Trees of Life pp 133-161 | Cite as

Selection, Drift, and the Aims of Evolutionary Theory

  • Timothy Shanahan
Part of the Australasian Studies in History and Philosophy of Science book series (AUST, volume 11)


According to textbook presentations of evolutionary theory, evolutionary change is a result of the interaction of a number of biological processes that together shift a population away from Hardy-Weinberg equilibrium. Among the factors typically mentioned are genetic mutation, gene flow (emigration and immigration), nonrandom mating, selection, and drift (“chance”).1 By constructing equations which factor in specific values for each of these processes, evolutionary biologists try to explain why a population follows a particular evolutionary trajectory. Hence, much of evolutionary biology is concerned with the empirical determination of values for each process, and the ways in which the various processes can and do interact with one another to produce evolutionary change.


Evolutionary Theory Selection Event Sampling Event Differential Survival Biological Entity 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Acknowledgements: I would like to express my sincere gratitude to Elliott Sober, Paul Griffiths, Susan Oyama, John Endler, Linda Zagzebski, and an anonymous Refseree for helpful suggestions on an earlier draft of this paper.Google Scholar
  2. 1.
    See, for example, D. Futuyma (1986), Evolutionary Biology, 2nd ed. (Sunderland, MA: Sinauer), pp. 85–87.Google Scholar
  3. 2.
    As John Endler has pointed out to me, most biologists do not think of selection and drift as distinct processes, but rather as ends of a continuum. This can be expressed quantitatively by noting that selection may take on any value from zero to a large value. “We speak of selection and drift as ends of a continuum, to focus our attention on the mechanisms and to show that they operate” (Endler, personal communication). Part of my aim in this paper is to present novel reasons in support of this view, and to draw out its implications for understanding evolutionary theory.Google Scholar
  4. 3.
    J. Beatty (1984), ‘Chance and Natural Selection’, Philosophy of Science 51:183–211.CrossRefGoogle Scholar
  5. 4.
    H. Kettlewell (1973), The Evolution of Melanism (Oxford: Oxford University Press).Google Scholar
  6. 5.
    Beatty (1984) ‘Chance and Natural Selection’, Philosophy of Science 51, p. 189.CrossRefGoogle Scholar
  7. 6.
    M. Scriven (1959), ‘Explanation and Prediction in Evolutionary Theory’, Science 130:477–482;CrossRefGoogle Scholar
  8. 6a.
    R. Brandon (1978), ‘Adaptation and Evolutionary Theory/ Studies in History and Philosophy of Science 9:181–206;CrossRefGoogle Scholar
  9. 6b.
    S. Mills and J. Beatty(1979), The Propensity Interpretation of Fitness’, Philosophy of Science 46:263–286;CrossRefGoogle Scholar
  10. 6c.
    J. Beatty (1984) [note 3]; E. Sober(1984), The Nature of Selection (Cambridge, MA: MIT/Bradford Press).Google Scholar
  11. 7.
    In this paper, I am treating drift as the “chance” factor in evolutionary theory. This is in keeping with most formal and informal presentations of the theory. The concept of “chance” has had and continues to have other meanings in evolutionary biology, such as coincidence, ignorance of causes and as “accident”. I have explored these different notions elsewhere; space prohibits me from doing so here.Google Scholar
  12. 8.
    The class of issues at stake in considering the selection/ drift distinction all concern cases in which factors other than selection coefficients (assignments of fitness values) affect the fates of organisms, populations, and species. According to Mayts “founder principle,” for example, a new species may originate from the genetic divergence of a peripheral isolate of a population. Which individuals form the isolate, and which survive the initial displacement, may have very little to do with their selection coefficients, but may have very great evolutionary consequences in the long run.Google Scholar
  13. 9.
    Beatty (1984) [note 3], p. 192. A sampling event may be “indiscriminate” with regard to some properties, but not with regard to certain others. Balls drawn out of an urn by a blindfolded man have been sampled indiscriminately with regard to colour. But perhaps not with regard to location (top or bottom of urn). In the example, the twins were sampled indiscriminately with regard to presence of an X chromosome (they both possessed at least one), but perhaps not with regard to location (higher or lower on the ridge). One might contend that if there has been sampling at all, then there must be some properties in virtue of which some were taken out, and others left in. The questions then concern (a) the nature of such properties, and (b) the role they play in (i) evolution and (ii) evolutionary theory.Google Scholar
  14. 10.
    A. Rosenberg, A. (1978), The Supervenience of Biological Concepts’, Philosophy of Science 45:368–386.CrossRefGoogle Scholar
  15. 11.
    Mills and Beatty (1979) [note 6].Google Scholar
  16. 12.
