Determinism, Realism, and Probability in Evolutionary Theory Determinism, Realism, and Probability in Evolutionary Theory Author(s): Marcel Weber Source: Philosophy of Science, Vol. 68, No. 3, Supplement: Proceedings of the 2000 Biennial Meeting of the Philosophy of Science Association. Part I: Contributed Papers (Sep., 2001), pp. S213-S224 Published by: The University of Chicago Press on behalf of the Philosophy of Science Association Stable URL: http://www.jstor.org/stable/3080947 . Accessed: 07/10/2011 16:49 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. The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR to digitize, preserve and extend access to Philosophy of Science. http://www.jstor.org http://www.jstor.org/action/showPublisher?publisherCode=ucpress http://www.jstor.org/action/showPublisher?publisherCode=psa http://www.jstor.org/stable/3080947?origin=JSTOR-pdf http://www.jstor.org/page/info/about/policies/terms.jsp Determinism, Realism, and Probability in Evolutionary Theory Marcel Webertt Universitat Hannover Recent discussion of the statistical character of evolutionary theory has centered around two positions: (1) Determinism combined with the claim that the statistical character is eliminable, a subjective interpretation of probability, and instrumentalism; (2) In- determinism combined with the claim that the statistical character is ineliminable, a propensity interpretation of probability, and realism. I point out some internal prob- lems in these positions and show that the relationship between determinism, elimin- ability, realism, and the interpretation of probability is more complex than previously assumed in this debate. Furthermore, I take some initial steps towards a more adequate account of the statistical character of evolutionary theory. 1. Introduction. Probabilistic concepts and statistical reasoning are an in- tegral part of modern evolutionary theory. For example, an organism's fitness does not determine uniquely how many surviving offspring some individual will have, it only provides a statistical expectation. Similarly, only statistical generalizations can be made about the evolution of small populations that are subject to genetic drift. This statistical nature of evo- lutionary theory raises a number of foundational and philosophical prob- lems. One question is how the concept of probability should be interpreted. Another question one may ask is, why is evolutionary theory statistical? Scientists develop statistical theories for different reasons. In physics, this is evident in the contrast between classical statistical mechanics and quan- tum mechanics. The former treats systems of particles that obey deter- ministic laws of motion. However, the number of particles in these systems tSend requests for reprints to the author, Zentrale Einrichtung fiir Wissenschaftstheorie und Wissenschaftsethik, Universitait Hannover, Oeltzenstrasse 9,30169 Hannover, Ger- many; email: weber@ww.uni-hannover.de. :I wish to thank Roberta Millstein, Robert Brandon, Elliott Sober, and Paul Hoyningen- Huene for helpful suggestions. Philosophy of Science, 68 (Proceedings) pp. S213-S224. 0031-8248/2001/68supp-0018$0.00 Copyright 2001 by the Philosophy of Science Association. All rights reserved. S213 MARCEL WEBER is too large to follow them individually. Physicists, in this case, use statis- tical reasoning in order to reduce the complexity of systems which would otherwise be mathematically intractable. By contrast, quantum mechanics uses probabilistic concepts because there exist states of quantum systems which do not uniquely determine the outcome of measurements. Whereas statistical mechanics is statistical for epistemic reasons, quantum mechan- ics is believed to be (partly) statistical for ontological reasons, i.e., because it deals with indeterministic processes. What about evolutionary theory? In the recent literature in philosophy of biology, two contrary positions have been developed which address this set of issues. The first position holds that the evolutionary process is de- terministic. Therefore, the reasons for the statistical nature of evolutionary theory are epistemic; they lie in our cognitive limitations. Laplace's demon would have no need for evolutionary theory in its current, statistical forms; in other words, it is eliminable. Adherents to this position typically adopt a subjective or "ignorance" interpretation of probability. In addition, they are instrumentalists with respect to evolutionary theory, i.e., they claim that evolutionary theory falls short of representing the real causes of evo- lutionary change, although it may be useful for other reasons. This posi- tion has been defended by Horan (1994), Rosenberg (1994), and Graves et al. (1999), and will henceforth be referred to as GHR. The second position that has been