1. Introduction

There has recently been a renewed interest in the ontological status of the components of evolutionary change (natural selection, random genetic drift, mutation, gene flow, etc.). Elliott Sober (Reference Sober1984), Christopher Stephens (Reference Stephens2004), and Robert Brandon (Reference Brandon2006) have argued that the theory of evolution is best understood as a theory of forces, similar to Newtonian mechanics.Footnote ¹ Walsh (Reference Walsh2000), Matthen and Ariew (Reference Matthen and Ariew2002), and Walsh, Lewens, and Ariew (Reference Walsh, Lewens and Ariew2002) have argued that these components of evolutionary theory should not be construed as “forces” and that some, natural selection and drift in particular, should not even be considered causes of evolution.Footnote ² Hodge (Reference Hodge, Kruger, Gigerenzer and Morgan1987), Bouchard and Rosenberg (Reference Bouchard and Rosenberg2004), Reisman and Forber (Reference Reisman and Forber2004), Millstein (Reference Millstein2006), and Shapiro and Sober (Reference Shapiro, Sober, Machamer and Wolters2007) all argue for positions found somewhere in the middle, with natural selection and random genetic drift defended as (at least) causes of changes in trait frequencies but not necessarily as forces similar in kind to those found in Newtonian mechanics.

I am in the Newtonian camp. However, there is a problem for the Newtonian account that has yet to be addressed. To represent evolutionary forces as Newtonian vectors, we must represent them as “acting” in some type of evolutionary space (e.g., phenotypic or genotypic space) as opposed to acting in space and time as classical Newtonian forces do. I will argue that this shift from representation in space and time to representation in evolutionary parameter space has problematic consequences for the Newtonian interpretation because it opens the door for any cause of any event to be construed as a Newtonian force, thereby stripping Newtonian forces of their status as a unique kind of cause. In this article, I will develop this criticism and attempt an answer on behalf of the force interpretation. In answering this criticism, I hope to do two things: (1) illuminate the difference between forces and other types of causes and (2) answer the objection that random genetic drift is not an evolutionary force.

2. Evolution and Newtonian Forces

Newtonian forces are vectors; they are “pushes” and “pulls” with specifiable magnitudes and directions that originate from, or are grounded in, some physical source.Footnote ³ If we are to apply the concept of a Newtonian force to scientific theories outside of Newtonian mechanics, it will be helpful to generalize the conditions of forcehood. So, for any F, for F to be labeled a force, the following four criteria must be met:

My focus below will be DC, but a few words about the other criteria are in order.

The WALM position with respect to the forces of evolution is that the forces of evolution (in particular, natural selection and drift) are merely statistical trends and not causes of evolution. WALM argue that, since natural selection and drift do no causal work at either the population or individual level, there is no causal work for natural selection or drift to do; hence the denial of natural selection and drift as causes. According to WALM, natural selection and drift are not forces (at least in part) because they fail to satisfy CC.

The WALM position has given rise to several recent proposals that attempt to vindicate the causal efficacy of the components of evolution. Roberta Millstein (2006) argues that anyone who is committed to the existence of frequency-dependent selection is committed to population-level causation with natural selection as a cause. Shapiro and Sober (Reference Shapiro, Sober, Machamer and Wolters2007) defend natural selection and drift as population-level causes of changes in trait frequencies. Frederic Bouchard and Alex Rosenberg (Reference Bouchard and Rosenberg2004) argue that natural selection is causally efficacious with respect to individuals, not populations. And, Robert Brandon (Reference Brandon2006) has recently argued for the interpretation of natural selection as a force and evolutionary theory as a theory of forces (with all that entails about the causal efficacy of the components of evolutionary change).Footnote ⁶ These worries about the causal “status” of natural selection and drift are legitimate and contentious, but I will not discuss them in any greater depth, since the problem I hope to highlight will have to do primarily with the directionality criterion. I will assume for the rest of the article that natural selection, drift, mutation, migration, and so on satisfy CC.

