Journal of Philosophy, Inc.

The Principle of Drift: Biology's First Law Author(s): Robert N. Brandon Reviewed work(s): Source: The Journal of Philosophy, Vol. 103, No. 7 (Jul., 2006), pp. 319-335 Published by: Journal of Philosophy, Inc. Stable URL: http://www.jstor.org/stable/20619949 . Accessed: 24/03/2012 21:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Philosophy.

http://www.jstor.org

H?

?h

THE JOURNAL OF PHILOSOPHY

VOLUME cm, NO. 7, JULY 2006

THE PRINCIPLE OF DRIFT: BIOLOGY'S FIRST LAW*

Drift is to evolution as inertia is toNewtonian mechanics. Both are the "natural" or default states of the systems towhich they apply. Both are governed by zero-force laws. The zero-force

law in biology is stated here for the first time. The Newtonian analogy to evolutionary theory is fairly common in presentations of population genetics, especially in reference to the Hardy-Weinberg Law. In the philosophy of biology, Elliott Sober's1 is canonical. Recent literature in the philosophy of presentation has both criticized2 and defended3 the Newtonian analogy to biology evolutionary theory. Here I show exactly how the Newtonian analogy

is helpful. More interestingly, I show how the analogue of Newton's First Law, what I term The Principle of Drift, is, in a very important mental
way, non-Newtonian. Newtonian or not, it is one of the two funda

laws of biology

(along with the Principle of Natural

principle of inertia

Selection).

i. newton's

The focus of this paper is on what is, and what is not, an appropriate zero-force law in Evolutionary Theory. Before turning to that, I want to emphasize a few well-known points about the paradigm of zero

* Fred Nijhout, Grant Ramsey, Mark My thanks to Richard Lewontin, Dan McShea, and Alex Rosenberg for helpful comments on earlier drafts of this paper. I Rausher, also want to thank Greg Wray for help with references and Leonore Fleming for edito rial help. 1 Sober, Conceptual Issues in Evolutionary Biology (Cambridge: MIT, 1984, second M. Walsh, Tim Lewens, and Andre Ariew, "Trials of Life: Natural and Random Drift," Philosophy ofScience, lxix (2002): 452-73; Mohan Matthen and Andre Ariew, "Two Ways of Thinking about Fitness and Natural Selection," this xcix, 2 (February 2002): 55-83. journal, 3 See Christopher "Selection, Drift, and the 'Forces' of Evolution," Stephens, (2004): 550-70. Philosophy of Science, lxxi Selection 0022-362X/06/0307/319-35 ? 2006 The of Philosophy, Inc. edition). 2 See Denis

Journal

319

320

THE

JOURNAL

OF

PHILOSOPHY

state laws?Newton's First Law, also know as the Principle of Inertia. After reviewing these points I will explore what sort of explanations
Newton's First Law

Newton expressed the First Law in two parts, since he thought there was an absolute difference between objects at rest and objects a constant velocity. Of course relativity theory rejects this moving at distinction and therefore we often see simpler one-clause statements of the principle. But for our purposes, the Newtonian version is best. It states that:
(1) An object force. at rest will stay at rest unless acted on by a net external

supports.

(2) An object in motion with velocity v will remain in motion with

velocity v unless acted on by a net external force.

(Of course, relativity aside, the first clause can be seen as a special case of the second, where v = 0.) Newton's First Law represented a Aristotle major break from the tradition of Aristotelian physics. While would have agreed with the first clause stated above, he would have of constant disagreed with the second. For Aristotle themaintenance a force. (Put anachronistically, velocity requires the constant action of = for Newton F = ma, while for Aristotle F mv.) Thus Newton's First Law changes what calls for an explanation in terms of force. Or to put the point inmore Aristotelian terms, Newton fundamentally altered
the

All this is familiar. But let us now turn our attention to Newtonian an object is at rest at time t. explanations of inertial motion. Suppose rest. How do we explain this Over the time interval t-t' it remains at behavior? The acceleration, a, during this time interval is 0. Applying
Newton's Second Law, F= ma, we see that F= 0. Thus,

conception

of what

constituted

"natural"

motion.

the First Law the object remains at rest. This is elementary physics. We will see if it is elementary philosophy. to the covering-law is this? According What sort of explanation is to subsume it model of explanation,4 to explain a phenomenon under some one or more general laws. That is exactly what we did above, and so, based on thatmodel we would say that our explanation is a standard scientific explanation. But ifone considers thismodel of to be inadequate,5 as I do, then how do we think about explanation this explanation? That, not surprisingly, depends on the alternative
4 See,

according

to

for example, Carl Hempel, (New York: Free Aspects of Scientific Explanation Press, 1965). 5 and Statistical Relevance Statistical Explanation See Wesley Salmon, (Pittsburgh: World and theCausal Structure of the University Press, 1971), and his Scientific Explanation (Princeton: University Press, 1984).