rigorously defended holds that the evolutionary process itself is indeterministic. Thus, like in quantum me- chanics, there are ontological reasons for the statistical nature of evolu- tionary theory, which means it is ineliminable. Defenders of this position prefer a propensity interpretation of probability. For instance, they view fitness values as representations of irreducibly stochastic dispositions of individual organisms. Furthermore, they are realists about evolutionary theory. This position is shared by many philosophers of biology, but for simplicity I will refer to it as BC, after a recent defense by Brandon and Carson (1996). In this paper I want to show that GHR and BC both have internal problems. Even if determinism is true and the statistical character of evo- lutionary theory is in principle eliminable, neither GHR's instrumentalism nor their subjective interpretation of probability follow (Section 2). BC have also unduly simplified the relationship between determinism, realism, and the interpretation of probability. In addition, I argue that BC's thor- oughgoing indeterminism fails to give us a satisfactory account of evolu- tionary processes as we know them, even if strict determinism is false (Section 3). Finally, I shall take some initial steps towards a more adequate account of the statistical character of evolutionary theory (Section 4). 2. GHR and the Determinism-Instrumentalism Fallacy. Rosenberg's (1994) S214 DETERMINISM, REALISM, AND PROBABILITY instrumentalism about biology rests partly on his considerations of re- duction (see Weber 1996 for a critique) and partly on his account of prob- ability and the statistical nature of evolutionary theory. The essence of this account is captured by the following claim: [T]he only probabilities to which the theory [of evolution] is commit- ted are the subjective probabilities that agents of our cognitive powers require to apply the theory to actual processes among populations of interest to us. If people were a lot smarter, there would be propor- tionally less reason to appeal to such epistemic probabilities and mu- tatis mutandis less reason to treat the theory as statistical. (Rosenberg 1994, 59) Subjective probabilities are understood by Rosenberg in the customary (Bayesian) way, namely as degrees of belief (1994, 61). Thus, the theory of evolution does not tell us how nature really is, it merely tells us how much confidence we can have in certain predictions-hence instrumentalism. How are these rather strong claims, which most practicing biologists would probably reject, justified? A central premise in Rosenberg's argu- ment is his claim that biological processes are deterministic. He claims that, although the universe is fundamentally indeterministic because of quantum-mechanical effects, this indeterminism vanishes asymptotically as we move from the microphysical level (chemical bonds and below) to the macroscopic level of biological processes (1994, 61). For the sake of the present analysis, let us grant Rosenberg this premise, i.e., the deter- minism of biological processes. What interests me here is how he gets from this premise to his subjective interpretation of probability and, in turn, to his instrumentalist characterization of evolutionary theory, as presented above. The basic strategy of Rosenberg's argument is to show that an omnis- cient being would have no need for a statistical evolutionary theory. In a fictional example which is supposed to show this, Rosenberg (1994, 71ff.) considers a population of giraffes from which the individuals with the longest necks are continually removed by poachers. A group of conser- vation biologists, unaware of the illegal poaching, notes that the popula- tion moves away from the adaptive equilibrium in terms of neck length and attributes this change to genetic drift. Now, Rosenberg claims that if these biologists were in possession of the full information concerning the fate of long-necked giraffes, they would not attribute the observed changes to drift. Hence, drift is merely a "useful fiction." A statement that drift has occurred in a population means nothing but the fact that the real causes of evolutionary change are unknown. Some savvy being who has access to these real causes has no need for the theory of genetic drift. This example is generalized by Rosenberg to the effect that evolutionary models S215 MARCEL WEBER (except the principle of natural selection; see below) always merely reflect biologists' ignorance concerning the underlying causes of evolutionary change. Millstein (1996) has shown that Rosenberg's fictional example is flawed, for his alleged example of genetic drift is not a case of drift at all; it is a case of selection. A group of conservation biologists intelligent enough to notice that they are continually losing long-necked giraffes