Abstract
The unification approach, due to Friedman and Kitcher gives a functional definition of explanation � explanation is that which unifies data into basic categories of physical phenomena. Categorization according to some is reductionism, it is by �reducing� that we tend to unify and connect apparently diverse processes. This paper takes a different approach. We do support that explanatory unification is the fundamental �method� according to which science progresses but this very strength of unification is derived not from reductionism but by an appeal to holism within science. We would suggest a role for theoretical structures that tend to serve this �holistic� purpose.
One characteristic feature of theoretical explanation seems to be embedding diverse physical phenomena within abstract theoretical structures . These structures serve as frameworks or contexts within which the data is �explained�. Thus explanation can be contextual and bounded by theoretical frameworks. According to this view then, Hempel�s deductive nomological model would stand justified; explanations are preceded by theoretical laws or even structures.
On Theoretical Structures
Yet, what would one mean by structures and how would they have any explanatory relevance? Friedman (1981) argues that we explain the properties of gases by embedding them into the molecular theory of matter as we explain the properties of chemical compounds by fitting them into the atomic theory of matter. Theoretical structures are the foundations, the dough on which you fit your raisins and peanuts. The whole of science seems to be a practice in the art of �fitting in�, planetary motions in Newtonian gravity, Mendelian genetics in Evolutionary theory, theory of light in Electromagnetism, and the failed attempts to fit in quantum physics with relativity. When the fit misses a hit, problems arise and we are faced with picking up the �explanation� from the deluge of interpretations. For, explanations explain, but interpretations might not do so.
According to Friedman�s model, if P represents a theoretical structure with a set of defined mathematical relations and Q represents the observational substructure, or the data that needs explaining, then P would be an explanation of Q. So in deductive nomological terms, the explanandum Q is derived from the explanans P, a set of laws and description of facts. The laws consisting of mathematical relations would be the theoretical �structures� and these structures are the ones that are used in the process of fitting in to explain the explanandum. Structures are thus mathematical relations that tend to relate diverse phenomena.
In trying to offer a realistic picture to theoretical structures, Friedman distinguishes between the representational and the reductive capacities of structures. The positivistic approach of considering theoretical structures as mere representations of phenomena makes structures or laws powerless. Within this notion, observations are ontologically independent with structures serving only as epistemic props. To this of course, the realist would object. In order to give a realistic picture of structures, it seems according to Friedman and also to Kitcher, embracing reductionism is the only way out.
According to the reductivist stance, only by reducing theoretical structures to subparts which are comprised of observations, the theoretical hypothesis can be adequately confirmed (Friedman, 1981). Thus gases obeying molecular theory are substructures, observations that can be reduced to basic fundamental theories. Thus the molecular model of gas would be a reduction, which means structures must be reductive to have any ontological power. Observations are reduced to structures and basic laws in a more realistic world view. Kitcher has claimed that explanation consists of showing how to �reduce the number of types of facts we have to accept as ultimate or brute� (Cited in Jones, 1995). Explanation is about grouping, categorising, unifying and reducing all observations to finally reach the small set of premises, the brute facts that are fundamental to scientific theories. Kitcher�s view on reductionism seems to be more of a case of economy rather than Friedman�s reductive ontology. Kitcher appeals to unification via reduction, emphasizing like Friedman on the unifying power of theoretical structures to explain diverse phenomena.
Reductionism, according to these scholars would thus stand out as the stronger claim for explaining unification; it is only with reduction that structures can have any ontological power, a condition that cannot be satisfied by using structures as mere representations.
On Reduction and Unification
Morrison (2000) has given two common features of unified theories: (1) they embody a mathematical structure of phenomena that furnishes an abstract, general framework that unite diverse phenomena under a single theory; (2) the framework typically also contains a theoretical parameter or a concept that represents the unifying mechanism and that functions as the necessary element of the theoretical structure having unifying power .
Structures as we have suggested are mathematical relations that have unifying power by virtue of reductionism. Unification theory claims that explanations are deductive derivations of a statement of the fact to be explained . For an explanatory account to be successful, particular instances must be understood within the contextual general schemata and the derivation pattern used must be a subset of a range of derivation patterns . Jones suggests the importance of what he calls the �explanatory store� that is composed of the �smallest set of derivation patterns that together can be used to generate the largest amount of our total knowledge of the universe� (1995, p.22).
