Stanford Encyclopedia of Philosophy
This is a file in the archives of the Stanford Encyclopedia of Philosophy.

Reductionism in Biology

First published Tue May 27, 2008

Reductionism encompasses a set of ontological, epistemological, and methodological claims about the relation of different scientific domains. The basic question of reduction is whether the properties, concepts, explanations, or methods from one scientific domain (typically at higher levels of organization) can be deduced from or explained by the properties, concepts, explanations, or methods from another domain of science (typically one about lower levels of organization). Reduction is germane to a variety of issues in philosophy of science, including the structure of scientific theories, the relations between different scientific disciplines, the nature of explanation, the diversity of methodology, and the very idea of theoretical progress, as well as to numerous topics in metaphysics and philosophy of mind, such as emergence, mereology, and supervenience.

In recent philosophy of biology (1970s to the early 1990s), the primary debate about reduction has focused on the question of whether (and in what sense) classical genetics can be reduced to molecular biology. Another less prominent strand of discussion concerns whether evolutionary theory is inherently anti-reductionist because of the principle of natural selection. In the past decade, philosophical debate has shifted to also include whether developmental biology can be reduced to molecular genetics. Philosophical interest about reduction in biology is pervasive throughout the history of philosophy and science. Many contemporary debates have historical analogues, reflecting long-standing controversies among biologists about the legitimacy of reductionist research strategies and modes of explanation used by different life science subdisciplines.

1. Introduction

We reduce and reduce and reduce, but the habits of anthropomorphism and metaphor are hard to kill. (Margaret Drabble, The Sea Lady, 146)

Reduction (reductionism) encompasses several, related philosophical themes. At least three core types can be distinguished: ontological, methodological, and epistemic (Sarkar 1992; cf. Nagel 1998). Even though arguments for and against reductionism often include a combination of positions related to all three, these distinctions are significant because no straightforward entailment relations obtain between the different aspects of reduction (although tacit commitments about these relations are quite prevalent).

(i) Ontological reduction is the idea that each particular biological system (e.g., an organism) is constituted by nothing but molecules and their interactions. In metaphysics this idea is often called physicalism (or materialism), which assumes in a biological context that (a) biological properties supervene on physical properties (i.e., no difference in a biological property without a difference in some underlying physical property), and (b) each particular biological process (or token) is metaphysically identical to some particular physico-chemical process. This latter tenet is sometimes called token-token reduction, in contrast to the stronger tenet that each type of biological process is identical to a type of physico-chemical process. Ontological reduction in this weaker sense is a default stance nowadays among philosophers and biologists though the philosophical details remain controversial, such as whether there are genuinely emergent properties. Various conceptions of physicalism may yield different implications for ontological reduction in biology (cf. Dowell 2006). The denial of physicalism by vitalism, the doctrine that biological systems are governed by forces that are not physico-chemical, is largely of historical interest. Some authors have argued vigorously for the significance of metaphysical concepts in the discussion of reductionism in biology (Rosenberg 1978, 1985, 1994, 2006).

(ii) Methodological reduction is the idea that biological systems are most fruitfully investigated at the lowest possible level, and that experimental studies should be aimed at uncovering molecular and biochemical causes. A common characterization of this type of strategy is what has been termed “decomposition and localization” (Bechtel and Richardson 1993). While methodological reductionism is often motivated by the presumption of ontological reduction, this procedural recommendation does not follow directly from it. In fact, unlike the idea of ontological reduction, methodological reductionism can be quite controversial. It has been argued that exclusively reductionistic research strategies can be systematically biased so as to overlook salient biological features and that for certain questions a more fruitful methodology consists in integrating the discovery of molecular causes with the investigation of higher level features (Wimsatt 1980).[1]

(iii) Epistemic reduction is the idea that the knowledge about one scientific domain (typically about higher level processes) can be reduced to another body of scientific knowledge (typically concerning a lower and more fundamental level). While an endorsement of some form of epistemic reduction can be motivated by ontological reduction combined with methodological reductionism (e.g., the past success of reductionistic research in biology), the possibility of epistemic reduction does not follow from the conjunction of ontological and methodological reduction. Indeed, debates about reduction in the philosophy of biology have centered on this third aspect of reduction as the most controversial issue. Our discussion will therefore focus primarily on issues related to epistemic reduction. Prior to evaluating any reduction of one body of knowledge to another, a conception of those bodies of knowledge and what it would mean for them to be “reduced” must be explicated. A number of different models of reduction have been proposed as a consequence. Thus the debate about reduction in biology has not only revolved around whether epistemic reduction is possible, but on which notion of epistemic reduction is most adequate for actual science. Two basic categories can be distinguished: (a) models of theory reduction maintain that one theory can be logically deduced from another theory; and, (b) models of explanatory reduction focus on whether representations of higher level features can be explained by representations of lower level features.

The philosophical study of the notion and possibility of reduction links up with several core issues in the philosophy of science. As reduction involves relating (broadly speaking) one “body of scientific knowledge” to another, it presupposes an account of what the relevant epistemic units to be related are—theories, laws, models, concepts, or disciplines—and how they should be understood. Since the reduction of one theory to another is usually presumed to mean that somehow the reducing theory explains the principles of the reduced theory, reduction is also related to the issue of scientific explanation. For example, if a more fundamental theory fully reduces another theory or even reduces several theories, unification (of some sort) results. The situation where a historically earlier theory is reduced to a later theory is an intuitive case in which scientific progress might occur. Since arguments against the existence of this type of reduction have maintained that terms occurring in both the precursor and descendant theory have different meanings (“incommensurability”), reduction also relates to the issue of the meaning of scientific terms and conceptual change. Therefore, accounts of reduction depend on commitments made about other, coordinating philosophical notions and their adequacy depends in many cases on these related philosophical commitments. This has led some philosophers to pursue an account of reduction in biology that is as uncommitted as possible regarding the precise construal of general notions such as “law” and “explanation” (Sarkar 1998).

In philosophy of biology, reduction involves questions about how different biological disciplines and theories are related, and how biology itself is connected to physics and chemistry (e.g., is biology autonomous, possessing its own theoretical principles, explanations, and methods?). Over the past four decades, discussion has concentrated primarily on the question of whether classical genetics can be reduced to molecular genetics and biochemistry. This discussion was motivated by the increasing significance of molecular biology within the life sciences, and proceeded from previous accounts of reduction in general science, such as Nagel's (1961) model. A less central concern has been whether evolutionary biology and ecology are autonomous disciplines or reducible to molecular biology (Beatty 1990, Brandon 1996, Dupré 1993, Rosenberg 2006). One of the key issues in evolutionary contexts is the relationship between functional explanation (“teleology”) and causal-mechanistic explanation (Wouters 2005). The levels of selection debate includes questions about whether all evolutionary explanations can be cast in terms of genes and whether selection operates primarily or exclusively on the genetic level. The main debate about reduction in genetics has lost some intensity over the past ten years as a majority of philosophers developed a skeptical consensus regarding the possibility of reduction in a substantial sense (but see Waters 1990). Recently an analogous debate has been revived regarding whether developmental biology can be reduced to developmental genetics and molecular biology (Rosenberg 1997), reflecting recent progress in developmental genetics and growing philosophical interest in developmental biology. Much less attention has been given by philosophers to the question of whether biology can be reduced to physics (but see Rosenberg 2006). This has been an issue of concern to many biologists (e.g., Mayr 2004), both with respect to methodology (e.g., the adequacy of analyzing a system by dissecting it into component parts) and whether modes of explanation from physical science are inherently inappropriate for biological science, such as explanation by subsuming phenomena under universal laws.

The debate about reduction in biology has often been phrased in terms of “reductionism” vs. “anti-reductionism” (or, more archaically, “mechanism” versus “vitalism”). However, since different notions and models of reduction have been put forward, whether one takes a reductionist or anti-reductionist stance about a particular biological subdiscipline (e.g., classical genetics) actually depends on the notion of “reduction” being utilized (Wimsatt 1976a, Hull 1979). Furthermore, the reductionism versus anti-reductionism terminology has tended to create a false dichotomy between two extreme positions: reductionism as the idea that molecular biology can in principle fully explain all biological facts—making higher level biological theories dispensable—and anti-reductionism as the idea that higher level biological fields possess explanatory principles of their own and are fully autonomous in the sense of not benefiting from molecular biology. A variety of middle (and orthogonal) ground exists between these extremes and has motivated many of the efforts seen in alternative research programs in philosophy of science (see Section 5).[2]

2. Historical Background: Philosophical and Scientific

And what, asks Ailsa at random, as she flicks through the pages of Darwin, is morphology? Morphology is dead rabbits in formalin, says Humphrey. (M. Drabble, The Sea Lady, 225)

Undoubtedly, the growth and development of molecular biology over the past half century has made reductionism in biology a central issue (and DNA a household word). But it would be wrong to assume that the different aspects of reductionism only gain traction in the wake of the molecularization of biology juxtaposed with the discussion of reduction from a logical empiricist perspective (see Section 3.1). Besides a perennial concern with what makes life distinctive, we can distinguish at least two reductionist themes throughout history relevant to the life sciences and its philosophy: (1) the relation among different branches or domains of knowledge and (2) the relation between parts and wholes (Grene and Depew 2004, Magner 1994). These two themes link up in a complex fashion with both epistemic and ontological aspects of reduction. (Questions about methodological reduction tend to coalesce around new technologies that open up the possibility of pursuing reductionist research methods, such as making observations at a lower level with microscopy.) Additionally, these aspects of reductionism arise in the context of specific, enduring biological topics: the complex relations among different animals and plants in natural environments (“ecology”), the integrated relations among the parts and whole of an organism (“functional anatomy/physiology”), and the dynamic relations among the homogeneous components in the early stages of an embryo that eventually beget a unified whole organism containing heterogeneous parts in appropriate arrangement and connection (“development/reproduction”).