to poachers would conclude that there has been selection against long-necked giraffes, the selection pressure being exerted by the poachers. Thus in this example, evolutionary theory is not eliminated by the additional causal information. Instead, the case is merely moved from the domain of models of genetic drift into the domain of models of natural selection, which are also statis- tical. In order to show the eliminability of the theory of drift, Rosenberg would have to examine a genuine example of genetic drift. The reasoning behind Rosenberg's odd example, I suspect, lies in the fundamental di- chotomy he draws between selection and drift. Selection, in Rosenberg's metaphysics, is not subject to his biological instrumentalism, but is the very reason why biology can only be an instrumental science (1994, 127). Therefore, the concept of fitness is exempt from Rosenberg's subjective interpretation of probability; he views it as an indefinable, primitive term. However, this dichotomy seems odd, especially if one considers that se- lection in finite populations and drift are both stochastic processes which differ only in that the former is discriminate with respect to an organism's physical properties whereas the latter is indiscriminate (Beatty 1984). Of course, it could still be the case that an omniscient being would have no need for the theory of drift, even if Rosenberg's example fails to dem- onstrate this. Sober (1984, 126) has argued for the contrary. According to Sober, the statistical approach in evolutionary biology has led to signifi- cant generalizations about populations of organisms. These generaliza- tions unify a large number of highly heterogeneous phenomena that have very little in common except being instances of the same evolutionary models. They carve out natural kinds that would be invisible to Laplace's demon, who, therefore, wouldn't see the wood for the trees. Rosenberg (1994, 76ff.) responds to this line of reasoning by arguing that it actually supports his own, instrumentalistic position. Sober's ar- gument, according to Rosenberg, can only establish the indispensability of evolutionary theoryfor us as beings to whom its generalizations appear significant. The argument fails to establish that the kinds evolutionary generalizations pick out are natural kinds. Furthermore, it fails to establish that the probabilities featured in these generalizations are real in any sense. I suggest that two issues must be kept separate here. The first issue is whether evolutionary theory in its current statistical form has an explan- atory value which would be lacking in a non-statistical theory of the kind S216 DETERMINISM, REALISM, AND PROBABILITY that Laplace's demon would be able to produce (if determinism is true). The second issue is whether realism about current evolutionary theory is justified. The first issue has to do with the nature of scientific explanation, specifically, whether unification has explanatory value. Many philoso- phers of science think that it does, and if they are right, Sober's point concerning significant generalizations holds true. But this is not what Ro- senberg denies, if I understand him correctly. His point is that the kinds of explanations we find significant has something to do with what kind of beings we are. Sober and Rosenberg agree that the statistical generaliza- tions of evolutionary theory are explanatorily indispensable; they differ only in that Sober thinks they are indispensable in principle whereas Ro- senberg thinks they are only indispensable for us cognitively limited be- ings. Sober (1984, 127) admits that it is "difficult to bring this science fiction thought experiment to a decisive conclusion." I agree, but fortu- nately we don't have to bring it to a conclusion. For the main issue at stake is realism, and I want to suggest that, with respect to this issue, nothing follows from a theory's explanatory dispensability. A theory may be dispensable in the sense that an omniscient being would be able to understand the phenomena in question at a deeper level, but it is still possible that this theory correctly represents some aspects of reality. To put it differently, a theory may be indispensable merely for pragmatic reasons, i.e., for reasons which have to do with our cognitive abilities, but still be open to a realist interpretation. The fact that a theory falls short of giving us a complete account of some complex causal processes does not imply that this theory has no representational content whatsoever. A scientific realist is not committed to the thesis that even our best scientific theories provide complete descriptions of reality.1 Thus, even if Rosenberg is right (contra Sober) that smarter beings would have less reason to use a statistical approach, this doesn't imply that our cognitively limited