Relationalism in Political Theory and Research : The Challenge of Networked Politics and Policy-Making *

Political theory and research approaches are in most cases not explicit in all respects, but are usually shaped by implicit background theories and scientific worldviews, e.g. presuppositions and par adigms. Examples are idealism, materialism or institutionalism. A fairly new perspective in this regard is relationalism. This way of thinking, in its ontological orientation, conceives the world only in terms of relations. Research programmes with a relationalist orientation put this idea into practice with methods of social network analysis which are more and more used also in political science. Several variants of this new ism and type of network thinking are presented in this paper in a critical perspective. Relationalism is ultimately seen as fruitful perspective, but is seen as limited in its explanatory power, though. It should therefore always be combined with other approaches and perspectives.


Introduction
T he epistemic landscape of science is a complex matter.There are not just theories and data, but a wide range of additional cognitive entities that organise our images and representations of reality and provide them with meaning.Jeffrey Alexander (1982), for instance, just to give a prominent example, envisages a "scientific continuum" including entities such as presuppositions, models, concepts, definitions, methods and observations.Curiously, theories are missing from his list.If we add them, and look closer into the microcosm of theory, we discover even more variety.There is not just "theory", but multiple types and levels.For instance, we usually distinguish not only between theories of variable range, i.e. grand theories versus theories of the middle range or single-item theories, but we differentiate also with respect to exactness and precision.There are fuzzy or implicit metaphorical theories on the one hand, and more explicit and formalised theories on the other.A further distinction is between specific theories that relate to given objects of knowledge (such as the "state" or types of "political organisations") versus general background theories that organise knowledge and research processes in a more general way.Such frameworks or paradigms shape worldviews that provide and arrange a specific access to reality.This means that worldviews of this kind not only have ontological, but also epistemological and methodological implications.If we have a specific idea of the structure of the world, this also shapes the way we study and explore the world.
Since theory development in political science does not take place in isolation but has always been embedded in general philosophical discourses, also the worldviews in political science are influenced by competing philosophical perspectives.Best known is perhaps the old debate between materialism and idealism.Modern "-isms" such as individualism, pluralism, structuralism and institutionalism have increased the diversity of epistemic species also in the political and social sciences quite considerably.A rather new -ism in this context is relationalism, which originally spread in sociology, but now is also increasingly used in political science.Meanwhile, it has shaped a number of specific theories and particularly a new collection of methods -i.e.social network analysis -that are increasingly used also in political analysis.During recent decades these types of studies have gained growing popularity in the various subfields of political science.Since the 1980s, its methods have been used a lot in political sociology and public policy analysis, but during the last decade also increasingly in comparative politics and international relations.
The central question of this paper is how relationalism as a philosophical perspective can enhance political analysis, particularly the study of networked politics and policy-making.In a first step, this question will be approached by linking "relationalism" to policy network research.Thereafter, the aspirations and shortcomings of "relationalist thinking" will be discussed.The concluding section will formulate a sceptical position towards a "pure relational" perspective in which only "relations are at work".The paper calls for a pluralist ontological position, in which the world is governed not only by one single principle.