The aim of derivation involves enriching of the theoretical structure as also explaining the given data. Theories that unify a diverse range of phenomena have been thought to possess great explanatory power and a theory�s explanatory power has been analyzed in terms of its ability to unify. Thus Newton�s laws of motion have great explanatory power as they unify all motion by considering common laws, Newton�s universal gravitation unifies celestial and earthy motions, and Einstein�s general relativity unifies space-time, mass and energy. A unified theory is thus one that can explain several different phenomena using the same laws (Morrison, 2000). The grand unified theory that Einstein has searched for all his life, the search for a Theory of Everything, is the ultimate explanatory powerhouse that science seems to be seeking.
Yet, in our attempt to reach unification, we move through mathematical relations and theoretical structures, to which ultimately the phenomena are reduced to serve any unifying purpose. Reduction precedes unification, yet it seems by the very act of reducing to structures or relations, science strives to establish as order, a unity of things that can only be seen as the coming of the full circle.
There are two levels on which unification seems to work. Unification occurs at the level of the structures, as we see in equations, where �m� and �E� seem to have been related in the mass-energy equivalence E=mc2. Relations symbolize a sort of unification as diverse processes are expressed within precise mathematical limits. Newton�s universal gravitation and the relations between gravity and mass also express the same unification. Unification also seems to occur at the level of actually describing two diverse physical phenomena as in light and electromagnetism. Thus unification seems to be both intra-structural and inter-structural.
Our emphasis is on unification at all levels dependent on relations rather than reduction. Reduction is just a method to achieve unification but unification being the ultimate goal undermines the role of reductionism as a method. In fact according to our stance, reductionism is not even a necessary method; in so far as unification provides explanatory power to science, a holistic move towards �order� is that which drives science. We would suggest that although reductionism plays only a part in the explanation in the initial stage of �fitting in� several observations within one structure, the ultimate unifying power of theories is revealed only by appealing to what we call as Structural Holism.
On Causes and Explanations
Before we move on further with our thesis on holism, let us do a detour to discuss the differences between causal and unification accounts. As with unification, that serves an explanatory purpose, causation is also suggested as having explanatory power. Salmon (1984) argues for causation as explanation as scientific explanations frequently appeal to observable and unobservable causes. Thus scientific explanations require �causal� relations. He introduced direct and indirect causal relevance to imply that one event can be the direct cause of another or two events can be related to a common cause. In case of direct causal relevance, Salmon suggests that our current world picture �seems to incorporate the notion that causal influence is transmitted from one part of space-time to another by means of spatiotemporally continuous processes�(1984, p.210).
Salmon also discusses the notion common cause inferred in situations like in appearance of two tracks in a cloud chamber emerging from a single point. The principle of common cause is sometimes explanatory and sometimes inferential. The falling barometer reading explains correlation between barometer reading and atmospheric pressure, in case of crystal gazing for example, we infer existence of common cause which we have not yet located . Thus causation might be used to infer or to explain. In certain cases as in the correlation between the spectra of two images, we �inferred� the common cause, later the elliptical galaxy that was observed formed an important part of the explanation of that correlation (Salmon, 1984).
It could be suggested that in causation, inference precedes explanation, although this has been debated in recent times. There seems to be methodological issues that we can identify here. Causal accounts require inference as a process of explanation; unification accounts according to some require reduction as a process of explanation. Jones has suggested reductionism as an extreme form of unification and in this paper, we have argued for the primacy of unification via what we call structural holism.
Strevens (2004) have tried to unify the causal and unification approaches under a common banner of �kairetic� account. He tries to provide a kind of causal unification account to solve the primary problem of causal explanation, namely to identify which parts of the causal network are explanatorily relevant to the occurrence of the phenomenon in question.
Causal asymmetry that has been pinned down as �the� reason for supporting for causal accounts, as height of flagpole explains the length of the shadow but length of shadow does not account for height of the flagpole. Causation seems to be unidirectional at least in most cases. Unification on the other hand seems to depend on �mathematical relations� that may not place an emphasis on directionality. Yet causal unification does not spell out the role of theoretical structures and thus seems to miss out on the unifying power of �relations� in theory.