Aristotle is the foundation for both philosophical themes pertaining to reductionism. He addressed the relationship between domains of scientific knowledge in his concepts of superordinate and subordinate sciences (Posterior Analytics I). Each science has a distinctive subject genus, first principles, and set of predicates that are commensurable with the subject genus (McKirahan 1978, 1992, ch. 4–5). Though the relations among sciences are systematic, they are not necessarily reductive relationships as there is no most fundamental science. These connections are relevant to the other philosophical theme of the relations between parts and wholes because Aristotle utilized his science of elements and their potentials (e.g., as seen in On Generation and Corruption or Meteorology) to inform his description and analysis of organism components in the History of Animals and Parts of Animals. He treated three compositional “levels”: (i) composition in terms of elements and their potentials, (ii) composition of uniform parts such as blood and bone; and (iii) composition of non-uniform parts that include multiple uniform parts, such as the face or hands. The primary biological topic here is functional anatomy but development/reproduction is always in view and addressed directly in the Generation of Animals. Claims about “potentials” are buttressed by an account of their contributions to material properties composing uniform parts (Meteorology IV; see Lennox 2001, ch. 8; Popa 2005). Complex dispositions (e.g., fleshy, elasticity of the esophagus) are postulated from these material substrates that have relevance to, but do not determine the nature of, these organismal parts because functional strictures govern their nature (fleshy, elasticity of the esophagus is for the sake of repeated dilation during food ingestion). An esophagus outside of the purposive context of an organism will be an esophagus “in name only”. The teleological emphasis in Aristotle's biology means that he is often characterized as a kind of ontological anti-reductionist (e.g., the elemental potentials of lower levels of organization are for the sake of the uniform parts at higher levels of organization). This blends with epistemic issues related to particular topics, such as his rejection of Presocratic materialist (i.e., purely “mechanistic”) explanations for the origin of animal morphology during development. (Empedocles apparently claimed that vertebrae arise during ontogeny solely as a consequence of a unified cylindrical column physically breaking into the observable iterated units of the backbone; see Parts of Animals I.1 [640a].) But Aristotle was also attentive to issues of material composition and constraints arising from the nature of properties exhibited by animal parts, which often mark commitments to types of ontological and epistemic reduction. His concept of hypothetical necessity (Physics II; see Cooper 1987) permitted one kind of material to be put to use in many different organismal parts depending on the context (see below, Section 4.1). Galen, among many others subsequent to Aristotle, also explored many of these issues, whether it be from the perspective of how material complexes generate higher level properties (Mixtures) or the relations among parts and wholes in the context of a specific biological topic such as development (The Construction of the Embryo).

At the beginnings of the early modern period, William Harvey adopted an Aristotelian stance in the context of both functional anatomy and development (Lennox 2006). The relation between parts and wholes in ontogeny, including relevant material properties, is constantly in view (Harvey 1981 [1651]). René Descartes famously disputed these claims from the perspective of the “mechanical philosophy”. His mechanistic reductionism sought to explain the movement of the heart not in terms of its function but rather by appeal only to the matter in motion composing the circulating blood that is heated and thus expands akin to fermentation (Description of the Human Body). Descartes also made unsuccessful attempts at explaining embryology in terms of matter and motion alone; e.g., tissue and organ origination is accounted for via matter slowing down and aggregating in different regions of the embryo (Des Chene 2001, ch. 2; Smith 2006, Part II). Robert Boyle defended teleological explanations in anatomy and physiology (specifically, the topic of integrated relations among the parts and whole of an organism) by laying out how they were compatible with and supported by mechanistic explanations (Lennox 1983). Thus, he offered a version of epistemic reduction in line with his conception of mechanical philosophy, which included a mixed reductionist/non-reductionist methodology while simultaneously harboring reservations about ontological reduction. Immanuel Kant explicitly articulated the dialectic between a teleological outlook on organismal functioning (highlighting ecological relationships, functional anatomy, and development specifically) with a mechanistic understanding of causation modeled off of Newtonian mechanics (Critique of Judgment, Part II). Kant's maintenance of this tension between the reciprocal causal relations of parts in an organism and the linear-mechanism causality justified within his critical philosophy was a way of splitting apart epistemological and metaphysical issues, yielding a hybrid position with elements of epistemic anti-reductionism (organismal processes have to be understood in terms of their reciprocal contribution to a systemic goal) and ontological reductionism (all causation is linear-mechanical). In addition to this focus on the theme of reductive relations between parts and wholes, Kant also addressed the theme of the relation among different scientific domains in his discussion of sciences having their own distinct concepts and subject matter (Critique of Judgment, §68, 79).

Kant's work was influential, even if misinterpreted in various ways, and partially set the stage for the framing of a key reductionism related question of the early 19th century: teleology versus mechanism (Lenoir 1982). His discussion stressed the significance of methodological conclusions—it is necessary (a maxim or regulative principle) to take organisms as natural purposes if we are to investigate them scientifically. This encouraged speculations about new forces, akin to those found in Newtonian mechanics, which would explain biological phenomena mechanistically (e.g., the formative force explaining the development of the embryo; cf. Look 2006, Richards 2000). The Newtonian inspiration survived, with or without Kantian traces, well into the 19th century and beyond. Comparative anatomy (conceptualized broadly to include functional anatomy and embryology) exploited methodological suggestions from successful physical science, as when Geoffroy St. Hilaire appealed to Newton's musings that animals are governed by similar principles or laws as those found in mechanics (Le Guyader 2004). More typical is Richard Owen's discussion of “Matter” and its properties within living organisms, where he exhibits a clear anti-reductionist stance and postulates something akin to emergent properties.[3] In short, there was a diversity of mechanist (“reductionist”) and non-mechanist (“non-reductionist”) positions with respect to ontological, epistemic, and methodological aspects of reduction in the 19th century, similar to the philosophical diversity observable today (see Sections 3 and 5 below). Lenoir (1982) discusses in detail various 19th century accounts that combine reductionist and non-reductionist ideas (especially the research tradition he labels “teleomechanism”).

A related strand regarding reductionism winds its way through physiological investigations of the 19th century. Here the Cartesian mechanical-reductionistic program is rejoined with an emphasis on the animal machine (see Canguilhem in Delaporte 1994, ch. 5, 8, 10, 12; Coleman 1977, ch. 6). Methodological aspects of reductionism were seriously qualified because of their impact on experimental design despite clear metaphysical commitments to a form of ontological reduction (e.g., Bernard 1957 [1865]). But methodologically non-reductionist approaches can also be found in natural history and classificatory discourse, as seen in Lamarck's Zoological Philosophy (1984 [1809]), especially poignant given his materialist perspective on living beings.

In the late 19th and early 20th century the topic of development returned with a vengeance in the controversy over vitalism. A general movement towards a more mechanistic or materialistic interpretation of living systems was emerging at this time (e.g., Loeb 1912; cf. Allen 1975) and disagreements about explaining development between Wilhelm Roux and Hans Driesch loomed large (Maienschein 1991). But the famous example of vitalism in Driesch's interpretation of development and the autonomy of the organism should be seen as an issue of epistemology as much as metaphysics (Maienschein 2000). Instead of materialism versus vitalism (ontological reduction questions), the explanatory conflict involved the nature of differentiation in early ontogeny and to what degree it is prespecified. The theme of order and organization in living systems (especially during development) pervades many writings about methodological, epistemological, and ontological aspects of reductionism at this time (emphasizing both relations between physics and biology and between parts and wholes). Examples include Joseph Needham's Order and Life (1936), Kurt Goldstein's The Organism (1934/1963), E.S. Russell's The Interpretation of Development and Heredity (1930), D'Arcy Thompson's On Growth and Form (1917), Ludwig von Bertalanffy's Modern Theories of Development (1933), and J.H. Woodger's Biological Principles (1929). Philosophers of science who have had less influence on contemporary Anglo-American philosophy were keenly aware of these discussions. For example, Ernst Cassirer (1950, Part II), starting with Kant, explored reductionist themes related to causal explanations of development, including the debates about mechanism and vitalism surrounding Roux and Driesch, as well as the relations between physical science and biology. These topics also appeared in the pages of Philosophy of Science (e.g., Lillie 1934, 1942, 1948; Singer 1934, 1946), although not necessarily utilizing the specific terminology of reduction (cf. Byron 2007).

Ernest Nagel's general treatment of reductionism in both chapters 11 and 12 of The Structure of Science (1961) canvassed many different aspects of these perennial debates. Whereas the sections concerning theory reduction have usually been in view (connecting with the reductionist theme of relating domains of knowledge), Nagel also interacted with other biological literature dealing with claims about parts and wholes as well as emergence, thereby addressing the second reductionist theme that has been prominent since Aristotle (see Nagel 1961, 366ff). He highlighted the polysemy of “wholes”, “parts”, and “sums” inherent in claims about the success or failure in showing that the whole is just the sum of its parts (pp. 380ff), predictably rephrasing the discussion in epistemological rather than metaphysical terms. Additionally he treats the question of evolutionary emergence, i.e., whether genuinely new entities arise in the history of our universe and life on Earth, which was salient at this time (Goudge 1961). Nagel also dealt with the aforementioned theoretical biologists (e.g., Bertalanffy, Russell, and Woodger) on the topic of hierarchical organization and reduction in living systems, especially as seen in ontogeny (Nagel 1961, pp. 432ff). Nagel's contemporaries also treated reductionist themes, though their work has largely been overlooked (e.g., Morton Beckner's discussion of “organization” and “levels of analysis” as methodological and explanatory aspects of reduction; Beckner 1959, ch. 9).

