Sociobiology, Broad and Narrow

There are two different senses in which the term “sociobiology” is un-derstood. In the broad sense, it simply means all study of animal soci-ality and social behaviour from an evolutionary functionalist perspec-tive (see for example Alcock 2003). In this sense, the object of this dis-sertation is sociobiology. In the narrow sense, it refers to a particular school of doing so (call this classical sociobiology if you will), which hap-pened to be the school that started this line of research. Edward O.

Wilson (1975) was behind the name “sociobiology”⁸⁰ and was respon-sible for the first systematic collection of its methodological tools and for building a unified theory, but the tools as such came mostly from theoretical biologists such as Robert Trivers and William Hamilton.

The two central ideas of sociobiology were the gene’s eye perspective on adaptivity of behaviour and the use of game theory to analyse social situations, both of which have become fixed features of evolutionary approaches to social behaviour. Both approaches are usually consid-ered to be individualistic approaches, although they can be argued to be neutral in this regard – Elliot Sober and David Sloan Wilson, for example, have argued that some cases in which these tools are applied are cases of group selection (Wilson & Sober 1994; Sober & Wilson 1998; see also Okasha 2006 and Birch 2017). Furthermore, the sociobi-ological tools make no assumptions about the proximate or develop-mental mechanisms at all. I will now review some key aspects of

80 The term itself was introduced as early as the 1940s, but it did not gain a fixed reference and did not see any particular use before Wilson (Plotkin 2004: 105).

151

classical sociobiology, and also some other tools of sociobiology in the broad sense.

4.1.1. The New Synthesis

The study of behaviour has always been a part of biology, but the pi-oneers of modern animal behavioural biology, ethology, with evolu-tionary component as one of its central parts, were Karl von Frisch (1886–1982), Konrad Lorenz (1903–1989), and Nikolaas Tinbergen (1907–1988), who shared the 1973 Nobel Prize in Physiology or Medi-cine for this.⁸¹ The approach became an established research area within zoology by the early 1950s. As we have seen, the ethologists thought from the very beginning that there are several biological ways to approach behaviour, including evolutionary approaches. The bio-logical approach, generally speaking, was expanded to comprehend human behaviour by ethologists in the 1960s and 1970s (for example, by Konrad Lorenz (1966 [1963]) and his student Iranäus Eibl-Eibefeldt in a more academic context, and by Desmond Morris in a series of popular books; see Laland & Brown 2002; Plotkin 2004). This human ethology challenged the culture-centred view of many anthropologists of the time, replacing it with a zoologized view of humans. At the same time, they acknowledged the many peculiarities of human be-haviour (such as culture) and their own methodological limitations in studying humans properly (Laland & Brown 2002, 59–64)). However, early ethological work influenced some anthropologists, including Li-onel Tiger, Robin Fox (for example, Tiger & Fox 1971), and Donald Symons (1979).

Human ethology was arguably based on evolutionary function-alism, guided by the idea that animal behaviour can only be

81 Their studies, of course, built on existing tradition, starting from Charles Darwin himself (Darwin 1872), with most important prior advances arguably made by Oskar Heinroth, Charles Otis Whitman, and Julian Huxley (Burkhardt 1981; 2005).

152

understood in natural environments – laboratory experiments are in-adequate or at least insufficient – but it was not restricted to study of function. Ethologists were also interested in proximate causal mecha-nisms and developmental processes without an evolutionary perspec-tive, and much of their evolutionary attention was in building phylo-genetic trees, not in discovering adaptive functions. (See Laland &

Brown 2002; Plotkin 2004.) The tools of sociobiological analysis started to appear during the 1960s and 1970s (for example, Hamilton 1964a, 1964b & 1970; Smith 1964; Trivers 1971 & 1973; Maynard-Smith & Price 1973). Edward O. Wilson collected the methods in a sys-tematic theory in his field-coining book Sociobiology, subtitled The New Synthesis, referring to the expansion of the “old synthesis” of Darwin-ian evolutionary theory and MendelDarwin-ian genetics into the modern evo-lutionary biology (Wilson 1975). Wilson’s Sociobiology and Richard Dawkins’s The Selfish Gene (1976), which is a more philosophical take on the guiding principles of the new approach and its gene-centred ontology, popularized the approach inside and outside academia. The approach pushed proximate and developmental questions and phy-logenetic considerations into the background, concentrating on the function of behaviour. The first human applications of the approach appeared by the late 1970s, amid fierce controversy (see Segerstråle 2000). In retrospect, leaving the proximate and developmental ques-tions aside was an obvious step for the worse from the more tradi-tional approach of ethology, and this was partly guided by mistaken ideas such as very strong adaptationism (Gould & Lewontin 1979;