bi- ologists have failed completely in their attempts to provide a description of reality. This means that Rosenberg has not provided any compelling reasons to accept his instrumentalist account of biology, nor a subjective interpre- tation of probability, even if we accept his basic premise of determinism. Let us now have a look at another instrumentalist position, the one given by Horan (1994). Like Rosenberg, Horan starts with the assumption that biological processes are deterministic. On her account, the apparent randomness of processes like genetic drift is created by the indiscriminateness of its sampling from the breeding 1. Compare Waters' (1991) notion of "tempered realism." S217 MARCEL WEBER populations: Organisms are chosen in a way that is completely inde- pendent of how fit they are. If the sampling process is also determin- istic, then the statistical perspective introduced by the concept of "sampling" is not necessary. It is convenient and, given limitations in our knowledge of the vagaries of the environment, useful; however, it obscures the essentially deterministic character of evolutionary change. (Horan 1994, 85) Thus, Horan holds that even events like the establishment of a founder population by a natural disaster-one of the paradigmatic cases of the intrusion of chance in evolution-is a causally fully determined event which would have the exact same outcome under the same conditions. Similar to Rosenberg, Horan infers from this deterministic premise that "the statistical approach is instrumentally motivated but theoretically un- necessary" (1994, 78). I have argued above that whether or not the statis- tical approach is theoretically necessary, in other words, whether proba- bilistic theories of evolutionary change are in principle eliminable or not, has no consequences for the question of whether a realist interpretation of evolutionary theory is justified. These are separate issues. However, Horan has provided an additional argument for instrumentalism about evolutionary theory. Specifically, she has argued that population genet- ics-widely viewed as the core of Neo-Darwinian theory-fails to provide causal explanations of evolutionary processes. The main thrust of Horan's argument is that the equations of popu- lation genetics do not relate causes to their effects. They merely relate sets of effects to one another (1994, 88). For example, typical population ge- netics equations which give the change in gene frequency from one gen- eration to the next due to selection contain parameters called selection coefficients, which are supposed to represent the relative fitness of different genotypes. On the standard interpretation of selection theory, fitness dif- ferences are viewed as the cause of natural selection. So why do population genetics equations not relate causes to their effects? Because selection co- efficients, according to Horan (1994, 88), are themselves defined in terms of their effects. They specify the proportion of alleles that survive to the next generation.2 I think this argument is flawed. To see this, consider the following analogy: Newton's second law of motion relates a force F acting on a body of mass m to the body's acceleration a (F = ma). Now consider a simple mechanical model in which this law of motion is used to treat the motion of a mechanical system (say, a harmonic oscillator). With Horan, 2. Curiously, Horan's argument is reminiscent of the old "tautology" objection to the empirical content of selection theory. S218 DETERMINISM, REALISM, AND PROBABILITY we could argue that this model offers no causal explanation of the body's motion because the force Fis defined by it's effect, the acceleration a.3 But surely this would be absurd. The mathematical relations between physical quantities say nothing about whether or not a theory is causal. A theory is causal exactly if its dynamic equations have some causal physical system as a model. Nothing in the mathematical theory of population genetics stops us from giving it a causal interpretation, for instance, by saying that fitness differences cause gene frequency changes by natural selection.4 Se- lection coefficients only provide a measure of such fitness differences. I conclude that neither Rosenberg nor Horan have given us compelling reasons to accept their instrumentalist view of evolutionary theory, even if their deterministic premise is assumed to be true. Nothing follows from determinism for realism about biological theories, even if it is true that a statistical approach is chosen in evolutionary theory merely because hu- man beings lack the cognitive capacities for providing a full description of the complex causal processes which sustain life in every individual of a population. GHR's argument from determinism and eliminability to in- strumentalism is a fallacy, perhaps the result of latent reductionistic prej- udices according to which only theories which treat phenomena at the most fundamental level can be true. I want to end this section by pointing out that Rosenberg's and Horan's arguments could be used to show that classical statistical mechanics-which deals with fully deterministic sys- tems-cannot be interpreted realistically, or that it is not a causal theory. Foundational problems in statistical physics aside, probably very few physicists would accept such a verdict. So far, there are no reasons to accept such a verdict in evolutionary biology either, even for the deter- minist. But it is time now to visit the other camp, where determinism is out of fashion. 