THE

PRINCIPLE

OF

DRIFT

321

model one adopts. It iswell beyond the scope of this paper to argue over others. But elsewhere I have for one theory of explanation a argued for causal-mechanical model of explanation in evolutionary Since our ultimate concern is with evolutionary biology, biology.6 treats the above bit of let us ask the question of how that model

Newtonian physics. that If one thinks of Newtonian forces as Newtonian causes?and one might think that the explana seems reasonable?then certainly is noncausal. Our explanation above did tion of inertial phenomena not cite a force; therefore it did not cite a cause. Rather it cited the absence of causes. So how can it be a causal explanation? One re sponse to this question seems so unattractive that I will simply men tion itwithout laboring to refute it. One might say that Newton's First Law tells us what does, and what does not, require explanation. Inertial phenomena do not. Although there is something right about this response, and we will turn to that shortly, saying a whole class of

cannot be explained seems just wrong. well understood phenomena A second response is to say that the above reasoning is correct, that the explanation is noncausal, and that shows the limitations of the model of explanation. Although a number of causal-mechanical of science would be sympathetic to this response, I am philosophers not. Salmon (op. ext.) argues that the goal of explanatory science is the causal structure of the world. The Principle of Inertia seems to be a fundamental rather its Relativistic counterpart) (or of the causal structure of our world. Because of that, I would part to be a perfectly good causal consider our inertial explanation
explanation.

to uncover

mechanical

"special" explanations. That is the lesson of Newton's First Law. Thus we could mark this distinction by saying that noninertial phenom ena require "special-causal-explanations," while inertial phenomena require what I will call "default-causal-explanations."

But, returning to the point made above, there is something different about this sort of Newtonian explanation versus one that some noninertial phenomenon in terms of some net force explains on the object. The latter sort of phenomenon acting requires

In Newtonian mechanics, the Principle of Inertia clearly delimits what can, and cannot, be explained by default explanations. That is, it delimits the class of objects that fall under the zero-force law. We will see if there is any analogous law in biology.

6 See Brandon, Adaptation and Environment (Princeton: University Press, 1990), where I was inspired by Salmon's World. Scientific Explanation and theCausal Structure of the

322
II. THE The existence

THE

JOURNAL

OF

PHILOSOPHY

HARDY-WEINBERG of a zero-force

LAW

AS A ZERO-FORCE the existence

LAW of forces.

Are there forces of evolution? John A. Endler7 argued that the analogy to Newtonian than helpful. More physics was more misleading Dennis Walsh, Tim Lewens, and Andre Anew (op. cit), and recently, Mohan Matthen and Ariew separately (op. cit.) have argued against the force analogy. I will not try to respond to these arguments in a I have offered detailed point by point fashion; in part because criticisms of the "emergentist statistical interpretation" of evolution in the two last mentioned ary theory defended papers,8 but also I think a direct positive case can be made for considering because
certain evolutionary processes as forces.

law presumes

I have in mind are selection, Evolutionary Forces. The processes mutation, migration, and nonrandom mating. In physics a force is a vector quantity. That means ithas both a magnitude and a direction. Also various forces inNewtonian physics (for example, gravity, electrical in a standard quantitative unit (? force, friction) can all be measured mentioned above? From a population genetics point of view all can be in terms of their effects on gene frequencies measured (see caveat below). From the point of view of molecular evolution theywould be level of gene sequences. And one could measured on a finer grain?the in terms of phe take a purely phenotypic point of view and measure distributions. Let us ignore this embarrassment of riches and notypic focus on the population genetics perspective. Consider mutation and migration. Both are similar in that they are
la Newton). Is there a common currency to measure the processes

already net forces that could be broken down into their components. a Migration consists of immigration into population and emigration from it.Both can have effects on gene frequencies, which can then be

added together to get the force of migration; similarly formutation. In a simple system with one locus and two alleles, Ax and A2, there is an actual number of mutations from A\ to A2, and an actual number from A2 to Ai. The change in Aj is simply the second number minus the first. There is no problem in conceiving of these as forces. (Al on considering them as net forces rather though some might insist than as component forces: that would not militate that evolutionary theory is a theory of forces.) against

the view

7 Wild (Princeton: University Press, 1986). Endler, Natural Selection in the 8 Statistical See Brandon and Grant Ramsey, "What's Wrong with the Emergentist inMichael Ruse and David L. Interpretation of Natural Selection and Random Drift," to Philosophy of Biology (New York: Cambridge, Hull, eds., The Cambridge Companion forthcoming).

