., 2007c Richerson, 2002; Boyd et al., 2005; Fehr & Fischbacher, 2003; Gintis, 2000; Gintis et al., 2003; Henrich, 2004; Traulsen & Nowak, 2006 Cooperation Bowles, 2006; Taylor & Nowak, 2007; Traulsen & Hamilton, 1975; Lehmann et al., 2007b Nowak, 2006 Virulence Kohn, 2008; Sober & Wilson, 1998; Wilson, 2008; Frank, 1996b; Wild et al., 2009 Wilson & Wilson, 2007 Policing Sober & Wilson, 1998 Ratnieks, 1988; Wenseleers et al., 2004 There is no theoretical or empirical example of group selection that cannot be explained with kin selection. Here, we provide examples of situations where it has been argued that group selection gives a result that cannot be explained by kin selection, but where it was then shown that it can. More general theoretical overviews are provided elsewhere (Frank, 1986; Gardner et al., 2007a; Grafen, 1984; Hamilton, 1975; Queller, 1992a).

2005a; Bowles, 2006, 2009; Bowles et al., 2003; Boyd & competition occurs between groups. However, it differs in Richerson, 1990, 2002; Boyd et al., 2003; Gintis, 2003; that it refers to selection on a cultural trait, rather than a Gintis et al., 2003; Henrich, 2004). For example, as genetically determined trait. As with genetic group selection, discussed in Misconception 2, when groups compete for just because competition is occurring between groups, this territories, and territories are won by the groups with the does not mean that group level adaptations are expected to most cooperators. However, these models do not provide evolve (Gardner & Grafen, 2009). Consequently, while it is an alternative to inclusive fitness or kin selection — often argued that the group is the fundamental unit of individuals gain a direct fitness benefit through cooperating cultural evolution, or that cultural evolution is a group- because they increase the success of their group (including level process (Boyd & Richerson, 1985), there is no themselves), and an indirect fitness benefit in the cases formal basis for this. Finally, we also note that it has where the models also assume limited dispersal, which been suggested that there are even three different types of leads to significant relatedness between the individuals in a cultural group selection (Henrich, 2004)! group (see Misconceptions 5, 6 ,and 15). Another distinction is that kin selection, old group selection and 6.5.1.5. The various group selections. The above discus- new group selection are examining the level at which sion shows how the term group selection has been used to ultimate selective forces act, whereas the newer group mean three to six different things (Fig. 3), specifically, that selection is more proximate, saying that group competition (1) selection produces traits that maximize group fitness plays a causal role in mediating the fitness consequences of (old), (2) selection acts at multiple levels (new), or (3) cooperative behaviors. Brewer and Caporael (1990) define competition occurs between groups (newer). The various group selection to mean that the group is the selection forms of cultural group selection could be either subsumed environment for human evolution at the individual level, under newer, or form a new category (“even newer”)or which is analogous but not exactly equivalent to newer categories. This variable use of group selection has been group selection. possible because there is no formal theory of group selection (West et al., 2008, p.380–381; Gardner & Grafen, 2009), 6.5.1.4. Cultural group selection. The term group selec- which leads to authors confusingly switching between tion is also used when discussing “cultural group selection” different meanings (Palmer et al., 1997; Trivers, 1998a, or “gene-culture coevolutionary multilevel selection.” Cul- 1998b; West et al., 2007b, 2008). For example, several tural group selection is used to label situations when authors switch between the old and new group selection, differential group success results from the expression of using the new to justify the old (e.g., O'Gorman et al., 2008; different cultural traits (Boyd & Richerson, 2005, 2010; Fehr Robson, 2008; Sober & Wilson, 1998; Wilson et al., 2008), & Fischbacher, 2005b; Fehr et al., 2002; Gintis, 2003; Gintis while Bergstrom (2002) discusses all three types as if et al., 2003; Henrich, 2004; Henrich & Boyd, 2001; they are the same thing (old: pp. 85–86; new: pp. 71–72, Lehmann et al., 2008; McElreath & Henrich, 2006; 76–77, 80; newer: pp. 81, 85–86). Richerson & Boyd, 2005). This is analogous to the third A lack of an appreciation of the different types of group use described above, in that it is used to mean that selection has led to numerous sources of confusion. These S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 249 include (A) The new group selection approach has been or group selection (Sober & Wilson 1998)” and “Group used to justify old group selection thinking (e.g., Sober & selection can lead to the evolution of altruism only when Wilson, 1998; Wilson, 2008; Wilson & Wilson, 2007). groups are small and migration infrequent (Eshel 1972; Aoki (B) A new group selection approach is used to produce an 1982; Rogers 1990)” (see also Boyd & Richerson, 2002). equation that it is a form of Hamilton's rule, and so of general importance, but then, on the basis of old group 6.5.3. Misconception 11: Inclusive fitness or kin selection is selection thinking, it is suggested that this is unlikely to be a subset of group selection important for genetic traits or outside of humans (Bowles This is incorrect. Kin selection and group selection are et al., 2003, pp. 136-140; Boyd & Richerson, 1990, p. 340; just different ways of carving up the dynamics by which Henrich, 2004, pp. 15–16). This is analogous to saying that inclusive fitness maximisation is reached. Consequently, it is indirect fitness effects are thought to be generally no surprise that no group selection model has ever been unimportant, which is clearly incorrect. (C) The group constructed where the same result cannot be found with kin selection jargon hides links with other areas of evolutionary selection theory (Table 5). Although, while it is possible to theory. For example: (i) how the various group selection translate all group selection models into corresponding kin models with limited dispersal (e.g., Bowles et al., 2003; selection models, the reverse may not be true. One reason for Boyd & Richerson, 2002; Boyd et al., 2005; Traulsen & this is that it can be hard or impossible to incorporate many Nowak, 2006) relate to the inclusive fitness literature on important biological complexities into group selection the same issues (reviewed by Lehmann et al., 2007b; models (Queller, 2004). It is for this reason that group Queller, 1992b; West et al., 2002a, 2008); (ii) that some selection models have focused on the simplest possible models (e.g., Gintis, 2000; Wilson & Dugatkin, 1997) rely cases, whereas the inclusive fitness approach is also used to on greenbeard effects and, so, are unlikely to be of general develop specific models and provide predictions that can be importance, especially in humans (see Misconception 7), tested with empirical work (West et al., 2008). Another and (iii) it can obscure the various mechanisms by which reason is that the inclusive fitness approach has successfully within group competition can be repressed, such as integrated fundamental issues that have not been tackled in reciprocity, punishment, ostracism etc (e.g., O'Gorman the group selection literature, such as the theory of et al., 2008). reproductive value and gene-frequency change in class- structured populations (Frank, 1997, 1998; Taylor, 1990, 6.5.2. Misconception 10: Group selection can apply in 1996; Taylor & Frank, 1996; Taylor et al., 2007b). This has situations when inclusive fitness cannot explain cooperation proven particularly useful for dealing with issues such as (e.g., Arrow, 2007; Baschetti, 2007; Bergstrom, 2002; Boyd different forms of dispersal in spatially structured popula- et al., 2003; Fehr et al., 2002; Gintis et al., 2001, 2003; tions (West et al., 2002a). Henrich, 2004; Richerson & Boyd, Manuscript 1999) This is incorrect. The old group selection ideas only work 6.5.4. Misconception 12: Group selection leads to group in the extreme scenarios where there is negligible within adaptations (Reeve & Hölldobler, 2007; Sober & Wilson, group selection, which can occur via high relatedness or 1998; Wilson & Wilson, 2007; Wilson & Hölldobler, 2005; repression of competition (Fig. 4; Gardner & Grafen, 2009). Wynne-Edwards, 1962) In contrast, individuals are expected to maximise their As discussed in Section 2, Darwinism is a theory of the inclusive fitness irrespective of the relative strengths of process and purpose of adaptation. The purpose is that within-group versus between-group selection (Grafen, natural selection should lead to individuals appearing as if 2006a; Hamilton, 1975). New group selection is not an they were designed to maximize their fitness, and that this alternative to inclusive fitness—it is just a different way of fitness is inclusive fitness (Grafen, 2006a; Hamilton, 1964). looking at the dynamics of natural selection. Finally, the In contrast, since Wynne-Edwards, a number of workers newer group selection is also not in conflict with inclusive have argued that group selection will lead to “group fitness—it is a mechanism for providing direct and/or adaptations” that have been selected for because of their indirect fitness benefits. benefit for the good of the group, and that groups can be A recent example of the confusion that can arise here is viewed as adaptive individuals (superorganisms) in their provided by two quotes from the same paragraph of Boyd own right (Reeve & Hölldobler, 2007; Sober & Wilson, et al. (2005, p.215). It is first claimed that group selection 1998; Wilson, 2008; Wilson & Wilson, 2007; Wilson & works when interactions are not between relatives (this O'Brien, 2009; Wilson & Hölldobler, 2005; Wynne- misconception), but then, stated that group selection only Edwards, 1962). However, formal analysis has shown that favours altruism when groups are small and migration rare selection for group adaptations requires special circum- (i.e., which is when limited dispersal means interacting stances, with negligible within group selection (Fig. 4), such individuals will be highly related—see Misconceptions 5 as when (a) the group is composed of genetically identical and 6): “Cooperation among nonkin is commonly explained individuals (clonal groups, r=1), or (b) there is complete by one of two mechanisms: repeated interactions (Axelrod & repression of competition between groups (i.e., no conflict Hamilton 1981; Trivers 1971; Clutton-Brock & Parker 1995) within groups; Gardner & Grafen, 2009). 