On Structural Holism
Our stance on explanation is purely based on the unifying power of theoretical structures. Causation is in troubled waters within modern quantum physics and we propose an abandonment of causal accounts to accommodate unification as the explanatory principle of scientific theories and physical phenomena. Thus we are not concerned with how or why A causes B but whether A forms part of B�s structure; we are not concerned whether A and B are related by a common cause but by whether A and B have a common theoretical structure.
The mathematical relations in a theory unify diverse phenomena and the order that is revealed through this unification suggests a network of interconnections between events and between theories. This network of interconnections is a network of different structures in different theories to reveal a more and more general all encompassing framework that can contain within it all other theories. Structural Holism is the principle, we suggest that guides the unification accounts and relates widely diverse phenomena. We have discussed intra-structural (within relations) and inter-structural (between relations) unification and structural holism encompasses both. Maybe it is possible to distinguish between a phenomenal reductionism, or reduction of phenomena to structures but the ultimate guiding principle we suggest is the principle of holism.
So, what is structural holism? As we have suggested structural holism represents a unification principle that tends to connect different structures and theories like a web of networks. Causal unification as Strevens has proposed seeks to have a solution �in between� the causal and the unification accounts, yet as we have discussed, many believe this unification can only be given a realistic effectiveness by leaning on reductionism as the principle of categorization with structures embedding observations. The view that �explanation is fundamentally a matter of showing how numerous disparate conclusions about the world can be derived from a small set of premises� by the method of reductionism, is motivated by the need for a comprehensive theory of explanation determined by the role of theoretical structures that subsume all observational evidence. Kitcher�s reductivist account of explanatory unification has been criticised by scholars like Jones who advocate a more non-reductivist explanation based on unification theory.
We take on from here as we propose a fundamentally non-reductivist and holistic role for theoretical structures which form the foundations of explanations. Thus observational evidence is fitted in within structures and these different structures are connected via a principle of unification that we call �structural holism�. Thus structural holism is a step further from structural realism that proposes the immutability of structures in theory change. Structural holism implies a �connection� of structures that serves a common explanatory purpose. For instance, if some parts of structure or mathematical equations are carried over from one theory to the next, this �part� which remains as fundamental and explanatory is guided by a unifying holistic principle.
Structural Holism is thus the underlying principle according to which �the continuity or shift in form or structure� in theory change, takes place. John Worrall introduced structural realism using the case of transition in Fresnel�s elastic solid ether theory to Maxwell�s theory of electromagnetic field and argued that the important element of shift from Fresnel to Maxwell was a carrying over of structure. This was again based on the Russellian notion that we can only give a realistic interpretation of entities and processes with which we are acquainted by describing them in terms of their structural properties (Ladyman, 1998, p. 411).
Structural realism has, what Ladyman emphasizes an epistemic commitment to structures while not being fully ontologically committed to the content of the theory. The NMA which serves as the main driving force for structural realism demands an explanation for the predictive success of science. Considering structural realism, the success of science is dependent on how �relations� in a model or theory can fit in with the relations among phenomena in the world. In giving primary importance to structures, structural realism tend to overcome some problems of unobservables, as it is the structures and not the objects that are retained in explanations of theories.
Within the explanatory unification accounts, theoretical structures seem to play a pivotal role and these theoretical structures are considered not in isolation but within a larger group of derivation patterns. Thus, explaining light as a phenomenon within Maxwell�s theory of electromagnetism is not just subsuming light under a theoretical framework or structure but also explaining it by unifying several theoretical accounts. Bringing several theoretical accounts under a common pattern seems to be the goal of unification which justifies the search for a unified theory in science.
What we propose here, against Kitcher�s reductionism, is a difference in �method� or the �principle� according to which this unification works, although we retain Kitcher�s emphasis on unification as the central idea in explanation. Thus from our viewpoint the unification principle which provides explanations to observations is not one of reductionism, explanation does not involve reducing data to ultimate theoretical structures but rather explanations are provided by a principle of holism or connection between theoretical structures. Kitcher�s approach is thus vertical subsuming data under more and more ultimate theoretical structures. Structural holism has a horizontal approach and theoretical structures are considered as connected in a continual pattern. In case of theory change, structures that are carried over from one theory to the next are retained according to structural holism. Thus it is a holistic rather than a reductionist picture, the principle of structural holism that determines which structures are carried forward in theories and also which structures help to provide an explanation of the observations.