Despite Nagel's wide-ranging analysis, most subsequent philosophical discussion about reduction has concentrated on his account of theory reduction (see Section 3.1), especially whether it serves as an adequate general characterization of reductionism for all areas of scientific inquiry.[4] The biological case that has received the most scrutiny is the relationship between classical and molecular genetics, in part because the molecularization of genetics was palpable in this historical period. The manifold difficulties encountered in applying Nagelian theory reduction to genetics (see Section 4) has encouraged the growth of philosophy of biology as an independent disciplinary specialty, in part because these difficulties seemed to fit a problematic pattern—misconstruing biological reasoning with philosophical accounts of science forged on physical science examples. Philosophical issues in biology seemed to require distinct analyses that are more sensitive to empirical research in biology. Additionally, the rise of sociobiology and gene-centered evolutionary explanations provoked anti-reductionist stances among biologists on the basis of social and political commitments (e.g., Levins and Lewontin 1985).

The emergence of philosophy of biology and the concomitant attention to reductionism in situ within different life science contexts has lead many philosophers to revisit the perennial aspects of reductionism seen throughout history, many of which were simply tangential to the controversies over theory reduction. A number of current developments in reductionism debates are actually unconscious returns to neglected issues, albeit in different ways and in different contexts. That developmental biology has returned to the center of the discussion is not surprising because of its enduring status as a biological topic inviting reflection on reductionist themes. The same can be said for ecology and functional anatomy, although they are at present more peripheral for most philosophers. Social and political issues related to reductionist research in biology, especially as they pertain to humans (e.g., racial categorization using genetic profiles), continue to motivate philosophical analyses (see Section 5).

3. Models of (Epistemic) Reduction

He was not much of a diver. He was more of a laboratory man. Diving was out of fashion. … It was all DNA and genes and chromosomes and microbiology and eubacteria these days. Or so all those books had taught her. (M. Drabble, The Sea Lady, 8–9)

Most discussion in contemporary philosophy of science has centered on the issue of epistemic reduction (rather than ontological or methodological reduction, see Hoyningen-Huene 1989). The basic idea that the knowledge (or other epistemic units such as concepts) pertaining to one scientific domain can be reduced to another body of knowledge (or its epistemic units) has been spelled out in different ways. It is useful to group different models of reduction into two basic categories: (1) models of theory reduction maintain that a higher level theory can be logically deduced from a lower level theory, and, (2) models of explanatory reduction focus on whether representations of higher level features can be explained by representations of lower level features, typically by decomposing a higher level system into its parts (Sarkar 1992). Models of theory reduction were of primary interest in post-positivist philosophy of science that emerged in the 1960s and have received the most attention (Schaffner 1993). More recently, accounts of explanatory reduction have been developed to remedy the inadequacies of theory focused accounts, so that theory reduction has been largely displaced as the central issue by competing accounts of explanatory reduction.

3.1 Theory Reduction

According to theory reduction, theory TA reduces TB (e.g., Atomic physics reduces Biology) if TA logically entails TB. Given a deductive-nomological account of explanation (Hempel and Oppenheim 1965 [1948]), theory reduction as deduction from theoretical principles is an instance of explanation. In particular, TA logically entails and explains the laws of TB. Compare this with the logically weaker notion of reduction put forward by Kemeny and Oppenheim (1956), which includes instances of theory replacement. The Kemeny-Oppenheim model concerns the situation where the observable predictions made by TA include all observable predictions of TB (and possibly more). In this case TA can explain everything that TB explains, but TA need not encompass the theoretical principles of TB as in genuine theory reduction (e.g., TB may turn out to be empirically inadequate and eventually replaced by TA).

Many models of theory reduction derive from the account of reduction due to Nagel (1949, 1961). Working within the philosophical framework of logical empiricism, Nagel construed reduction as a logical relation between theories, where a theory is understood as a system of statements, containing laws, formulated in a first-order formal language. The reducing theory TA reduces the reduced theory TB if the laws of TB can be logically derived from TA, which Nagel called the “condition of derivability”. However, if TB contains scientific terms that do not occur in the language of the reducing theory TA (“organismal reproduction” is not an expression of biochemistry), derivability presupposes that the primitive expressions of TB (in particular its predicates) can be logically related to the language of TA. Nagel made this explicit by the “condition of connectability”: reduction presupposes that there are statements (e.g., conditionals) that contain expressions both from TA and TB, such that TA together with these statements entails the laws of TB. These connectability statements are often called bridge principles (or correspondence rules, or reduction functions), and different accounts of their nature have been put forward. Nagel rejected the idea that bridge principles are analytic statements but did not take a stand as to whether they are factual or conventional. Some philosophers have required that bridge principles are laws (rather than accidental generalizations), whereas others have maintained that they should express metaphysical identities between entities, properties, and processes (e.g., identifying cells with collections of molecules, in line with ontological reduction). Nagel himself construed reduction as an epistemic notion without ontological commitment; he did not even require that bridge statements be biconditionals, as mere conditionals when conjoined with TA can logically entail TB. (The Kemeny-Oppenheim notion of reduction / theory replacement is weaker than Nagelian reduction as it does not require such bridge principles connecting TA and TB.)

This account covers interlevel theory reduction. For instance, Oppenheim and Putnam (1958) argued for a hierarchical model of scientific disciplines, where each discipline is reducible to the theory on the next lower level (e.g., sociology to organismal biology, organismal biology to molecular biology, molecular biology to chemistry, and chemistry to microphysics). Another way to apply Nagel's general model is successional theory reduction, where a historically earlier theory is reduced to a later, more mature theory, so that reduction can be an instance of theoretical progress. In both cases Nagel held that theories needed to be indexed to a particular time: “The question whether a given science is reducible to another cannot in the abstract be usefully raised without reference to some particular stage of development of the two disciplines” (Nagel 1961, 361). In this context, reduction (and especially Nagel's connectability condition) was prominently criticized by Paul Feyerabend (1962, 1965a, 1965b) and also by Thomas Kuhn (1962). Feyerabend argued that an earlier and later theory may use the same theoretical term with a different meaning (e.g., “mass” in classical and relativistic mechanics). This incommensurability of meaning makes the connectability of these theories' expressions, and thus the logical derivation of one theory from another, impossible. Whether or not genuine incommensurability obtains is controversial (see, e.g., Sankey 1994).

Nagel's model was revised and applied to a specifically biological context—the reduction of classical genetics to biochemistry—by Kenneth Schaffner (1967, 1969, 1976; see Schaffner 1993 for his mature account). Schaffner's “General Reduction Model” addresses a key drawback of Nagel's account, namely the fact pointed out by Feyerabend that even though the theory to be reduced may be approximately true, it may contain empirically false statements, especially if it is a historical precursor of the more mature reducing theory (e.g., the inheritance “laws” of classical genetics have exceptions). If the theory to be reduced (TB) contains false statements then it cannot possibly be deduced from TA. For this reason, Schaffner considers instead TB*, which is a corrected version of the theory to be reduced in that TB* makes more adequate predictions than TB. On this model, reduction occurs if TA and TB* can be related by reduction functions (bridge principles) and TB* follows logically from TA enriched by the reduction functions. (While Nagel focused on the laws of TB being deducible from TA, Schaffner requires that all statements of TB/TB* are deducible from TA.) Schaffner's central tenet about reduction in biology is that the success of molecular biology shows that classical genetics and traditional fields of experimental biology are in the process of being reduced to biochemistry. While this reduction has not been completed yet, a logical derivation of traditional experimental biology from a (completed) theory of biochemistry is in principle possible and will eventually be achieved. Schaffner's account has an ontological aspect in that he interprets the reduction functions as synthetic statements expressing identities between the entities and processes of the two domains related (in line with Sklar 1967 and Causey 1972a, 1972b), though he does not argue that these identities are metaphysical necessities. The Nagel-Schaffner approach to theory reduction assumes a syntactic account of theory structure such that they are axiomatized systems formalizable in first order predicate calculus. A variety of other models of theory reduction exist, some of which use a semantic (i.e., theories as families of models) rather than a syntactic account of scientific theories (see Woodger 1952, Suppes 1957, Hooker 1981, Balzer and Dawe 1986a,b).

Schaffner's account provided the motivation and point of departure for much of the reduction debate in philosophy of biology, which was conducted primarily in the 1970s and 80s and centered on the issue of whether (and in which sense) classical genetics can be reduced to molecular genetics or biochemistry. Three criticisms pertaining specifically to the Nagel-Schaffner model bear mentioning. (Section 4 reviews the main objections to different models of reduction more generally.)

(a) Michael Ruse (1971) and David Hull (1974) have argued that the reduction of classical genetics in Schaffner's sense does not obtain and that the relation between classical and molecular genetics is rather one of theory replacement. Although Schaffner's account permits a correction of the theory to be reduced (TB* instead of TB), it is not clear when a modified version TB* (corrected “classical genetics”) still counts as being largely similar to TB (classical genetics). This leaves the ambiguity of whether one should conclude that TB was reduced because TB* is similar enough or that TB has been replaced by another theory because the two are so dissimilar (Ruse 1976, Hull 1976). Schaffner himself only requires that TB and TB* are “strongly analogous” (1993, 429), without offering a specification of how much similarity is required for reduction rather than replacement (Winther forthcoming). William Wimsatt (1976a) emphasizes that while interlevel, compositional reductions are transitive, successional reductions through history are intransitive, so that a continuum between genuine successional reduction and replacement obtains. Any sequence of successive reductions can be expected to evolve into an instance of replacement given enough time.

(b) Nagel and Schaffner rely on a syntactic construal of theories, which assumes that a theory is set of statements in a formal language, including laws (Kitcher 1984). Yet it has been argued that molecular biology as the reducing theory does not consist of a small body of laws, so that a covering-law model of explanation and reduction is not applicable (Culp and Kitcher 1989, Kitcher 1984, Sarkar 1998). If molecular biology or other biological fields are inadequately represented as consisting of a clearly delineated theory reconstructed as a set of statements in a formal language, then an account of reduction that presupposes this construal of scientific theories is problematic. The Nagel-Schaffner account may also fail to view scientific theories as dynamic entities that change over time (Hull 1976, Wimsatt 1979).