250 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 It is useful here to distinguish adaptation and design from attractive, it is highly misleading to portray multilevel dynamics of how selection leads to design. The dynamics of selection as a means to unify the economic and social selection can be examined with either an individual sciences by suggesting that our self-regarding preferences (inclusive fitness or kin selection) or group selection (broadly matching the predictions of classical rational choice approach. However, only the individual level approach theory) are explained by biological individualist selection provides a general model of adaptation. The idea that while population-level (principally cultural) evolutionary individuals strive to maximise their inclusive fitness holds processes explain why we have pro-social preferences (e.g., irrespective of the intensity of selection operating within and Shennan, 2002). between groups (Section 2; Fig. 4). In contrast, as discussed above, group adaptations or maximization of fitness at the 6.6. Strong Reciprocity group level are only expected in the extreme case where In recent years, there has been much attention to the there is no within group selection. suggestion that cooperation in humans can be explained by “strong reciprocity,” which is defined as a predisposition to 6.5.5. Misconception 13: Most evolutionary biologists view help others and to punish those that are not helping (Bowles group selection as hotly debated, completely wrong, or that & Gintis, 2004, 2008; Boyd et al., 2003; Fehr & Gächter, there is some ulterior motive for the lack of attention given 2002; Fehr & Fischbacher, 2003; Fehr & Rockenbach, 2004; to it (Baschetti, 2007; Sober & Wilson, 1998; Traulsen & Fehr et al., 2002; Gintis, 2000, 2003; Gintis et al., 2003, Nowak, 2006; Wilson & Wilson, 2007) 2005a). This literature has contributed to 10 misconceptions, “ This misconception is encapsulated in phrases such as I numbers 1, 2, 4–8 and 14–16. It is useful here to divide the believe that this is a hold-over of US ideologies, which have work on strong reciprocity into four areas—what the ” been strongly individualist and anti-collectivist (Baschetti, empirical data show, what it is argued the empirical data “ 2007), or vigorous criticism and a general denial of such show, what the theoretical models show, and what it is ” ideas (Traulsen & Nowak, 2006). We cannot stress enough argued the theoretical models show. A major source of that this is incorrect. While the old group selection idea does confusion is that all of these four areas are in disagreement not hold (selection does not maximize fitness at the group with each other, and that there are several inconsistencies level except under the very special circumstances described between the different papers on this topic. in Misconception 12), the new or newer ideas are able to A number of elegant economic experiments have explain the dynamics of natural selection. Indeed, many suggested that people have a propensity to cooperate and researchers who normally focus on the kin selection punish individuals who do not cooperate (Burnham & approach, including ourselves, use multi level selection Johnson, 2005; Fehr & Fischbacher, 2003; Gächter & methods when they are the most appropriate tool for solving Herrmann, 2009). Importantly, this includes one-shot games, the problem (Frank, 1998; Gardner et al., 2007a; Gardner & without the possibility for repeated interactions, where Grafen, 2009). The reason that most evolutionary biologists, individuals would gain a greater financial reward from not both theoretical and empirical, do not use the group selection cooperating or punishing. This is a clear demonstration that approach, or use it very little, is that they find it less useful people do not always behave in ways that maximise their (Frank, 1998; Queller, 2004), and if they express negative economic payoffs, even if they are given perfect knowledge. views, it is because it has generated more confusion than It has been argued that strong reciprocity provides an insight (reviewed in detail by West et al., 2007b, 2008). Put explanation for this behaviour (see Misconception 4 for ’ another way, the method isn t wrong per se, it is more that it quotations). is often misused and misinterpreted. The inclusive fitness approach has received more 6.6.1. Misconception 14: Human cooperation in economic attention because it is easier to develop general models and games requires the novel evolutionary force of apply them to real biological situations. It is for this reason strong reciprocity that (a) the group selection debate only takes place over There is a large empirical literature showing that when simple models and has not stimulated empirical work, and humans play anonymous one-shot economic games, they (b) all the major developments in social evolution theory cooperate more than would be expected if they were purely have been pioneered and led by the inclusive fitness self-interested (Ledyard, 1995). From a proximate perspec- approach, and not group selection (Section 2.2; West et al., tive, it has been argued that this is because individuals value 2008). In contrast to this empirical progress spurred by the the success of others as well as their own, showing prosocial inclusive fitness approach, group selection thinking appears preferences (Fehr and Schmidt, 1999). From an evolutionary to be easy to misapply, leading to incorrect statements about perspective, it has been argued that this proximate how natural selection operates, as shown by research in mechanism cannot be explained by standard evolutionary many areas such as animal behaviour (reviewed by Dawkins, explanations of cooperation, such as kin selection and 1976), microbiology (reviewed by West et al., 2006a), reciprocity, and requires the a novel explanation of strong parasitology (reviewed by Herre, 1993) and agriculture reciprocity (Fehr & Gächter, 2002; Fehr & Fischbacher, (reviewed by Denison et al., 2003). While inherently 2003; Fehr & Rockenbach, 2004; Fehr et al., 2002; Gintis S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 251 et al., 2003, 2005b). For example, it has been claimed that from a number of experiments which show that players human behaviour “cannot be rationalized as an adaptive trait taking part in one shot games, in which there are no future by the leading evolutionary theories” (Fehr et al., 2002). interactions, still adjust their level of cooperation in However, the empirical data are open to multiple response to artificial cues, such as the presence of eye- explanations and do not support this claim. First, in some spot pictures on computer desktops (Bateson et al., 2006; cases, an equally valid explanation for the data is that Burnham & Johnson, 2005; Haley & Fessler, 2005)or humans are antisocial, rather than prosocial. In the ultimatum interactions with individuals which do not influence the game, the expected strategy is for individuals to make game (Houser & Kurzban, 2002; Kurzban et al., 2007). minimal offers and for these to be accepted. If there is a The idea here is that these cues trigger responses that have chance that minimal offers will be rejected (punished) then arisen in response to situations outside of the laboratory, individuals are expected to make larger offers (Gale et al., where whether or not they are being observed will matter. 1995). Consequently, the larger than minimal offers that are Further support comes from cultural differences in observed in experiments may just reflect the fact that experimental games (Gächter & Herrmann, 2006; Henrich individuals expect small offers to be punished. In this case, et al., 2006; Henrich et al., 2005), which appear to reflect the unexpected behaviour is the rejection of small offers, and differences in how the game is perceived to relate to so, we might conclude that the data show that humans have a everyday events (Binmore, 2006). To put it another way, tendency to punish at a level greater than that expected from “Experimental play often reflects patterns of interaction selfish interests. Note that our purpose here is not to argue found in everyday life” (Henrich et al., 2005, p. 798) and that humans are particularly pro- or antisocial, just that it is not just the game set up imposed by the experimenter. easy to give multiple explanations for the data. Furthermore, even in laboratory settings, behaviours such as Second, higher than expected levels of cooperation can be punishment can provide a direct benefit if longer periods of explained by individuals making mistakes in laboratory interactions are allowed for (Gächter et al., 2008). settings. The previous interpretation of economic games is The data discussed in the previous two paragraphs based upon the implicit assumption that if individuals do not suggest that humans have a psychology which can “misfire” play perfectly, then this does not lead to a systematic bias in in laboratory settings. While it might be argued that the the level of cooperation (Kümmerli et al., 2010). This is a possibility that humans don't always behave perfectly is no problem, because when the predicted behaviour is to not surprise, the more important point is that such imperfect cooperate at all (e.g., in standard public goods games), then behaviour can lead to a systematic bias towards higher than any deviations from perfection would automatically be expected levels of cooperation. Future work must address perceived as greater than expected cooperation (Houser & this issue, through the use of appropriate controls and by Kurzban, 2002). Kümmerli et al. (2010) tested this exercising greater caution when interpreting the absolute possibility, by allowing individuals to play modified level of cooperation in particular treatments (Kümmerli versions of public goods games, where 100% cooperation et al., 2010). It would be useful to test whether there is a bias was the strategy that would maximise their personal financial to accept evidence for humans being “extra cooperative,” game. They found that while this led to an increased level of without sufficient basis, due to a bias towards positive cooperation, it did not lead to full cooperation (see also evidence or because this is a nice result to get. Houser & Kurzban, 2002; Laury & Holt, 2008; Saijo & This “misfire” idea has been argued to be incorrect in Nakamura, 1995). If the logic from previous studies (e.g., several papers, where it is labelled the “big mistake” or Fehr & Schmidt, 1999) was applied to this result, then it maladaptation hypothesis (Boyd & Richerson, 2002; Fehr & would give a utility function that is negatively influenced Henrich, 2003; Gintis et al., 2003; Henrich, 2004). The by the success of others (an antisocial preference). Given implicit idea here is that humans should always behave that a simultaneous positive and negative regard to others is perfectly. However, this hypothesis is clearly falsified by the not possible, these data instead suggest that individuals numerous examples of how proximate mechanisms which have a tendency to avoid both full defection and full have been previously favoured by natural selection lead to cooperation (Haselton & Nettle, 2006). behaviours that do not maximise fitness under certain Third, another possible explanation is that higher levels of conditions. For example, the mismatch between real danger cooperation are normally favoured and that this leads to a and our fear of snakes and spiders versus automobiles, psychology that results in cooperation in one-shot experi- various aspects of the porn industry, rises in obesity, or the ments. The idea here is that, even if they are given perfect decline in reproductive rate can be associated with better information, individuals find it hard to disassociate them- living conditions (Hagen & Hammerstein, 2006). It is even selves from the real world, and so, cooperation occurs as a clearly falsified in the context of economic games, where, as byproduct of the fact that is normally favoured (Bateson discussed above, individuals show variation in behaviour in et al., 2006; Binmore, 2006; Burnham & Johnson, 2005; response to misleading “cues” of being observed, such as Hagen & Hammerstein, 2006; Haley & Fessler, 2005; Levitt eye-spots on computers. It can be misleading to call such & List, 2007; Nowak et al., 2000; Trivers, 2004; West et al., imperfect behaviour a maladaptation or a mistake, in the 2007b). Experimental support for this suggestion comes sense that it may be the optimal state, just that the benefits 252 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 of improving a behaviour have to be balanced or traded off direct and indirect benefits can occur and so whether it is against the costs (Partridge & Sibley, 1991; Stearns, 1992). mutually beneficial or altruistic will depend upon parameter The general point here is that maximisation of fitness does values (Lehmann et al., 2007c). Similarly, punishment can not imply perfect behaviour in every possible situation, and provide a direct or indirect benefit by reducing competition that the selective regimen needs to be considered, as has been for the actor or their relatives, respectively. shown frequently in the animal behaviour literature (Davies, Overall, the relative importance of direct and indirect 1992; Herre, 1987; Herre