Download Interplay of Parts & Wholes: Holism vs Reductionism in Science & Social Science and more Exams Holism in PDF only on Docsity!
AGAINST AND FOR HOLISM:
A REVIEW AND REJOINDER
TO
D. C. PHILLIPS*
Stanley Bailis
San Francisco State University
INTRODUCTION
Those of us who justify interdisciplinary work by appeal to holism must pay attention to well-made cases against it. D. C. Phillips' case is one. He maintains that holism, taken seriously, can't work. Phillips sees in holism three distinguishable theses about complex entities and ways of dealing with them more appropriately than can be done by the traditional scientific method. The first, which he calls Holism 1, contends that wholes emerge from the association of their parts and can neither be investigated adequately nor explained by mechanistic analysis. Holism 2 identifies the reality of wholes in powers of organization that cannot be explained by reduction to the properties of their parts. Holism 3 is a call for conceptual terms referring to wholes--a call that sometimes denies that such terms can be rigorously developed under the prevailing scientific image of nature. Certain positive aspects of these theses have merit, in Phillips' view, but their negative claims about the scientific method are untenable. For him, the method itself is so reasonable and moderate that all scientists, even holists, are bound to use it. The futility of holism follows from its efforts to reject a method that it cannot avoid. The literature of holism is large and difficult because it is expressed, often vaguely, in the specialized jargons of many different disciplines. Nevertheless,
Holistic Thought in Social Science (Stanford, CA: Stanford Univ. Press), 1976, 149 pp.
Phillips presents an ordered account of it that clearly delineates holism's main theses and applications as they have been developed since the mid-nineteenth century in several fields--philosophy, biology, psychology, sociology, anthropology, historiography, political science, and general system theory. His treatment is closely reasoned, carefully documented, and based on a thorough grasp of the doctrines he finds wanting. It is, consequently, worth going along with him to see in more detail what he thinks holism is and has been, what he finds wrong with it, and what's left of it when he's done, it is also worth considering the possibility of a rejoinder to his case. For if his case is logically necessary, holism is not. THE NATURE OF HOLISM The Theses Holism 1 had ancient origins but came to modern prominence following the Romantic movement when it was formulated by certain neo-idealists in philosophy and in biology by the organicists. Done for the most part between 1880 and the later 1920 's, this work generally reflects the theory of internal relations developed earlier by Hegel and his disciples: For things to exist as parts, rather than as isolated entities, the wholes to which they belong must also exist. Things connected as parts in a whole have "relational properties" that are different from their characteristics in isolation. Parts and wholes are reciprocally influential, so that neither term can change without altering the other. From these broad notions, neo-idealists like F. H. Bradley, A. E. Taylor and J. McTaggart derived five more specific ideas that form the philosophical core of Holism 1. First, the analytic method is inappropriate for the study of complex wholes and their parts. Analysis separates parts from the whole in order to investigate them as things in isolation, thereby depriving parts of their relational properties and wholes of their coherence. Second, because relational properties are among the defining attributes of parts, parts cannot be understood as such outside of the wholes to which they belong. Third, wholes are more than the sum of their parts considered as isolates because wholes include the relational properties of their parts. Fourth, wholes determine the properties of their parts in the sense that altering the whole (e.g. by adding, deleting or substituting a part within it) necessarily alters relationships obtaining among parts. Fifth, parts are dynamically inter-dependent in the sense that changes in the properties of any one will change the relational properties of all the others and, consequently, the nature of the whole itself. Organicists advanced similar ideas in biology, chiefly because they doubted the ability of mechanistic analysis to explain the discoveries about vital processes that were being made on every hand: Evolutionists noted spontaneous variations in individual members of a species that appeared to s pread in blended or dilut ed f orm among s ubs equent generat ions. Microscopists revealed the internal complexity of the cell and its nucleus, making
emergent characteristics of wholes. For these characteristics only exist on the level of the intact whole; and parts, which exist on subordinate levels, do not contain informa- tion sufficient to describe fully much less to deduce the whole. At bottom, wholes are as real and as much possessed of properties as are their parts. Such properties manifest principles according to which wholes effect the organization of their parts. These principles must be sought on the level of reality at which they operate, and they alone are sufficient to explain the characteristics of wholes emergent at that level. On these grounds, proponents of Holism 2 challenge the sufficiency if not the validity of analytic method and the doctrine of mechanism--the first because it proceeds by decomposing wholes; the second because it seeks nothing of explanatory signifi- cance at the level of wholes. Reductionism is rejected