THE

PRINCIPLE

OF

DRIFT

323

mating that do not involve differential reproduction among different genotypes. (So, for example, we are not here talking about positive assortative mating where rare genotypes have lower reproductive success because of difficulty in finding mates.) Nonrandom mating has an immediate effect on genotype frequencies, not gene fre quencies.

we are mating is different. I mention itnow, because to consider the Hardy-Weinberg Law, and nonrandom mat about can cause a population to deviate ing is one of the processes that from H-W equilibrium. Nonrandom mating has two basic forms: me say positive assortative mating and negative assortative mating. Let that here we are solely concerned with forms of nonrandom explicitly Nonrandom

(Consider a population with two alleles, Ai and A2, at mate with each frequencies p and q. If the homozygotes preferentially other, p and q will remain unchanged but there will be a decrease in the number of heterozygotes.) Nonrandom mating can have selective effects (for example, sexual selection or selection against inbreed on ing), but from the point of view of the H-W law it is its effect that matters. This contradicts the statement genotypic frequencies in terms of made above that all of these processes can be measured effects on gene frequencies. But we can achieve consistency by on genotypic measuring mutation and migration in terms of effects of view that is in From a population genetic point frequencies. convenient, but possible. So I consider this a minor technical point and will henceforth ignore it. to the view that these evo Selection presents a new challenge are forces. People tend to think of directional lutionary processes selection when they think of selection (this point will be explored and defended below). The other two patterns of selection are stabilizing and disruptive. These terms come from quantitative genetics so are

would

usually defined in terms of phenotypic distributions. Thus directional selection is when one extreme in the distribution is favored (for example, the taller the fitter). Stabilizing selection iswhen a single point in the distribution is favored (for example, 5m tall is the fittest). selection occurs when 2 (or more) Disruptive points in the dis are favored (for example, 2m tall and 6m tall are the fittest). tribution But these terms have come to be used more generally and can be translated into genetic terms. For instance in a two-allele model, if A\ is rare and A\A^ is fitter than AiA2, which is fitter than A2A2, then we

unproblematic for directional selection. But how about the other two forms?What is the direction of stabilizing selection? If a population, or some subpart of it,finds itself to the leftof the selected point, then

call that directional selection. But remember that a force is and a direction. This seems totally supposed to have a magnitude

324

THE

JOURNAL

OF

PHILOSOPHY

the direction of the force is to the right. But if the population is on the right of that point then the direction is to the left. Is our force analogy in trouble? No, or rather, only if it is problematic to think of gravity as a force. In our solar system two objects of equal mass and equal distance from the Sun, but on opposite sides of the Sun, will experience a force of equal magnitude but opposite direction. That is to two subpopulations exactly analogous equidistant from the se lected point in stabilizing selection but on opposite sides (assuming the selection gradient is the same on both sides). That gravitational force is relative to position is obvious in Newtonian mechanics. So that a similar relativity exists with respect to stabilizing and disruptive selection is not problematic. Thus, although it has not been exactly straightforward, we have made a positive case for treating certain evolutionary processes as that drift has yet to be mentioned. TheHardy-WeinbergLaw. There are two importantly different ways of stating the H-W law. One statement of it lumps drift together with mutation, migration, and the rest. The other does not. Here are the forces. Notice
two versions: H-Wi: If a exists with population then p and q respectively, two alleles, A\ and A2, with frequencies in a the population single generation

of A\A\ = q, and genotypic p and frequencies frequencies p2; = = that there is no selection, AiA2 2pq; and A2A2 q2?provided or nonrandom mutation, migration, mating, drift. H-W2: If an

will settle into genie and genotypic equilibrium with gene

two alleles, exists with Aj and A2, with infinite population and q respectively, then in a single generation the p frequencies and genotypic will settle into genie with population equilibrium of A\A\ = gene p and q, and genotypic frequencies frequencies

p2; AjA2
selection,

= 2pq; and A2A2

migration,

mutation,

or nonrandom

q2?provided

that there is no

mating.

direction. Drift has a magnitude that can be probabilistically predicted prior to the fact, and can be quantitatively accessed after the fact.9But drift definitely does not have a direction. (That, of course, iswhy it is called drift.) Given a population consisting of two selectively neutral
9 "The Difference between See Brandon, Biology and Philosophy, xx (2005): 153-70.