et al., 2001; Krebs & McCleery, fitness benefits will depend upon the details and parameter 1984; Pompilio et al., 2006; Wehner, 1987). Evolutionary values of a model (Gardner & West, 2004; Lehmann et al., theory does not predict that humans (or any other organism) 2007c). Specifically, whether cooperation and punishment should behave as perfect maximising agents in every are favoured as either mutually beneficial or altruistic situation in which they can be placed. behaviours depends upon parameters such as group size and the dispersal rate (Lehmann et al., 2007c). For 6.6.2. Misconception 15: The theoretical models on strong example, decreasing group size makes cooperation and reciprocity provide a novel solution to the problem of punishment more likely to provide a direct benefit because cooperation, that are outside of the usual inclusive fitness the actor gains a greater share of the group benefit from explanations (Bowles & Gintis, 2004; Fehr & Rockenbach, cooperation, and a greater benefit from the reduced 2003, 2004; Fehr & Fischbacher, 2005b; Gintis, 2000) competition that follows from punishment. A general It has been claimed that the theoretical models of strong point here is that the earlier models of strong reciprocity reciprocity do not rely on “explanatory power of inclusive were analysed with a simulation approach and then fitness theory” and “cannot be explained by inclusive explained with verbal arguments. Since then, multilocus fitness” (Bowles & Gintis, 2004) and that they can explain population genetic methodology has been used to provide the evolution of cooperation and punishment, even when analytical solutions that allow the underlying selective they do “not yield future economic benefits for the altruist” forces to be formally analysed, showing that these earlier (Fehr & Rockenbach, 2003) “it is implausible to expect that verbal arguments were incorrect (Lehmann et al., 2007c). these costs will be repaid” or “even though as a result they Considering Fig. 2, the strong reciprocity models have receive lower payoffs than other group members” (Bowles & involved selective forces that occur on multiple branches Gintis, 2004). However, this is not possible—a trait will not [e.g., non-enforced direct benefits; enforced direct benefits be selected for unless it provides an inclusive fitness benefit (punishment); indirect benefits by limited dispersal], as (see Section 2). One source of confusion here is the jargon well as a branch outside the tree that isn't even cooperation used in the strong reciprocity modeling literature, in that the (spiteful green beards). strategies that are referred to as altruistic are not necessarily altruistic as they can lead to an increase in personal fitness 6.6.3. Misconception 16: The claims made in the empirical (Misconception 1). and the theoretical strong reciprocity literature The other source of confusion is that while the impression are compatible is given that the strong reciprocity models do not rely upon We return to our point that there are four contradictory standard direct and indirect fitness benefits, more formal aspects of strong reciprocity. First, the empirical results show analyses have shown that they do, it is just that this was not that humans cooperate at higher levels than expected in some made explicit (Gardner & West, 2004; Lehmann et al., situations, and punish individuals who do not cooperate. 2007c). Cooperation can provide a direct benefit because it That is a clear and repeatable result. Second, it has been provides a benefit to everyone in the group, including the claimed that this propensity can be explained by strong focal cooperator through reducing the chance of group reciprocity. However, strong reciprocity is a proximate extinction or increasing the chance of success in between mechanism and not a solution to the ultimate problem of why group competition (analogous to models of “group augmen- humans cooperate (Misconception 4). Third, it has been tation”). Cooperation can provide an indirect benefit because claimed that the theoretical models of strong reciprocity can these models assume limited dispersal, which leads to a explain cooperation and punishment in one-shot encounters significant relatedness between the individuals interacting and that they provide a novel solution to the problem of within the group (Misconception 6), for example, r≈0.1 in cooperation that is outside of inclusive fitness theory. Fourth, groups of size 50 if the migration rate is 0.1 (Lehmann et al., the theoretical models of strong reciprocity actually show 2007c). This extent to which relatedness can build up how competition between groups and limited dispersal can appears to be frequently ignored in the strong reciprocity lead to direct and/or indirect benefits to cooperation theoretical literature—for example, Bowles & Gintis (2004) (Misconceptions 1, 4–6, 8 and 15). These models therefore assume group sizes of 20, where relatedness will be higher, are easily understood from an inclusive fitness context and but claim that “there are many unrelated individuals, so do not predict cooperation in one-shot encounters. In order altruism cannot be explained by inclusive fitness” (Bowles to predict cooperation in one-shot encounters, it would & Gintis, 2004). Note that we are not saying that in be necessary to develop more mechanistic models, which their model strong reciprocity is always altruistic, as both allowed for factors such as a trade-off between the S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 253 complexity of a strategy and its cost, and could hence predict bias, where individuals copy others, from either their own or misfiring (Misconception 14). other groups, on the basis of some arbitrary payoff or The potential confusion that can arise from these estimate of success (Bergstrom, 1995; Boyd & Richerson, contradictions is illustrated in how two sentences from the 1985, 2002; Boyd et al., 2003; Henrich, 2004). In this case, abstract of a single paper can contradict each other (Gintis in comparison with genetic selection, cultural selection is et al., 2003). Specifically, it is first claimed that strong less favourable for cooperation, and even leads to selection reciprocity cannot be explained by standard evolutionary for harming behaviours (Feldman et al., 1985; Lehmann models, then soon followed by a second sentence that claims et al., 2007c, 2008). The reason for this is that by helping strong reciprocity is evolutionarily stable (which means it neighbours and, hence, achieving a lower payoff, a helping can be explained by evolutionary theory): “strong reciprocity individual makes it less likely that they will be imitated. is a predisposition to cooperate with others and to punish Conversely, harming can be selected for because it those who violate the norms of cooperation, at personal cost, decreases competition with neighbours, who will then be even when it is implausible to expect that these costs will be less likely to be chosen (Lehmann et al., 2008). Lehmann repaid.” and “We show that under conditions plausibly et al. (2008; 2007c) argue that earlier papers came to the characteristic of the early stages of human evolution, a small different conclusion that such imitation could favour number of strong reciprocators could invade a population of cooperation because: (i) Boyd et al. (2003) did not compare self regarding types, and strong reciprocity is an evolutionary the situation with genetic evolution, they just claimed it stable strategy.” Confusion also arises because of incon- would be less likely to favour cooperation; (ii) Boyd and sistencies between papers. For example, compare the first Richerson (2002) made the additional assumption that there quote in this paragraph with “strong reciprocity must have was some other mechanism driving the initial spread of the promoted individual fitness, or it could not have evolved. trait, so that it exceeded a certain threshold frequency at Our contention is that strong reciprocity enhanced relative which it became beneficial (through avoidance of punish- fitness because groups with a high frequency of altruism ment) and, hence, was no longer altruistic. survived and prospered at a higher rate than groups with a Our aim here is not to argue whether cultural evolution low frequency of altruism” (Gintis et al., 2008, p. 248). makes it easier or harder for cooperation to evolve. This is an exciting and active area of research with much to be done. 6.7. Cultural evolution Instead, we merely wish to emphasise that this provides another example of the need to formally determine how Up until now, we have focused on genetic evolution. theoretical models are working, and their relation to existing However, humans are clearly unique in the extent to which theory. Considering the example given in the previous behaviour can be transmitted culturally, and the possible paragraph, Boyd & Richerson (2010) and Boyd et al. role of cultural evolution also needs to be considered. (Submitted) have argued that the difference between their Culture is information capable of affecting an individual's and Lehmann's results are due to differences in whether small behaviour that is acquired from other members of their or large fitness consequences were allowed for, whereas species through teaching, imitation, and other forms of Lehmann replies that both small and large effects were social transmission or social learning (Boyd & Richerson, examined in Lehmann et al. (2007c), and no assumptions 1985). Cultural traits can therefore be transmitted horizon- were made about the size of fitness effects in Lehmann & tally between individuals of the same generation. This Feldman (2008b). The advantage of this debate is that it contrasts with genetically inherited traits that are generally makes such assumptions explicit and so will clarify when only passed vertically from parent to offspring, with cultural evolution either favours or disfavours cooperation, notable exceptions in bacteria (Smith, 2001; West et al., but also why. More specifically, what forms of cultural 2006a). It is often suggested that cultural evolution is able mechanisms would be favoured by genetic selection, and to explain cooperation in cases where genetic selection how would these influence selection for cooperation? cannot (Bergstrom, 1995; Boyd & Richerson, 1985, 2002, 2005, 2006, 2010; Boyd et al., 2003; Fehr et al., 2002; Gintis, 2003; Henrich, 2004; Henrich & Boyd, 2001; 7. Discussion McElreath & Henrich, 2006; Richerson & Boyd, 2005). One reason for this is that cultural traits can be transmitted In the preceding sections we have provided a general horizontally within groups, which could lead to cultural review of social evolution theory, the potential solutions to relatedness r being higher than genetic r. the problem of cooperation, and some common misconcep- However, recent theory by Lehmann et al. (2007c, 2008; tions. Here, we return to the specific questions surrounding Lehmann & Feldman, 2008b) has questioned whether cooperation in humans: (1) Why do humans cooperate? (2) cultural evolution will automatically make it easier for Are humans special, and if so, why? Throughout, our focus cooperation to evolve. Consider the case of when imitation is on why humans behave as they do, rather than what they occurs through adaptive learning mechanisms such as ought to do, i.e., positive, not normative or regulative, “pairwise payoff comparison” or “prestige” or “success” science (Friedman, 1953). 