The magnitude and physical source criteria are less controversial. The magnitude of natural selection is represented by the selection coefficient, gene flow by the migration rate, drift by population size, and mutation by the mutation rate. Furthermore, these parameters are determined physically. The selection coefficient is determined by facts about fitness that are determined by facts about the individuals in a population, the environment they inhabit, and the interactions among them (among other factors). The parameter for population size is determined (obviously) by the number of individuals in a population. Mutation rate is determined by the average number of mutations of alleles per million gametes. And the migration rate is determined by the interactions that constitute the transfer of genes between populations (presumably, these relationships will be primarily reproductive in nature), as well as population size and initial gene frequencies. The components of evolutionary change clearly satisfy MC and PSC.

The problematic criterion for the purpose of this article is DC. Even friends of the force interpretation, such as Brandon (Reference Brandon2006), ultimately reject drift as a force because it lacks a specifiable direction and therefore fails to satisfy DC. The generality of the directional criterion raises a prima facie problem for the Newtonian interpretation of evolutionary biology, one that I will attempt to solve in what follows.

3. Directionality and Spatial Pluralism

What could it mean to speak of the direction of natural selection, drift, mutation, or gene flow? The meaning of “direction” in Newtonian mechanics is straightforward—it is a direction in space and time. Newtonian forces push and pull physical bodies in specifiable spatiotemporal directions. But, in an important sense, evolutionary forces do not “act” like physical Newtonian forces. Evolutionary forces push and pull populations of organisms (not bodies) in evolutionary space, not in space and time. Let us consider natural selection as a test case to make this point clear.

In the simplest case, if a particular phenotype, genotype, or gene is fittest (and its fitness is not dependent on the frequency of the phenotype/genotype/gene in the population), then we can expect to see higher frequencies of that phenotype in future generations.Footnote ⁷ In this simple case, the direction of natural selection is toward that area of evolutionary space where that phenotype, genotype, or gene's frequency in the population is greater.Footnote ⁸ If the fitness of a given phenotype is dependent on its frequency in the population, then it is possible that natural selection could maintain the frequency of that phenotype at an equilibrium well below fixation (this is the case of frequency-dependent selection). The details of the different kinds of selection are unimportant. What is important is that the direction in which natural selection pushes a population depends on facts about the fitness of specific phenotypes, genotypes, or genes. If it is a case of frequency-dependent selection, then natural selection will maintain the population at an equilibrium point; if it is like the simple case described above, then it will push a population toward that area of evolutionary space where the frequency of the fittest phenotype, genotype, or gene is greatest. Similar stories can be told about mutation and gene flow (e.g., mutation pushes a population, albeit weakly, toward that part of evolutionary space where that mutation is more prevalent). This type of reasoning is representative of how Brandon (Reference Brandon2006) and others argue that natural selection, mutation, and gene flow have directions.

The line of reasoning leading to the conclusion that natural selection has a direction fails in the case of drift, and it is for this reason that Brandon (and others) argues that drift does not have a direction. Defenders of the force interpretation, such as Sober (Reference Sober1984) and Stephens (Reference Stephens2004), have argued that drift does have a direction; the direction of drift is to march a population toward homozygosity (and therefore away from heterozygosity). Brandon's objection to the directionality of drift is as follows. Stating that drift is the only force acting on a population will allow us to predict that the population will march toward homozygosity, but it does not tell us whether the resulting population of homozygotes will be homozygous dominant or homozygous recessive. All that we know is that the most likely outcome is that the population will either march to that area of evolutionary space where the dominant allele is at fixation or to that area where the recessive allele is at fixation. The objection then is that drift has a disjunctive direction (it will push a population here or there), but since it does not allow us to predict which disjunct will (probably) be satisfied, drift does not really have a direction at all. If we cannot predict which disjunct will be satisfied, then drift does not really have a direction and so drift is not a (Newtonian) force.