IfH-Wx is to be considered a zero-force law then drift needs to be a force just like the other evolutionary processes. However, I think that there are overwhelming reasons for not considering it a force. and Consider first the idea that a force has both a magnitude

Selection

and Drift: A Reply

to Millstein,"

THE

PRINCIPLE

OF

DRIFT

325

= = .5,we can predict that one of alleles, Ax and A2, at frequencies p q the two alleles will go to fixation (in the absence of evolutionary forces), but we cannot predict which will. That means drift has no direction. Christopher Stevens (op. cit.) argues that drift does have a direction at the genotypic level. He suggests that itpredictably leads to loss of heterozygosity and increase of homozygosity. That is a prediction, but a prediction without a direction. It does not saywhich

of A\A\ or A2A2 will increase in frequency. In physics that would be like saying that a 20-Newton force is acting on object A. Such a statement either makes no sense (the magnitude, but not direction has been specified) or is incomplete (oops, I meant a 20-Newton

downward force). Notice thatwe were able to specify the direction of inH-Wi. the other evolutionary processes mentioned Second, and perhaps more importantly, drift isnot a "special" force in evolution; it is the default position. By that Imean that it ispart and parcel of a constitutive process of any evolutionary system?namely, the sampling process. Sampling takes place in a number of ways. I will not try to exhaustively name all of them, because I cannot, and I no one can. That is, I suspect there are as of yet undiscovered suspect sampling processes going on in evolutionary systems. But here is a partial list of sampling processes. Evolution requires reproducing entities that form lineages, parent-offspring lineages. So in many cases there is sampling of parents (not all potential parents actually

In sexual diploid organisms there is game tic sampling in reproduce). the formation of a new diploid cell (fertilized zygote). There is also chromosomal sampling in the formation of gametes. There is further sampling of parts of chromosomes in crossover events during meiosis. At the ecological level there is habitat sampling when organisms are
dispersed over

that leads to drift, sampling, is a necessary part of any evolutionary system. It is not some new process added to the basic evolutionary
In contrast, mutation, scenario. migration and nonrandom mating are

heterogeneous

selective

environments.10

The

process

process.

clearly separable processes, which may, or may not, be a part of some (Asking if drift is a force is just like asking if inertia is a force in Newtonian mechanics. The lesson of Newton's First law is that it is not.) Selection, I have argued11 is both conceptually and from drift.When nonequal probabilities empirically govern the
of evolutionary

particular

distinguishable

10 See Brandon, Adaptation and Environment, especially chapter Environment in the Theory of Natural Selection." 11 See Brandon, "The Difference between Selection and Drift."

2, "The Concept

326

THE

JOURNAL

OF

PHILOSOPHY

sampled entities, selection becomes possible. However the possibility of drift does not go awayjust because the equiprobable distribution is no in force. Drift becomes longer impossible only in the highly condition that I have labeled a state ofMaximal Probability implausible Difference?that is,when all of the fitnesses of the competing entities equal either 1 or 0, with both values being present. (See Figure 1.) Selection, like drift, depends on the sampling process. However, unlike drift, ithas a direction and is separable from drift. Thus I think it is reasonable to treat selection as a "special" evolutionary force (on a par with mutation, and so on), while it is to clearly unreasonable think of drift that way. I will offer more support for this position shortly. For now, let us tentatively adopt the position that drift is not a force. It follows thatH-Wi is not a zero-force law. For it to be true to mix it needs forces with one non legitimate evolutionary force (drift). H-Wi is useful in many ways. It is true. It just is not a
zero-force law.

H-W2 does not run afoul of the above problem. Itmentions legitimate evolutionary forces in its proviso-clause. However,

only it is

Selection Necessary

Selection Possible

Selection Impossible

Drift
Impossible 1: The heavy Figure the infinite number bility Differences The of these arrows Possible horizontal

Drift

line, with

dotted

center

section,

leftto theEquiprobable Distribution (EP?all

selection emanating and drift from the different areas indicate the

of possible fitness distributions = fitness = 0 or 1, with (MPD?all

from Maximal some

represents Proba on the

of

of descriptions the distribution

fitness the same) on the right.

the modalities falling under

of both)

descriptions.

THE

PRINCIPLE

OF

DRIFT

327

unclear how itapplies to real populations, all of which are finite. I am not criticizing H-W2 on the basis of its employing an idealization. That is normal and useful in science. But when we talk about frictionless planes in physics and get results that apply in that idealized situation, we ultimately want to be able to apply it to real world situations. So how do we apply H-W2 to finite populations? One answer is thatwe do not. I presume that most would be unsatisfied with this. A second answer is that in applying H-W2 to finite populations we transform it back intoH-Wi, that is,we stick drift back into the proviso-clause. But
then, as we have seen, we no

level distinguish selection from drift. For instance, at the molecular we compare synonymous substitutions (substitutions in the third a same amino acid), with place in codon that end up coding for the
nonsynonymous selection is substitutions acting.12 to see where, and with what strength,