254 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 7.1. Why do humans cooperate? how and why cultural evolution models influence the evolution of helping (Section 6.7). The discussion surrounding Misconceptions 1 and 6 make it clear that cooperation in humans could have originally 7.2. Are humans special? evolved due to either (or both) direct and indirect fitness benefits. Direct benefits could have arisen for a number of It is frequently assumed that the form of cooperation reasons including more cooperative groups being more in humans is special (Boyd & Richerson, 2002; Boyd successful, through competition with other groups or et al., 2003; Fehr & Fischbacher, 2003, 2005b; Fehr & avoiding group extinctions (group augmentation), all the Rockenbach, 2004; Henrich, 2004). For example “The usual reciprocity arguments, avoidance of punishment and nature and level of cooperation in human societies is other mechanisms. Indirect benefits are likely because unmatched in the animal world” (Quervain et al., 2004)or reasonable estimates of migration rates and group sizes for “Human cooperation represents a spectacular outlier in the early hominids suggest there would have been appreciable animal world” (Fehr & Rockenbach, 2004)or“Human relatedness between interacting individuals (Lehmann et al., altruism goes far beyond that which has been observed in 2007c). Indeed, a synergy between direct and indirect the animal world” (Fehr & Fischbacher, 2003). Indeed, this benefits is also likely—as discussed in Section 5.3, direct assumption has even been taken as a starting point, that benefits are often more likely to become important when cooperation in humans requires different evolutionary cooperation is already favoured due to indirect benefits. (ultimate) forces, rather than something that must be A possible question is what were/are the relative demonstrated: “What are the ultimate origins behind the importance of direct and indirect fitness benefits in rich patterns of human altruism described above? It must be explaining cooperation in humans? However, we suggest emphasized in the context of this question that a convincing that this question is so unanswerable to be almost pointless. explanation of the distinct features of human altruism The relative importance of direct and indirect fitness benefits should be based on capacities which are distinctly human— depends upon the exact parameter values of theoretical otherwise, there is the risk of merely explaining animal, not models, with the same model being able to lead to mutually human, altruism.” (Fehr & Fischbacher, 2003). In this beneficial or altruistic cooperation depending on the values section we critically assess the different ways in which taken by its parameters (Lehmann et al., 2007c). Researchers human cooperation may be special. We are not denying that are unlikely to be able to obtain sufficiently good parameter humans could be special, but want to determine, from an estimates about ancestral humans to address this problem evolutionary perspective, exactly why. with sufficient confidence. This is clearly illustrated by the Do humans have especially high levels of altruism (Fehr extent to which the last 40 years of research have been & Fischbacher, 2003, 2005b; Warneken et al., 2007)? No, a unable to resolve the relative importance of direct and number of organisms have higher levels of altruism than indirect fitness in cooperative breeding vertebrates, where humans, ranging from social amoebae and bacteria to ants the empirical and experimental opportunities are much and cooperative breeding vertebrates. In both social amoebae greater (Clutton-Brock, 2002; Cockburn, 1998; Griffin & and the social insects, a number of individuals completely West, 2002; Jennions & Macdonald, 1994). forgo the chance to reproduce to help others, which We stress here that our aim when discussing the various represents the most extreme possible form of altruism. In misconceptions has not been to argue against the possible social amoebae and bacteria, these are the stalk cells, which importance of factors such as punishment, between-group hold up spore cells so that they can be dispersed (Bonner, competition or cultural evolution. Instead, our main aim has 1967; Gilbert et al., 2007; Velicer et al., 2000). In social been to point out that there is often a large disparity between insects these are the sterile workers that give up the chance to what it is claimed is shown by a particular data set or reproduce for themselves and instead help to raise the theoretical model, and what is actually shown. Key examples offspring of the queen or queens (Bourke & Franks, 1995; have included claiming that (1) relatedness is not important Hamilton, 1972). In cooperative vertebrates, helping is in a particular model, but then, assuming a population sometimes mutually beneficial, and sometimes altruistic, structure that leads to an appreciable relatedness between depending upon the species (Griffin & West, 2003). An interacting individuals, i.e., relatedness is there, just extreme example at the altruistic end of the continuum is the unacknowledged (Misconceptions 5–7, 15); (2) an altruistic long tailed tit, where helpers never reproduce and so group-beneficial trait is being modelled, when actually the cooperation has been favoured purely by indirect fitness trait can be mutually beneficial (Misconceptions 1, 2 and 15), benefits (MacColl & Hatchwell, 2004; Russell & Hatchwell, or even spiteful and costly at the group level (Misconception 2001). In contrast, in humans, direct fitness benefits are often 7); (3) proximate data provides an answer to an ultimate likely to play a greater role, and cooperation is more likely to question (Misconception 4). Similar examples can be found be mutually beneficial than altruistic. elsewhere, such as discussions on how and when selection Are humans special because cooperation occurs be- favours hostility between groups (compare Choi & Bowles, tween nonrelatives (Boyd & Richerson, 2002; Boyd et al., 2007 with Lehmann & Feldman, 2008a), or the debate over 2003; Fehr & Fischbacher, 2003; Fehr & Rockenbach, S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 255 2004; Henrich, 2004)? No, cooperation between nonrela- lack of evidence for it playing a role in explaining tives occurs in a range of organisms. Many forms of cooperation in other organisms (Section 5.2.2). Impor- cooperation occur between nonrelatives in birds and tantly, this fine-tuning of behaviour can be done in mammals (Clutton-Brock, 2002). In cooperative breeding response to both previous experience (learning) and vertebrates, there are several examples where non-relatives observations of others (social learning). Furthermore, this cooperate, the indirect fitness benefits of cooperation appear has allowed the extreme division of labour that is observed to be negligible and it is thought that cooperation is driven in human societies. by direct fitness benefits (Clutton-Brock, 2002; Cockburn, While many organisms have impressive proximate 1998; Emlen, 1997; Griffin & West, 2002; Jennions & mechanisms for enforcing cooperation, humans can have Macdonald, 1994; Krebs & Davies, 1997). Even in social both more complex and diverse systems. Mechanisms such insects such as ants and wasps, there are some examples as direct and indirect reciprocity can be important in where nonrelatives come together for mutually beneficial humans, whereas they are thought to be beyond the cooperation (Bernasconi & Strassmann, 1999; Queller et al., cognitive abilities of most other animals (Stevens & 2000). However, perhaps the most extreme examples of Hauser, 2004; Stevens, Cushman, & Hauser, 2005). More cooperation between nonrelatives are the various examples complex and unique mechanisms to enforce cooperation of cooperation between species, termed mutualisms (Herre have arisen in humans, such as contracts, laws, justice, et al., 1999; Sachs et al., 2004; West et al., 2007a). For trade and social norms, leading to incredible feats such as example, between cleaner fish and their clients on the the extreme division of labour that keeps large cities or tropical reef, fig trees and fig wasps, plants and their nations going (Binmore, 1994, 1998, 2005b; Boyd & mycorrizae or rhizobia root symbionts, or the various Richerson, 1992; Seabright, 2004; Young, 2003). These symbionts that live within animal hosts. Finally, we also mechanisms allow direct benefits to be obtained from note that cooperation between non-relatives has also played cooperation in situations where cheating would otherwise a key role in some of the major evolutionary transitions, be favoured. To put this into game theoretic terms, such such as the incorporation of symbiotic bacteria that became mechanisms allow more efficient equilibria to be reached mitochondria, in the transition to eukaryotes (supplemen- than would ever be possible in less cognitively advanced tary material; Queller, 2000). species. Cultural evolution allows a potential way in which Are humans special because we enforce cooperation different mechanisms or strategies could be tested with mechanisms such as punishment? No, enforcement (Binmore, 2005b; Boyd & Richerson, 1985), and deter- occurs across a range of taxa from plants to animals mining how this influences cooperation remains a major (Section 5.2.2). For example, clients chase and attack outstanding task (Section 6.7). cleaner fish that do not cooperate (Bshary & Grutter, 2002; The above discussion suggests that humans are special Bshary & Grutter, 2005), soya bean plants cut off the because our cognitive abilities mean we are particularly oxygen supply to rhizobia that do not supply them with efficient enforcers, which has expanded the range of nitrogen (Kiers et al., 2003), dominant meerkats attack and situations in which cooperation can be favoured. However, evict subordinates who try to breed (Young et al., 2006), we stress that we are not saying that humans have the best honey bees destroy (police) eggs laid by workers (Ratnieks cognitive abilities for all behaviours related to cooperation. & Visscher, 1989) and ineffective pollinators are punished For example, considering indirect fitness benefits, while by a range of plant species (Goto et al., 2010; Jander & social amoebae and social insects are able to adjust their Herre, 2010; Pellmyr & Huth, 1994). behaviour in response to direct cues of genetic relatedness In contrast, what appears to be special about cooperation (Boomsma et al., 2003; Mehdiabadi et al., 2006), humans in humans is the proximate factors involved. Humans are must rely on indirect learnt cues such as childhood co- able to assess the local costs and benefits of cooperative residence (Lieberman et al., 2003). Overall, the general point behaviour, and adjust their behaviour accordingly (Fehr & appears to be that, as with other aspects of the mental powers Gächter, 2002; Fehr & Rockenbach, 2003; Fehr & and moral sense, the difference in cooperative behaviours Fischbacher, 2003, 2004; Henrich et al., 2005; Semmann between humans and other animals is “one of degree and not et al., 2004; West et al., 2006b). Consequently, human of kind” (Darwin, 1871, pp. 104–106). cognitive abilities allow individuals to be highly flexible in Supplementary data associated with this article can the level of cooperation they perform in response to be found, in the online version, at doi:10.1016/j. whether there is the possibility for punishment (Fehr & evolhumbehav.2010.08.001. Gächter, 2002), cues of reciprocity (Bateson et al., 2006; Semmann et al., 2004), whether they are competing locally Acknowledgments or globally for resources (West et al., 2006b), and competition between groups (Burton-Chellew et al., 2010; We thank Jean-Baptiste André, Abigail Barr, Michele Puurtinen & Mappes, 2009). In many of these cases, Belot, Ken Binmore, Rob Boyd, John Brenner, Terry human behaviour does appear to be special. For example, Burnham, Pat Churchland, Raymond Duch, Herb Gintis, the importance of reciprocity in humans contrasts with the Joel Guttman, Dominic Johnson, Laurent Lehmann, Ruth 256 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 Mace, Daniel Rankin, Francis Ratnieks, Meredith Rolfe, Boomsma, J. J., Nielsen, J., Sundstrom, L., Oldham, N. J., Tentschert, J., Mark Schaffer, Thom Scott-Phillips, Steve Stearns, Robert Petersen, H. C., et al. (2003). Informational constraints on optimal sex allocation in ants. Proceedings of the National Academy of Sciences of Trivers, Jack Vromen, and Peyton Young for comments, the United States of America, 100, 8799–8804. discussion