(c) Hull (1976), Wimsatt (1976a, 1976b), and Sahotra Sarkar (1992, 1998) have objected that Schaffner focuses too much on formal considerations about reduction, rather than substantive issues. His account construes reduction as a logical relation that is “in principle” possible even though it is not occurring in practice. If reduction in this sense is not occurring in biology, then significant questions arise as to whether it best fits biological knowledge and practice. Nagel claimed that premature attempts at reduction can retard scientific progress, thus implicitly decoupling methodological demands from epistemological ones (Nagel 1961, 362–3). Schaffner (1974, 1993) acknowledges that reductionism in his sense has been peripheral to the practice of molecular biology, but maintains that his formal model of reduction captures the reduction relation between theories. Wimsatt (1979) suggests that Nagel's and Schaffner's focus on in-principle considerations and the context of justification, to the exclusion of in-practice considerations and the context of discovery, leads to misunderstandings about the nature of reductionism debates, including historical and current controversy among scientists (see Section 2). Reduction may obtain between classical and molecular genetics, but not in the sense of Nagel's or Schaffner's logical empiricist approach (Hull 1974).

3.2 Explanatory Reduction

Explanatory reasoning patterns of life scientists, including reduction, are now center stage in many parts of philosophy of biology, such that theory reduction is no longer the primary focus. Various models of explanatory reduction focus on reductive explanation and differ from theory reduction (which is also a kind of explanation, at least on the deductive-nomological notion of explanation) in two important ways:

(a) While theory reduction assumes that reduction is a relation between theories, models of explanatory reduction permit other features as the relata of a reductive explanation, such as fragments of a theory, generalizations of varying scope, mechanisms, and even individual facts. One approach pursued along these lines is due to Kenneth Waters (1990, 1994, 2000). Instead of construing theory relations in terms of the Nagel-Schaffner model, he identifies a key principle of inference operating in classical genetics and molecular genetics, the “difference making principle” (gene differences cause differences in phenotypes), which can be ascribed to genes in both classical and molecular genetics. An explanatory reduction is achieved between these because their causal roles correspond. Another example of explanatory reduction is due to Marcel Weber (2005). In line with models of theory reduction, Weber maintains that the explanans essentially involves biochemical laws. But he assumes that the explanandum is a fact or basic generality about biological mechanisms (rather than a genuine law or a whole theory), such as the mechanism of action potentials in neurotransmission. By permitting the explanans to make reference to higher level structures such as axons, he does not require that it contain exclusively biochemical expressions.

(b) A characteristic difference between models of theory reduction and explanatory reduction stems from the fact that the former—focusing on the epistemic notion of deduction—do not explicitly incorporate the ontological idea (which is historically ubiquitous) that a reduction explains the whole in terms of its parts. Models of explanatory reduction, in contrast, typically assume that reductive explanation is causal explanation, where a higher level feature is explained by the interaction of its constituent parts. Stuart Kauffman (1971) gave an early expression of the idea that the search for explanations in experimental biology proceed by decomposing an overall system into interacting parts. Wimsatt (1976a) also recognized that this more naturally mapped onto the language usage of biologists, especially the practice of discovering and elucidating mechanisms. Mechanism approaches have now developed into a robust alternative to theory reduction (see Section 5), but given that many philosophers of biology and neuroscience conceive of explanation as the causal description of mechanisms, accounts of explanatory reduction can rely on the notion of a mechanistic explanation (e.g., Delehanty 2005).

Models of explanatory reduction avoid the problems facing theory reduction mentioned at the end of Section 3.1. Since explanatory reduction is not committed to the relata of an explanation being theories involving laws, it can defend the possibility of reduction without formally reconstructing a clearly delineated theory of molecular biology or classical genetics, showing that this theory contains laws (according to a particular notion thereof). Its focus on reduction as explanation of a whole in terms of its parts captures a feature of actual molecular explanations. Whereas accounts of theory reduction have to acknowledge that reduction in their sense has not yet been achieved and must appeal to the “in principle” possibility of reduction, accounts of explanatory reduction comport well with the piecemeal nature of actual scientific research. Experimental and molecular biology can offer reductive explanations even if most facts have not yet been explained and existing explanations still require more details to be filled in. Current biology effectively explains individual facts and generalities of smaller scope, and the explanans can be a mere fragment of a theory (cf. Schaffner 2006 on “patchy and fragmentary” reductions). Finally, given that mechanistic explanations in experimental biology are preceded by and closely tied to the discovery of mechanisms, explanatory reduction provides an account of reduction that can be related to scientific discovery, while approaches of theory reduction have to separate the context of justification (reduction and explanation) from the context of discovery.

An early model of explanatory reduction was put forward by William Wimsatt (1976a). Wimsatt assumes that a reduction is a causal explanation of an individual event (rather than a generality), involving a “compositional redescription” of a higher level state of affairs into some of its component features. This model is related to Wesley Salmon's (1971) account of statistical explanation, which uses the notion of screening off to explicate the idea of one property being statistically more relevant (and thus more explanatory) than another one. Wimsatt replaces Salmon's statistical relevance with the idea of causal relevance (as explanatory reduction in biology involves the search for causally relevant factors), and he develops the notion of “effective screening off”. This involves the idea that the compositional redescription and explanation in terms of a lower level property has a greater “cost” of explanation (e.g., is more complicated) than the description in terms of the higher level property that is effectively screened off. A consequence of the epistemic “cost” requirement is that an explanatory reduction that effectively screens off a described higher level feature does not make it irrelevant and thus eliminable (as in Salmon's account) because it can be retained for pragmatic purposes.

Sahotra Sarkar's (1998) detailed discussion of reductionism and genetics offers several relevant conceptual clarifications. Focusing on substantive issues about reduction in genetics (especially representation, see Section 4.3), rather than formal considerations in a theory of reduction, Sarkar develops an account of explanatory reduction that is not committed to a particular account of explanation. Two such basic types of reduction are “abstract hierarchical reduction” and “strong reduction” (or spatial hierarchical reduction), where the latter unlike the former requires that entities of a lower hierarchical level are physical parts of higher level entities. (Each type of reduction can come in stronger and weaker forms as different kinds of approximations are involved in different reductive explanations.) Sarkar argues that explanations in classical genetics, which explain phenotypic patterns in terms of classical genes, are abstract hierarchical reductions. The entities involved in this case are alleles, loci, linkage groups, genotypes, and phenotypes, which stand in hierarchical relations (e.g., alleles A and a form genotype Aa), with individual alleles occupying the lowest level. It is an abstract rather than a spatial hierarchy; the assumption that A and a map to Aa at the next higher level does not require Aa to be a physical entity that has A and a as its spatial parts. Furthermore, the genotypic entities in this hierarchy are governed by various transformations and regularities (principles of inheritance) such as segregation, linkage, and mutation. Phenotypic patterns of inheritance are explained by genotypic features in the well-known fashion of classical genetics. While this abstract hierarchy can be interpreted as a spatial hierarchy by viewing classical genes as being parts of chromosomes, Sarkar's point is that the various practices in classical genetics such as segregation and linkage analysis did not require a physical interpretation of genetic maps, and that the explanations in terms of an abstract hierarchy can be genuine reductions since they invoke a lower and more fundamental hierarchical level.

Regarding spatial hierarchical reduction, Sarkar argues that this regularly occurs in molecular biology, where certain parts of experimental biology are reduced to molecular biology, more precisely macromolecular physics. The various components of cells form nested part-whole relations, resulting in a spatial hierarchy. Many mechanisms of classical genetics can be reductively explained by the physical principles governing the behavior of macromolecules (though approximations are still necessary). Explanatorily relevant principles for the physics of macromolecules include weak interactions, hydrogen bonding, hydrophobic bonding, allosteric transformations, lock-key fit, or the idea that structure determines function. Sarkar's central tenet is that the elucidation of the molecular mechanisms of replication, recombination, and cell division provide reductive explanations of the principles of classical genetics. While classical genetics offered an account of gene transmission and could also rely on cytological explanations, these accounts did not yield a complete explanation of how and why recombination occurs. Molecular biology filled in this gap and “corrected” classical genetics.

In some ways the most unique defense of explanatory reduction is found in the recent work of Alex Rosenberg (2006), which departs from his earlier critical focus on theory reduction (1978, 1985, 1994). Rosenberg's argument has multiple components. First, nothing less than strict laws (universal, exceptionless, spatio-temporally unrestricted) are required for explanation and the only candidate law in biology is the principle of natural selection (PNS). Second, why-necessary explanations are better than how-possible explanations in historical sciences such as biology, but why-necessary explanations are only available at the molecular level because structure becomes decoupled from function above this level. Therefore, all how-possible explanations in “functional biology” (i.e., non-molecular biology), even those invoking the PNS, and any descriptions from functional biology involving higher levels of organization get explained (why-necessarily) by the PNS operating on the molecular level (often occurring at some relatively distant point in evolutionary history).[5]

4. Problems with Reductionism

You can't learn everything from the laboratory, that's what he used to say. The whole is more than the sum of its parts, he told us. The whole behaves differently from the parts, and has different properties. That's what he taught us, and he was right. It's out of fashion to say these days, when we spend our time scrutinizing the interactions of eukaryotic microbes, but it's true, nevertheless. It's still true. (M. Drabble, The Sea Lady, 140–1)

Accounts of theory reduction and explanatory reduction face several potential problems. Some critiques of (epistemic) reductionism emphasize that the effect of molecular processes strongly depends on the context in which they occur, so that one molecular kind can correspond to many higher level kinds (Section 4.1). A prominent and significant challenge to reductionism stems from the fact that higher level biological structures and processes can be realized by different kinds of molecular processes, so that many molecular kinds can correspond to one higher level kind (Section 4.2). A point that has not been stressed in previous critiques of reduction is the fact that biological processes must be represented before they can be explained (Sarkar 1998). Attending explicitly to representation highlights two features that have been relatively neglected in debates about explanatory reduction: temporality and intrinsicality (Section 4.3).