outright because it misrepre- sents the explanatory problem at issue: Taking the emergent characteristics of wholes as given, reductionists treat them as constituted by and therefore reducible to the properties of interacting parts. For them, the problem is to discover the interaction laws and relational attributes of parts from which given emergent characteristics can be deduced. In Holism 2 , the problem is to discover the organizational principles that govern and therefore explain both the whole's emergent characteristics and the interaction patterns of its parts. Phillips identifies Holism 3 in statements calling for the development of scientifically useful terms that describe the organizational features of wholes, their emergent characteristics, and the relational properties of their parts. This does not require full acceptance of Holism 1 or 2, nor does it conflict with analysis and reductionism. David Easton, for example, has demonstrated the usefulness of treating political life as if it were an open system possessing inputs and outputs, interactions and feedback loops. Such terms occasion descriptions stressing the interdependency of political phenomena and the organizational dynamics by which their operate. But Easton insists that his strategy is essentially conceptual, not an assertion that the system he posits is physically real. In this he breaks with Holism 2, and he goes on to break with Holism 1 by asserting that analysis must follow conceptualization. Analysis is our only means of identifying the real things whose relationships interest us and, consequently, of providing an empirical basis for theories that can explain the phenomena we've described. One need not admit the integrity and reality of wholes in order to make use of concepts about them. But Holism 3 has been used more ambitiously by proponents of Holism 1 and 2 who want to overhaul the conceptual bases of science: The prevailing image of nature in science reflects the assumptions of analysis, mechanism and reductionism. Hence, nature is conceived as if it were composed of truly elemental things and forces that can be examined separately for their abiding characteristics and explained by the fundamental laws they obey--laws that can also account for aggregates emerging from the interplay of nature's elements. This image should be abandoned and nature should instead be conceived as a
hierarchy of organizational levels so arranged that wholes at any given level can be parts of higher level wholes and have lower level wholes for parts. As wholes become parts of higher level wholes, they acquire new relational properties that are relevant to their operation at that level and non-existent at levels below it. There simply are no relevant matters "more fundamental" than those which exist as a function of the whole in question. So the scientific study of any whole must be guided by conceptions developed for the level of organization at which it is taken to exist. Phillips points to arguments for this version of Holism 3 in Weiss' work, but he elaborates the thesis primarily in terms of general system theory (GST) and its acknowledged founder, the biologist, Bertalanffy. Rooted in organicism, GST defines its basic object of study, the system, as a complex of interacting components. Wholes are systems, and while wholes must differ substantively at different levels, there are nevertheless general principles that hold for all systems regardless of what their components and binding forces or relations are. it is the purpose of GST to discover these principles and to express them in a "logico-mathematical" formalism that can be used across the sciences to effect the study of wholes as such. There is here no resort to matters more fundamental than the whole itself. Rather, the whole is described and also account for aggregates emerging from the interplay of nature's elements. This image should be abandoned and nature should instead be conceived as a hierarchy of organizational levels so arranged that wholes at any given level can be parts of higher level wholes and have lower level wholes for parts. As wholes become parts of higher level wholes, they acquire new relational properties that are relevant to their operation at that level and non-existent at levels below it. There simply are no relevant matters "more fundamental" than those which exist as a function of the whole in question. So the scientific study of any whole must be guided by conceptions developed for the level of organization at which it is taken to exist. Phillips points to arguments for this version of Holism 3 in Weiss' work, but he elaborates the thesis primarily in terms of general system theory (GST) and its acknowledged founder, the biologist, Bertalanffy. Rooted in organicism, GST defines its basic object of study, the system, as a complex of interacting components. Wholes are systems, and while wholes must differ substantively at different levels, there are nevertheless general principles that hold for all systems regardless of what their components and binding forces or relations are. It is the purpose of GST to discover these principles and to express them in a "logico-mathematical" formalism that can be used across the sciences to effect the study of wholes as such. There is here no resort to matters more fundamental than the whole itself. Rather, the whole is described and explained directly in terms of the formal theory of systems. Mismatches between theory and observation are grounds for changing the theory here as they are in other approaches to science. The difference is that observations of wholes, not of other things, are being used to determine the theory's worth.