    I am here assuming that the/an epistemic interpretation of probability is correct, that is, that probability claims are judgments as to the likelihood of a given event in light of evidence we have of past events of this sort. Probability claims, on this view, are always relevant to a body of evidence known by one or more inquirers. On a propensity interpretation of probability, according to which certain events have a probability of occurring quite independently of any evidence had by any inquirers, it could be assumed that the two twins had different probabilities of being struck my lightning before the momentous train of events that transpired that day on the mountain began. We could then say that the twins differed in their (objective) propensities for being struck by lightning, and hence differed in fitness with regard to lightning strikes. The twin that survived had a greater fitness than the one that died, and hence this was a selection event. This way of looking at the twin case would allow one to preserve the link between fitness differences and selection. Note, however, that it would result in the same judgment of the twin case: on either interpretation of probability, this case turns out to be an example of selection.Google Scholar
  17. 13.
    Fitness is a theoretical concept which has its use in evolutionary theory — to capture generalizations — but is not part of the causal explanation of a particular selection event. It is not the case that A outsurvived B because A had greater fitness. If A outsurvived B, it was because A possessed some set of properties, P, that B lacked. One can say that A outsurvived B because of A’s superior fitness, but this is just a shorthand technique which leaves entirely open precisely why A fared better than B. For this one needs to identify relevant property differences. To see how selection can discriminate between organisms which have identical fitnesses, consider the following example. Organisms A and B have the characteristics listed below (numbers represent arbitrary values of components of fitness — i.e., specific prop-In this case, both organisms have identical fitness. But suppose that a disease epidemic breaks out, killing Organism B (with low resistance) but sparing Organism A (with high resistance). I would argue that this is an example of selection, despite the lack of overall fitness differences between the two organisms. Fitness is a useful predictive tool, but is not a causal factor in selection events.Google Scholar
  18. 14.
    Sober (1984) [note 6], pp. 88–96.Google Scholar
  19. 15.
    L. Darden and J. Cain (1989), ‘Selection Type Theories’, Philosophy of Science 56:106–129.CrossRefGoogle Scholar
  20. 16.
    This is not to deny that fitness differences may be conducive to (or correlated with) selection events. The greater the fitness differences among organisms, the greater the chance that they differ significantly in underlying properties. The greater the difference in underlying properties, the greater the potential for selection. The important point to note here is that it is underlying property differences, not fitness differences, that are causally significant with respect to selection.Google Scholar
  21. 17.
    Mills and Beatty (1979) [note 6], p. 268.Google Scholar
  22. 18.
    Some recent studies on body size as an ecological and evolutionary factor include: R.H. Peters (1983), The Ecological Implications of Body Size (Cambridge: Cambridge University Press); W.A. Calder in (1984), Size, Function,and LifeHistory (Cambridge, Mass.: Harvard University Press);CrossRefGoogle Scholar
  23. 18b.
    K. Schmidt-Neilsen (1984), Scaling: Why is Animal Size So Important? (Cambridge: Cambridge University Press);CrossRefGoogle Scholar
  24. 18b.
    M. LaBarbera (1989), ‘Analyzing Body Size as a Factor in Ecology and Evolution’, Annual Review of Ecology and Systematics 20:97–117.CrossRefGoogle Scholar
  25. 19.
    Elsewhere [T. Shanahan (1990a), ‘Evolution, Phenotypic Selection, and the Units of Selection’, Philosophy of Science 57:210–225] Ihaveargued at length that selection can operate on properties that may be quite temporary in their duration. Position in a dominance hierarchy, holding of a territory, and possession of mates are all highly significant factors determining reproductive success. Yet each of these can change in value many times during the life of an individual. Such characteristics may not Refslect any underlying disposition for dominance, resource holding power, or sexual charisma, but rather may simply be a function of history: Whoever stakes out a territory (or builds a harem) first may enjoy a competitive advantage simply in virtue of beingßrst. Displacement from the position of advantage and later reintroduction often results in a disfavored position. The point is that characteristics need not be stable properties of an individual to be highly significant for selection.CrossRefGoogle Scholar
  26. 20.