4.1 Context-Dependence of Molecular Features (one-many)

One key challenge for reduction stems from the fact that the effect of a molecular entity or mechanism may strongly depend on the context in which it occurs (Hull 1972, 1974, 1976; see also Wimsatt 1979, Gilbert and Sarkar 2000, Laubichler and Wagner 2001). Classical geneticists were fully aware of the fact that a phenotype is brought about by the interaction of several classical genes—the same allele may lead to two different phenotypes if occurring in two individuals with a different overall genotype (Waters 2004). Although one-many phenomena have been known for some time, it is now a robust empirical generalization that a molecular pathway may have different effects in different cellular contexts, so that the same pathway can be involved in different functions in different species or in different parts of an individual (Gilbert and Sarkar 2000, Laubichler and Wagner 2001, Burian 2004). Even the amino acid sequence produced by a molecular gene may depend on DNA elements outside of this gene and non-genetic factors, so that a gene can code for distinct products in different cells or different states of a cell (Stotz 2006). Thus there is a one-many relation between molecular kinds and higher level kinds: a molecular mechanism can causally lead to or be part of different higher level states depending on the context.[6] A common conclusion drawn from this by biologists objecting to reductionism is that the parts and molecular components of an organism have to be understood in terms of the organized whole in which they occur (recall Kant's regulative principle; Section 2). While paying attention to context and organismal organization is a valid methodological recommendation, it is less clear why this is an argument against epistemic reduction. Proponents of theory reduction such as Nagel (1961) and Schaffner (1976) have replied that a molecular reduction surely can take the relations of parts and the context of molecular processes into account, e.g., by specifying the relevant context as initial conditions in the molecular premises from which the higher level state is to be deduced (cf. Frost-Arnold 2004).

Context-dependence is primarily a problem for models of theory reduction, whereas accounts of explanatory reduction need not be similarly affected. The reason is that models of explanatory reduction can take the organismal context for granted without being committed to reducing it molecularly. Science can avail itself of causes as difference makers relative to a given causal context. Both experimental investigation and explanation can focus on one among many causes, relegating everything else to the context or background (which is often held fixed in experimental studies). For example, if the loss in function of a particular gene leads to an abnormal phenotype in a knock-out experiment, this gene is one causal factor implicated in the development of the normal phenotype. Explanations appealing to this gene—as a salient causal factor relative to a context—are genuine explanations even if the other genes involved in the phenotype are unknown and the cellular context of the gene has not yet been understood (cf. Waters 1994). Since accounts of explanatory reduction usually focus on explanations as they occur in experimental biology, they can take the context for granted without providing a specification of it, provided that its relevance is acknowledged. Weber (2005) argues that reductive explanations essentially involve molecular features, in particular physico-chemical laws, but he allows that the explanans typically makes reference to higher level structural features (such as axons or cells). These act as boundary conditions for the reductive explanation though the explanatory force is carried by the physico-chemical laws. If molecular biologists discover that the same mechanism produces different effects in distinct contexts and only one of these effects is the target of inquiry, a part of the mechanism's context becomes epistemically relevant but can be included in the explanation (Delehanty 2005).

In contrast, an account of theory reduction such as the Nagel-Schaffner approach is committed to deduce a description of a higher level phenomenon from purely molecular premises. In order to deduce even a single higher level statement (a fortiori a total higher level theory on Schaffner's account), the premises must contain (in addition to the lower level laws) a specification of the total lower level context—every feature that has some causal impact. This includes features internal to an organism and some part of its environment. While explanatory reduction can appeal to particular causes relative to a background that is not molecularly specified, a logical deduction as required by theory reduction can go through only if the total causal context is contained in the premises.[7] A molecular characterization of the total context is in principle possible due to ontological reduction coupled with supervenience: for any higher level state there is always a total molecular configuration that determines this higher level phenomenon. However, this entails a retreat to the “in principle” possibility of a molecular derivation because it is not just the case that the molecular specification of the total context cannot currently be given but that it will not even be forthcoming in the future (Hull 1972, 1974, 1976).[8] In this respect accounts of explanatory reduction have a clear advantage over models of theory reduction, as the former are committed neither to formal derivation nor the assumption that the explanans of the reductive explanation involves only molecular notions.

Jason Robert's (2004) treatment of explanation and methodology in developmental biology highlights what is probably the main valid point stemming from the context-dependence of molecular features. While acknowledging the legitimacy of reductionistic research strategies that isolate molecular causes relative to a background that is held fixed for experimental purposes, Robert emphasizes that this methodology does not license an inference to an exclusively reductionistic position maintaining that all explanatorily relevant causes are molecular. That developmental genetics uncovers various developmental differences between individuals that are due to differences in gene expression does not entail that genes are the only factors relevant in ontogeny. As genetic research holds other cellular and organismal factors fixed, the causal influence of these non-genetic resources is simply not assessed by this research method. It begs the question against those who maintain that some explanations of development must invoke higher level features to simply assert that the organismal context can always be molecularly described. For example, an appeal to developmental genetics research success alone is inadequate for reducing cellular context to molecular causes because the discovery of genetic causes occurs against a fixed organismal background (Robert 2004). The dependence of molecular features on a cellular and organismal context poses problems for theory reduction approaches maintaining that higher level phenomena can be deduced from purely molecular premises. (It is also a problem for claims that the embryo can be computed from a purely molecular specification of the fertilized egg, as endorsed by Rosenberg 1997; see Keller 2002, ch. 9.) Robert's argument is significant in putting necessary boundaries on the kinds of explanatory inferences that can be drawn from reductionist research strategies. What remains missing from many philosophical discussions critical of reduction is an explicit account of methodological strategies and epistemological frameworks that both capture actual scientific reasoning and offer an alternative to reductionism (Love 2006, forthcoming; see Section 5).

4.2 Multiple Realization of Higher Level Features (Many-One)

Apart from the one-many relation between molecular and higher level kinds, Hull (1972, 1974, 1976) also points to the existence of a many-one relation as a challenge for reductionism. This turns out to be a problem for both theory reduction and explanatory reduction (though to differing degrees). Hull's point is that the same higher level phenomenon (e.g., a Mendelian trait such as coat color) can be produced by several different molecular mechanisms, or that a higher level phenomenon is realized by different kinds of molecular configurations (multiple realization). In defense of his model of theory reduction, Schaffner (1976) replies that it is sufficient to specify one such molecular situation. From this one can logically derive the higher level situation that is determined by the molecular constellation. However, one has to be clear about the nature of the explanandum in a reductive explanation (i.e., what is the consequent of a reductive deduction). Is it a type of higher level phenomenon (described by higher level predicates) or merely a token phenomenon (an instance of a higher level phenomenon)? Theory reduction is the idea that a whole theory (such as classical genetics) can be reduced, or at least that its laws can be reduced. The Mendelian principle of segregation, for instance, is a law describing a phenomenon occurring in all sexually reproducing organisms, not just a token phenomenon, so this general principle—covering various lower level realizations—would have to be explained/deduced molecularly.

Consider the example that Schaffner (1993) offers in support of the possibility of reduction: the Mendelian concept of “dominance”. On his account, given certain classical genes a and b, the property “allele a is dominant (relative to allele b)” can be expressed in molecular terms as follows. Assume that the genotype aa yields phenotype A and genotype bb corresponds to phenotype B, while ab yields A (because of dominance). We ascertain that gene a consists of DNA sequence α and gene b consists of DNA sequence β, and that phenotype A is brought about by protein א while phenotype B is brought about by protein ב. This yields the following supposed reduction:

Allele a is dominant (relative to b) iff the presence of two copies of DNA sequence α molecularly produces protein א [thus phenotype A], the presence of two copies of sequence β produces protein ב [phenotype B], and the presence of one copy of α and one copy of β produces א [phenotype A]. (Schaffner 1993, 442)

While Schaffner explicitly claims to have shown that the predicate (the concept of) “dominance” can be reduced to the language of biochemistry, his evidence actually shows how a single instance of dominance (dominance of specific allele a over allele b) is brought about by states on the molecular level. But epistemic reduction requires reducing higher level concepts and, since Mendelian dominance can be brought about by different kinds of molecular situations, this higher level predicate corresponds to many molecular predicates, making any reductive translation of higher level concepts very complex. (And, in addition to translating such concepts into molecular terms, theory reduction also requires deducing higher level generalizations involving such concepts.) A similar issue can apply to explanatory reduction. Biology attempts to explain kinds of phenomena occurring in different contexts and organisms, so that reductive explanations also must molecularly explain kinds of phenomena in addition to single instances.

