The Dialectics of Multi-Level Selection in Evolutionary Biology

Introduction

Dialectical thinking in biology has famously been put forward by the eponymous book of Levins and Lewontin (1985). Much of the book was dedicated to the interpenetration between organisms and their environments. In particular, chapter 3, “The Organism as the Subject and Object of Evolution”, originally published by Lewontin (1983), explored how organisms simultaneously change and respond to their environments. This contrasted with two one-way causal models that commonly infuse biology: (1) on developmental timescales the phenotype is determined by the genotype, and (2) on evolutionary timescales the phenotype is determined by the environment, through the selection of genotypes. Levins and Lewontin’s take on the reciprocal causation between organisms and their environments contributed to the elaboration of dialectical accounts of development (e.g. Oyama, Gray, and Griffiths 2001; Sultan 2015) and of selection pressures (e.g. Odling-Smee, Laland, and Feldman 2003; Svensson 2018).

However, one of Lewontin’s most cited and influential contributions is strikingly absent from the Dialectical Biologist. In “The Units of Selection”, Lewontin (1970) argues that Darwinian evolution by natural selection can in principle apply to any level of biological organisation, provided that heritable phenotypic variability cause differences in fitness. This was a major step towards the emergence of what is called today Multi-Level Selection (MLS) theory (e.g. Okasha 2006; Wilson and Sober 1989), according to which several levels of biological organization can simultaneously be the target of selection.In the 60’s and 70’s the debate focused on group selection, which may counterbalance selection at the level of organisms in some cases. In his book ”Adaptation and Natural Selection”, Williams (1966) famously argued against the importance of group selection in nature,although he acknowledged that it was sound, in principle. He pointed out that Lewontin and Dunn (1960) and Lewontin (1962) ”has produced what seems to me to be the only convincing evidence for the operation of group selection” (Williams 1966, p. 117), which will be detailed in Section 5.

Given the importance of Lewontin’s contribution to MLS theory, it is surprising that it has not be discussed in the Dialectical Biologist. Inspired by Engels’s Dialectics of Nature, Levins and Lewontin derive in the concluding chapter of the Dialectical Biologist some general ”dialectical principles” that apply to the organic world, often in the light of the reciprocal causation between organisms and their environments. These principles include (Levins and Lewontin 1985, p. 273-280):

1. ”a whole is a relation of heterogeneous parts that have no prior independent existence as parts”
2. ”the properties of parts have no prior alienated existence but are acquired by being parts of a particular whole”
3. ”the interpenetration of parts and wholes is a consequence of the interchangeability of subject and object, of cause and effect”
4. ”because elements recreate each other by interacting and a recreated by the wholes of which they are parts, change is a characteristic of all systems and all aspects of all systems”
5. ”persistence and equilibrium are not the natural state of things but require explanation, which must be sought in the actions of the opposing forces”

The present contribution aims to sketch some dialectical aspects of MLS theory, following the above principles.

Parts and wholes

The first and second principles lie at the core of MLS theory, according to which properties of parts result, in part, from selection pressures exerted on the wholes they form. In the case of the evolution of altruism¹ for instance, behaving altruistically is an individual property that is negatively selected at short time scales by individual-level selection inside each group, and positively selected at longer time scales during selection among groups. Selection among groups occurs because containing many altruists is a beneficent group-level property. MLS theory is at odds with the naive version of holism which asserted that individual traits evolved for the preservation of the species (“Arterhaltend”, Lorenz 1963), in line with the old concept of Balance of Nature. As Dawkins (1981) ironically put: “Should we then not expect lions to refrain from killing antelopes, ‘for the good of the mammals’ ?”. As indicated by its name, MLS theory considers that several levels of selection are acting simultaneously, with no predetermined outcome. This point will be further elaborated in Section 5.

In contrast to MLS theory, according to the “gene-eyes view” (Ågren 2021) championed by Dawkins (1976) and others, selection acts at the level of genes only. Altruistic alleles² can be selected if they spread more efficiently indirectly through the bearer’s relatives (“kin selection”), following Hamilton’s rule (Hamilton 1964). According to the gene-eyes view, the group-level property “containing many altruists” is therefore explained by selection at the level of genes. Even the individual-level property “behaving altruistically” is explained by lower-level processes at the gene-level. In this reductionist move, genes constitute the ontological basement of evolutionary theory. MLS theory instead denies the existence of a fundamental unit of selection. This commitment of MLS theory stands in line with the concluding chapter of the Dialectical Biologist:

MLS theory is precisely focusing on the relations between selection pressures acting on adjacent levels of organization.

In the context of group selection, it has repeatedly been shown that MLS theory and kin selection theory are mathematically equivalent (Lehtonen 2016; Lion, Jansen, and Day 2011; Smith 1998; van Baalen and Rand 1998). Therefore, selection coefficients can be attributed to genes in order to predict accurately the evolutionary dynamics. However, the levels-of-selection question is not at what level should fitness be counted (bookkeeping), but at what level(s) fitness is caused (Gould 2002; Walsh 2004). As Sober and Lewontin (1982) put it: ”To say that objects differ in fitness is not yet to say why they do so”. Fitness differences between individuals are therefore caused by selection acting simultaneously among groups and among individuals which illustrates the second principle: ”the properties of parts have no prior alienated existence but are acquired by being parts of a particular whole”.