Differences aside, these arguments resemble Holism 1 in that they clearly present networks of interpersonal relations as organic wholes whose parts can neither be separately understood nor used by themselves to explain the characteristics of the whole. As well, they both advance concepts devoted to the study of social wholes--Holism 3. And both are plainly opposed to methodological individualism--a major form of reductionist thinking in the social sciences. Individualists like the historian J. W. N. Watkins and the economist F. A. von Hayek do accept the reality of emergent phenomena in social life--of institutions, for example, or social structures or cultures that must emanate from enduring relationships among persons. They even grant that partial explanations of some emergents in terms of others are possible, as when inflation is attributed to full employment. But, they insist, full explanation requires such phenomena to be deduced from laws of individual behavior and statements about the properties of specific persons. For social wholes are not physical entities but conceptual constructions inferred from the observable activities of people. As such, wholes cannot be said to have any motive force of their own--the sorts of interests and aims to which explanations of human behavior must ultimately appeal. Nor, for the same reason, can wholes be directly observed and made the subjects of empirical laws that are needed in scientific explanations. Thus resort to individuals is necessary if full explanations are wanted. As we have seen, Phillips takes this sort of reasoning to be logically com- patible with Holism 1. Accordingly, he doubts the sufficiency of organicist rejections of methodological individualism like those advanced by Dewey and Durkheim. For a stronger expression of Holism 2 , he turns to the more recent argument of M. Mandelbaum, who declares outright that concepts referring to forms of social organization cannot be completely reduced to concepts referring to the thoughts and actions of particular persons. So, for example, we can't explain in individualistic terms why people who are as individuals different nevertheless behave similarly toward a bank teller, nor why the teller behaves similarly toward them despite their individual differences. At some point we must appeal to the fact that the parties to this interaction are customers and an employee of the bank. That the parties all know this is a psychological fact. But what they know about is a social fact that is necessary to the explanation of their behavior. In effect, societal things and the facts about them are as real and fundamental as individuals and the psychological facts about them. Reductionists reject this reasoning. For Phillips, their opposition is the basis for separating Holism 2 from Holism 1 which reductionists can accept. Enough has already been said about general system theory to suggest its relevance here. Phillips regards it as little more than a modern restatement of Dewey's ideas. But he does note that the generality of the system concept invites application to a wider range of relationships than those which can be characterized by interpersonal or as having to do with the organism/environment
relation per se. Eastern's use of it in political science, mentioned earlier, illustrates this point as does another example of GST Phillips examines--A. Koestler's discussion of language as a hierarchically ordered system in which initially generalized and implicit intentions emerge, through differentiation, as specific ideas in articulate speech. These examples deal with relations among systemic functions, like information processing, rather than with persons as such. Combined with the references in Philiips' notes and bibliography, they serve to indicate that GST has been associated with all three forms of holistic thinking in the social sciences since the 1950 's. The concepts of structure and function are made vague by a welter of varying definitions and usages in the social sciences. But Phillips clarifies matters sufficiently to show how both concepts may have holistic connotations. Thus structuralists usually argue that human social behaviors are not random but ordered or related--partly in terms of the biological needs they serve, but more importantly by virtue of the sociocultural rules they obey. And while these rules may vary substantively across groups, they are themselves related in ways that are regular enough to be expressed as universal laws. In this argument, "structure" refers to higher level or universal relations that hold as laws for lower level relations--for the rules in a given society that hold in turn for yet lower order relations embedded in the interactions of specific categories of persons. Different brands of structuralism locate the origins of universal