A final answer is thatwe do not revert toH-Wi, we leave the provisos alone, and we assert H-W2 is true of finite populations. Then H-W2 If no evolutionary forces act on basically says of finite populations: a population remain then its gene and genotypic frequencies is logically equivalent to the following: If a popu unchanged. That then some evo lation changes in gene or genotypic frequencies on it.Assuming, as we are for the moment, lutionary force is acting that I am right in not classifying drift as an evolutionary force, then these statements are false. A population with no evolutionary forces a that changes over gen acting on itwill drift. (And so, population or may not be experiencing evolutionary forces.) erational time may This truth is exactly what guides contemporary work on trying to

longer

have

a zero-force

law.

is H-W2 applied in a straightforward manner to finite populations not only false in what it says, it is also misleading inwhat it suggests. The first gloss on H-W2, stated in the paragraph above, basically says: If no forces then no change. That does not strictly imply the fol lowing: If no change then no force; but itdoes suggest it. It suggests that lack of change is a sign of lack of force. (This suggestion is clear in Sober's Newtonian analogy (op. cit.).) But in fact, the opposite is true: If no change then a set of forces must be acting. Again this truth is important in that it licenses the sorts of inferences biologists

12 to Detect Selection in Populations "Methods See, for example, Martin Kreitman, to the Human," Annual Review of Genomics and Human with Applications Genetics, I P. Bielawski, "Statistical Methods for Detecting (2002): 539-59; Ziheng Yang andjoseph Trends in Ecology and Evolution, xv (2000): 496-503; Michael Molecular Adaptation," and Stephen P. Wooding, in the Human Bamshad Selection "Signatures of Natural Genome," Nature Reviews Genetics, iv (2003): 99-111.

328
need to make

THE

JOURNAL

OF

PHILOSOPHY

is either: (a) A true zero-force law inapplicable to any real population; or (b) Applicable, but then equivalent toH-Wi, and so not a zero-force law; or (c) Applicable, but false. And so the Hardy-Weinberg Law fails to be a zero-force law in evolutionary biology.
HI. THE

this shortly). Thus H-Wg

in understanding

equilibrium

situations

(more on

Is there a zero-force law in evolutionary

OF DRIFT AS A ZERO-FORCE LAW

theory?

PRINCIPLE

The
and

answer

unstated.

is "yes." But the law in question has heretofore been I will state itfirst, and then defend its truth, itsusefulness,
as a zero-force law.

its status

The Evolutionary Principle of Drift: (A) A population at equilibrium will tend to driftfrom thatequilibrium
acted on by an evolutionary force. at rest will tend to start moving (A population external force.) unless unless acted on by an

on some net evolu t, caused (B) A population evolutionary trajectory by to from the extrapolated force F, will tend depart path tionary on F alone or or based (in either direction magnitude predicted even F if no other force unless intervenes, both) evolutionary continues to act. to stay in motion, on an external by but change its in motion will tend (A population acted unless trajectory, continually

force.)

Principle of Drift (PD) is stated in two clauses, just like Newton's version of his Principle of Inertia. I have stated each clause first in more precise biological language and then, in parentheses, to that facilitates a comparison in a more physics-like language The Newton's In fact, is clearly non-Newtonian. in that it requires the constant action of a force it is Aristotelian to keep a population in the same state of motion. But Aristotle First Law. Clause (B)

would

not be happy with Clause (A), which is both non-Newtonian and non-Aristotelian. So in this way the PD is quite unlike Newton's First Law. However, in another way, there is a strict parallel toNewton's First Law. Recall our discussion above where we saw how Newton's law sets out inertia as the default state of an object. I argued that Newtonian are importantly different from New explanations of inertial states of non-inertial states. The latter require the tonian explanations citation of "special" Newtonian causes, while the former are what I The PD serves exactly the same called default-causal-explanations. It clearly sets out the default state of a population, or in other purpose.

THE

PRINCIPLE

OF

DRIFT

329

words, it is a principle of default evolutionary action. In the absence of evolutionary forces drift occurs, or, more precisely drift will tend to this, drift can (see Figure 1 above). Do not misunderstand occur in the presence of evolutionary forces as well. So, for certainly instance, in clause (B) above, a population continually acted on by net evolutionary force, F, will quite frequently depart from the trajectory predicted from F alone. F needs to be very strong relative to the occur

that enables us tomake this distinction. Is the PD true? I think its truth is patently obvious once one of it is removed. One might think that potential misunderstanding the PD takes a stand on an empirical issue, the truth of which is in doubt. That is,one might think that the PD is somehow committed to some particular claim about the relative importance of drift versus selection. But this is a misunderstanding. Newton's First Law says nothing about the relative frequency of inertial objects versus ob jects experiencing a net force in the universe. Indeed, think of the paradigm example of a system to which Newton's Laws apply?our solar system. All of the major objects in that system are experiencing constant acceleration due