Subsequent discussion based on ideas in metaphysics and philosophy of mind has clarified how this many-one relation between molecular and higher level kinds bears on epistemic reduction. In his general discussion of reduction in psychology and social science, Jerry Fodor (1974, 1975, 1997) distinguishes between token-token reduction and type-type reduction (or between token physicalism and type physicalism). Token-token reduction is the idea that each token higher level process is metaphysically identical to a token physico-chemical process (the weak sense of ontological reduction that has nearly universal assent: see Section 1). The dispute centers on how much more than token-token reduction is possible. A substantially stronger idea is that each type of higher level phenomenon corresponds to one type of lower level phenomenon. Type-type reduction yields a very bold version of epistemic reductionism, as in this case higher level concepts and knowledge about higher level phenomena could be immediately reduced to molecular notions. Yet Fodor argues that type-type reduction (regarding mental and social kinds) is empirically false. Consider a law or causal principle of a higher-level science relating two kinds S and T, such as ∀x(SxTx), which can be read as “all Ss are Ts” or as “all events of type S bring about events of type T” (some laws or principles might be of a more complex logical form, but we use the simplest case). As a matter of fact, a higher level kind is often realized by several different kinds on the physical level (different instances of the higher level kind are of a different type on the molecular level). If kind S is realized by the lower level kinds M1, M2, …, Mi, the bridge principle (reduction function) is of a disjunctive form:

x (Sx ↔ (M1xM2x ∨ … ∨ Mix))

T is likewise multiply realized, say by the molecular kinds N1, N2, …, Nj. Fodor's point is that the lower level basis of the higher level law ∀x(SxTx) consists in several laws (reducing different higher level instances), such as:

x(M1xN3x), ∀x(M2xN1x), ∀x(( M3xM4x) → N2x), …

Thus, while on the higher level there is one unified law ∀x (SxTx), its translation into molecular terms is a disjunctive statement:

x ((M1xM2x ∨ … ∨ Mix) → (N1xN2x ∨ … ∨ Njx))

This disjunctive statement by itself is not a law (it has the appearance of being a gerrymandered combination of laws), so that a higher level law cannot be reduced to a lower level law. Token-token reduction obtains, but since the natural kinds (to which laws apply) of the higher level science cross-cut the kinds of the lower level science, type-type reduction fails and the law-based explanation on the higher level cannot be replaced by a unified law-based explanation on the lower level. Steven Kimbrough (1978) argues that this situation also applies to the relation between classical and molecular genetics: ontological (token-token) reduction is possible, but not epistemic (type-type) reduction.[9]

The multiple realizability of higher level kinds is viewed by many philosophers of mind and social science as an obstacle to reduction (Kincaid 1997), and critics of reductionism in the philosophy of biology use the many-one relation between molecular and higher level phenomena in a similar way. Rosenberg's earlier discussions criticizing theory reduction in light of multiple realization (1978, 1985, 1994) used Donald Davidson's (1970) notion of supervenience, which is another way to spell out the idea of ontological without epistemic reduction (non-reductive physicalism). Higher level facts supervene on physical facts if any difference between two higher level situations implies a difference in the physical basis. Yet this is consistent with two distinct physical situations corresponding to the same higher level state (many-one). Since biological properties such as fitness supervene on physical properties, fitness can be nothing more than a combination of physical properties, even though no such property complex is necessary for having a particular fitness value. Higher level theories abstract away from irrelevant variation in lower level features and thereby arrive at generalizations. From this perspective, theory reduction of classical to molecular genetics is impossible as the many–many relation between classical and molecular kinds makes any reductive account radically disjunctive. At the same time, the notion of supervenience implies that classical and molecular genetics are not incommensurable, and that molecular genetics can shed light on the exceptions of classical genetics.

Another prominent critique of reductionism in genetics is due to Philip Kitcher (1984). He argues that the point is not that the complexities of any reduction (due to the many-one relations between molecular and higher level kinds) exceed our cognitive capacities. Rather, reductionism fails because a molecular account does not adequately explain since it does not involve the natural kinds that underlie the appropriate causal relations. A principle of classical genetics (e.g., segregation) is explained by cellular processes (e.g., the behavior of chromosomes in meiosis). Cytology picks out higher level kinds as the causally relevant kinds, and rephrasing these cellular processes in molecular terms would lead to an account that mentions various combinations of molecular kinds, while failing to exhibit the unified kinds that are the causes of the process being studied. A key aspect of Kitcher's argument is that an epistemic account of reduction fails for ontological reasons, by not capturing the natural kind structure of chromosomal phenomena studied by classical genetics.

Schaffner's reply to the multiple realization objection (1993, 463–6) holds that the same DNA sequence can be present in many individuals, so it is not just a token phenomenon that is studied by molecular biology, but molecular generalizations apply to restricted types. This is true but fails to address the core issue. Critics of reductionism do not argue that each higher level type corresponds to several lower level tokens (which is trivially true) or that molecular biology only studies token phenomena. Rather, the point is that a higher level type corresponds to several lower level types, and for each of these lower level kinds a different molecular account obtains. Thus, a unified (as opposed to disjunctive) causal explanation of the higher level phenomenon appears impossible.

Waters (1990) challenges the commitment to unification that implies there is explanatory loss in appealing to the multiply realized “gory details” of molecular biology. In some cases, disjunctive explanations at a lower level seem to explain better than unified explanations at higher levels. Elliott Sober (1999) extends this line of argument by pointing out that while an explanation in terms of higher level kinds can be more unified or general by encompassing heterogeneous lower level kinds, this same heterogeneity exhibits important differences in scientifically interesting properties. Only the lower level science can account for such differences and therefore, in this sense, offers a deeper explanation than the higher level science. As a result, explanations are not better or worse along a one-dimensional scale, but there are several epistemic virtues an explanation can possess—generality as well as depth. Sometimes a lower level explanation is better relative to one epistemic property, while a higher level explanation is preferable relative to another epistemic quality, such as unification.[10]

It is natural for contemporary accounts of explanatory reduction to focus on mechanisms and the study of model organisms. One reply to the multiple realization objection from this perspective explicitly argues for token-token reduction, discussing how the causal description of a token mechanism explains (Delehanty 2005). Weber (2005) acknowledges that higher level features are multiply realizable, yet this does not hamper the reducibility of token systems or individual organisms: “the fact that chemotactic behavior is multiply realizable does not affect the reductionistic explanation of this organism's behavioral biology” (2005, 48). However, this stratagem ignores the key point of contention. Token-token reduction is uncontroversial among both reductionists and anti-reductionists, and it is a relatively trivial claim that follows from physicalism. The focus on token phenomena or individual organisms fails to address an important epistemological feature of scientific knowledge: some scientific descriptions and explanations pertain to a single organism only, but (in general) descriptions and explanations cover types of phenomena. It is an intellectual aim of biology to offer general explanations covering many organisms and the use of model organisms is predicated on this aim. Any version of epistemic reductionism has to address this feature of scientific knowledge and explanation to be an adequate account of reduction.

Consider recent discussions of explanatory reduction in developmental biology. Explaining development—as understood by many developmental biologists—means to have an account of the development (ideally, all of ontogeny) exhibited by the individuals belonging to a larger group of organisms (e.g., understanding the developmental principles that govern vertebrate development). As the same morphological structure may develop in different species by means of different developmental processes and based on the action of different genes, multiple realization obtains and is relevant for such explanations (Laubichler and Wagner 2001). Even during the development of a single organism, a structure may be present across time and have a stable developmental-functional role, yet its underlying molecular and cellular basis may change (Brigandt 2006). Certain entities above the molecular level can be more robust so that when some developmental processes change or break down these higher level entities are not modified or removed. Knocking out a gene need not have any impact on development due to genetic redundancy, while eliminating a particular cell or group of cells may dramatically interfere with normal development. A higher level causal connection can be more salient in that the regular operation of the mechanism strongly depends on the presence of this kind of causal relation. This is not to deny that such a higher level cause is realized by molecular entities and their interactions (i.e., supervenience obtains). But if a higher level cause is less sensitive to disturbance or intervention than certain molecular causes, the more powerful mechanistic explanation is the one that appeals to the more robust causal connections. This is a reason why some biological sciences preferentially invoke higher level entities and processes, explaining in terms of higher level natural kinds and causes (if the latter are multiply realized or causally vary on the molecular level).

4.3 Representation, Temporality, and Intrinsicality

Sarkar's (1998) account of explanatory reduction draws attention to the requirement of representation: natural phenomena must be symbolized, embodied, pictured, or designated through media such as equations, scale miniatures, or abstract diagrams. Every reductive explanation involves a representation of the systems or domains to be related by reduction. (Note that this is not isomorphic to the question of how theory structure is represented; see, e.g., Schaffner 1996.) Almost all of the discussion surrounding one-many (Section 4.1) and many-one problems (Section 4.2) for theory and explanatory reduction assume or presume particular representations that directly impact the arguments offered (e.g., they involve approximations of different kinds). Sarkar distinguishes three representational criteria for a reductive explanation: (i) fundamentalism: the explanation of the phenomena relies entirely on features that pertain to a more fundamental realm; (ii) abstract hierarchy: the system is represented as a hierarchy, where the lower levels are (usually) regarded as the more fundamental; (iii) spatial hierarchy: the abstract hierarchy is rendered physical by the requirement that the entities on the lower level are (spatial) parts of the entities at the higher levels (see Section 3.2). Whether or not there are one-many or many-one relations between different hierarchical levels turns on the nature of the hierarchy invoked and how it is characterized. The decomposition of a system is not a univocal enterprise and, depending on the principles utilized, can generate both competing and complementary sets of part representations from the same system (Kauffman 1971, Wimsatt 1974). Certain basic governing principles are often visible, such as functional versus structural decompositions (Bechtel and Richardson 1993, Winther 2006). Therefore, prior to determinations of whether reductive explanations succeed or fail, questions of representational choice and adequacy need to be explicitly addressed.

These representation issues are not idle with respect to differing explanations, both past and present, because the same line of argument about the non-independence of organismal parts constituted reasons for treating the organism from a non-reductionist standpoint (cf. Aristotle, Parts of Animals II.3; Owen, Hunterian Lectures [1837] in Sloan 1992, 213–4). Individuation of parts and their explanatory relationship to wholes is critically dependent on the principles utilized in representing them. Wimsatt (1974, 1986, 1997) has argued that reduction and emergence are compatible within the context of explaining organizational complexity in living systems, going so far as to argue that reductionist methodology can identify emergence when specific conditions of aggregativity are delineated. Different failures of aggregativity result when different conditions are not met, leading to a subtle taxonomy of compositional relations, as well as highlighting that different decompositions yield distinct fulfillments and/or violations of those conditions that can be methodologically exploited by scientists for the purpose of causal discovery.