    A beautiful example of nonheritable phenotypic differences that have great selective significance: Sterility or fertility in Hymenoptera is determined by environmental conditions during growth (e.g., kind of food given), and is mediated physiologically by hormonal titers. See CD. Michener (1974), The Social Behavior of the Bees (Cambridge: Harvard University Press). As Wcislo (1989, p. 157) points out, “The feedback relationships between behavior and demographic factors, and social organization and life-history traits, imply that social structure determines which reproductive opportunities will be available to individuals”Google Scholar
  27. 20a.
    (W.T. Wcislo [1989], ‘Behavioral Environments and Evolutionary Change’, Annual Review of Ecology and Systematics 20:137–169).CrossRefGoogle Scholar
  28. 20b.
    For more on this point, see S.A. Altmann and J. Altmann(1979), ‘Demographic Constraints on Behavior and Social Organization’, in LS. Bernstein and E.D. Smith (eds.), Primate Ecology and Human Origins (New York: Garland), pp. 47–63,Google Scholar
  29. 20c.
    T. Shanahan(1990b), ‘Group Selection and the Evolution of Myxomatosis’, Evolutionary Theory 9:239–254.Google Scholar
  30. 21.
    According to Sober, “The sort of environment an organism inhabits is part of its phenotype” (Sober 1984 [note 6], p. 119). Butif so, then how can an organism interact with its environment? At most an organism can interact with part of its own phenotype! If the environment is a part of the phenotype, then selection is impossible, since selection is an interaction between phenotypic properties and critical factors in the environment.Google Scholar
  31. 22.
    The determination of the relevant environment for a given selection event is conceptually as well as empirically problematic. Environments cannot be distinguished along sharp boundaries. In any case, there is (as yet) no completely non-arbitrary way to individuate environments, so claims that the organisms occupy the same or different environments must necessarily be inconclusive. Fortunately, selection does not require a common environment. What is required for selection is rather common “critical factors” in the environment, i.e., a common selective agent (Darden and Cain, 1989 [note 15]). Selection is not an interaction between an organism’s fitness and its environment. Just as not all components of an organism’s fitness are relevant to a selection event, so too not all components of an organism’s environment are relevant. The “environment,” like “fitness,” is causally inert.Google Scholar
  32. 23.
    A word of clarification: Strictly speaking, it is not that there is no biological distinction between selection and drift, because if one of these processes is more likely to lead to adaptations than the other, then there is a biological distinction. Rather, there is no hard and fast biological distinction: the two processes lie on a continuum. What differentiates selection and drift is not the nature of the events transpiring per se, but the kinds of effects or results one can expect from the process when different kinds of conditions obtain, e.g., whether selection takes place on widely exemplified properties, whether it is in a consistent direction, whether it results in adaptations, etc. In a previous discussion of selection and drift [T. Shanahan (1989), ‘Beatty on Chance and Natural Selection’, Philosophy of Science 56:484–489.] I argued that selection and drift are clearly distinct, and are distin-guished on the basis of whether sampling is on the basis of fitness differences or not. I now think that this is mistaken, or at least too simple. Evolutionary theory distinguishes selection and drift in terms of whether or not fitness differences are thought to be causative in differential biological success. But strictly speaking fitness differences have no causative power, and thus this way of distinguishing selection and drift as distinct processes in nature fails.Google Scholar
  33. 24.
    Sober (1984) [note 6, p. 115] argues that “Separating selection and drift yields concepts that are needed to mark important biological distinctions.” As I understand his argument, two populations may be characterised by identical sets of selection coefficients. Yet, if they differ in size they may experience quite different evolutionary careers. (In a smaller group the chances of random fixation of genes is greater.) On this view, the concept of “drift” is a way of capturing the importance of population size for evolutionary change. I would interpret this claim as follows: “Drift” is a concept used to fill in the space left between the predictions of abstract theory and the facts of concrete biological reality. Selection coefficients are educated guesses about the likely effects of certain properties in specified environments. When factored into a population genetics equation which assumes infinite population size, a prediction can be made about the change in gene frequencies in a population. But as selection coefficients are at best guesses (based on past correlations between properties and effects), they can be off the mark in actual biological scenarios. Thus, the concept of “drift’ is introduced to account for changes in populations that deviate from those expected on the basis of selection coefficients. My claim is that the events described as “drift’ are not different in kind from events described as “selection,” but differ only in that drift events are not predictable to the extent that selection events are. Drift events are, by definition, the residue left over when populational changes fail to accurately Refslect selection coefficients. Selection and drift are distinguished in theory even though they are not entirely distinct in nature.Google Scholar