Kitcher 1985) and unwarranted behaviourism (which also reflected the differences in approaches between European and American psy-chological as well as zoological traditions; see Segertråle 2000; Laland

& Brown 2002), but an important motivation behind sociobiology was the emergence of new theoretical tools, such as kin selection and game-theoretical models, to understand social behaviour from the evolutionary point of view. A charitable reading of this new approach

153

is that its proponents got a bit overexcited about the new methods and approaches.⁸²

The most central theoretical idea of the approach was the gene’s eye view, or gene-selectionism, which switches attention from the fit-ness of individuals to the fitfit-ness of genes. Although this shift of atten-tion is away from individuals, it was built on direct criticism of group selection models. The early proponents of group selection (for exam-ple, Wynne-Edwards 1962) thought that the adaptation of behaviour is sometimes for the good of the group. The individualists (for example, Maynard-Smith 1964 and Williams 1966) pointed out that even if in-dividuals behaving for the good of the group make the whole group fitter, behaviour that makes the individual fitter than other individu-als in the group (where all individuindividu-als get the benefits of being in the good group) will still be selected. This means that individual benefits trump group benefits every time they are in conflict. The logic in the gene-selectionism, however, is that it is the genes of the individuals that matter. The genes that are associated with behaviour that makes those very genes better, no matter in which individual, is selected. If a gene is associated with behaviour that promotes the fitness of the cop-ies of the same gene in other individuals, it can be selected.⁸³ One cru-cial aspect of this move is to distinguish the different functions of the copying entity and the phenotypic entity in the logic of the selection pro-cess, as explicated by Dawkins (1976; see also 1982) and elaborated by David Hull (1980, 1981 & 1988a). The copying entity (replicator) is the proper carrier of fitness, while the phenotypic entity (Dawkins’s vehi-cle, Hull’s interactor) interacts causally with the environment in ways that determine the fitness of the replicators. I have already discussed some complications to this idea, raised by the complexity of develop-mental processes that are left black boxes in this image. The whole

82 Furthermore, as I explained in the previous chapter, black-boxing proxi-mate, developmental and evolutionary-historical mechanisms makes some sense, although this was a mistake nevertheless.

83 What exactly “gene” is referring to here is a more complicated question that will be returned to later.

154

approach has been since questioned (for example, Griffiths & Gray 1994; Oyama, Griffiths & Gray 2000; Godfrey-Smith 2009). I will return to the problems and the remaining insights of this distinction later.

4.1.2. Kin Selection

The key notion of gene-selectionism is the concept of inclusive fitness.

It was introduced by William Hamilton (1964a & 1964b), based on pre-vious work by Ronald Fisher (1930) and John Haldane (1932 & 1955).

Inclusive fitness consists of two parts: direct fitness, which refers to the positive effects the trait has on the organism’s own reproductive ca-pacity, and indirect fitness, which refers to the positive effects the trait has on the fitness of other organisms who share the same genetic basis for the trait. Both promote the selection of the underlying genetic basis that the trait is connected to. This is the basis for the effect that John Maynard Smith (1964) coined kin selection and Robert Trivers (1985) proclaimed to be the most important idea in theoretical evolutionary biology since natural selection itself: helping your kin to increase the fitness of your own genes. Formally put, when

ai is the direct positive fitness effect the trait has on the individual i, b^ij is the positive fitness effect i has on the individual j,

c^ij is the negative fitness effect on i for having the effect on j,⁸⁴ rij is the multiplier from the degree of relatedness between i and j, and

wⁱ is the inclusive fitness of i,

the inclusive fitness can be calculated from the equation wⁱ = aⁱ – c^ij + Σ r^ij b^ij

84 The behaviour may be costly, in which case there is a loss in absolute fitness.

But even if there is no such loss, merely helping someone has a negative fit-ness effect on relative fitfit-ness. I will return to this (and to this distinction) later.