Subjects and objects

According to the third principle, subjects and objects are interchangeable. The usual theory of evolution by natural selection instead favours linear causation, where populations of organisms respond to external selection pressures dictated by the environment. Selection pressures are conceived as independent causes that remain unaltered during the process of adaptation. Standard evolutionary theory has therefore been qualified as “externalist” (Godfrey-Smith 1996; Laland et al. 2013). Dialectical accounts of biology like the Developmental Systems Theory and Niche Construction Theory (Odling-Smee, Laland, and Feldman 2003; Oyama, Gray, and Griffiths 2001), instead rely on the reciprocal causality between organisms and their environments. In particular, Niche Construction Theory stresses that selection pressures are not independent causes, since they can be modified by the activity of the populations subject to them.

MLS theory also reach the conclusion that selective forces act both as subjects and objects. In the Descent of Man, Darwin famously provided the first sketch of what is called today group selection. He wrote:

Inspired by Engels’s law of the negation of the negation, historian of science Patrick Tort interpreted this as the ”reversing effect” of evolution (Tort 2012). During this (multilevel) process, natural selection favors the evolution of social traits in human populations. Social traits lead humans to take care of each other, thereby negating the strength of selection, in return. Here again, natural selection is both a subject and an object, a cause and an effect. In the Dialectical Biologist, Levins and Lewontin write: ”the entities that are the objects of laws of transformation become subjects that change these laws”.

In comparison to Richard Levins and Richard Lewontin, John B.S. Haldane has received less attention (but see Gouz 2012). However, Haldane expressed related views almost 50 years before the publication of the Dialectical Biologist (Haldane 1937) and was also influenced by Engels’s Dialectics of Nature, to which he wrote a preface (Haldane 1940). Together with Sewall Wright (1945), Haldane was one of the early population geneticists to support what was later called MLS theory, with a reasoning similar to that of Darwin:

In the last sentence of the above quote, by the way, Haldane points out that the strength of selection between individual humans might become high again following the disappearance of ”small endogamous groups”. Since the development of large societies is the product of mankind, this further stresses that natural selection can be altered by its own target.

In addition to group selection, Haldane also tackled species selection. While he was participating to the defense of Madrid (modestly, according to him) in 1937 during the Spanish civil war, he wrote:

This assertion might be questionable from a paleontological point of view, but the matter is elsewhere: Haldane points out that short-term selection between individuals can be negated by long-term selection between species, an argument that has also been raised to explain the evolution of sex (Gouyon, Vienne, and Giraud 2015).

Change

The fourth principle claims that change must be sought in “all aspects of all systems”, that processes are ontologically prior to objects. The idea that natural selection is both a cause and an effect, as outlined in the previous section, reveals that natural selection itself is a historical product subject to change, especially in a MLS context where higher levels of selection modify previous selection regimes.

On the other hand, the centrality of change might appear at odds with MLS theory, which is grounded on the existence of biological objects belonging to several levels of selection. However, during the last decades the levels-of-selection problem has shifted from a synchronic to a diachronic perspective that aims to explain the evolutionary origin of selection levels (Okasha 2006). This move started in the 90’s with the book of Maynard Smith and Szathmáry (1995) on Major Evolutionary Transitions, also called Evolutionary Transitions in Individuality (ETIs). The study of ETIs focuses on the processes leading to the emergence of higher levels of selection, instead of taking them for granted. Higher levels of biological organization resulting from an ETI therefore become an explanandum. Following Lloyd and Wade (2019): ”multilevel selection precedes and attends, rather than follows, an evolutionary transition.”

Change therefore infuse all aspects of MLS theory, from the nature of the levels to natural selection itself. During the transition towards multicellularity for instance, previously autonomous cells belonging to the same lineage have formed multicellular organisms (Michod 2006). During this process, selection between cells is weakened and ultimately negated by selection between organisms. The emergence of organisms change the properties of natural selection, which further strengthens this new level of organization.

Struggle of opposites

Levins and Lewontin’s fifth principle stresses that processes are fueled by actions of opposing forces, paraphrasing Lenin (1925)³: “Development is the struggle of opposites.” and inspired by the law of the unity and conflict of opposites formulated by Engels (1873-1886). Although this has not been formulated explicitly by advocates of MLS theory, the struggle of opposites is one one its central aspects, as in the example provided by Lewontin (1970) and summarized by Williams (1966) as follows (emphasis mine):

As emphasized in italics, selection pressures acting at the level of genes are opposed to those acting at the level of organisms. Although this contradiction can be resolved from a gene-centered perspective, considering that the short-term fitness of genes need not be aligned with their long-term fitness (Bourrat 2015), the causal power of the genic and organismic selection levels are nevertheless opposed. This is also particularly apparent in the current notion of ”intragenomic conflict” that is now used to designate cases like the t-allele (Gardner and Úbeda 2017).