  34. 25.
    Such properties will consequently not be of great interest to many evolutionary biologists. A distinction needs to be made between those entities which function is the process of selection, and those which function in processes of adaptation. Those biological entities that are the objects of phenotypic selection are units of selection. The objects of natural selection (which requires heritability) are units of adaptation. Units of adaptation are always units of selection, but units of selection are not always units of adaptation (Shanahan 1990a [note 19]).Google Scholar
  35. 26.
    Rosenberg (1988) suggests two alternative interpretations of drift: (1) Drift is a cover for unknown nonevolutionary (i.e., non-adaptational, non-selective) forces. (2) Drift is a cover for selective (i.e., adaptation-producing) forces of which we are ignorant. Rosenberg seems to pRefser (1). My position is neither (1) nor (2), but a third (hybrid) position: Drift is a cover for selective forces of which we are ignorant [as in (2)], but it is also non-adaptational [as in (1)]. “Drift” is a term used to cover selective events which have a low probability of leading to adapta-tional change. “Drift” Refsers to non-adaptational selective events.Google Scholar
  36. 27.
    According to Pierre Laplace, “We regard a thing as the effect of chance when it offers to our eyes nothing regular or indicative of design and when we are moreover ignorant of the causes which have produced it. Thus chance has no reality in itself; it is only a term fit to designate our ignorance concerning the manner in which the different parts of a phenomenon are arranged among themselves and in relation to the rest of Nature” (P.S. Laplace, ‘Memoire sur la probabilite’ des causes pour les evenements’, Oeuvres completes, VIII, 27–65; quoted in K.M. Baker (1975), Condorcet: From Natural Philosophy to Social Mathematics (Chicago: University of Chicago Press).Google Scholar
  37. 28.
    When one is considering something like the origin of Homo sapiens, which occurred only once in the universe, one needs both general principles and plenty of specific details pertaining to early proto-hominid environments, etc. In a case like this, one might well choose to know everything there is about the origin and evolution of this one species, rather than settle for general principles that apply to Homo and lots of other primate groups as well.Google Scholar
  38. 29.
    And scientific values determine scientific practice. As Rosenberg (1988, p. 189) points out, “The question of whether evolutionary phenomena are statistical or not, is a different one from the question whether our best theory of these phenomena is unavoidably statistical” (p. 188). If it turns out that the phenomena are deterministic but that we frame our theory in statistical terms because doing so is pragmatically expedient, “then the best theory we can frame about evolution will turn out to be a useful instrument, but not a complete account of evolution itself” A. Rosenberg, A. [1988], Is the Theory of Natural Selection a Statistical Theory ?’ in M. Matthen and B. Linsky (eds.), Philosophy and Biology (Calgary: University of Calgary Press), pp. 187–207.Google Scholar
  39. 30.
    Sober (1984), ‘Explanation and Prediction in Evolutionary Theory’, Science 130, p. 117.Google Scholar
  40. 31.
    Sober (1984), ‘Explanation and Prediction in Evolutionary Theory’, Science 130, p. 134.Google Scholar
  41. 32.
    According to the Hardy-Weinberg Equation, in an infinite population drift (sampling error) is ruled out. At the other extreme, in a “population” of two organisms, one might say that whatever happens to these is, according to the theory, a case of drift (sampling error). Predictive power of the theory is proportional to population size: The larger the population, the less potential for sampling error; the smaller the population, the greater the potential for sampling error. Explanatory accuracy is inversely proportional to population size: The smaller the population, the greater the potential for determining precisely why a given sampling event occurred. Predictive power and explanatory accuracy are inversely related in evolutionary biology. Rosenberg (1988) [note 29] makes a similar point, when he points out that usefulness and realism are inversely related.Google Scholar
  42. 33.
    S.J. Gould, S.J. (1989), Wonderful Life: The Burgess Shale and the Nature of History (New York: W.W. Norton).Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 1992

Authors and Affiliations

  • Timothy Shanahan
    • 1
  1. 1.Department of PhilosophyLoyola Marymount UniversityUSA

Personalised recommendations