size for it to dominate drift.13 Drift is the default effective population state of evolutionary systems. We can explain evolutionary change due to drift. The drift pro ducing potential of the sampling processes that are constitutive of the evolutionary process is a fundamental part of the causal structure of our world. Thus such explanations are causal but differ from evo lutionary explanations that cite specific evolutionary forces. Just as in the Newtonian is a default-causal case, the first sort of explanation and the second a special-causal-explanation. It is the PD explanation

clause (A). It says that in the absence of evolutionary a in particular, selection) population previously at will start to drift. The empirical debate over the relative equilibrium importance of drift versus selection is irrelevant to this claim. By hypothesis clause (A) removes selection. Then, by hypothesis, the existing variants are selectively neutral. From that it follows that drift will occur. forces (and

to gravitational forces. Likewise, the PD is with the claim that selection and other perfectly compatible evolutionary forces are the primary movers of the vast majority of populations. Consider

13 See Brandon and H. Fred Nijhout, "The Empirical Non-equivalence of Genie of Selection: A (Decisive) Refutation of Genie Selectionism Genotypic Models Pluralistic Genie Selectionism," in Philosophy of Science. forthcoming

and and

330

THE

JOURNAL

OF

PHILOSOPHY

tionable metaphysical that the evolutionary assumption?namely, process is fundamentally indeterministic. First, the question of deter minism versus indeterminism for the whole world, or for some part of it (the evolutionary process), is an empirical question, not a meta

Consider clause (B). It says that in the absence of evolutionary forces (and in particular, selection) a population previously moving in state space with trajectory t will depart from that trajectory. Again, the truth of this is fairly obvious. If the trajectory were due to one or more of the special evolutionary forces, then removing those forces would tend to change the trajectory. So the truth of the Principle of Drift is not hostage to any particular resolution of the empirical selectionist/neutralist debate. However, some philosophers might think its truth is on a ques dependent

I believe, and have defended the view, that the physical question. is indeed indeterministic,14 but I think that the evolutionary process only assumption that has been at work in describing the Principle of Drift is that drift is a real phenomenon. Drift happens. Drift is from other sorts of evolutionary change. No biologists distinguishable doubt this. And that is all I have assumed. The (misguided) meta determinist needs to explain how drift can exist in a fun physical damentally deterministic world. (Since that is a fantasy world I leave that exercise to him or her.) to think that I have defended It is also a misunderstanding the Principle of Drift at the expense of the H-W Law. I have argued that theH-W Law isnot a zero-force law, but I have not denied itsutility. In fact its strength relative to the Principle of Drift flows from its limi tations. The H-W Law is explicitly couched in the terms of the pop

ulation

sexual organisms. That allows it, for genetics of diploid to be used to show that assortative mating is occurring in a instance, population when genotypic frequencies depart from the H-W predic tion. The PD is not couched in this specific language and could not be used for this particular inference. But, as John Beatty15 pointed out some time ago, this very specificity shows that theH-W Law isno law at all. It depends on conditions that are evolutionarily derived, and not at all necessary for the evolutionary process. The evolutionary forces

14 of Evolutionary Character and Scott Carson, "The Indeterministic See Brandon Proof for Determinism 'No Hidden Variables but No Room Either," Theory: No (1996): 315-37. Philosophy of Science, lxiii 15 in P.D. Beatty, "What's Wrong with the Received View of Evolutionary Theory?" of and R.N. Giere, eds., PSA 1980, Volume 2 (East Lansing, MI: Philosophy Asquith Science Association, 1980), pp. 397-426.

THE

PRINCIPLE

OF

DRIFT

331

is a law of nature. One way to think about the lawfulness of biological generalizations is to ask if theywould be exportable to life elsewhere in the universe. Were I asked to take Evolutionary Theory to some sentient beings elsewhere in the universe I would pack the Principle of Natural Selection. Where pack the Principle also pack my math H-W Law because organisms. If I did

the H-W into play could change and render H-W totally to life on Earth (for example, if all lineages were to go inapplicable asexual). The Principle of Drift is different. It is not predicated on on the fact that evolution consists of diploidy or sex. It is based solely one or more sampling processes. In thatway it isperfectly general and that brought

in importantly, it does not provide appropriate null hypotheses evolution. It suggests evolutionary stasis as a null hypothesis. But this is just wrong. In the absence of evolutionary forces drift occurs. So drift is the appropriate null hypothesis. a zero-force law. The PD Finally, and relatedly, the H-W law is not is. It is in virtue of this fact that it does
IV. CONSEQUENCES

there is life there is natural selection. I would also of Drift. Along with a few personal items I would books. And that is all. I would not bother with the Iwould not necessarily expect to find diploid sexual find them Iwould realize that (p + q)2 = p2 + 2pq + recreate H-W. q2 and would thereby So although my point is not to criticize the H-W Law, it is clear that it is not fundamental, unlike the PD. Likewise, and perhaps more

null hypothesis.