relations differently. In the case Phillips examines most closely--Levi-Strauss' anthropological account of the matrilineal avunculate--they originate in mental functioning that is the same for all people. But structuralists who are differently inclined, say toward economic universal, may argue differently about origins while remaining thoroughly holistic. What is necessary is a focus on total systems of relations as opposed to their elements. Origins aside, the laws of structure can only be discovered where they are actually expressed--in the forms of relations among social categories of persons that comparison shows to be regular across groups. Variations at lower levels of observation can be deduced from (explained by) these laws. Since the reverse is not true, this argument is often used to reject reductionism (Holism 2 ) and to call for wholesale conceptual reform (Holism 3 ) along lines that are largely organicist in nature (Holism 1 ). The concept of function is commonly used to explain a system's characteristics in terms of how they help to satisfy its needs. Links between characteristics and needs may be intended by a system's members (manifest functions), or not (latent functions). But a functionalist inquirer must have a theory that accounts for the system's needs, and some means of showing that the characteristics in question are the best or only way of meeting them. The inquirer need not assume that all characteristics have functions or that all functions are only beneficial. Func t ionalis m is of t en as s oc iat ed wit h ev olut ion t heory , as when a s pec ies ' c harac t eris t ic s are s hown t o hav e been s elec t ed f rom v arious
particularistic, it cannot discover them or laws of any other sort. Historicists, arguing somewhat holistically along lines suggested by J. S. Mill, regard synchronic laws as derivative corollaries of diachronic iaws that do control the succession of systemic states. Indeed, it is one of the purposes of social science to discover the laws according to which earlier states of society produce later ones. Many holists tend to disregard this distinction and the dispute that goes with it. H. Spencer, for example, explained the presence of chiefs in groups having more than a hundred members both synchronically as a function or organizational necessity and diachronically as the result of an evolutionary law by which all systems move from homogeneous to heterogeneous or differentiated states. Dewey made similar arguments to explain specific psychological functions as systemic necessities at any given time and as evolutionary effects of on-going adaptation. Levi-Strauss regarded the two approaches as complementary ways of doing the same sort of thing. Behind such views is an argument offered early in the 18th century by G. W. Leibnitz that simultaneous and successive causation are continuous. The separateness or independence of things is, in causal terms, illusory at any moment and overtime; the universe is an interrelated unit and must be explained as such. The resemblance between this line of reasoning and the claims of both Holism 1 and the strong form of Holism 3 is evident. The concern with relations in the study of human behavior takes yet another form in Gestalt psychology, developed by M. Wertheimer, W. Kohler and K. Koffka after 1910. Gestalts are mental unities which represent experienced objects and situations as wholes rather than as collections of separate elements. Perception, for example, is not a mechanical response of sensory receptors to the separate properties of elements comprising a stimulus object or situation: The same elements (say, two spatially separated flashing lights) will be differently perceived (as one light moving between two points, or as two lights flashing successively) depending upon their own relations (like the time interval between flashes) and relations internal to the perceiver (like the coordination of responses of different receptors with each other and with experientially based conceptions). On these grounds, and in keeping with Holism 1 , Gestaltists argue that any psychological process must be understood as a relationship of reciprocal influences holding between relations internal to persons and to the objects or situations they experience. Having formed a concept to deal with this matter, Gestaltists manifest at least the mild form of Holism 3. Such are Phillips' main examples of holistic thought in the social sciences. Certainly his survey is not exhaustive, even when the many cases he mentions in passing or in notes are added to those he examines in detail. But Phillips is, after all, operating as a philosopher interested in the forms and uses of arguments. Typification and criticism, not exhaustiveness, are his goals. Having reviewed his examples, we can turn to his critique.