155

where sigma refers to the sum of the fitness effects from all relevant individuals. The value of the relatedness multiplier is the same as the degree of the shared genes (when only those genes in relation to which there is variation in the population in the first place are considered) on average – in diploidic organisms, the multiplier between the par-ents and offspring, as well as between siblings, is 0.5, between first cousins 0.25, and so on. These multipliers, however, presuppose that the parents are not related at all, and the population is infinite. Relax-ing these unrealistic idealizations decreases the multiplier. In a popu-lation consisting of only one family of two generations, the multiplier would always be zero. According to Hamilton’s rule, which is the cen-tral equation of kin selection, altruistic behaviour of i towards j can get selected, if rb – c > 0, that is, if rb > c. (Futuyma 1998, 595–596.)

I will return to kin selection later, but two observations should already be made here. Relatedness as such is not a causal factor. What is important is that the gene to be selected is associated with behaviour that benefits individuals who have that same gene, which includes the individual themselves and some others. The explanatory power of Hamilton’s rule does not come from relatedness, and even less from the overall shared genome, but from the likelihood of the gene in ques-tion existing in the other individuals, which is higher the more closely related they are. There is a correlation between the amount of shared in-heritance due to relatedness and the probability of sharing the gene. The first is not explanatory, which makes “kin selection” an unfortunate phrase.⁸⁵ Another important point is that the two causal mechanisms in kin selection (the direct fitness benefits and the indirect fitness) are very dissimilar factors. Direct fitness has to do with the individual’s reproductive capacity compared to other individuals, whereas indi-rect fitness makes sense only through the structure of the population and therefore requires a different mechanistic explanation even if the two factors could be modelled in the same equation. This has one

85 This is not exactly so straightforward, but it is sufficient for now. I will dis-cuss kin selection in greater detail in the final chapter.

156

particularly important consequence. Although sociobiology was in-terpreted as an individualist account of social evolution by its propo-nents and critics alike, this is not the only possible interpretation of these tools. Even if genes were the only replicators that matter (Daw-kins 1989 [1976]), both individuals and groups might still be the real-izers of the interactions that the genes are selected for – in other words, the vehicles or interactors (Wilson & Sober 1994). Hamilton himself, for one, thought kin selection was a form of group selection (Hamilton 1970) after being convinced by George Price (1970) on this point. Elliot Sober and David Sloan Wilson (for example, Wilson & Sober 1994; So-ber & Wilson 1998) have made a strong argument for this, too. I will return to this issue in more detail in the final chapter of the dissertation.

4.1.3. The Evolutionary Game Theory

The other central theoretical element in sociobiology is the use of evo-lutionary game theory, developed especially by John Maynard Smith (Maynard Smith & Price 1973; Maynard Smith 1982). Social interaction can be modelled as a game where different options for behaviour have different fitness consequences depending on how the other(s) in the situation behave. The games are repeated and the number of rounds is unknown, since the focus is on the behavioural traits (or patterns of behaviour), not individual instances of social behaviour. The overall fitness of any behavioural disposition depends on the frequency of encounters with the various types of behaviour, and in an evolving population (with heritable behavioural dispositions) the dynamical selection process (that is, the very selective environment that evolves in the process), too, can be modelled as a series of games – this addi-tion of the dynamic aspect is what differentiates evoluaddi-tionary game theory from classical game theory. It is used specifically to understand conflicts and cooperation, but not only that. The central idea is that if the formal system has an evolutionarily stable strategy (ESS; Maynard Smith 1983), that is, no other alternative strategy can replace it, the sys-tem will ultimately reach it. The strategies are obviously idealizations,

157

not descriptions of actual behavioural traits, and they concentrate solely on the emerging fitness effects. Furthermore, the games are about consequences, not what the “players” are “aiming for” in the situations.

This means that a social setting may have a certain game-theoretical structure measured in fitness without the concrete social interaction (defined with the individuals’ preferences, for example) having an iso-morphic game-theoretical structure. The illustrations usually simplify this by concentrating on context where something concrete with a fit-ness consequence (for example, resources or other direct conse-quences of the behaviour) is “in play”. The structure of the evolution-ary game may correlate with the structure of behavioural-level aims, but not always, and these structures cannot be collated. Still, they are useful tools for understanding the dynamics in interactions.