3. The Fallacy Again, and a Problem in BC. Brandon and Carson (1996) have challenged Horan's and Rosenberg's assumption of an underlying determinism in biological processes. They have offered two different ar- guments against determinism. The first argument tries to show that quan- tum mechanical indeterminism can, at least in principle, "percolate up" 3. It could be objected that F = ma is an empirical law (which is controversial) whereas the relationship between selection coefficients and gene frequency change is definitional. However, it has proven to be difficult to decide which parts of a set of dynamic equa- tions should be viewed as empirical and which parts as definitions of the theoretical magnitudes. On some of the most current accounts of the structure of theories, this question becomes irrelevant, since all dynamic equations are viewed as defining a set of models which are then applied to the world as wholes (Giere 1988, 84). 4. Sober (1984, 88ff.) holds fitness to be causally inert for other reasons. However, he shows how selection theory can be viewed as causal nevertheless. S219 MARCEL WEBER to the macroscopic level of biological populations. At the same time, Bran- don and Carson (1996, 320) claim that the possibility of quantum- mechanical indeterminism with population-level effects is insufficient to ground an "autonomous evolutionary indeterminism" which, on their view, is required if one wants to be a realist about evolutionary theory. For this reason, they present a second argument for biological indeter- minism. The basic strategy of this argument is to infer the falsity of de- terminism from its alleged theoretical unfruitfulness and lack of empirical support (1996, 329-333). Both of these arguments are problematic, as Millstein (1997, Chapter 2) and Graves et al. (1999) have shown. My aim here is not to offer more criticism of Brandon and Carson's argument for indeterminism, nor do I want to take sides on this issue. Instead, I want to show, first, that BC, too, have assumed too simple a relationship be- tween determinism, realism, and the interpretation of probability. Second, I want to point out an internal problem in BC's overall position. One of the main points of disagreement between GHR and BC is whether the probabilities posited in evolutionary models are "epistemic" or whether they are "real." In other words, do these probabilities merely reflect scientists' ignorance or their subjective degrees of belief, or do they represent some properties which exist in biological organisms and their environment independently of scientists' beliefs? Brandon and Carson (1996, 336) exhibit some modesty in conceding that they have not refuted the instrumentalist view. What they think they have established is the following: What we have shown is that if one is a realist in one's attitude toward science-that is, if one thinks that a primary aim of doing science is to develop theories that truly describe the mechanisms producing the phenomena, and if one takes theoretical fruitfulness and experimental confirmation as evidence for the reality of theoretical entities-then one should conclude that ET [evolutionary theory] is fundamentally indeterministic. If, however, one is a metaphysical determinist with respect to ET-that is, one who has decided for reasons outside of science that the process of evolution is deterministic-then one should conclude, along with Rosenberg and Horan, that ET is an instrumen- tal science. (Brandon and Carson 1996, 336) The first part of this claim, that a realist about evolutionary theory is committed to indeterminism, has already been criticized by Millstein (1997) and Graves et al. (1999). According to Millstein, the only rational attitude for the realist to take towards the question of indeterminism vs. determinism is that of an agnostic. The second part of the claim is once again what I have called the determinism-instrumentalism fallacy. Brandon and Carson (1996, 326) seem to think that probabilities ex- S220 DETERMINISM, REALISM, AND PROBABILITY press either subjective degrees of belief or real, indeterministic