In addition to individuation, two other issues directly related to representation can be mentioned: temporality and intrinsicality. In most discussions of epistemic reduction, both theoretical and explanatory, no explicit distinction has been drawn between constitutional relations or arrangement (spatial) and causal ones (temporal). One source of this focus on spatial relations is the objection to reductionism from structural organization (Polanyi 1968; cf. Delehanty 2005, Frost-Arnold 2004, Schaffner 1993, ch. 9.2), which is related to the “context” objection (Section 4.1). It also derives from the fact that the Nagelian model of theoretical reduction treats the theoretical structures to be related at a designated time and because the physical science models used as a template for thinking about reduction emphasized spatial composition questions rather than causal relations. As Schaffner has remarked: “Biological theories are usually given in the form of a series of temporal (and frequently causal) models. In physics, time is usually eliminated by making it implicit in differential equations, whereas in biology a temporal process … is the rule” (1993, 83–4; cf. Kellert 1993, 93). Nagel presciently recognized differences between spatial and temporal reasoning in his discussion of teleological explanation (in part because he was reading the literature on explaining ontogeny by theoretical biologists; see above, Section 2).

The contrast between structure and function is evidently a contrast between the spatial organization of anatomically distinguishable parts of an organ and the temporal (or spatiotemporal) organization of changes in those parts. What is investigated under each term of the contrasting pair is a mode of organization or a type of order. In the one case the organization is primarily if not exclusively a spatial one, and the object of the investigation is to ascertain the spatial distribution of organic parts and the modes of their linkage. In the other case the organization has a temporal dimension, and the aim of the inquiry is to discover sequential and simultaneous orders of change in the spatially ordered and linked parts of organic bodies. (Nagel 1961, 426)

Kitcher also glimpsed the potential significance of temporality for reductive explanation in the context of embryology:

Because developmental processes are complex and because changes in the timing of embryological events may produce a cascade of effects at several different levels, one sometimes uses descriptions at higher levels to explain what goes on [later] at a more fundamental level. (Kitcher 1984, 371)

A very different distinction (agreed upon by diverse participants) utilizing time has arisen to distinguish between the historical succession of theories via reduction (“diachronic reduction”) and current attempts to relate parts to wholes, such as in explanatory reduction or interlevel theory reduction (“synchronic reduction”; see Section 3.1).

Synchronic reduction is mereological explanation, in which the behavior of more composite items described in reduced theories is explained by derivation from the behavior of their components by the reducing theory. Thus, reduction is a form of explanation. Diachronic reduction usually involves the succession of more general theories which reduce less general ones, by showing them to be special cases which neglect some variables, fail to measure coefficients, or set parameters at restricted values. As the history of science proceeds from the less general theory to the more general, the mechanism of progress is the reduction of theories. (Rosenberg 2006, 28)
The first distinction to note is that between synchronic and diachronic conceptions of reductionism. The term reduction is often used to refer to the relation between a theory and its historical successor. … This is … what I mean by diachronic reductionism. My concern, on the other hand, is solely with synchronic reductionism, that is to say, with the relations between coexisting theories addressed to different levels of organization. (Dupré 1993, 94–5)

What this distinction misses is the possibility of diachronic aspects of part-whole or inter-level relations. This is one of the core reasons why development is a persisting biological topic for reductionism. During ontogeny there are causal interactions over time among parts and activities (both homogeneous and heterogeneous) to bring about new parts and activities (both homogeneous and heterogeneous), eventually generating a unified adult organism.

Temporality can be added as a fourth criterion on representations for reductive explanations: (iv) temporal hierarchy: the abstract hierarchy is temporally indexed by the requirement that the entities and their behaviors on the lower level are temporally prior to the entities and their behaviors at the higher levels. Although temporal or process hierarchies have not received the same attention as spatial hierarchies (such as in Wimsatt's aggregativity criteria), their importance is seen in explanations that invoke higher level features (e.g., tissue interactions during ontogeny) at an earlier time to causally explain lower level features at a later time (e.g., gene expression patterns), which violates the temporality criterion. For example, a recent study of the developmental origin of aortic arch asymmetry involves a mixture of reductionist and non-reductionist elements according to the criterion of temporality (Yashiro et al. 2007). The explanation involves several events in a sequential process. First, fundamental level properties (gene expression) change non-fundamental level properties (arterial structure), which alters blood flow dynamics (thereby fulfilling the temporal hierarchy condition). These new dynamics induce a change in gene expression (thereby violating the temporal hierarchy condition). This altered gene expression then produces the non-fundamental property of interest, aortic arch asymmetry (thereby fulfilling the temporal condition). Thus, depending on how time is represented and utilized, an explanation may be reductive, non-reductive, or both. This is a good example of how ontological and epistemic reduction can yield strikingly different answers. A biologist using a higher level feature at an earlier time to explain a lower level feature at a later time (a non-reductionist explanation by the condition of temporality) is not denying that this higher level feature is composed of lower level features (e.g., cells and molecules), nor the possibility that the higher level feature was caused by a different set of lower level features at some more distal point in time.

Another example is found in protein folding within molecular biology. Functional proteins are folded structures composed of amino acid components linked together into a linear chain. If we ask whether the folded protein is mereologically composed of its amino acid parts given current representations in molecular biology, then the affirmative answer seems to favor explanatory reduction with respect to spatial hierarchy. But if we ask whether the linear amino acid chain folds into a functional protein (a causal process with a temporal dimension) purely as consequence of its linked amino acid parts, then the answer is much less clear. Recent empirical studies have emphasized the necessary role of other folded proteins to assist in the proper folding of newly generated linear amino acid chains (Frydman 2001). That the linked amino acid components alone are insufficient causally, even if they are sufficient constitutionally, allows for a more explicit appreciation of the significance of temporality, especially because the relations of interest concern molecular biological phenomena (as opposed to appeals to higher levels of organization from cytology or anatomy).

A final representational feature is the distinguishing of “internal” from “external”. Wimsatt (1980) drew attention to the fact that every investigation must divide a system from its environment and that methodological reductionism favors attributions of causal responsibility to “internal” parts of a system rather than those deemed external (see also Wilson 2004). In the protein folding case it appears that the failure of temporally indexed explanatory reduction involves causal powers of something extrinsic to (an instance of) the process under scrutiny (i.e., the folding of a single amino acid chain). The presumption of atemporal, compositional relationships has encouraged this distinction to be ignored because nested part-whole relations are predicated on a prior individuation of a system from its environment. This line of thought relates to the context-dependency objection but emphasizes the importance of functional dependency relations uncovered by attending to temporality in addition to structural organization.

Recognizing temporal aspects of reductionism alongside other representational issues like intrinsicality, which are not captured by a focus on mereology alone, supplements the evidential base for an argument in favor of talking about different kinds of reductionism rather than in terms of one unified account of reduction or overarching dichotomies of “reductionism” versus “anti-reductionism” (or “holism”). Once we incorporate distinctions regarding different types of reduction (ontological, epistemological, and methodological), the different interpretations of these types of reduction (e.g., theoretical versus explanatory epistemological reductions), the different representational features involved in these reductions (spatial, temporal, etc.), and the distinct biological topics to which philosophers have applied them (e.g., development or functional anatomy), it is more difficult to conceptualize a single, adequate conception of reduction that will do justice to the diversity of phenomena and reasoning practices in the life sciences. The multiplicity and heterogeneity of biological subdisciplines only reinforces this argument and suggests to some that we should move beyond reductionism entirely.

5. “Alternatives” to Reductionism: Moving Beyond Dichotomies

The brutality and the reductionism of the microbiological victory had not destroyed all speculation. Ethology and ethologists had been savaged, but they had not perished. (M. Drabble, The Sea Lady, 150)

One of the ubiquitous features of modern biology that seems prima facie incompatible with many methodological and epistemological reductionist theses is the proliferation and flourishing of biological subdisciplines, molecular and otherwise. Despite the purported “hegemony of molecular biology” (Kitcher 1999), biological subdisciplines focused on higher levels of organization have not disappeared. As Fodor (1997) has felicitously put it: “Special sciences: still autonomous after all these years”. For some time this feature of biological research, along with many of the problems associated with theoretical reduction discussed above (especially its inapplicability to scientific practice, but also difficulties in identifying clear-cut and distinct “levels” to reductively relate), have prompted philosophers to reject “reduction” as the appropriate relation among concepts and explanations from different domains.[11] Instead, relations of coordination driven by a commitment to unifying, integrating, or synthesizing aspects of scientific knowledge have been offered. Lindley Darden and Nancy Maull were at the forefront of this discussion with the concept of an inter-field theory that decoupled the aims of unification from those of reduction (Darden and Maull 1977, Maull 1977; cf. Burian 1993, Grantham 2004). The emphasis is similar to that of explanatory reduction in moving the focus away from theories (and theoretical terms) as the only epistemic units of interest but is different in emphasizing new relata such as coordination, integration, synthesis, or reciprocal interaction as opposed to reduction. One particular phenomenon of interest here is the demand for multidisciplinary research because the explanatory task involves a relation of mutual epistemic coordination that is in some respects not reductive; i.e., there is an implicit rejection of the “fundamentality” of one particular discipline producing the most empirically adequate explanations (Love forthcoming). Disciplinary relations are not determined by some abstract account of levels of organization (e.g., the so-called “layer cake” model; Oppenheim and Putnam 1958), but instead are determined by the structure of the problem being investigated (cf. Bechtel 1986, 1993). This outlook also meshes with sociological analyses of disciplinary organization (Gerson 1998).

More recently, “mechanisms” approaches have been widely discussed (Bechtel 2006, Craver 2005, Darden 2006, Glennan 1996, Machamer et al. 2000). It is common among philosophers to view experimental biology (especially) to be fundamentally concerned with the discovery of mechanisms (Darden and Craver 2002, Tabery 2004), where mechanisms are understood as akin (though not equivalent) to machines with interconnected, organized parts operating to produce regular or expected outcomes. The explanatory project envisioned from this perspective involves adequately describing these mechanisms, including both their entities and activities, which are inherently inter-level. This gives the appearance of a mismatch with accounts of epistemic reduction that involve relating descriptions of entities at one level with those from another level. A mechanisms perspective is historically motivated as some of the language associated with reduction in the early modern period also includes the terminology of mechanism or its cognates (see Section 2). “Mechanism” has also been advocated by those discussing reductionism in the context of evolutionary biology (Brandon 1996).