PHILLIPS' CRITIQUE
Phillips' contention that holism can't work rests on three lines of argument. First, the holists are wrong to suppose that an object can only be understood when all of its properties are known. Second, this supposition has led holists to misconstrue the concerns and capabilities of analysis, mechanism and reductionism with respect to complex entities. Third, the same erroneous supposition admits logical absurdities into holism that deprive it of a method and of the advantages it claims over mechanistic analysis and reductionism. The bases for these claims warrant review. During the nineteenth century it was common to speak about the "nature" of a thing, as if things had absolute identities that were determined by all of their properties. It was on this basis that holists, particularly the organicists, saw differences of identity in differences of even a single relational property. By contrast, twentieth century discussions of such matters tend to distinguish between the defining and accompanying properties of things. Defining properties grant identity to an object, allowing it to be recognized for what it is. Accompanying properties may be present or absent without altering the identity and recognizability of a thing. Since it is possible for relational attributes to be accompanying rather than defining properties, a thing can retain its identity while being observed apart from the full panoply of its possible relationships. This lends much practical legitimacy to the method of analysis and to explanatory doctrines that depend on it. The widespread acceptance of this distinction since 1950 reflects the influence of Wittgenstein's ideas about how we use terms to characterize things. A term refers to a cluster or family of characteristics only some of which need be present in a thing for the term to be applied to it. It is the cluster which is defining, and the presence of even different subsets of its elements in several objects allows all of them to be called by the same term. This amounts to classifying nonidentical objects as alike--a decisively analytic move that assigns identity to things apart from the full set of their properties. But since objects classified according to some of their attributes do not lose the rest of their traits merely by virtue of being grouped together, we have here a way of discovering the whole range of characteristics-- including the relational properties--possessed by the members of a given class. Because this applies to complex entities, the relational concerns of holism are not lost, and an orderly basis for generalizing about classes of things is gained. Twentieth century holists, particularly proponents of GST, have continued to suppose that a thing can only be understood when all of its properties are known. Accordingly, they have persisted in the rejection of any method or doctrine that attempts to separate objects for closer study. In Phillips' view, this amounts to a misreading of the concerns and capabilities of analysis, mechanism and reductionism.
which seem beyond us at present--even problems of the sort that holists claim only their approach can handle. To present the absurdity contained in holism, Phillips invokes an argument ad- vanced by William James and Bertrand Russell against the Hegelian doctrine of internal relations: If all things are related, and if relational properties must be known for a thing to be understood, then we must know everything in order to know something. Hence, we can never know anything at all. Let us examine this point further. Suppose we wish to understand a complex entity E composed of three elements A, B and C. According to Holism 1 , E emerges from the relational properties of its elements. Hence, it is pointless to examine A, B and C separately because the relational properties of these elements will be missed. Nor will it do to examine the subsidiary combinations AB, AC, and BC since AB is affected by C, AC by B, and BC by A. In effect, nothing will do to understand E save understanding ABC. But this is the same as saying that E can only be understood by understanding E, a fully circular argument. In order to break this circle, some holists grant that elements and subsidiary combinations may usefully be studied as a way of determining the relational properties needed to explain emergent characteristics--so long as the latter are the primary objects of inquiry. But this is, as we have seen, what analysts, mechanists and reductionists do when they treat relational properties as