propensi- ties. If determinism is true, then probabilities are "epistemic," if it is false, then probabilities are genuinely stochastic propensities. But surely this dichotomy is too simple. For instance, there is a classical interpretation of probability which is neutral with respect to determinism/indeterminism: the frequency interpretation, according to which probabilities represent the actual or limiting frequency of an event in a series of like events. The frequency interpretation, too, makes probabilities real; it is an objective interpretation. Another objective interpretation of probability which is applicable to deterministic systems has been given by Mackie (1973, 181). On Mackie's interpretation, for instance, the probability of a certain out- come in a game of dice represents the number of initial states which, to- gether with the laws which govern the motion of dice, determine this out- come. The propensity interpretation, as it was introduced by Popper (1959), is also intended to be applicable to such deterministic chance setups (although, for Popper, the problem of the interpretation of quantum the- ory was a main motivation behind the propensity interpretation). Proba- bilities, thus understood, represent a property of chance setups that may have deterministic laws of working. Thus, contrary to what BC (and GHR) say, a determinist about biological processes can be a realist with respect to evolutionary theory and the probabilities it posits. Of course, it is still possible that, although realism about evolutionary theory is com- patible with determinism conceptually, determinism has been refuted em- pirically by biological research. I did not want to investigate the latter problem here. My aim so far was to show that the relationship between realism, determinism, and the interpretation of probability is more com- plex than both sides of the debate have assumed. Thus, it is a mistake to think that probabilities either have to represent subjective degrees of beliefs or irreducibly stochastic propensities. There are other possibilities. Of course, this by itself does not undermine BC's claim that probabilities in evolutionary theory do represent genuinely probabilis- tic propensities. Now I want to show that this is problematic, too. The main problem with BC's position is that it is committed to the view that all applications of the concept of probability in evolutionary theory are manifestations of indeterminism. Even if there exists variation in bio- logical systems which is not explainable by "hidden variables," there can be no doubt that there is also variation which is caused by some hidden variables, for example, micro-environmental differences in moisture, nu- trient concentration, pH, etc., which affect the growth of plants. Consider a colony of genetically identical plants that share the same environment. An investigation reveals that, in a given year, the individual plants produce different numbers of seeds. On the usual interpretation of the concept of fitness, these differences in seed production are not to be taken as an S221 MARCEL WEBER indication that the plants differ in fitness. Rather, biologists would cal- culate the average seed production to obtain an estimate of the genotype's fitness in the given environment. Now let us further assume that the vari- ation in the number of seeds is not the result of genuinely indeterministic events during the development of the plants. Instead, the differences are caused by an uneven distribution of nutrients in the soil, which is typical for this type of environment. Even if BC (1996, 333) are right in arguing that it is methodologically unsound to assume that there are always such hidden variables, surely it is not unscientific to assume that such hidden variables exist sometimes. Does this mean that probabilities are rendered epistemic in cases where there are hidden variables? This seems wrong. In my example, one could still say that the probability distribution for the number of seeds produced says something about the relationship between plants of that genotype and their environment; they are therefore not merely "epistemic." However, the probability distribution, in this case, does not represent a genuine stochastic propensity of the kind postulated in quantum mechanics; rather, it expresses the statistical effect of the action of the hidden variables. In my hypothetical (but realistic) example, the source of probability is not indeterminism but the heterogeneity of the environment. The problem is thus not only that BC's account cannot differentiate between cases where biological variation is caused by genuine indetermin- istic processes and cases where variation is caused by hidden variables. Because they view all probabilities featuring in evolutionary theory as expressions of irreducibly stochastic propensities, their position renders many cases of evolution incomprehensible. Even if objective chance exists in biological processes, it is unlikely that it can explain all the variability that renders evolutionary processes stochastic. Evolutionary theory, I maintain, is applicable to intra- and inter-populational variability regard- less of whether this variability is a consequence of objective chance or the action of complex hidden variables. Thus, even if strict determinism is false, BC's position fails to account for those applications of evolutionary concepts where biological variability is not caused by genuinely indeter- ministic events but by some complex hidden variables, which are likely to abound. 