Complexity has also been a source for the articulation of non-reductionistic epistemological accounts of biological reasoning (Wimsatt 1974). Sandra Mitchell (2003) has argued that different forms of biological complexity block any straightforward reductionist perspective. Multiple (idealized) causal models must be integrated from different disciplinary approaches to adequately represent this complexity. She specifically draws attention to self-organization, which refers to how system behaviors or patterns emerge from non-specific interactions among lower-level components (Camazine et al. 2001, Kauffman 1993), such as is observed in the collective behavior of social insects. (Self-organizing phenomena are also relevant to reductionism and emergence in physical science; see, e.g., Goldenfeld and Kadanoff 1999, Whitesides and Ismagilov 1999.) Although decomposing a system into parts is important for claims about self-organization, there is no localization of causal responsibility to these components (Bechtel and Richardson 1993, ch. 9). Instead the system behavior is a function of the way the components interact. For example, gene networks and human social networks may exhibit isomorphic patterns of self-organization due to their interactive structure, which is unrelated to whether genes or humans are the components. Thus, self-organizing phenomena occupy a strange place in the reductionism discussion because system components are used to explain the behavior of the system but it is the nature of their interactions (not their specific characteristics) that generate patterns of behavior, which are often referred to as “emergent” properties of the system.

Another distinct strand of discussion about reductionism is found in critiques of scientific methodology and explanation from the perspective of feminist and/or social epistemology (e.g., Longino 1990). A core idea in these discussions is that reductionism is involved in a positive feedback loop with other factors (e.g., socioeconomic status or perspectives on gender) that are mutually reinforcing (see, e.g., Keller and Longino 1996 or Levins and Lewontin 1985). Thus, reductionism (whether ontological, epistemic, or methodological) can be understood as a manifestation of more diverse commitments (e.g., political) than much of the philosophy of biology literature on reductionism would suggest. This line of criticism also indicates a tighter connection between reductionism in biology and the social sciences, which encourages explicitly treating their interrelations. Elisabeth Lloyd (2002) has argued that appropriate medical research should not be restricted simply to molecular biological investigation because higher levels of social organization that are culturally sanctioned have unrecognized causal effects of health. As the social sciences and medicine attempt the application of different life science reasoning strategies, similar philosophical issues predictably transfer into this context, but with the potential for more socially damaging consequences.

A shared feature of many of these alternatives to reductionism is an emphasis on pluralist dimensions of epistemology in biology that situate methodological and explanatory reduction as one among many necessary components (Dupré 1993, Kellert et al. 2006, Longino 2000, 2002, Mitchell 2003). In different ways these positions highlight how scientific investigation of biological phenomena demands diverse epistemic tools without necessarily having to tie it all together back into fundamental physics, macromolecular chemistry, or even molecular biology. Thus, ontological reduction becomes more or less unhinged from epistemology because there is no need to demonstrate how each and every domain of scientific inquiry is anchored back in physical stuff. Scientists are epistemologically profligate, even if they are largely metaphysically frugal (see Dupré 1993 and van Fraassen 2002 for differing perspectives on science and metaphysics). The plurality of biological research motivates a philosophical pluralism about biology (cf. Kellert et al. 2006), both in terms of the many meanings available for different methodological and epistemological aspects of reduction, and the need to have more than these aspects of reductionism in order to secure empirical adequacy with respect to biological phenomena.

It is clear that these different kinds of “alternatives” to reductionism successfully capture some features left untreated or ignored by theory reduction.[12] But an open question remains whether they supplant discussions of reduction or bear some complex relation to them (cf. Craver 2005). Reasons in favor of the latter position include the continued use of “reductionist” language (for and against) in biological research, as well as the fact that a near universal reason for rejecting reduction and choosing an alternative (such as mechanisms) is the problematic nature of theory reduction, which only represents one of many aspects of reduction. One interpretation of some alternatives to reduction is that they are actually responding to the complexity of reduction that has become more apparent over the past two decades (e.g., seeing explanations as inherently inter-level in nature). Thus, for example, mechanisms approaches are another way to treat the heterogeneity of explanatory reduction that was largely ignored in the literature on theory reduction. This heterogeneity is a pervasive feature of the history of biology; these alternatives are responding in part to long-standing controversies among life science investigators about the legitimacy of reductionist research strategies and modes of explanation (see Section 2). The compatibility or conflict of these various alternatives and different explanatory reduction accounts is not yet known (but see Bechtel 2006, 40–44).

6. Future Prospects

Arguably, the discussion of reductionism in biology is becoming more rather than less philosophically interesting. This is a consequence of recognizing the diverse conceptual landscape carved out over the past two decades, one that is much larger than that conceived of when only Nagelian theory reduction was in view. Biological science is now more specialized than ever but disciplinary proliferation brings with it issues relevant to any analysis of reduction: using the same terms differently, disparate methodologies, different explanatory norms, and divergent interests in “levels” of biological organization. It also includes an acknowledgment that certain enduring biological topics like development still engender difficult questions about reductionism. Philosophers need to recognize that a criterion of adequacy on accounts of reductionism in biology involves recovering why scientists make frequent pronouncements about the failure of reductionism in different areas of life science inquiry.[13]

There is also increasing contact between neuroscience and psychologically oriented investigations. A potential prospect related to the contact between molecular neurobiology and psychology is the injection of epistemic considerations into philosophy of mind discussions. More attention to issues of representation, decomposition, and temporality could alter the nature of these debates. At the same time, a more explicit evaluation of metaphysical components imported from philosophy of mind into philosophy of biology is warranted. Mediated by the work of Jaegwon Kim, Rosenberg incorporated the concept of supervenience from Davidson three decades ago (Rosenberg 1978);[14] and Rosenberg has recently redeployed Kim's (1998) causal exclusion argument in support of reductionism in biology (Rosenberg 2006, see also Sachse 2007). Discussions of the status of downward causation and realization in biological systems are relevant for a better understanding of the intersection between epistemological and metaphysical components of reduction (Craver and Bechtel 2007, Robinson 2005, Wilson and Craver 2007). For example, an interest in the context sensitivity of realization in philosophy of mind (Wilson 2004, ch. 6) invokes issues pertaining to the context objection, individuation, temporality (especially causation versus constitution), and intrinsicality. These analyses can then be imported into biological contexts to argue for or against particular aspects of reductionism (see, e.g., Wilson 2005).

One specific area where more work needs to be done is relating discussions of different kinds of hierarchies to the questions of reductionism. Many have discussed the diversity of biological hierarchies (Grene 1987, Korn 2002, 2005), some from areas of biological research that are for the most part ignored when reductionism comes into view, such as paleontology and systematics (Valentine and May 1996, see also Grantham 2004). Paying attention to temporality encourages the exploration of functional or control hierarchies in more detail (Salthe 1985, 1993, Wimsatt 2002).

All of these possibilities suggest a more robust recognition of disciplinary heterogeneity in biological investigation relevant to reductionism. Although ecology has received some treatment (Dupré 1993, ch. 5; Mikkelson 2004), issues of decomposition and representation have not yet received wide attention (cf. Lafrancois 2006). Another important area related to explanatory reduction and mechanisms in experimental biology that has been interpreted from an anti-reductionist perspective is cancer biology (Moss 2002, Soto and Sonnenschein 2005). Because of the diversity of explanations found in these different disciplines, the nature of scientific explanation returns with a vengeance (cf. Schaffner 2006). Although only a few reductionists demand explanation to be strictly nomological (Rosenberg 2006, Weber 2005), the subtle interplay between explanation and reduction will likely have to be explicitly considered when ranging over diverse biological subdisciplines. Additionally, as more attention is given to the diversity of investigative reasoning (or “scientific practice”) in and among these disciplinary contexts, the interplay between different aspects of methodological and epistemic reduction will no doubt become more salient.

A final prospect concerns whether disparate discussions of reduction that are occurring with respect to different sciences will interact fruitfully. Although discussions of reduction in philosophy of physics have largely diverged from those in philosophy of biology, some connections have been drawn. For example, Sarkar has approached reductionism in quantum mechanics using his account of reductionism that is more well-known in its application to biological contexts (Jaeger and Sarkar 2003). The potential intersection between these discussions also arises around the relations between parts and wholes or temporality, which have also come under scrutiny from philosophers focused on the physical sciences (Hüttemann 2004, Rueger 2000). Another area worthy of more attention is reductive explanation in chemistry, which largely has been ignored by philosophers of biology (see, e.g., LePoidevin 2005, Ramsey 1997, Scerri 1994).

In all these cases it seems clear that debates about reductionism in biology have not reached a denouement but rather portend vigorous philosophical discussion as the heterogeneity of issues related to its ontological, epistemological, and methodological aspects are brought to bear on perennial biological topics. The task of philosophers focused on reductionism in biology will be to attend to these promiscuities of reasoning and seek to develop accounts of reduction that offer a more general perspective on biological knowledge and scientific inquiry.


Other Internet Resources

[Please contact the authors with suggestions.]

Related Entries

biology: notion of individual | developmental biology | developmental biology: epigenesis and preformationism | emergent properties | gene | genetics: molecular | life | mereology | molecular biology | multiple realizability | natural selection: units and levels of | physicalism | physics: intertheory relations in | scientific explanation | scientific progress | scientific unity | supervenience | teleology: teleological notions in biology


Tom Doyle, Andreas Hüttemann, Ken Waters, Rob Wilson, and an anonymous referee provided useful suggestions on earlier versions of this entry. Thanks to Naomi Scheman for the suggestion to look at Margaret Drabble's The Sea Lady. Ingo Brigandt's work on this essay was funded with an Izaak Walton Killam Memorial Postdoctoral Fellowship by the Killam Trusts of Canada.