and/or supplying reprints, and the Leverhulme Bourke, A. F. G., & Franks, N. R. (1995). Social Evolution in Ants. Trust, Royal Society, European Research Council and Princeton. New Jersey: Princeton University Press. Biotechnology, and Biological Sciences Research Council Bowles, S. (2006). Group competition, reproductive levelling and the – for funding. evolution of human altruism. Science, 314, 1569 1572. Bowles, S. (2009). Did warfare among ancestral hunter-gatherers affect the evolution of human social behaviors? Science, 324, 1293–1298. References Bowles, S., & Gintis, H. (2004). The evolution of strong reciprocity: cooperation in heterogeneous populations. Theoretical Population Alcock, J. (2005). Animal Behavior Eighth Edition. Sunderland, MA: Biology, 65,17–28. Sinauer Associates. Bowles, S., & Gintis, H. (2008). Cooperation. In L. Blume, & S. Durlauf Alexander, R. D. (1974). The evolution of social behavior. Annual Review (Eds.), The New Palgrave Dictionary of Economics. MacMilan. of Ecological Systematics, 5, 325–383. Bowles, S., Choi, J. K., & Hopfensitz, A. (2003). The co-evolution of Alexander, R. D. (1987). The biology of moral systems. New York: Aldine individual behaviors and social institutions. Journal of Theoretical de Gruyter. Biology, 223, 135–147. Amann, E., & Yang, C. L. (1998). Sophisitication and the persistence of Boyd, R., & Richerson, P. J. (1985). Culture and the Evolutionary Process. cooperation. Journal of Economic Behavior & Organization, 37,91–105. Chicago: University of Chicago Press. Aoki, K. (1982). A condition for group selection to prevail over Boyd, R., & Richerson, P. J. (1990). Group selection among alternaive counteracting individual selection. Evolution, 36, 832–842. evolutionarily stable strategies. Journal of Theoretical Biology, 145, Arnold, J. E. (1993). Labor and the rise of complex hunter-gatherers. 331–342. Journal of Anthropological Archaeology, 12,75–119. Boyd, R., & Richerson, P. J. (1992). Punishment allows the evolution of Arrow, H. (2007). The sharp end of altruism. Science, 318, 581–582. cooperation (or anything else) in sizable groups. Ethology and Aumann, R. (1981). Survey of repeated games. In V. Bohm (Ed.), Essays in Sociobiology, 13, 171–195. Game Theory and Mathematical Economics in Honor of Oskar Boyd, R., & Richerson, P. J. (2002). Group beneficial norms can spread Morgenstern (pp. 11–42). Mannheim: Bibliographisches Institut. rapidly in a stuctured population. Journal of Theoretical Biology, 215, Aumann, R., & Maschler, M. (1995). Repeated Games with Incomplete 287–296. Information. Cambridge, Mass: MIT Press. Boyd, R., & Richerson, P. J. (2005). The Origin and Evolution of Cultures. Avilés, L. (1993). Interdemic selection and the sex ratio: a social spider Oxford: Oxford University Press. perspective. American Naturalist, 142, 320–345. Boyd, R., & Richerson, P. J. (2006). Culture and the evolution of the human Axelrod, R. (1984). The Evolution of Cooperation. New York: Basic Books. social instincts. In S. Levinson, & N. Enfield (Eds.), Oxford: Berg. Axelrod, R., & Hamilton, W. D. (1981). The evolution of cooperation. Boyd, R., & Richerson, P. J. (2010). Transmission coupling mechanisms: Science, 211, 1390–1396. cultural group selection. Philosophical Transactions of the Royal Baschetti, R. (2007). Evolutionary, neurobiological, gene-based solution of Society B, 365, 3787–3795. the ideological qpuzzleq of human altruism and cooperation. Medical Boyd, R., Richerson, P. J., & Henrich, J. (Submitted). Rapid cultural Hypotheses, 69, 241–249. adaptation can facilitate the evolution of large-scale cooperation. Bateson, M., Nettle, D., & Roberts, G. (2006). Cues of being watched Boyd, R., Gintis, H., Bowles, S., & Richerson, P. J. (2003). The evolution of enhance cooperation in a real-world setting. Biology Letters, 2, 412–414. altruistic punishment. Proceedings of the National Academy of Sciences Becker, G. S. (1974). Altruism, egoism, and genetic fitness: economics of the United States of America, 100, 3531–3535. and sociobiology. Journal of Economic Literature, 14, 817–826. Boyd, R., Gintis, H., Bowles, S., & Richerson, P. J. (2005). The evolution Bergmüller, R., Bshary, R., Johnstone, R. A., & Russell, A. F. (2007). of altruistic punishment. In H. Gintis (Ed.), Moral Sentiments and Integrating coopeative breeding and cooperation theory. Behavioural Material Interests: The Foundations of Cooperation in Economic Life Processes, 76,61–72. (pp. 215–227). Cambridge, MA: MIT Press. Bergstrom, T. C. (1995). On the evolution of altruistic ethical rules for Brewer, M. B., & Caporael, L. R. (1990). Selfish genes vs. selfish siblings. The American Economic Review, 85,58–81. people: sociobiology as origin myth. Motivation and Emotion, 14, Bergstrom, T. C. (1996). Economics in a family way. Journal of Economic 237–243. Literature, 34, 1903–1934. Brockhurst, M. A., Buckling, A., & Gardner, A. (2007). Cooperation peaks Bergstrom, T. C. (2002). Evolution of social behavior: individual and group at intermediate disturbance. Current Biology, 17, 761–765. selection. Journal of Economic Perspectives, 16,67–88. Brown, S. P. (2001). Collective action in an RNA virus. Journal of Bernasconi, G., & Strassmann, J. E. (1999). Cooperation among unrelated Evolutionary Biology, 14, 821–828. individuals: the ant foundress case. Trends in Ecology & Evolution, 14, Bshary, R. (2002). Biting cleaner fish use altruism to deceive image-scoring 477–482. client reef fish. Proceedings of the Royal Society of London B, 269, Binmore, K. (1994). Game Theory And The Social Contract Volume 1: 2087–2093. Playing Fair. Cambridge, MA: MIT Press. Bshary, R., & Grutter, A. S. (2002). Asymetric cheating opportunities and Binmore, K. (1998). Game Theory And The Social Contract Volume 2: partner control in a cleaner fish mutualism. Animal Behaviour, 63, Just Playing. Cambridge, MA: MIT Press. 547–555. Binmore, K. (2005a). Economic man or straw man? Behavioral and Brain Bshary, R., & Schäffer, D. (2002). Choosy reef fish select cleaner fish that Sciences, 28, 817–818. provide high-quality service. Animal Behaviour, 63, 557–564. Binmore, K. (2005b). Natural Justice. Oxford: Oxford University Press. Bshary, R., & Grutter, A. S. (2005). Punishment and partner switching cause Binmore, K. (2006). Why do people cooperate? Politics, Philosophy & cooperative behaviour in a cleaning mutualism. Biology Letters, 1, Economics, 5,81–96. 396–399. Binmore, K. (2007). Game Theory: A Very Short Introduction. Oxford: Burnham, T. C., & Johnson, D. P. (2005). The biological and evolutionary Oxford University Press. logic of human cooperation. Analyse & Kritik, 27, 113–135. Bonner, J. T. (1967). The Cellular Slime Molds. Princeton: Princeton Burt, A., & Trivers, R. (2006). Genes in Conflict: The Biology of Selfish University Press. genetic Elements. Cambridge, Massachusetts: Harvard University Press. S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 257

Burton-Chellew, M. N., Ross-Gillespie, A., & West, S. A. (2010). Fehr, E., & Fischbacher, U. (2005b). Human altruism—proximate patterns Cooperation in humans: competition between groups and proximate and evolutionary origins. Analyse & Kritik, 27,6–47. emotions. Evolution and Human behavior, 31, 104–108. Fehr, E., Fischbacher, U., & Gächter, S. (2002). Strong reciprocity, human Choi, J. K., & Bowles, S. (2007). The coevolution of parochial altruism and cooperation and the enforcement of social norms. Human Nature, 13, war. Science, 318, 636–640. 1–25. Clutton-Brock, T. H. (2002). Breeding together: kin selection, reciprocity Feldman, M. W., Cavalli-Sforza, L. L., & Peck, J. R. (1985). Gene-culture and mutualism in cooperative animal societies. Science, 296,69–72. coevolution: models for the evolution of altruism with cultural Clutton-Brock, T. H. (2009). Cooperation between non-kin in animal transmission. Proceedings of the National Academy of Sciences of the societies. Nature, 462,51–57. United States of America, 82, 5814–5818. Clutton-Brock, T. H., & Parker, G. A. (1995). Punishment in animal Field, A. J. (2001). Altruistically inclined? The behavioral sciences, societies. Nature, 373, 209–216. evolutionary theory, and the origins of reciprocity. Michigan: University Cockburn, A. (1998). Evolution of helping behavior in cooperatively of Michigan Press. breeding birds. Annual Review of Ecological Systematics, 29, 141–177. Fisher, R. A. (1918). The correlation between relatives on the supposition of Colwell, R. K. (1981). Group selection is implicated in the evolution of Mendelian inheritance. Transactions of the Royal Society of Edinburgh, female-biased sex ratios. Nature, 290, 401–404. 52, 399–433. Cornwallis, C., West, S. A., & Griffin, A. S. (2009). Routes to cooperatively Fisher, R. A. (1930). The Genetical Theory of Natural Selection. Oxford: breeding vertebrates: kin discrimination and limited dispersal. Journal Clarendon. of Evolutionary Biology, 22, 2245–2457. Fisher, R. A. (1941). Average excess and average effect of a gene Cornwallis, C., West, S. A., Davies, K. E., & Griffin, A. S. (2010). substitution. Annals of Eugenics, 11,53–63. Promiscuity and the evolutionary transition to complex societies. Foster, K. R., Wenseleers, T., & Ratnieks, F. L. W. (2006). Kin selection is Nature, 466, 969–972. the key to altruism. Trends in Ecology & Evolution, 21,57–60. Crozier, W. R. (2001). Blushing and the exposed self: Darwin revisited. Frank, R. H. (1987). If Homo economicus could choose hos own utility Journal for the Theory of Social Behaviour, 31,61–72. function, would he want one with a conscience? The American Darwin, C. (1859). On the Origin of Species by Means of Natural Selection, Economic Review , 593–604. or, the Preservation of Favoured Races in the Struggle for Life. London, Frank, S. A. (1986). Hierarchical selection theory and sex ratios. I. General UK: John Murray. solutions for structured populations. Theoretical Population Biology, Darwin, C. (1871). The Descent of Man and Selection in Relation to Sex. 29, 312–342. London: John Murray. Frank, S. A. (1995a). Mutual policing and repression of competition in the Davies, N. B. (1992). Dunnock Behaviour and Social Evolution. Oxford: evolution of cooperative groups. Nature, 377, 520–522. Oxford University Press. Frank, S. A. (1995b). George Price's contributions to evolutionary genetics. Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford University Press. Journal of Theoretical Biology, 175, 373–388. Dawkins, R. (1979). Twelve misunderstandings of kin selection. Zeitschrift Frank, S. A. (1996a). Policing and group cohesion when resources vary. fur Tierpsychologie, 51, 184–200. Animal Behaviour, 52, 1163–1169. Dawkins, R. (1982). The extended phenotype. Oxford: Oxford University Frank, S. A. (1996b). Models of parasite virulence. Quarterly Review of Press. Biology, 71,37–78. de Waal, F. (2008). Putting the altruism back into altruism: the evolution of Frank, S. A. (1997). Multivariate analysis of correlated selection and kin empathy. Annual Review of Psychology, 59, 279–300. selection, with an ESS maximization method. Journal of Theoretical Denison, R. F., Kiers, E. T., & West, S. A. (2003). Darwinian agriculture: Biology, 189, 307–316. when can humans find solutions beyond the reach of natural selection. Frank, S. A. (1998). Foundations of Social Evolution. Princeton: Princeton Quarterly Review of Biology, 78, 145–168. University Press. Diggle, S. P., Griffin, A. S., Campbell, G. S., & West, S. A. (2007). Frank, S. A. (2003). Repression of competition and the evolution of Cooperation and conflict in quorum-sensing bacterial populations. cooperation. Evolution, 57, 693–705. Nature, 450, 411–414. Frank, S. A. (2010). A general model of the public goods dilemma. Journal El Mouden, C., West, S. A., & Gardner, A. (2010). The enforcement of of Evolutionary Biology, 23, 1245–1250. cooperation by policing. Evolution, 64, 2139–2152. Friedman, M. (1953). The methodology of positive economics. In M. Emlen, S. T. (1997). Predicting family dynamics in social vertebrates. Friedman (Ed.), Essays in Positive Economics (pp. 3–43). Chicago: In J. R. Krebs, & N. B. Davies (Eds.), Behavioural Ecology (pp. 228–253). University of Chicago Press. Oxford, UK: Blackwell. Fudenberg, D., & Maskin, E. (1986). The folk theorem in repeated games with Eshel, I. (1972). On the neighbor effect and the evolution of altruistic traits. discounting or with incomplete information. Econometrica, 54,533–554. Theoretical Population Biology, 3, 258–277. Gächter, S., & Herrmann, B. (2009). Reciprocity, culture and human Fehr, E., & Schmidt, K. (1999). A theory of fairness, competition and cooperation: previou insights and a new cross-cultural experiment. cooperation. Quarterly Journal of Economics, 114, 817–868. Philosophical Transactions of the Royal Society of London Series B, Fehr, E., & Gächter, S. (2002). Altruistic punishment in humans. Nature, 364, 791–806. 