There is a response to Brandon and others who deny the directionality of drift (e.g., WALM) that might vindicate drift as a force, but before considering it, let me briefly return to our discussion of the space in which evolutionary forces operate. Obviously, phenotypic space (or any type of evolutionary space) is not space and time. It is a mathematical construct that describes, in the above case, the genetic constitution of a population of organisms. Furthermore, the “movement” of populations due to evolutionary forces is movement in the mathematical space that represents the evolutionary dynamics of a given population. The criteria for being a Newtonian force must be sufficiently liberal with the allowed notions of space, direction, and magnitude in order for the Newtonian interpretation to apply to the case of evolutionary biology (otherwise, we could never start the process of generalizing the conditions and applying it to areas of science beyond Newtonian mechanics). The problem that arises from extending the force interpretation to evolutionary biology is that the liberality of the force criteria threatens to license force-talk for almost any cause of any effect.

For example, the criteria seem to entail that opium possesses not merely a dormitive virtue but a dormitive (Newtonian) force (i.e., opium causes drowsiness in virtue of a dormitive force).Footnote ⁹ After all, opium causes drowsiness; the dormitive virtue that causes drowsiness has a direction (toward drowsiness in a suitably defined fatigue-space); the dormitive virtue can be assigned some roughly specifiable strength (some doses of opium cause fatigue faster than others); and it certainly has a physical source (the physico-chemical properties of opium). Dormitive “virtues” satisfy all four Newtonian force criteria. Insofar as saying that opium causes drowsiness in virtue of a dormitive Newtonian force is problematic, because appealing to a dormitive force (or virtue) seems to be a poor, if not entirely empty, explanation of the causal efficacy of opium, there is a worry that the title of “force” ceases to be an interesting title at all. Given the Newtonian interpretation's commitment to a pluralistic view of representing direction in different “spaces,” presumably any instance of a cause producing an effect could be described in force language (as we did with the opium example). Therefore, absent a principled argument against the acceptability of certain sorts of spatial representations (e.g., fatigue-space), the proponent of the force interpretation seems forced to accept force-talk in the case of dormitive virtues and any other putative case of causation. The forces in Newtonian mechanics or evolutionary biology are forces, but so are dormitive virtues. Molière would not be pleased.

I think that there is a solution to this problem but not one that will render forces different in kind from other causes; rather, it is one that will set forces apart from other causes and explain their privileged role in certain areas of scientific theorizing. Furthermore, this solution will help to answer Brandon and WALM's dismissal of drift as a force. I turn to this now.

4. Unification, Mathematical Precision, and Forces

If forces are to be ontologically interesting (i.e., are “more” than mere causes), additional criteria may need to be added to the conditions above that either rule out “dormitive force” cases as legitimate instances of Newtonian forces or explain our intuition that Newtonian force explanations in physics and natural selection and drift explanations in evolutionary biology are explanatory, whereas dormitive virtue explanations are not. Though I will not go into great detail, I will mention two such criteria that appear in the literature on forces.

Bigelow, Ellis, and Pargetter (Reference Bigelow, Ellis and Pargetter1988) and Wilson (Reference Wilson and Borchert2006) claim that a distinctive feature of forces is their ability to unify disparate phenomena. Appeal to forces in laws or in explanations of disparate phenomena “has explanatory power as a result of revealing what is common among widely differing phenomena” (Bigelow et al. Reference Bigelow, Ellis and Pargetter1988, 623). On this view, forces are just a special type of cause that are able to connect disparate phenomena. Gravitational forces, for example, can be instantiated by many different types of bodies (planets, organisms, etc.), but every instantiation is an instantiation of the very same force (Bigelow et al. Reference Bigelow, Ellis and Pargetter1988, 624). Forces, according to Bigelow et al., are distinguished from other causes by their generality and unifying power. Interestingly, similar sentiments can be found in Newton's own writings. In Query 31 of the Opticks, Newton details the unifying power of the gravitational force by discussing the optical, chemical, and physical phenomena to which gravitational forces are relevant (Reference Newton1953, 377–397). According to Newton, the unifying power of the gravitational force sets it apart from other causes. If unifying power is an essential feature of forces, then this may serve to distinguish it from other types of causes. If this is indeed a distinctive feature of forces as opposed to mere causes, then we might want to insist further that to be a Newtonian force, it must be able to unify a wide variety of seemingly disparate phenomena (Bigelow et al. Reference Bigelow, Ellis and Pargetter1988, 628–629).Footnote ¹⁰