initial conditions to deductive arguments that account for the same sorts of phenomena. To make this concession, is to grant that holism has no method of its own--it must resort to analysis to get data--and that it lacks the advantages claimed for it regarding the explanation of complex entities. Proponents of Holism 2 attempt to avoid these difficulties by constructing the explanatory problem differently. The goal is not to explain E in terms of the relational properties of A, B and C. Rather, it is to explain these properties by appeal to the prin- ciples according to which the organizational power of E operate. But this supposes that E itself can be isolated from the yet more complex systems to which it belongs. Mechanists and reductionists do make this analytic assumption, but serious holists cannot without violating an essential premise of their position. If all things are related, and E is a thing, then it must have relational properties that are necessarily missed when it is examined in isolation. Again, either everything must be known in order for anything to be understood, or holism has no way of working that isn't used by those it attacks. The strong form of Holism 3, being an extension of Holism 1 and 2, has much the same difficulties. But, because it is presented as a call for conceptual reform, it includes an additional source of trouble. If the images invoked by Holism 3 are purely conceptual, they can hardly be said to have physical powers of organization. At some point, reference must be made to real things. But if real things really are embedded in a hierarchy of relationships, then none of them can
be understood outside of the whole to which they belong--which is to say, again, that the entirety of nature must be known before its parts can be understood. The mild form of Holism 3 escapes these problems by granting the legitimacy of employing analytic, mechanist and reductionists strategies in the empirical pursuit of holistic conceptions. But this, of course, is only to grant again that holism as such has no way of working of its own and no advantages over analysis, mechanism and reductionism. Virtually all of the uses of holism in social science must fall by the same arguments: Dewey is compelled to grant the usefulness of information about individual objects. Neither Durkheim nor Mandelbaum can reject reductionism logically by arguments that require the very sorts of information from which reductionists can also deduce emergent social phenomena. GST is unable to specify its subject matter and make predictions about it without engaging in the same separations of objects that its central premise rejects. Holistic versions of structuralism, functionalism, historicism and Gestaltism are in the same boat. All of them must either be immobilized by their assumptions or resort to analysis and do what their opponents do too. Plainly, Phillips' critique is an espousal of the positions that holism rejects. Holism has solved no problems not solved as well by analysis, mechanism and reductionism; and problems that did for a time resist solution by the latter approaches also eluded holism. In this sense, holism is as unnecessary as it is unworkable--holists must either resort to analysis or accept the absurdity of needing to know everything in order to know anything. In a postscript, Philips briefly identifies those features of holism which remain acceptable under his critique: The holists have raised pertinent questions about ways of investigating organic wholes and about the effects of analytic decomposition upon such inquiries. They have rightly called attention to the importance of dynamic relations between parts and to the difficulties of predicting the emergent properties of wholes. And they have been right to call for the development in science of concepts referring to wholes. These are substantial concessions. But Philips grants them, taking his critique to have dismissed those points of holism which are antithetical to traditional science--the rejection of analysis, mechanism and reductionism; the ideas that the whole determines the nature of its parts and that the nature or identity of an object consists of all of its attributes, including all of its relational properties; and the call for a total reconceptualization of nature in holistic terms. In his view, what is left when these points are gone is safe. REJOINDER Phillips' critique can be answered in ways that argue for the continued use of holism both generally and as a justification of interdisciplinary work.