415, 137–140. Gächter, S., & Herrmann, B. (2006). The limits of self-governance when Fehr, E., & Rockenbach, B. (2003). Detrimental effects of sanctions on cooperators get punished: experimental evidence from urban and rural human altruism. Nature, 422, 137–140. Russia. Centre for Decision Research and Experimental Economics. Fehr, E., & Fischbacher, U. (2003). The nature of human altruism. Nature, Discussion Paper 13. Available at: http://ideas.repec.org/p/cdx/dpaper/ 425, 785–791. 2007-11.html Fehr, E., & Henrich, J. (2003). Is strong reciprocity a maladaptation? In P. Gächter, S., Renner, E., & Sefton, M. (2008). The long-run benefits of Hammerstein (Ed.), Genetic and Cultural Evolution of Cooperation punishment. Science, 322, 1510. (pp. 55–82). Cambridge, MA: MIT Press. Gale, J., Binmore, K. G., & Samuelson, L. (1995). Learning to be imperfect: Fehr, E., & Rockenbach, B. (2004). Human altruism: economic, neural, and the ultimatum game. Games and Economic Behavior, 8,56–90. evolutionary perspectives. Current Opinion in Neurobiology, 14, 784–790. Gardner, A. (2008). The Price equation. Current Biology, 18, R198–R202. Fehr, E., & Fischbacher, U. (2004). Third-party punishment and social Gardner, A. (2009). Adaptation as organism design. Biology Letters, 5, norms. Evolution and Human Behavior, 25,63–87. 861–864. Fehr, E., & Fischbacher, U. (2005a). Altruists with green beards. Analyse & Gardner, A., & West, S. A. (2004). Cooperation and punishment, especially Kritik, 27,73–84. in humans. American Naturalist, 164, 753–764. 258 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

Gardner, A., & Grafen, A. (2009). Capturing the superorganism: a formal Grafen, A. (2006a). Optimisation of inclusive fitness. Journal of Theoretical theory of group adaptation. Journal of Evolutionary Biology, 22, Biology, 238, 541–563. 659–671. Grafen, A. (2006b). A theory of Fisher's reproductive value. Journal of Gardner, A., & West, S. A. (2010). Greenbeards. Evolution, 64,25–38. Mathematical Biology, 53,15–60. Gardner, A., West, S. A., & Buckling, A. (2004). Bacteriocins, spite and Grafen, A. (2007a). An inclusive fitness analysis of altruism on a cyclical virulence. Proceedings of the Royal Society of London, Series B, 271, network. Journal of Evolutionary Biology, 20, 2278–2283. 1529–2535. Grafen, A. (2007b). The formal Darwinism project: a mid-term report. Gardner, A., West, S. A., & Barton, N. H. (2007a). The relation between Journal of Evolutionary Biology, 20, 1243–1254. multilocus population genetics and social evolution theory. American Grafen, A. (2007c). Detecting kin selection at work using inclusive fitness. Naturalist, 169, 207–226. Proceedings of the Royal Society of London, Series B, 274, 713–719. Gardner, A., Arce, A., & Alpedrin, J. (2009). Budding dispersal and the Grafen, A. (2009). Formalizing Darwunism and inclusive fitness theory. sex ratio. Journal of Evolutionary Biology, 22, 1036–1045. Philosophical Transactions of the Royal Society of London Series B, Gardner, A., West, S.A., and Wild, G. (Submitted). The genetical theory of 364, 3135–3141. kin selection. Journal of Evolutionary Biology. Grafen, A., & Archetti, M. (2008). Natural selection of altruism in enelastic Gardner, A., Hardy, I. C. W., Taylor, P. D., & West, S. A. (2007b). Spiteful viscous homogeneous populations. Journal of Theoretical Biology, 252, soldiers and sex ratio conflict in polyembryonic parasitoid wasps. 694–710. American Naturalist, 169, 519–533. Griffin, A. S., & West, S. A. (2002). Kin selection: fact and fiction. Trends Gauthier, D. (1986). Morals by Agreement. Oxford: Clarendon Press. in Ecology & Evolution, 17,15–21. Gilbert, O. M., Foster, K. R., Mehdiabadi, N. J., Strassmann, J. E., & Griffin, A. S., & West, S. A. (2003). Kin discrimination and the benefit Queller, D. C. (2007). High relatedness maintains multicellular of helping in cooperatively breeding vertebrates. Science, 302, cooperation in a social amoeba by controlling cheater mutants. 634–636. Proceedings of the National Academy of Sciences of the United Griffin, A. S., West, S. A., & Buckling, A. (2004). Cooperation and States of America, 104, 8913–8917. competition in pathogenic bacteria. Nature, 430, 1024–1027. Gintis, H. (2000). Strong reciprocity and human sociality. Journal of Hagen, E. H., & Hammerstein, P. (2006). Game theory and human Theoretical Biology, 206, 169–179. evolution: a critique of some recent interpretations of experimental Gintis, H. (2003). The hitchhiker's guide to altruism: gene-culture games. Theoretical Population Biology, 69, 339–348. coevolution, and the internalization of norms. Journal of Theoretical Haig, D. (2002). Genomic Imprinting and Kinship. New Brunswick, NJ: Biology, 220, 407–418. Rutgers University. Gintis, H., Smith, E. A., & Bowles, S. (2001). Costly signalling and Haldane, J. B. S. (1932). The causes of evolution. New York: Longmans. cooperation. Journal of Theoretical Biology, 213, 103–119. Haley, K. J., & Fessler, D. M. T. (2005). Nobody's watching? Subtle cues Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (2003). Explaining affect generosity in an anonymous economic game. Evolution and altruistic behavior in humans. Evolution and Human Behavior, 24, Human behavior, 26, 245–256. 153–172. Hamilton, W. D. (1963). The evolution of altruistic behaviour. American Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (2005a). Moral Sentiments and Naturalist, 97, 354–356. Material Interests: The Foundations of Cooperation in Economic Life. Hamilton, W. D. (1964). The genetical evolution of social behaviour, I & II. Cambridge, Massachusetts: MIT Press. Journal of Theoretical Biology, 7,1–52. Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (2005b). Moral sentiments and Hamilton, W. D. (1970). Selfish and spiteful behaviour in an evolutionary material interests: origins, evidence, and consequences. In H. Gintis model. Nature, 228, 1218–1220. (Ed.), Moral sentiments and material interests: The foundations of Hamilton, W. D. (1971). Selection of selfish and altruistic behaviour in some cooperation in economic life (pp. 3–39). Cambridge, Massachusetts: extreme models. In J. F. Eisenberg, & W. S. Dillon (Eds.), Man and MIT Press. beast: Comparative social behavior (pp. 57–91). Washington, DC: Gintis, H., Henrich, J., Bowles, S., Boyd, R., & Fehr, E. (2008). Strong Smithsonian Press. reciprocity and the roots of human morality. Social Justice Research, Hamilton, W. D. (1972). Altruism and related phenomena, mainly in social 21, 241–253. insects. Annual Review of Ecological Systematics, 3, 193–232. Goto, R., Okamoto, T., Kiers, E. T., Kawakita, A., & Kato, M. (2010). Hamilton, W. D. (1975). Innate social aptitudes of man: an approach from Selective flower abortion maintains moth cooperation in a newly evolutionary genetics. In R. Fox (Ed.), Biosocial anthropology discovered pollination mutualism. Ecology Letters, 13, 321–329. (pp. 133–155). New York: Wiley. Grafen, A. (1982). How not to measure inclusive fitness. Nature, 298, Hamilton, W. D. (1996). Narrow roads of gene land: I Evolution of social 425–426. behaviour. Oxford: W.H. Freeman. Grafen, A. (1984). Natural selection, kin selection and group selection. In Hamilton, W. D., & May, R. (1977). Dispersal in stable habitats. Nature, J. R. Krebs, & N. B. Davies (Eds.), Behavioural ecology: An 269, 578–581. evolutionary approach (pp. 62–84). Oxford, UK: Blackwell Scientific Hammerstein, P. (2003). Genetic and cultural evolution of cooperation. Publications. Cambridge: MIT Press. Grafen, A. (1985). A geometric view of relatedness. Oxford Surveys in Hardin, G. (1968). The tragedy of the commons. Science, 162, 1243–1248. Evolutionary Biology, 2,28–89. Harvey, P. H., Partridge, L., & Nunney, L. (1985). Group selection and the Grafen, A. (1990a). Biological signals as handicaps. Journal of Theoretical sex ratio. Nature, 313,10–11. Biology, 144, 517–546. Haselton, M. G., & Nettle, D. (2006). The paranoid optimist: an Grafen, A. (1990b). Do animals really recognise kin? Animal Behaviour, integrative evolutionary model of cognitive biases. Personality and 39,42–54. social. Psychology Review, 10,47–66. Grafen, A. (1991). Modelling in behavioural ecology. In J. R. Krebs, & Helanterä, H., & Bargum, K. (2007). Pedigree relatedness, not greenbeard N. B. Davies (Eds.), Behavioural ecology, an evolutionary approach genes, explains eusociality. Oikos, 116, 217–220. (pp. 5–31). Oxford: Blackwell. Helanterä, H., & Sundström, L. (2007). Worker policing and nest mate Grafen, A. (1999). Formal Darwinism, the individual-as-maximizing-agent recognition in the ant Formica fusca. Behavioral Ecology and analogy and bet-hedging. Proceedings of the Royal Society of London, Sociobiology, 61, 1143–1149. Series B, 266, 799–803. Henrich, J. (2004). Cultural group selection, coevolutionary processes and Grafen, A. (2002). A first formal link between the Price equation and an large-scale cooperation. Journal of Economic Behavior & Organization, optimization program. Journal of Theoretical Biology, 217,75–91. 53,3–35. S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 259

Henrich, J., & Boyd, R. (2001). Why people punish defectors. Journal of can explain human cooperation in public goods games. Proceedings of the Theoretical Biology, 208,79–89. National Academy of Sciences USA, 107, 10125–10130. Henrich, J., McElreath, R., Barr, A., Ensminger, J., Barrett, C., & Bolyanatz, Kurzban, R., DeSciolo, P., & O'Brien, E. (2007). Audience effects on A, et al. (2006). Costly punishment across human societies. Science, moralistic punishment. Evolution and Human behavior, 28,75–84. 312, 1767–1770. Lack, D. (1966). Population Studies of Birds. Oxford: Clarendon Press. Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., & Gintis, H., et al. Laury, S. K., & Holt, C. A. (2008). Voluntary provision of public (2005). “Economic man” in cross-cultural perspective: Behavioral goods: experimental results with interior Nash equilibria. Handbook of experiments in 15 small-scale societies. Behavioral and Brain Sciences, Experimental Economics Results, Volume I (pp. 792–80). Amsterdam: 28, 795–855. Elsevier. Henrich, N., & Henrich, J. (2007). Why humans cooperate: a cultural and Ledyard, J. O. (1995). Public goods: a survey of experimental research. In evolutionary explanation. New York: Oxford University Press. J. H. Kagel, & A. E. Roth (Eds.), The handbook of experimental Herre, E. A. (1987). Optimality, plasticity and selective regime in fig wasp economics. Princeton: Princeton University Press. sex ratios. Nature, 329, 627–629. Lehmann, L., & Keller, L. (2006). The evolution of cooperation and Herre, E. A. (1993). Population structure and the evolution of virulence in altruism. A general framework and classification of models. Journal nematode parasites of fig wasps. Science, 259, 1442–1445. of Evolutionary Biology, 19, 1365–1378. Herre, E. A., Machado, C. A., & West, S. A. (2001). Selective regime and fig Lehmann, L., & Feldman, M. W. (2008a). War and the evolution of wasp sex ratios: Towards sorting rigor from pseudo-rigor in tests of belligerance and bravery. Proceedings of the Royal Society of London, adaptation. In S. Orzack, & E. Sober (Eds.), Adaptionism and optimality Series B Proc Biol Sci, 2008, 2877–2885. (pp. 191–218). Cambridge: University Press. Lehmann, L., & Feldman, M. W. (2008b). The co-evolution of Herre, E. A., Knowlton, N., Mueller, U. G., & Rehner, S. A. (1999). The culturally inherited altruistic helping and cultural transmission under evolution of mutualisms: exploring the paths between conflict and random group formation. Theoretical Population Biology, 73, cooperation. Trends in Ecology & Evolution, 14,49–53. 