There are different games to capture different social situations, but the classic of this approach (and the one that will be examined later) is the formalization of Robert Trivers’s (1971) theory of reciprocal altruism. The starting point for this is the intuitive idea that mutual help makes the individuals participating in the interaction fitter than individuals who do not participate. The possibility of free riding (tak-ing help and not reciprocat(tak-ing) should, however, make such tenden-cies unlikely to evolve. This situation is a classic case of prisoner’s di-lemma. If we have only defecting and cooperative strategies, it is always better to defect, no matter what the other player chooses to play.⁸⁶ However, if reciprocity is an option, it may be the winning strategy.

Reciprocation can be modelled as a Tit-for-Tat strategy (TFT), where the player cooperates first, and then reacts to however the partner played the previous round – keep cooperating if they did, otherwise

86 For example, if cooperation costs one unit (of whatever has robust effects for fitness) but gives three units of benefit for the partner, defecting against another defector leaves you with zero units, whereas cooperating would only cost you a unit, and defecting against a co-operator pays three units, whereas cooperating would only grant two units. The only ESS in this setting is defect-ing with zero gain, even though everyone cooperatdefect-ing would give two units for everyone.

158

defect. This strategy does worse than defecting on the first round, but if there are enough reciprocal players and/or pure co-operators, it may do better overall. If it invades the population, it becomes ESS.⁸⁷ (Fu-tuyma 1998, 584–586; Hargreaves Heap & Varoufakis 1995, 197–198;

Sober & Wilson 1998, 79–80.)

Once again, the game-theoretical models have usually been in-terpreted as individualist models, and understandably, so – they model individuals and their behavioural strategies in social situations.

But, once again, there are alternative interpretations in which game-theoretical models are sometimes interpreted as descriptions of group-level selection processes (see Sober & Wilson 1998). Further-more, it is not so clear what the abstract strategies and games are mod-elling in the real world. I will get back to these issues, too.

4.1.4. Group Selection

Not all early models of social evolution were individualistic. David Sloan Wilson had already presented his trait group model during the emergence of sociobiology (Wilson 1975), although much of the foun-dational theoretical work and its popularization was done much later in the 1990s, notably as a collaboration between Wilson and Elliot So-ber (Wilson & SoSo-ber 1994; SoSo-ber & Wilson 1994 & 1998; see also SoSo-ber 1980a & 1984 and Wilson 1989). Other pioneers of both theoretical and empirical work include, for example, Charles Goodnight and Lori

87 TFT does not automatically do better than defecting – it does not necessarily invade the population. But if it does, it is able to become fixed. There are also a lot of ways to make reciprocal strategy better in realistic ways, given some cognitive capacities. For example, in observer-TFT the player chooses the strat-egy of the first round based on previous observations about the partner’s strategies (Pollock & Dugatkin 1992), and in a variation of this the player uses others’ attitudes as a clue for this (Castro et al 1998). In strong reciprocity (Gintis 2000a) the defectors are punished – this is a problematic case, since even if this can make defecting unfavourable, punishing may be costly. (This may be a reason why, for example, lions tolerate free-rides.)

159

Stevens (Goodnight, Schwartz & Stevens 1992; Stevens, Goodnight &

Kalisz 1995; Goodnight & Stevens 1997). Much of the later discussion on individualism and holism has been about what counts as group se-lection – whether kin sese-lection and the evolution of reciprocal altruism are really forms of group selection, for example, as mentioned above (see Sober & Wilson 1998; Okasha 2006; Goodnight 2012; Birch 2016).

Wilson’s trait group model, however, is an unambiguous case of group selection in which the group level differences direct selection and override the invasion of selfish phenotypes from within. Even if a selfish (individual-benefitting) type is always fitter than the altruis-tic (group-benefitting) type within every group, all individuals in a group with more altruists are fitter, which may make altruists fitter in the population overall. If there is a mechanism that keeps this struc-ture constant, group selection takes place. However, the issue of indi-vidualism and holism (as defined in the introduction) is partly about how we should interpret the sociobiological (in the broad sense) mod-els as modmod-els of causal mechanisms in evolution. The fitness consequences that follow certain patterns (specified by the models) are a necessary condition for selection but modelling only the consequences does not say anything about the mechanisms that cause these patterns. This is a matter of relevant causal processes on all three levels discussed:

proximate, developmental, and evolutionary. Group selection will be the topic of the final chapter.

4.1.5. Biological Markets

The Biological Markets Theory put forward by Ronald Noë and Peter Hammerstein (1994, 1995 & 2016) is a major new development in the

The Biological Markets Theory put forward by Ronald Noë and Peter Hammerstein (1994, 1995 & 2016) is a major new development in the