4. Outlook. I have shown that both of the recently defended major posi- tions which try to account for the statistical character of evolutionary theory are ridden with problems, regardless of the truth of determinism. The time is ripe to develop alternative accounts that avoid the pitfalls I have pointed out. I suggest that viable alternatives should satisfy the fol- lowing criteria of adequacy. First, an account of the statistical character of evolutionary theory should be open to a realist interpretation at least S222 DETERMINISM, REALISM, AND PROBABILITY to the extent that the best theories of physics and chemistry are. To allow realism about the latter but not about evolutionary theory seems odd, considering that evolutionary theory is one of the most highly confirmed theories in the history of science (Brandon and Carson 1996, 316). Second, a viable alternative should be compatible with determinism. If a non- negligible fraction of biological variability is caused by hidden variables, which is a safe bet, then we need an account which can explain how the combined action of these hidden variables gives rise to the stochastic char- acter of the evolutionary process. Should it then turn out that objective chance plays a role as well (which is far from clear at present; see Millstein 1997), this would not undermine such an account. Clearly, compatibility with determinism is going to be the hard part. But we should do at least as good as philosophers of physics (e.g., Sklar 1993) who have developed realist accounts of the foundations of classical statistical mechanics, which is a fully deterministic theory. The crux of the matter is going to lie in the concept of probability. Perhaps it is time to admit that the propensity interpretation has outlived its usefulness, because it makes a mess of the distinction between objective chance as it occurs in quantum mechanics and chance as it occurs in de- terministic systems. I suggest that the most helpful way to think about probability in evolutionary theory is in terms of ensembles, an approach which has also proven to be fruitful in statistical mechanics (Sklar 1993). For example, to say that an allele has a probability ofp to go to fixation by random drift means that in a fraction of p in a (fictional) ensemble of populations the allele will go to fixation.5 Thus, evolutionary probabilities can be viewed as frequencies, though not as frequencies in a series of trials, but frequencies in ensembles of populations. Obviously, the interesting question is going to be how the relevant ensembles are to be individuated. I will deal with this question elsewhere. My goal here is to motivate alter- native accounts of the statistical character of evolutionary theory, not to actually present such an alternative. It should be easier now, since we know where the pitfalls are. REFERENCES Beatty, John (1984), "Chance and Natural Selection", Philosophy of Science 51: 183-211. Brandon, Robert N. and Scott Carson (1996), "The Indeterministic Character of Evolution- ary Theory: No 'No Hidden Variables Proof' But No Room for Determinism Either", Philosophy of Science 63: 315-337. 5. This ensemble approach is mentioned by participants of this debate in various places (e.g., Rosenberg 1994, 68; Brandon and Carson 1996, 323). However, they don't seem to be aware of its potential for solving the conceptual problems at hand. S223 S224 MARCEL WEBER Giere, Ronald N. (1988), Explaining Science: A Cognitive Approach. Chicago: University of Chicago Press. Graves, Leslie, Barbara L. Horan, and Alexander Rosenberg (1999), "Is Indeterminism the Source of the Statistical Character of Evolutionary Theory?", Philosophy of Science 66: 140-157. Horan, Barbara L. (1994), "The Statistical Character of Evolutionary Theory", Philosophy of Science 61: 76-95. Mackie, John L. (1973), Truth, Probability, and Paradox. Oxford: Clarendon Press. Millstein, Roberta L. (1996), "Random Drift and the Omniscient Viewpoint", Philosophy of Science 63 (Proceedings): S10-S18. (1997), The Chances of Evolution: An Analysis of the Roles of Chance in Microevo- lution and