506–516. Houser, D., & Kurzban, R. (2002). Revisiting kindness and confusion in public Lehmann, L., Keller, L., & Sumpter, D. J. T. (2007a). Inclusive fitness goods experiments. The American Economic Review, 92,1062–1069. explains the evolution of helping and harming behaviors on graphs. Howard, N. (1971). Paradoxes of rationality: theory of metagames and Journal of Evolutionary Biology, 20, 2284–2295. political behavior. Cambridge, Massachusetts: MIT Press. Lehmann, L., Feldman, M. W., & Foster, K. R. (2008). Cultural Hughes, W. O. H., Oldroyd, B. P., Beekman, M., & Ratnieks, F. L. W. transmission can inhibit the evolution of altruistic helping. American (2008). Ancestral monogamy shows kin seelction is the key to the Naturalist, 172,12–24. evolution of eusociality. Science, 320, 1213–1216. Lehmann, L., Keller, L., West, S. A., & Roze, D. (2007b). Group selection Jander, K. C., & Herre, E. A. (2010). Host sanctions and pollinator cheating and kin selection. Two concepts but one process. Proceedings of the in the fig tree-fig wasp mutualism. Proceedings of the Royal Society of National Academy of Sciences of the United States of America, 104, London, Series B, Proc R Soc B, 277, 1481–1488. 6736–6739. Jennions, M. D., & Macdonald, D. W. (1994). Cooperative breeding in Lehmann, L., Rousset, F., Roze, D., & Keller, L. (2007c). Strong-reciprocity mammals. Trends in Ecology & Evolution, 9,89–93. or strong-ferocity? A population genetic view of the evolution of Kandori, M. (1992). Social norms and community enforcement. The Review altruistic punishment. American Naturalist, 170,21–36. of Economic Studies, 59,63–80. Leigh, E. G. (1971). Adaptation and Diversity. San Francisco: Freeman, Kiers, E. T., Rousseau, R. A., West, S. A., & Denison, R. F. (2003). Cooper and Company. Host sanctions and the legume-rhizobium mutualism. Nature, 425, Leigh, E. G. (1983). When does the good of the group override the 78–81. advantage of the individual? Proceedings of the National Academy of Killingback, T., Bieri, J., & Flatt, T. (2006). Evolution in group-structured Sciences of the United States of America, 80, 2985–2989. populations can resolve the tragedy of the ccommons. Proceedings of Leigh, E. G. (1991). Genes, bees and ecosystems: the evolution of a the Royal Society of London, Series B, 273, 1477–1481. common interest among individuals. Trends in Ecology & Evolution, 6, Knight, C. (1998). Ritual/speech coevolution: a solution to the problem 257–262. of deception. In J. R. Hurford (Ed.), Approaches to the Evolution of Levitt, S. D., & List, J. A. (2007). What do laboratory experiments Language. Cambridge: Cambridge University Press. measuring social preferences reveal about the real world? Journal of Knight, C. (2008). “Honest fakes” and language origins. Journal of Economic Perspectives, 21, 153–174. Consciousness Studies, 15, 236–248. Lieberman, D., Tooby, J., & Cosmides, L. (2003). Does morality have a Kohn, M. (2008). The needs of the many. Nature, 456, 296–299. biological basis? An empirical test of the factors governing moral Kokko, H., Johnstone, R. A., & Clutton-Brock, T. H. (2001). The evolution sentiments relating to incest. Proceedings of the Royal Society of of cooperative breeding through group augmentation. Proceedings of London, Series B, 270, 819–826. the Royal Society of London, Series B, 268, 187–196. Lieberman, E., Hauert, C., & Nowak, M. A. (2005). Evolutionary dynamics Krebs, J. R., & McCleery, R. H. (1984). Optimization in behavioural on graphs. Nature, 433, 312–316. ecology. In J. R. Krebs, & N. B. Davies (Ed.), Behavioural ecology, Luce, R., & Raiffa, H. (1957). Games and Decisions. New York: Wiley. 2nd edn. Oxford: Blackwell. MacColl, A. D. C., & Hatchwell, B. J. (2004). Determinants of lifetime Krebs, J. R., & Davies, N. B. (1987). An introduction to behavioural fitness in a cooperative breeder, the long tailed tit Aegithalos caudatus. ecology. Second Edition. Oxford, England: Blackwell Scientific Journal of Animal Ecology, 73, 1137–1148. Publications. MacLean, R. C., & Gudelj, I. (2006). Resource competition and social Krebs, J. R., & Davies, N. B. (1993). An introduction to behavioural ecology. conflict in experimental populations of yeast. Nature, 441, 498–501. Third Edition. Oxford, England: Blackwell Scientific Publications. Mailah, G. J., & Samuelson, L. (2006). Repeated games and Krebs, J. R., & Davies, N. B. (1997). Behavioural ecology. An evolutionary reputations: Long-run relationships. New York: Oxford University Press. approach. 4th Ed. Oxford: Blackwell Scientific. Manski, C. F. (2000). Economic analysis of social interactions. The Journal Kruuk, H. (2003). Niko's nature. Oxford: Oxford University Press. of Economic Perspectives, 14, 115–136. Kümmerli, R., Gardner, A., West, S. A., & Griffin, A. S. (2009). Limited Mateo, J. M. (2002). Kin-recognition abilities and nepotism as a function of dispersal, budding dispersal and cooperation: an experimental study. sociality. Proceedings of the Royal Society of London, Series B, 269, Evolution, 63, 939–949. 721–727. Kümmerli, R., Burton-Chellew, M. N., Ross-Gillespie, A., & West, S. A. Maynard Smith, J. (1964). Group selection and kin selection. Nature, 201, (2010). Resistance to extreme strategies, rather than prosocial preferences, 1145–1147. 260 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

Maynard Smith, J. (1976). Group selection. Quarterly Review of Biology, Paley, W. (1802). Natural Theology. London: Wilks & Taylor. 51, 277–283. Palmer, C. T., Fredrickson, B. E., & Tilley, C. F. (1997). Categories and Maynard Smith, J. (1982). Evolution and the theory of games. Cambridge: gatherings: group selection and the mythology of cultural anthropology. Cambridge University Press. Evolution and Human Behavior, 18, 291–308. Maynard Smith, J. (1983). Models of evolution. Proceedings of the Royal Panchanathan, K., & Boyd, R. (2004). Indirect reciprocity can stabilize Society of London, Series B, 219, 315–325. cooperation without the second-order free rider problem. Nature, 432, Maynard Smith, J. (1995). Life at the edge of chaos? The New York review, 499–502. 28–30. Partridge, L., & Sibley, R. (1991). Constraints in the evolution of life Maynard Smith, J., & Price, G. R. (1973). The logic of animal conflict. histories. Philosophical Transactions of the Royal Society of London Nature, 246,15–18. Series B, 332,3–13. Maynard Smith, J., & Szathmary, E. (1995). The major transitions in Pellmyr, O., & Huth, C. J. (1994). Evolutionary stability of mutualism evolution. Oxford: W.H. Freeman. between yuccas and yucca moths. Nature, 372, 257–260. Mayr, E. (1961). Cause and effect in biology. Science, 134, 1501–1506. Pfeiffer, T., Schuster, S., & Bonhoeffer, S. (2001). Cooperation and McElreath, R., & Henrich, J. (2006). Modeling cultural evolution. In R. I. M. competition in the evolution of ATP-producing pathways. Science, 292, Dunvar, & Barret (Eds.), Oxford Handbook of Evolutionary Biology. 504–507. Mehdiabadi, N. J., Jack, C. N., Farnham, T. T., Platt, T. G., Kalla, S. E., Pfennig, D. W., Collins, J. P., & Ziemba, R. E. (1999). A test of alternative Shaulsky, G., et al. (2006). Kin preference in a social microbe. Nature, hypotheses for kin recognition in cannibalistic tiger salamanders. 442, 881–882. Behavior Ecology, 10, 436–443. Michod, R. E., & Hamilton, W. D. (1980). Coefficients of relatedness in Platt, T. G., & Bever, J. D. (2009). Kin competition and the evolution of sociobiology. Nature, 288, 694–697. cooperation. Trends in Ecology & Evolution, 24, 370–377. Milinski, M., & Wedekind, C. (1998). Working memory constrains Pompilio, L., Kacelnik, A., & Behmer, S. T. (2006). State-dependent human cooperation in the prisoner's dilemma. Proceedings of the learned valuation drives choice in an invertebrate. Science, 311, National Academy of Sciences of the United States of America, 95, 1613–1615. 13755–13758. Pradel, J., Euler, H. A., & Fetchenhauer, D. (2009). Spotting altruistic Milinski, M., Semmann, D., & Krambeck, H. J. (2002). Donors to charity dictator game players and mingling with them: the elective assortation of gain in both indirect reciprocity and political reputation. Proceedings classmates. Evolution and Human behavior, 30, 103–113. of the Royal Society of London, Series B, 269, 881–883. Provine, W. B. (2001). The origins of theoretical population genetics. Mock, D. W., & Parker, G. A. (1997). The evolution of sibling rivalry. Chicago: Chicago University Press. Oxford: Oxford University Press. Puurtinen, M., & Mappes, T. (2009). Between-group competition and Mulder, R. A., & Langmore, N. E. (1993). Dominant males punish helpers human cooperation. Proceedings of the Royal Society of London, for temporary defectrion in superb fairy wrens. Animal Behaviour, 45, Series B, 276, 355–360. 830–833. Queller, D. C. (1992a). Quantitative genetics, inclusive fitness, and group Murray, M. G. (1985). Figs (Ficus spp.) and fig wasps (Chalcidoidea, selection. American Naturalist, 139, 540–558. Agaonidae): hypotheses for an ancient symbiosis. Biological Journal Queller, D. C. (1992b). Does population viscosity promote kin selection? of the Linnean Society, 26,69–81. Trends in Ecology & Evolution, 7, 322–324. Nettle, D. (2009). Evolution and genetics for psychology. Oxford: Oxford Queller, D. C. (1992c). A general model for kin selection. Evolution, 46, University Press. 376–380. Nowak, M. A. (2006). Five rules for the evolution of cooperation. Science, Queller, D. C. (2000). Relatedness and the fraternal major transitions. 314, 1560–1563. Philosophical Transactions of the Royal Society of London Series B, Nowak, M. A., & Sigmund, K. (1998). Evolution of indirect reciprocity by 355, 1647–1655. image scoring. Nature, 393, 573–577. Queller, D. C. (2004). Kinship is relative. Nature, 430, 975–976. Nowak, M. A., & Sigmund, K. (2005). Evolution of indirect reciprocity. Queller, D. C., & Goodnight, K. F. (1989). Estimating relatedness using Nature, 437, 1291–1298. genetic markers. Evolution, 43, 258–275. Nowak, M. A., Page, K. M., & Sigmund, K. (2000). Fairness versus reason Queller, D. C., Zacchi, F., Cervo, R., Turillazzi, S., Henshaw, M. T., in the ultimatum game. Science, 289, 1773–1775. Santorelli, L. A., et al. (2000). Unrelated helpers in a social insect. Nowak, M. A., Tarnita, C. E., & Antal, T. (2010). Evolutionary dynamics in Nature, 405, 784–787. structured populations. Philosophical Transactions of the Royal Society Quervain, J. F. d., Fischbacher, U., Treyer, V., Schellhammer, M., of London Series B, 365,19–30. Schnyder, U., Buck, A., et al. (2004). The neural basis of altruistic O'Gorman, R., Sheldon, K. M., & Wilson, D. S. (2008). For the good of the punishment. Science, 305, 1254–1258. group? Exploring group-level evolutionary adaptations using multilevel Rapoport, A., & Chammah, A. M. (1965). Prisoner's Dilemma. Ann Arbor: selection theory. Group Dynamics: Theory, Research, and Practise, 12, University of Chicago Press. 17–26. Ratnieks, F. L. W. (1988). Reproductive harmony via mutual policing by Oates, K., & Wilson, M. (2001). Nominal kinship cues facilitate altruism. workers in eusocial Hymenoptera. American Naturalist, 132, 217–236. Proceedings of the Royal Society of London, Series B, 269, 105–109. Ratnieks, F. L. W., & Visscher, P. K. (1989). Worker policing in the Okasha, S. (2004). Multilevel selection and the partitioning of covariance: a honeybee. Nature, 342, 796–797. comparison of three approaches. Evolution, 58, 486–494. Ratnieks, F. L. W., Foster, K. R., & Wenseleers, T. (2006). Conflict Okasha, S. (2006). Evolution and the levels of selection. Oxford, UK: resolution in insect societies. Annual Review of Entomology, 51, Oxford University Press. 581–608. Orlove, M. J., & Wood, C. L. (1978). Coefficients of relationship and Reeve, H. K., & Hölldobler, B. (2007). The emergence of a superorganism coefficients of relatedness in kin seelction: a covariance form for the rho through intergroup competition. Proceedings of the National Academy formula. Journal of Theoretical Biology, 73, 679–686. of Sciences of the United States of America, 104, 9736–9740. Owren, M. J., & Bachorowski, J. A. (2006). The evolution of emotional Richerson, P. J., & Boyd, R. (1999). The evolutionary dynamics of a crude expression: a “selfish-gene" account of smiling and laughter in early superorganism. Human Nature, 10, 253–289. hominids and humans. In T. J. Mayne, & G. A. Bonanno (Eds.), Emotions: Richerson, P. J., & Boyd, R. (2005). Not by genes alone. Chicago: Current Issues and Future Directions. New York: The Guilford Press. University of Chicago Press. Palameta, B., & Brown, W. M. (1999). Human cooperation is more than by- Richerson, P. J., & Boyd, R. (1999). Complex societies: The evolutionary product mutualism. Animal Behaviour, 57,F1–F3. origins of a crude superorganism. Human Nature. S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262 261

