Home > Positioning > Persons > Sober

Elliott Sober (1948–)

Sober is a philosopher of biology whose work has centred on two questions: what natural selection is (its logical structure, its relationship to other evolutionary forces, its explanatory reach) and at what level it operates (genes, organisms, groups, or all three simultaneously). His most influential contribution, with David Sloan Wilson, is the revival of multilevel selection theory — the argument that natural selection can and does operate at multiple levels simultaneously, and that the question is not which level is “real” but how selection at different levels interacts in particular cases. Unto Others: The Evolution and Psychology of Unselfish Behavior (1998) reopened a question that George C. Williams’ Adaptation and Natural Selection (1966) had closed for a generation, reframing it in terms that gene-centric selectionists and group-selectionists could both engage with.

Life

Born 6 June 1948. Undergraduate at the University of Pennsylvania (BA in philosophy). PhD in philosophy at Harvard (1974). Appointed to the Department of Philosophy at the University of Wisconsin-Madison, where he has spent his entire career. Hans Reichenbach Professor of Philosophy and Vilas Research Professor. Sober’s institutional base is philosophy, not biology, but his work engages directly with biological theory and is cited by biologists as well as philosophers — a cross-disciplinary position he shares with David Hull and a small number of others in the philosophy of biology.

His first book, The Nature of Selection (1984), established his approach: formal analysis of evolutionary concepts using the tools of analytic philosophy (causal modelling, probability theory, philosophy of science) applied to real biological questions. The collaboration with D. S. Wilson on Unto Others (1998) brought him to a wider biological audience. His later work has extended into the philosophy of statistics, the logic of phylogenetic inference, and the general philosophy of science.

Multilevel selection

The units-of-selection debate — at what level does natural selection operate? — has been one of the central conceptual controversies in evolutionary biology since the 1960s. The classical position, articulated by Williams (1966) and extended by Dawkins (1976), is that selection operates at the level of the gene: genes are the replicators, organisms are the vehicles, and what looks like group-level adaptation is better explained by individual or gene-level selection. Wynne-Edwards’ group-selectionism (1962) — the claim that organisms restrain their own reproduction for the good of the group — was the target Williams dismantled, and after Williams the burden of proof shifted decisively against group selection.

Sober and D. S. Wilson’s Unto Others (1998) reframed the question. Their argument is not a return to Wynne-Edwards’ naive group selection but a multilevel framework: selection can operate simultaneously at multiple levels (gene, organism, group), and the outcome depends on the relative strengths of selection at each level. A trait that is costly to the individual but beneficial to the group can spread if between-group selection (groups containing more altruists outcompete groups containing fewer) is strong enough to override within-group selection (selfish individuals outcompete altruists within each group).

The formal structure: in a population of groups, each containing a mix of altruists and selfish individuals, the proportion of altruists decreases within every group (within-group selection favours selfishness) but the groups with more altruists grow faster (between-group selection favours altruism). The global outcome depends on the balance. If group structure is strong — groups are distinct, migration between them is limited, and group-level competition is significant — then between-group selection can maintain altruism despite within-group disadvantage.

Sober and Wilson argued that this framework is not merely a mathematical possibility but describes real biological systems: social insects, microbial communities, and cases of human cooperation. The claim is not that group selection is always important but that it is sometimes important, and that a priori exclusion of group-level selection on theoretical grounds (as Williams and Dawkins advocated) is a mistake.

The nature of selection

The Nature of Selection (1984) analyses what natural selection is — its logical structure, not its biological details. Sober distinguishes selection of (what is selected — individual organisms, or traits within organisms) from selection for (what property is being selected for — a trait’s direct effects on fitness, as opposed to traits that are merely correlated with it). A trait can be selected without being selected for: green beetles may be more likely to be eaten by birds (selection against greenness), but if greenness is correlated with slower running speed, the actual selection is for speed, not against colour.

The distinction matters for adaptationist reasoning: to claim that a trait is an adaptation is to claim it was selected for, not merely that it was selected. The conflation of the two is one of the methodological errors that Gould and Lewontin criticised in “The Spandrels of San Marco” (1979).

Sober also distinguishes natural selection from genetic drift as evolutionary forces. Both change gene frequencies; they differ in structure. Selection is a directional force — it pushes populations toward higher fitness. Drift is a random process — fluctuations in gene frequency due to finite population size, with no directional tendency. The two interact: in small populations, drift can overpower selection; in large populations, selection dominates. The distinction is conceptually clean; measuring the relative contributions of selection and drift in any particular case is empirically difficult.

Where Sober stops

The multilevel selection framework is formal and general — it specifies the conditions under which group-level selection can override individual-level selection. Whether those conditions obtain in any particular biological system is an empirical question that the framework identifies but does not answer. Critics have argued that the framework is too permissive — that almost any population structure can be redescribed in multilevel terms, making the question of whether group selection “really” operates undecidable. Steven A. Frank and Andy Gardner have argued that multilevel selection and inclusive fitness (Hamilton’s framework) are mathematically equivalent — different accounting methods for the same underlying processes. If this equivalence holds, the debate is not about biology but about bookkeeping: which framework is more useful for asking particular questions. Sober has acknowledged the formal equivalence in some formulations while maintaining that the frameworks differ in what they foreground — that multilevel selection makes group-level processes visible in a way that inclusive fitness does not.

The extension of multilevel selection to human cooperation has been pressed on its empirical basis. Sober and Wilson argue that human cultural group selection — groups with cooperative norms outcompeting groups without them — is a significant force in human evolution. The claim depends on the strength of cultural between-group variation relative to within-group variation, a ratio that is empirically contested. Whether human groups are sufficiently distinct (culturally and genetically) for group-level selection to operate with significant force is debated by both evolutionary biologists and anthropologists. The formal framework identifies the relevant variables; the empirical values remain uncertain.

Sober’s distinction between selection-of and selection-for clarifies adaptationist reasoning but does not resolve the adaptationist debate itself. The distinction tells you what a correct adaptationist claim would look like; it does not tell you whether any particular adaptationist claim meets the standard. The methodological problem Gould and Lewontin identified — that plausible adaptive stories are easy to construct and hard to test — remains after Sober’s analysis, restated more precisely but not dissolved.

Key works

See also: Williams · Dawkins · Hamilton · Wynne-Edwards · Darwinism