Macroevolution. Ph.D. Dissertation. Minneapolis: University of Minnesota. Popper, Karl R. (1959), "The Propensity Interpretation of Probability", British Journalfor the Philosophy of Science 10: 25-42. Rosenberg, Alexander (1994), Instrumental Biology or the Disunity of Science. Chicago: Uni- versity of Chicago Press. Sober, Elliott (1984), The Nature of Selection: Evolutionary Theory in Philosophical Focus. Cambridge, MA: MIT Press. Sklar, Lawrence (1993), Physics and Chance: Philosophical Issues in the Foundations of Sta- tistical Mechanics. Cambridge: Cambridge University Press. Waters, C. Kenneth (1991), "Tempered Realism About the Force of Selection", Philosophy of Science 58: 553-573. Weber, Marcel (1996), "Fitness Made Physical: The Supervenience of Biological Concepts Revisited", Philosophy of Science 63: 411-431. Article Contents p. S213 p. S214 p. S215 p. S216 p. S217 p. S218 p. S219 p. S220 p. S221 p. S222 p. S223 p. S224 Issue Table of Contents Philosophy of Science, Vol. 68, No. 3, Supplement: Proceedings of the 2000 Biennial Meeting of the Philosophy of Science Association. Part I: Contributed Papers (Sep., 2001), pp. i-vii+S1-S543 Front Matter [pp. i-v] Introduction [p. vii] Metaphysics and Methodology of Science Refusing the Devil's Bargain: What Kind of Underdetermination Should We Take Seriously? [pp. S1-S12] Is Structural Realism Possible? [pp. S13-S24] Are Realism and Instrumentalism Methodologically Indifferent? [pp. S25-S37] Deflationary Metaphysics and the Natures of Maps [pp. S38-S49] Reference and Resemblance [pp. S50-S61] The Reemergence of 'Emergence' [pp. S62-S75] The Recently Claimed Observation of Atomic Orbitals and Some Related Philosophical Issues [pp. S76-S88] Causal Loops and the Independence of Causal Facts [pp. S89-S97] Foundations of Probability Undermining Undermined: Why Humean Supervenience Never Needed to Be Debugged (Even If It's a Necessary Truth) [pp. S98-S108] Probability and Symmetry [pp. S109-S122] Bayesian Methodology A Bayesian Account of Independent Evidence with Applications [pp. S123-S140] Statistical Model Selection Criteria and Bayesianism [pp. S141-S152] Bayesian Statistics in Radiocarbon Calibration [pp. S153-S164] Old Evidence and New Explanation III [pp. S165-S175] Theory-Ladenness and the Neurology of Perception The Theory-Ladenness of Observation and the Theory-Ladenness of the Rest of the Scientific Process [pp. S176-S186] Reentrant Neural Pathways and the Theory-Ladenness of Perception [pp. S187-S199] Perceptual Symbols and Taxonomy Comparison [pp. S200-S212] Philosophy of Biology and Cognition Determinism, Realism, and Probability in Evolutionary Theory [pp. S213-S224] Modeling Evolution in Theory and Practice [pp. S225-S236] Pluralism and Sex: More Than a Pragmatic Issue [pp. S237-S250] Model Organisms as Models: Understanding the 'Lingua Franca' of the Human Genome Project [pp. S251-S261] Group-Level Cognition [pp. S262-S273] Quantum Mechanics Quantum Logic Is Alive ∧ (It Is True ∨ It Is False) [pp. S274-S287] Explaining Information Transfer in Quantum Teleportation [pp. S288-S300] Realism and Underdetermination: Some Clues from the Practices-Up [pp. S301-S312] Statistical Mechanics Statistical Mechanics and the Asymmetry of Counterfactual Dependence [pp. S313-S324] Infinite Systems in SM Explanations: Thermodynamic Limit, Renormalization (Semi-) Groups, and Irreversibility [pp. S325-S344] Relativity and Fields Definition, Convention, and Simultaneity: Malament's Result and Its Alleged Refutation by Sarkar and Stachel [pp. S345-S357] Can the Bundle Theory Save Substantivalism from the Hole Argument? [pp. S358-S370] The Principles of Gauging [pp. S371-S381] Quantum Gravity Why Quantize Gravity (Or Any Other Field for That Matter)? [pp. S382-S394] Singularities and Scalar Fields: Matter Theory and General Relativity [pp. S395-S406] Black Hole Remnants and Classical vs. Quantum Gravity [pp. S407-S423] Against the Excesses of Quantum Gravity: A Plea for Modesty [pp. S424-S441] Theories, Models and Analogies Simulations, Models, and Theories: Complex Physical Systems and Their Representations [pp. S442-S454] Extrascientific Uses of Physics: The Case of Nonlinear Dynamics and Legal Theory [pp. S455-S466] Science and Values Science, Biases, and the Threat of Global Pessimism [pp. S467-S478] Racism and Human Genome Diversity Research: The Ethical Limits of "Population Thinking" [pp. S479-S492] Ecological Stability, Model Building, and Environmental Policy: A Reply to Some of the Pessimism [pp. S493-S505] Science and Values: Are Value Judgments Always Irrelevant to the Justification of Scientific Claims? [pp. S506-S518] History and Philosophy of Science Counterfactual Histories: The Beginning of Quantum Physics [pp. S519-S530] Can a Historian of Science Be a Scientific Realist? [pp. S531-S541] Back Matter [pp. S543-S543]