Robson, A. J. (1990). Efficiency in evolutionary games: Darwin, Nash and Taylor, P. D. (1996). Inclusive fitness arguments in genetic models of the secret handshake. Journal of Theoretical Biology, 144, 379–396. behaviour. Journal of Mathematical Biology, 34, 654–674. Robson, A. J. (2008). Group selection. In S. Durlauf, & L. E. Blume (Eds.), Taylor, P. D., & Frank, S. A. (1996). How to make a kin selection model. The New Palgrave Dictionary of Economics. Second Edition. Palgrave Journal of Theoretical Biology, 180,27–37. Macmillan. Taylor, P. D., Day, T., & Wild, G. (2007a). Evolution of cooperation in a Rogers, A. R. (1990). Group selection by selective emigration: the effects of finite homogeneous graph. Nature, 447, 469–472. migration and kin structure. American Naturalist, 135, 398–413. Taylor, P. D., Wild, G., & Gardner, A. (2007b). Direct fitness or inclusive Rose, M. R., & Lauder, G. V. (Eds.), (1996). Adaptation. San Diego: fitness: how shall we model kin selection. Journal of Evolutionary Academic Press. Biology, 20, 301–309. Rousset, F. (2004). Genetic structure and selection in subdivided Tinbergen, N. (1963). On aims and methods of ethology. Zeitschrift fur populations. Princeton, NJ: Princeton University Press. Tierpsychologie, 20, 410–433. Rousset, F., & Roze, D. (2007). Constraints on the origin and maintenance Traulsen, A., & Nowak, M. A. (2006). Evolution of cooperation by of genetic kin recognition. Evolution, 61, 2320–2330. multilevel selection. Proceedings of the National Academy of Sciences Russell, A. F., & Hatchwell, B. J. (2001). Experimental evidence for kin- of the United States of America, 103, 10952–10955. biased helping in a cooperatively breeding vertebrate. Proceedings of Trivers, R. (2004). Mutual benefits at all levels of life. Science, 304, the Royal Society of London, Series B, 268, 2169–2174. 964–965. Russell, A. F., & Wright, J. (2008). Avian mobbing: byproduct mutualism is Trivers, R. (1985). Social Evolution. Benjamin/Cummings, Menlo Park, CA. not reciprocal altruism. Trends in Ecology & Evolution. Trivers, R. L. (1971). The evolution of reciprocal altruism. Quarterly Sachs, J. L., Mueller, U. G., Wilcox, T. P., & Bull, J. J. (2004). The Review of Biology, 46,35–57. evolution of cooperation. Quarterly Review of Biology, 79, 135–160. Trivers, R. L. (1974). Parent-offspring conflict. American Zoologist, 14, Saijo, T., & Nakamura, H. (1995). The “spite” dilemma in voluntary 249–264. contribution mechanism experiments. Journal of Conflict Resolution, Trivers, R. L. (1998a). Think for yourself. Skeptic, 6,86–87. 39, 535–560. Trivers, R. L. (1998b). As they would do to you. Skeptic, 6,81–83. Schloss. (2002). Emerging accounts of altruism. In S. G. Post (Ed.), van Rooy, R. (2003). Being polite is a handicap: towards a game theoretical Altruism and Altruitsic Love (pp. 212–242). Oxford: Oxford analysis of polite linguistic behavior. In M. Tennenholtz (Ed.), Proc. of University Press. the 4th Conference on Theoretical Aspects of Rationality & Knowledge Schuessler. (1989). Exit threats and cooperation under anonymity. Journal (pp. 45–58). of Conflict Resolution, 33, 728–749. Velicer, G. J., Kroos, L., & Lenski, R. E. (2000). Developmental cheating in Scott-Phillips, T. C. (2007). The social evolution of language, and the the social bacterium Myxococcus xanthus. Nature, 404, 598–601. language of social evolution. Evolutionary Psychology, 5, 740–753. Wade, M. J. (1985). Soft selection, hard selection, kin selection, and group Scott-Phillips, T. C. (2008). Defining biological communication. Journal of selection. American Naturalist, 125,61–73. Evolutionary Biology, 21. Warneken, F., Hare, B., Melis, A. P., Hanus, D., & Tomasello, M. (2007). Seabright, P. (2004). The company of strangers: A natural history of Spontaneous altruism by chimpanzees and young children. PLOS economic life. Princeton: Princeton University Press. Biology, e184,5. Semmann, D., Krambeck, H. J., & Milinski, M. (2004). Strategic investment Wedekind, C., & Milinski, M. (2000). Cooperation through image scoring in reputation. Behavioral Ecology & Sociobiology, 56, 248–252. in humans. Science, 288, 850–852. Sharp, S. P., McGowan, A., Wood, M. J., & Hatchwell, B. J. (2005). Wehner, R. (1987). ‘Matched filters’—neural models of the external world. Learned kin recognition cues in a social bird. Nature, 434, 1127–1130. Journal of Comparative Physiology A, 161, 511–531. Shennan, S. (2002). Genes, memes and human history. London: Thames and Wenseleers, T., Helantera, H., Hart, A., & Ratnieks, F. L. W. (2004). Worker Hudson. reproduction and policing in insect societies: an ESS analysis. Journal Sherman, P. W. (1977). Nepotism and the evolution of alarm calls. Science, of Evolutionary Biology, 17, 1035–1047. 197, 1246–1253. West, S. A. (2009). Sex allocation. Princeton: Princeton University Press. Silk, J. B. (2002). Kin selection in primate groups. International Journal for West, S. A., & Gardner, A. (2010). Altruism, spite and greenbeards. Parasitology, 23, 849–875. Science, 327, 1341–1344. Smith, J. (2001). The social evolution of bacterial pathogenesis. Proceedings West, S. A., Pen, I., & Griffin, A. S. (2002a). Cooperation and competition of the Royal Society of London, Series B, 268,61–69. between relatives. Science, 296,72–75. Smuts, B. (1999). Multilevel selection, cooperation, and altruism. Human West, S. A., Griffin, A. S., & Gardner, A. (2007a). Evolutionary Nature, 10, 311–327. explanations for cooperation. Current Biology, 17, R661–R672. Sober, E., & Wilson, D. S. (1998). Unto others: The evolution and psychology West, S. A., Griffin, A. S., & Gardner, A. (2007b). Social semantics: of unselfish behavior. Harvard, MA: Harvard University Press. altruism, cooperation, mutualism, strong reciprocity and group selection. Soltis, J., Boyd, R., & Richerson, P. J. (1995). Can group-functional Journal of Evolutionary Biology, 20, 415–432. behaviours evolve by cultural group-selection? An empirical test. Cur- West, S. A., Griffin, A. S., & Gardner, A. (2008). Social semantics: How useful rent Anthropology, 36, 473–493. has group selection been? Journal of Evolutionary Biology, 21,374–385. Spence, A. M. (1973). Job market signalling. Quarterly Journal of West, S. A., Kiers, E. T., Pen, I., & Denison, R. F. (2002b). Sanctions and Economics, 90, 225–243. mutualism stability: when should less beneficial mutualists be tolerated? Stearns, S. C. (1992). Evolution of life histories. Oxford: Oxford University Journal of Evolutionary Biology, 15, 830–837. Press. West, S. A., Griffin, A. S., Gardner, A., & Diggle, S. P. (2006a). Social Stevens, J. R., & Hauser, M. D. (2004). Why be nice? Psychological evolution theory for microbes. Nature Reviews Microbiology, 4, constraints on the evolution of cooperation. Trends in Cognitive 597–607. Sciences, 8,60–65. West, S. A., Gardner, A., Shuker, D. M., Reynolds, T., Burton-Chellow, M., Stevens, J. R., Cushman, F. A., & Hauser, M. D. (2005). Evolving the & Sykes, E. M., et al. (2006b). Cooperation and the scale of competition psychological mechanisms for cooperation. Annual Review of Ecology, in humans. Current Biology, 16, 1103–1106. Evolution, and Systematics, 36, 499–518. Westneat, D. F., & Fox, C. W. (2010). Evolutionary Behavioral Ecology. Taylor, C., & Nowak, M. A. (2007). Transforming the dilemma. Evolution, Oxford: Oxford University Press. 61, 2281–2292. Whitlock, M. C., Davis, B. H., & Yeaman, S. (2007). The costs and beenfits Taylor, P. D. (1990). Allele-frequency change in a class structured of resource sharing: reciprocity requires resource heterogeneity. Journal population. American Naturalist, 135,95–106. of Evolutionary Biology, 20, 1772–1782. 262 S.A. West et al. / Evolution and Human Behavior 32 (2011) 231–262

Wild, G., Gardner, A., & West, S. A. (2009). Adaptation and the evolution psychological science. Current Directions in Psychological Science, of parasite virulence in a connected world. Nature, 459, 983–986. 17,6–9. Wiley, R. H., & Rabenold, K. N. (1984). The evolution of cooperative Wilson, E. O. (1975b). [ROMAN]. Cambridge, Massachusetts: Harvard breeding by delayed reciprocity and queuing for favorable social University Press. positions. Evolution, 38, 609–621. Wilson, E. O., & Hölldobler, B. (2005). Eusociality: origin and Wilkinson, G. S. (1984). Reciprocal food sharing in the vampire bat. 308, consequences. Proceedings of the National Academy of Sciences of 181–184. the United States of America, 102, 13367–13371. Williams, G. C. (1966). Adaptation and natural selection. Princeton: Wolf, J. B., Brodie, E. D. I., & Moore, A. J. (1999). Interacting phenotypes Princeton University Press. and the evolutionary process II Selection resulting from social Wilson, D. S. (1975a). A theory of group selection. Proceedings of the interactions. American Naturalist, 153, 254–266. National Academy of Sciences of the United States of America, 72, Woolfenden, G. E. (1975). Florida scrub jay helpers at the nest. The Auk, 143–146. 92,1–15. Wilson, D. S. (1977). Structured demes and the evolution of group Wright, S. (1931). Evolution in mendelian populations. Genetics, 16, advantageous traits. American Naturalist, 111, 157–185. 97–159. Wilson, D. S. (2008). Social semantics: towards a genuine pluralism in the Wynne-Edwards, V. C. (1962). [ROMAN]. Oliver and Boyd: Edinburgh. study of social behaviour. Journal of Evolutionary Biology, 21, Young, A. J., & Clutton-Brock, T. H. (2006). Infanticide by subordinates 368–373. influences reproductive sharing in cooperatively breeding meerkats. Wilson, D. S., & Colwell, R. K. (1981). The evolution of sex ratio in Biology Letters, 2, 385–387. structured demes. Evolution, 35, 882–897. Young, A. J., Carlson, A. A., Monfort, S. L., Russell, A. F., Bennett, N. C., Wilson, D. S., & Dugatkin, L. A. (1997). Group selection and assortative & Clutton-Brock, T. H. (2006). Stress and the suppression of interactions. American Naturalist, 149, 336–351. subordinate reproduction in cooperatively breeding meerkats. Pro- Wilson, D. S., & Wilson, E. O. (2007). Rethinking the theoretical foundation ceedings of the National Academy of Sciences of the United States of of sociobiology. Quarterly Review of Biology, 327–348. America, 103, 12005–12010. Wilson, D. S., & O'Brien, D. T. (2009). Evolutionary theory and Young, H. P. (2003). The power of norms. In P. Hammerstein (Ed.), Genetic cooperation in everyday life. In S. A. Levin (Ed.), Games, groups, and Cultural Evolution of Cooperation. Cambridge: MIT Press. and the global good. Berlin: Springer-Verlag. Zeder, M. A. (2009). The Neolithic macro-(r)evolution: macroevolutionary Wilson, D. S., Vugt, M. V., & O'Gorman, R. (2008). Multilevel theory and the study of culture change. Journal of Archaeological selection theory and major evolutionary transitions: Implications for Research, 17,1–63.