Section 4 presented how new selection levels can emerge. Michod (2006) suggested that during an ETI fitness is ”transferred” from the lower to the higher level, for instance from the cell to the multicellular organism⁴. This implies an intermediate situation of fitness duality where both levels bear fitness, leading to MLS. This has some similarities with the concept of ”dual power”⁵ in the soviet literature. For instance, Lenin wrote in his ”April Theses” that:

The notion of fitness transfer also has an analogy in Trotsky (1930): ”The political mechanism of revolution consists of the transfer of power from one class to another.” Once an ETI is complete, selection at the lower level becomes negated by the emergence of the lower level, as in the case of worker policing in eusocial insects (Ratnieks and Visscher 1989). Lenin (1917, May 20) thought that ”dual power cannot last long”, which might be true in the history of social revolutions, as well as in some ETIs. However, during the history of life many MLS situations do not lead to complete ETIs where selection at the lower level rarely occurs, but instead remain under the regime of fitness duality.

Quantity and quality

Although they don’t incorporate it in their “official” list of principles, Levins and Lewontin also mention on several occasions Engels’s law of the passage of quantitative changes into qualitative changes, it self derived from Hegel. Trotsky previously noted that the Darwinian theory of the origin of species is an instance of this law, where the continuous transformation of species can lead to the formation of a new species⁶:

In the context of MLS theory, Svensson (2018) suggested that the ETI research program can also be interpreted as a case of transformation of quantity into quality, since ETIs represent discrete stages that supervene on a continuous evolutionary process. With respect to ETIs, Okasha (2005b) suggested a three-step model: (1) collective fitness is first defined as average particle fitness, (2) collective fitness is not defined as average particle fitness any more, but is still proportional to it, and (3) collective fitness is neither defined as nor proportional to average particle fitness. Although it is still unclear whether this model accurately represents an ETI (e.g. Bourrat 2025), it nevertheless illustrates how quantitative evolutionary changes in the way particles interact may lead to the qualitative emergence of collective fitness. Moreover, an ongoing debate in MLS theory questions the ontological status of the levels of selection (Boucher 2023; Bourrat 2021; Wilson 2007). Discrete selection levels might either result from mere conventions, or constitute real entities. The dialectical notion of transition from quantity to quality would rather support the realist stance.

Conclusion

I do not contend that the similarities between several dialectical aspects of social history and MLS theory outlined here correspond to deep ontological homologies. It would for instance make little sense to hold that the state of dual power in 1917 is precisely the same thing as dual fitness during an ETI. Rather than homologies, I believe these similarities constitute interesting analogies, that may spring from two complementary aspects of dialectics: habits of thought and ontological properties.

The dialectical principles sketched by Levins and Lewontin, and before them by Engels, cannot serve as a magic wand. Instead, dialectical principles should be used undogmatically as epistemic tools that can themselves be altered during this process. Dialectical principles reveal habits of thought shared by several biologists both interested in MLS theory and in Marxism (which does not entail lifetime adhesion): John B.S. Haldane, John Maynard-Smith (one of Haldane’s students)⁷ , Richard C. Lewontin, Stephen Jay Gould, Elliott Sober. This pedigree may help to explain why MLS theory bears many aspects of dialectical thinking.

On the other hand, dialectical principles may reflect real properties of the world. Unlike other Marxists like e.g. György Lukács (1923), Levins and Lewontin adopt in the Dialectical Biologist a realist position regarding dialectical principles, that are conceived as foundational ontological facts (Deeg 2023). Many MLS theorists were not aware of dialectical essays, or did not perceive any relation between these two fields. The fact that MLS theory is absent from the Dialectical Biologist constitutes a striking example. Therefore, the dialectical characteristics of MLS theory may partly result from real aspects of the biological world. Following Trotsky (1999): ”You can’t just foist dialectics on facts, but must derive it from the facts, from their nature and their development.”

¹ Altruism may have a distinct meaning in evolutionary biology and in the social sciences or in everyday life. In short, altruism is a behaviour that decreases the fitness of its bearer and increases that of the beneficiaries.
² It is a common line of thought in evolutionary biology to suppose that some alleles increase an individual’s propensity to behave altruistically. This assumption is relaxed in cultural evolutionary models (e.g. Bowles 2009).
³ Written in 1915 in Bern, published in Moscow in 1925 after Lenin’s death. Translation: Marxists Internet Archive
⁴ See Takacs, Doulcier, and Bourrat (2024) for an alternative account of ETIs
⁵ The Russian term dvoevlastie may also be translated as ”double sovereignty”. https://www.marxists.org/archive/trotsky/1930/hrr/ch11.htm
⁶ This anagenetic perspective of speciation has later been challenged by e.g. Eldredge and Gould (1972)
⁷ The complex and changing position of John Maynard-Smith regarding MLS theory has been analysed by Okasha (2005a)