mechanists and reductionists must be tacitly accepting tenets of holism that Phillips rejects. Insofar as analysis seeks the isolable properties of objects, it cannot be regarded as providing information about their relational properties. The seeking itself is certainly in line with the main purposes of analysis--describing a system's initial conditions in order to determine (1) the identity of its unknown components by showing that they behave like objects specified in laws, or (2) the applicability of laws by showing that they correctly represent the behavior of system components which are known to be like the objects they mention. Such purposes require knowledge of isolable properties because it is in terms of them that objects are identified or defined in laws. It is true, of course, that these properties affect the relational characteristics that objects acquire in given contexts. But it is likewise true that isolable properties, being independent of particular contexts, are different from relational properties which are context-dependent. These points about the analytic study of objects as system components pertain as well to the study of systems as such. The analysis of a system requires it to be isolated--unaffected by factors not being considered explicitly. Without isolation it cannot be told whether factors identified in laws are those responsible for the system's observed behavior. This is why analysts invoke assumptions like ceteris paribus or mutatis mutandis and try to control their subject matters by physical or statistical means. It is partly because holists doubt that an object or system can be both isolated and known that they question the effectiveness of analytic method. But their larger doubts are based on the idea that the analytic search for properties in isolation leads away from relational properties. It would be wrong, however, to insist that there is no way of using analytic methods to discover the relational properties of parts. By systematically observing an object in different contexts, we can determine which of its characteristics remain the same and which alter, hence which define its identity as a separable object and which define its identity as a part of each context, in effect, contexts are used as independent variables in an essentially descriptive process of inquiry. Admittedly, this procedure is not analysis in the usual sense because it involves no laws. But it does use the analytic method of manipulating some variables to see how others behave, and it does lead to knowledge of the isolable properties of objects. That it also yields knowledge of relational properties must be counted as something of an advantage. However, to gain this advantage it is necessary to regard the contexts or independent variables as wholes and to know them fully in advance. Otherwise we should not be entitled to treat the altering properties of objects as relational properties of parts, and we should not be able to specify the contextual differences involved.
Accepting these conditions amounts to accepting at least one of two holistic tenets which Phillips rejects--that wholes (the differing observational contexts) determine the relational properties of their parts, and that the whole must be known for its parts to be understood. Regarding the first tenet it may be objected that, since it is not the whole but only some of its properties that are varied in order to create differing observational contexts, it is not the whole which determines. To raise this objection, however, is to invoke the second tenet by claiming knowledge of the whole--in this case, that it has specific properties which can vary while the rest remain constant. Phillips circumvents these issues--the analytic emphasis on properties in isolation, and the tacit acceptance of unwanted tenets--by placing relational properties in the province of laws. Laws do state the forms of relationships holding between properties of objects. But since these forms must remain the same in different contexts in order to be law-like, it is reasonable to question whether they can always be found for relational properties which are context-dependent. This question can be raised about the use of laws in both mechanist and reductionist treatments of wholes. Mechanist laws express the regular forms of relationships between defining properties of basic entities and the forces which affect them. The entities mentioned are basic in the sense of being elementary units from which a wide variety of things more complex than themselves are made. Their properties must be defining in the sense of being stable enough to permit entities of the same class to be recognized in different contexts. The forms of relationships between these properties must be regular in the sense of remaining the same in different contexts. The specific values of these properties, like the values of forces operating upon them, vary with surrounding conditions. But the categories of properties and the forms of their relations must remain the same in all contexts if information about them is to have explanatory significance. It is the apparent restriction of mechanistic laws to defining properties and fixed relationships which brings holists to doubt their appropriateness for the study of wholes. The parts of complex systems, even when they are basic entities, display properties that are not among the defining attributes assigned to them in mechanistic laws and that are consequently not covered by the relationships advanced in such laws. Holists address this matter by supposing that objects, even basic entities, acquire properties from their association with each other where "association" refers not only to the collection of objects present but also to the manner in which they are organized. These are relational properties, of course, and it is because they do not always appear in mechanistic laws that holists have doubts. Phillips doesn't dismiss this matter. Rather he considers it in terms of two tenets of holism which he takes to be fully compatible with mechanism--that concepts of wholes facilitate discussion of the properties of complex systems, and that the parts of a whole are dynamically interdependent. Thus, confronted