Structural holism is thus a unifying principle in science that, 1) determines which structures are retained in theory change, 2) determines how these structures can explain concerned data, and 3) determines how one theoretical structure relates to another to give a common unified picture of scientific progress.
The unifying principle of holism determines which structures are carried over as only those structures that are coherent even within theory change, are retained. Thus �coherence� seems to be a determinant here. This principle also influences which data can be fitted into which structure and whether the data is consistent with the theory. �Consistence� seems to act as a catalyst between theory and observation. The unifying picture of structural holism tends to unify structures or theories that are congruent to give a unified picture of the whole and we suggest that science aims for this unified picture. Coherence, Consistence and Congruence are features of Structural Holism that serve its triple purpose.
Conclusion
In considering structural holism, we do assume that a part of theories are retained as the structural realists would have it. We only go a step further by suggesting that the �principle� behind this retention of structures is a principle of structural holism that plays an explanatory, unifying and predictive role within science.
Bibliography
Achinstein, Peter
The nature of explanation / Peter Achinstein.
New York : Oxford University Press, 1983.
Friedman, Michael
Explanation and Scientific Understanding
Journal of Philosophy 71 (1974): 5--19.
Friedman, Michael
Theoretical Explanation in
Reduction, time and reality: studies in the philosophy of the natural sciences / edited by Richard Healey.
Cambridge : Cambridge University Press, 1981.
Jones, Todd
Reductionism and the Unification Theory of Explanation
Philosophy of Science, 62, pp.21-30, 1995.
Jones, Todd
Unification, Reduction, and Non-Ideal Explanations
Synthese
An International Journal for Epistemology, Methodology and Philosophy of Science
112 (1): 75-96, July 1997
Kitcher, Philip
The advancement of science : science without legend, objectivity without illusions / Philip Kitcher.
New York : Oxford University Press, 1993.
Kitcher, Philip
Two Approaches to Explanation
Journal of Philosophy, 82 (11) (1985) pp. 632-639
Kitcher, Philip
Explanatory Unification
Philosophy of Science 48:507-531. (1981)
Kitcher Philip
�Explanatory Unification and the Causal Structure of the World.�
In P. Kitcher and W. Salmon (eds.),
Scientific Explanation
Minneapolis: University of Minnesota Press, 1989, pp. 410-505. 1989
Scientific explanation / edited by Philip Kitcher and Wesley C. Salmon.
Minneapolis : University of Minnesota Press, c1989.
Ladyman, James
What is structural realism? / James Ladyman.
1998.
Studies in History and Philosophy of Science; v29A, 1998, pp.409-29
Theories of explanation / edited by Joseph C. Pitt.
New York ; Oxford : Oxford University Press, 1988.
Morrison, Margaret
Unifying scientific theories : physical concepts and mathematical structures / Margaret Morrison.
Cambridge : Cambridge University Press, 2000.
Psillos, Stathis
Causation and explanation / Stathis Psillos.
Chesham : Acumen, 2002.
Scientific theories / edited by C. Wade Savage.
Minneapolis : University of Minnesota Press, c1990.
Salmon, Wesley C.
Scientific explanation and the causal structure of the world / Wesley C. Salmon.
Princeton, N.J. : Princeton University Press, c1984.
Wallace, William A.
Causality and scientific explanation / by William A. Wallace.
Ann Arbor : University of Michigan P., 1972-74.
Wright, G.H. von
Explanation and understanding / by Georg Henrik von Wright.
London : Routledge and K. Paul, 1971.
The structure of scientific theories.
Urbana ; London : University of Illinois Press, 1977.
2nd ed. / edited with a critical introduction and an afterword by Frederick Suppe.
Strevens, Michael
The Causal and Unification Approaches to Explanation Unified Causally
Nous
Volume 38, Issue 1, Page 154 - March 2004