provide

the appropriate

Exactly how Evolutionary Theory relates toNewtonian mechanics may seem an esoteric topic in philosophy of biology. It is not. It has real I have explored the biological consequences of the PD consequences. elsewhere.16 Here the focus will be on themore purely philosophical.

perspective have suggested other definitions, such as, evolution is change in phenotypic distributions, or evolution is change in develop mental programs. But from the perspective of this paper, all such definitions are entirely misleading. is a heterogeneous Change

First, the very concept of evolution suggests the view that stasis is the default condition of evolutionary systems and that change iswhat is fundamental to evolution. Thus evolution is often defined as change in gene frequencies. Those objecting to a narrow population genetics

16 See

Complexity:

Brandon and Nijhout; and see D. McShea the Zero-Force Expectation for Biology"

and

Brandon,

"Diversity

and

(in preparation).

332

THE

JOURNAL

OF

PHILOSOPHY

fitness in terms of evolutionary change. (For Fisher, fitness is the per a capita rate of increase of type.) Recently Matthen and Ariew {op. cit.) have argued that evolutionary biologists mustmeasure fitness in terms of evolutionary change. But ifone of the primary ways selection acts is in stabilizing selection, then these views are entirely wrong-headed. at the Strong stabilizing selection (or what is called purifying selection results in no change (in gene frequencies, or at the molecular level) molecular level, in DNA sequence). And indeed, at the molecular the strongest evidence of the past action of selection is evolu level (As, for example, in the PAX6 tionary conservation of DNA sequence. in virtually identical forms in fruit fly,zebrafish, mice, and gene found humans.18) But the Fisherian is conceptually precluded from seeing this as strong selection, since nothing is changing. Further, it should be noted against the argument that biologists must/should measure fitness in terms of evolutionary change, that in fact, biologists mea sure fitness in that way only very rarely and only as a last resort.19 Finally, the views put forward in this paper have implications for what is perhaps the most widely debated question in philosophy of whether or not there are laws in biology. I used to biology?namely, think that biology had exactly one law?the Principle of Natural
Selection.201 now see it has two.

produce long-term stasis other than strong stabilizing selection. this privileging of change, the failure to see the fun Second, causes of change (default versus special), and damentally different the failure to see stasis as a strong signal of selection, together have led to empirically inadequate and explanatorily empty conceptions of fitness. Thus, starting with R.A. Fisher,17 many have sought to define

category in evolution. Itmay be driven by special evolutionary forces, or itmay result from no net forces at all. Stasis, on the other hand, seems to be quite homogeneous in its cause. There is no way to

Perhaps laws in evolutionary mental

more

interesting

than counting the number of funda the thoroughly theory is recognizing

17 Natural Selection (New York: Oxford, 1930). Fisher, The Genetical Theory of 18 See, for example, R. Quiring, U. Walldorf, U. Kloter, and WJ. Gehring, "Homology of the Eyeless Gene of Drosophila to the Small Eye Gene inMice and Aniridiain Humans," Science, cclxv (1994): 785-89. 19 H.E. Hoekstra, and J.G. Kingsolver, Brandon and Ramsey; See Endler; J.M. P. Gibert, and P. Beerli, "The Hoekstra, D. Berrigan, S.N. Vignieri, CE. Hill, A. Hoang, The American Naturalist, in Natural Selection Populations," Strength of Phenotypic clvii (2001): 245-61. 20 in Peter Clark of Biology," and Alex Rosenberg, See Brandon "Philosophy pp. and Katherine 147-80. Hawley, eds., Philosophy of Science Today (New York: Oxford, 2003),

THE

PRINCIPLE

OF

DRIFT

333

probabilistic nature of those laws. I have argued extensively else where21 that the Principle of Natural Selection is an instantiation of (a specially structured version of) what is called the Principle of Direct Inference in Probability Theory.22 This is the principle that allows one

to actual fre to infer from statistical probabilities (or chances) For example, ifcoin A has a probability of heads = 0.7 (in quencies.23 a certain chance set-up) while coin B has a probability of heads = 0.3 (in the same chance set-up), then ifwe toss both coins 10 times (in the same set-up) then (probably) coin A will land on heads more than coin B. That is an application of the Principle of Direct Inference. Consider the PNS: If organism A is fitter than organism B in envi ronment E, then (probably) A will have more offspring in E than B. of the same Clearly, this is a perfectly analogous application

principle. In this paper we have seen how the Principle of Drift is purely the result of the finite sampling processes that are part and parcel of the evolutionary process. Thus, for those who are reduction (I am not), these two laws point to probability theory isticallyminded as the reductive foundation of evolution, not to physics or chemistry.