Bigelow et al. also mention that forces in physics are not just forces with a magnitude but with a mathematically specific magnitude. They claim that “a force is not just a ‘force’, but a force with 3.14 × 10⁻¹⁵ newton” (628–629). If this is true, then my condition above can be amended to claim that having some assignable strength is not enough to satisfy the magnitude criterion. Rather, the assignable strength must be mathematically precise (in some sense like the example just given). This may well exclude the ascription of a dormitive force to opium if the magnitude of that force can be given a weak qualitative rendering but not a mathematically precise one. What counts as being “mathematically precise” or precise enough is not a discussion I will enter into here; but the general thrust of the point is understandable without a detailed account of criteria for mathematical precision.

The important point is that there seem to be resources (1) to exclude dormitive force cases and (2) to set forces apart from other (mere) causes. But even if these criteria do not succeed in excluding dormitive virtue cases (because dormitive virtues are unifying or because we are able to attribute mathematically precise magnitudes to them, or both) or in setting forces apart from other causes, we could still posit a continuum of forces with maximally precise and unifying forces on one end and mathematically vague and weakly unifying forces on the other. But it should be noted that even then we would be able to distinguish between the types of forces by appeal to the explanatory superiority of the (more) mathematically precise and unifying forces (whose explanatory superiority is a function of their mathematical precision and unifying power). Dormitive forces would still be forces but their explanatory inferiority and mathematically imprecise manner of unifying sleep-inducing phenomena would place them at the opposite end of the continuum from paradigm cases like those in Newtonian physics.

This provides the friend of the force interpretation with a strategy to rebut the dismissals of drift as a force of evolution. One can grant that if drift is the only force acting on a population, then all that we know is that one disjunct of its disjunctive direction (homozygous dominant or homozygous recessive) will be satisfied, and so drift does not have as specific a direction as do the other evolutionary components already covered. But even if drift's direction is not as specific as the directions of natural selection, mutation, and migration, the statement that drift moves populations away from the area of evolutionary space where heterozygosity is present (or where the frequency of heterozygotes in the population is high) is certainly not an empty claim. It is substantial that drift moves populations away from a specific area of evolutionary space. One might not know exactly where in evolutionary space the population will move, but we have a good idea where it will not be. Drift's lack of directional specificity might make it less useful for prediction in some cases, but whether this is true depends on our predictive and explanatory interests. If we are interested in predicting where in evolutionary space a population composed of a mix of heterozygotes and homozygotes will be pushed under the influence of drift alone, we will have at least a general idea of where to look (i.e., we know that it is highly probable that the population will not occupy that area of evolutionary space where heterozygosity is present in the population). Drift's lack of directional specificity is due to drift's direction being logically weaker than the directions of the other evolutionary forces. That is, drift's direction emerges only when we logically weaken the description of evolutionary space from talk of particular genotypes to talk of the logically weaker properties of homozygosity and heterozygosity. If my suggestion above is right and forces lie on some continuum, then it may be true that drift is less “forcelike” than the other components of evolutionary change, but not to any great degree. Drift still has unifying power and the magnitude attributable to drift can be made as mathematically precise as one wants. Once we recognize the consequences of extending the Newtonian interpretation to other sciences, we see that on a continuum of forces, drift is not as forcelike as natural selection, mutation, and migration, but it is much closer to its evolutionary brethren than it is to the (intuitively) explanatorily impotent dormitive virtue cases.

5. Conclusion

The strategy of extending the Newtonian interpretation to evolutionary biology leads to a new conception of forces that (1) demarcates forces from other causes and (2) makes room for drift as a Newtonian force similar in kind though importantly different from the other evolutionary forces. Evolutionary forces may be more or less forcelike depending on how they score on all of the considered criteria. Regardless of the specifics, it seems clear that they are closer to the Newtonian end of the spectrum than to the end comprised of “dormitive virtue” cases.