Nothing in these arguments denies the value of mechanistic analysis and reductionism as ways of investigating and explaining natural phenomena. Rather the arguments assert that because these approaches are concentrated upon properties of objects which either remain the same or vary determinately across contexts, they do not always deal effectively with context-dependent character- istics. It is with respect to their focus on defining properties that laws advanced by mechanists and reductionists differ from those suggested in holistic versions of structuralism and functionalism. In effect, holists emphasize the fact that objects identified as the same in terms of their defining properties can display altogether different relational properties depending upon how they are organized. This emphasis does imply that the organizational characteristics of wholes determine the relational properties of objects considered as parts. And the implication does lead directly to the claim of absurdity. The charge of absurdity asserts that two tenets of holism--wholes determine the nature of their parts, parts are identified by their relational properties--lead to circularity and either to the contradiction of accepting analysis or to the impossible situation of needing to know everything in order to know anything. The tenets are held to be circular because they can be taken to mean that the whole determines itself. This interpretation follows from the claim that "whole" and "nature of its parts" are the same things. More specifically, "whole" refers to the full set of relations obtaining among the objects which are its parts; and "nature of its parts" refers to those properties of objects which, deriving from their relations with each other, identify them as parts. By these definitions, a whole is not itself without all of its parts, and no object is a part without the relational properties it gets from the whole. Accordingly, "whole" and "nature of its parts" refer to the same things and the tenets assert the circular argument that the whole determines itself. A different interpretation is possible, however. To begin with, insofar as "nature of its parts" refers specifically to the relational properties of objects determined by the whole, the reference need not be taken to imply that such objects have no other properties--properties that are not determined by the whole but that do influence what relational properties the whole can effect. In this sense, the whole does determine the nature of its parts, but it does not completely determine itself. Consider, for example, a barnyard population of hens which, as a whole, is arranged into a dominance hierarchy. The whole need not be seen as determining all the properties of each bird which make each a hen. It need only be seen as determining the relational property of position in the hierarchy that each bird has. In this sense, the whole does determine the relational properties or "nature" of its parts, But insofar as the parts have properties affecting their "natures" that are not determined by the whole--as being a hen affects inclusion
in the hierarchy without being determined by it--the whole does not determine itself. It is true that the relational properties of all the parts must be known if the whole is to be specified or known. This may seem to constitute a circle insofar as "whole" and "relational properties of all the parts" are taken to refer to the same things. But it must be recognized as well that the whole has properties not possessed by any of its parts. Specifically, the whole has organizational characteristics that are not identical to the relational properties of any of its parts. Suppose, for example, that our dominance hierarchy of hens is organized so as to be completely transitive and irreversible. We may say of this organization, in purely formal terms, that one of its members must dominate all the others, that one must dominate none of the others, that each of its remaining members must dominate some but not all of the others, that none of its members may dominate its dominator or its dominator's dominator and so on. While it is true that every hen in the population must have some relational properties corresponding to these charac- teristics of the whole, it is also true that no member has all of them. In this sense, "whole" and "relational properties of all the parts" do not refer to the same things. By dealing with particular wholes, these arguments and examples may seem to beg the larger questions embedded in the charge of absurdity: If the whole determines the nature of its parts, and the entirety of nature is itself a whole, then parts can hardly be said to have properties not determined by the whole--unless we wish to invoke extra-natural causes. This is so, but it is also true that the whole of nature has organizational characteristics not possessed by any of its parts. Thus, even in this extreme case, holism need not be regarded as completely circular in that "whole" and "nature of its parts" do not refer to the same things. But even if holism isn't circular, its commitment to the image of nature as a whole brings the "contradiction" and "impossibility" issues to bear: if nature is a whole, then the study of any of its parts by itself--for example, the study of a particular whole--amounts to acceptance of the analytic strategy of examining objects in isolation. Holists reject this strategy because it deprives parts of relational properties which must be known if the whole is to be understood. The study of particular wholes therefore contradicts the grounds on which holists reject analysis. Holists do single out and investigate particular wholes. But it is not clear that in doing so they either accept analysis or contradict the grounds on which they reject it. For one thing, holists need not accept the analytically necessary practice of identifying their subject matters by likening them to objects specified in laws. We have already seen why, according to Phillips, analysts must do this. By contrast, holists may identify their subject matters conceptually, as Phillips approvingly indicates that Easton did by likening political life to the concept of an