"The Difference between to Millstein," Selection and Drift: A Reply Biology and xx (2005): 153-70. These issues involving the interpretation of probability Philosophy, and the principle of direct inference are enormously complex, important, and well beyond the scope of this paper. I would only add thatmy view of them seems to be just what evolutionary theory requires.

Isaac Levi, "Direct Inference," this journal, 1 (January 1977): 5-29. My view is lxxiv, that an objectivist theory of probability requires principles to connect observation of to objective probabilities relative frequencies and to connect objective probabilities with are no different from any other sort of events. On this view probabilities observable theoretical entity, for example, black holes. Ifwe are meaningfully to posit black holes we need to go from observables to black holes and vice versa; similarly for principles are on this view. Given these objective probabilities. Subjective probabilities superfluous but they do no work. My view of principles we can reconstruct subjective probabilities, of objective probabilities is similar to Karl Popper's evolutionary theory's requirement view of quantum mechanics. See Popper, "The Propensity of Proba Interpretation x (1959): 25-42. bility," The British Journal for the Philosophy ofScience, My view of objective in the works of Hans Reichenbach, Salmon, and Hacking, probability finds precedent I follow Salmon's idea of objectively homogeneous ref among others. In particular, erence classes as being the solution to the reference class I have specifically problem. these ideas in the context of evolutionary biology and have argued that my developed notion of "selective environmental solves the reference class problem in homogeneity" of this argument, see Brandon, evolutionary biology. For the most recent development

21 See Brandon, and Evolutionary Studies in History and "Adaptation Theory," Philosophy of Science, ix (1978): 181-206, and Adaptation and Environment. 22 See Ian Hacking, 1965), Logic of Statistical Inference (New York: Cambridge, pp 1-12. are other There instead of connecting interpretations of this principle, which see connect them with subjective probabilities; objective probabilities with outcomes,

334

THE

JOURNAL

OF

PHILOSOPHY

V. BIOLOGY'S

FIRST

LAW

Newton's

First Law is so called not because itwas the first law of discovered or articulated. Here "First" does not relate to physics temporal priority. Rather it describes a sort of logical, or better, nomological priority. Newton's Law of Inertia describes the default, or "natural," state of Newtonian systems. It describes what happens
no net Newtonian forces are acting on such a system. It is, as we

when

biology's first law.24 But from a temporal point of view the Principle of Drift isnew, or at least relatively new. However, in the paper that introduced the Hardy without law, G.H. Hardy talks about a drift phenomenon, Weinberg

have seen, a zero-force law. I have shown that The Principle of Drift is to Newton's First Law in these ways. That is why I call it analogous

using thatword.25 This is probably the first discussion of evolutionary drift.26 Itwould be an interesting historical project to see just how and when drift as the default state of evolutionary systems became em bedded into evolutionary thinking. My point is that, unlike the Prin ciple of Natural Selection, which one might reasonably trace back to

It has been first Darwin, the Principle of Drift is relatively modern. named and articulated in this paper. However, contemporary evo lutionary biologists regularly make the inferences it licenses. So I do not claim to have made a new biological discovery here. What I have done is to systematize these important inferences into a single law, a law that is fundamental to contemporary
VI. CONCLUSIONS

evolutionary biology.

similar ismodern evolutionary theory toNewtonian mechanics? Taken in isolation, that question is at best silly, at worst meaningless. Is the answer 17? Taken in the context of recent literature in the can say that it is not as similar toNewtonian philosophy of biology we as Christopher Stevens thinks, but much more so than mechanics either Walsh, Lewens, and Ariew, or Matthen and Ariew would allow. How I have found surprising is how useful is the comparison. What for a zero-force law. The Newton's First Law provides the model of Drift, which is stated here for the first time, is in one Principle to maintain its For a population important way non-Newtonian.

24 to suggest that evolutionary biology In calling this biology's first law I do not mean to suggest something I do mean that virtually all is the whole of biology. However, lies at the center of agree with, namely, that evolutionary biology biologists would biology. 25G.H. Hardy, "Letter to the Editor," Science, xxvm (1908): 26 Richard Lewontin, personal 2005. communication, 49-50.

THE

PRINCIPLE

OF

DRIFT

335

trajectory through state-space the constant action of evolutionary forces is required. But, the Principle of Drift is likeNewton's Principle of Inertia in that it is a zero-force law, and so it does provide law fails to be a appropriate null-hypotheses. The Hardy-Weinberg zero-force law. It fails not just because it is not a law, but also because
itmixes

dom mating) appropriate

evolutionary

forces

with a nonforce

(selection,

null-hypothesis

thus it fails to provide for evolutionary scenarios. (drift). And

ROBERT N.

migration,

mutation,

nonran

the

BRANDON

Duke University