Policy Networks as a Relational View on Politics and Policy-making
A large part of relational research is focused on networks in public policy.Such policy networks are constellations of public and private actors which generate and implement public policies at all political levels -from the local community to politics at the global level.Like in other research areas of social network analysis, the study of political networks aims at the systematic description and explanation of "network effects" and the influence and constraints of "relational structures" in the political realm and their modi operandi (Kenis, 2007;Provan, 1995).An increasing number of studies have taken this orientation.Some years ago, we compiled a biography of publications on political networks (Schneider, 2007) which included more than a thousand titles from the 1970s up to 2007.Among them were about 200 quantitative studies, about 550 articles on qualitative research, the rest was methodological literature.Since the early 1990s, qualitative and quantitative network studies have expanded continuously.
Already in the early phase of growth of this new research field a debate on its epistemological status began.Tanja Börzel complained about the conceptual variety and confusion in this research area (Börzel, 1998), and other scholars criticised the metaphorical orientation of this literature (Dowding, 1995;Pappi, 1998).Can policy network analysis be based on a theory or a model?Is it a new paradigm, a specific approach or just a bundle of specialised methods?A number of scholars used "policy networks" as a core concept of a new theory on the governance of modern societies with a more refined view on the relationship between state and society (Mayntz, 1996;Rhodes, 2006).For other scholars the study of networks and social relations in policy-making was just a the-ory-unspecific application of analytical methods and research techniques that are compatible to a great variety of theories and concepts (Knoke, 2011;Pappi, 1993).Even other researchers called for greater conceptual exactness and a more systematic integration into foundational social theories (Dowding, 1995).
Interestingly, a rather similar debate took place during the 1980s in sociology, when complaints arose on the poverty of theory in social network analysis (Burt, 1980;Granovetter, 1979).But times have changed, meanwhile a kind of "catch-up-theorising" has taken place and an increasing number of scholars now is talking of "network theory" (Borgatti, 2011), and some schools of thought even propagate the emergence of a "network paradigm" and the establishment of "network science" (Börner, 2007).
Even more pretentious, are efforts aiming at the philosophical foundation of network research in a new -ism, i.e. relationalism.Already in the 1980s Barry Wellman emphasised that network analysis is not just a method but a comprehensive approach implying a specific view on the world (Wellman, 1983).More than a decade later Emirbayer tried to promulgate such an idea by his "manifesto for a relational sociology" (Emirbayer, 1997).In this perspective social network analysis was conceived of as, "… not primarily a theory or even a set of complicated research techniques, but rather a comprehensive new family of analytical strategies, a paradigm for the study of how resources, goods, and even positions flow through particular figurations of social ties" (Emirbayer, 1997, p. 313).
In the following years, the idea of relationalism attracted an increasing number of followers and a veritable academic industry has emerged under this rubric.This paper deals with the promises and shortcomings of this new philosophical perspective.

Between Background Knowledge, Theory and Observation
In order to examine the relationship between political network analysis and relationalism, both epistemic entities first have to be discussed at a philosophical level.In Alexander's "scientific continuum" quote above, the scientific production of knowledge is conceived as a spectrum of increasing abstraction from empirical reality to metaphysics (Alexander, 1982).Starting by observation, it proceeds over methods via increasing abstraction to concepts, models, and general presuppositions, which form the most abstract level.The latter includes implicit assumptions and background knowledge for any form of theory building.Fruitful in this perspective is that the naïve, positivist view of theory building is overcome, and the multiple layers in the knowledge production are recognised.However, despite these improvements, this perspective overlooks the multi--dimensionality of theorising.
Drawing on the work of the Canadian philosopher of science Mario Bunge, it is more convincing to expand Alexander's "continuum" into a two-dimensional space (Bunge, 1996;Bunge, 1998).In addition to the dimension of increasing abstraction there is also a dimension of increasing exactness.These two dimensions are depicted in Figure 2. Empirical knowledge can be, on the one hand, intuitive and vague, without exact definition and explicit specification of the assumptions that underlie measurement and observation.
For instance, if measurement is based on a clear indicator concept, we know in a more precise way how certain facts are represented in the data.Furthermore, at the level of theory this perspective distinguishes between vague and implicit background knowledge and clearly explicated or exact theories.Theories should be hypo-deductive conceptual systems, or at least coherent and logical conceptual networks.Paradigms are mental figurations at a higher level, including families of theories with similar architectures, design principles, and world views (Bunge, 1996;1998;2003).
At the empirical level an important distinction should be made between facts and data.The latter are representations of facts which are matched with models or theories at a more abstract level.Axiomatic, mathematised theories try to produce maximal exactness but imply also "farness" from empirical reality.In the following section we try to arrange the different meta-theoretical components of network analysis and relationalism in this two-dimensional space.

From Network Metaphors to Network Science
To conceive society as an organism, or the state as a ship was the starting point of famous political metaphors and grand theories on the state and society (take for instance Plato's republic).Also for network thinkers, there is a broad array of images available (Brandes, 2009).
Everybody who has looked at a spider's web or a fisherman's net can easily comprehend a set of nodes and links.A network thus depicts a specific arrangement of nodes and links (edges) as the smallest units of relational configurations.On the other hand, there are the notions of the web, net and braid (the translation of the German word "Geflecht"; see for instance the concept "Flechtwerk" of Werner Sombart and the notion of "Politikverflechtung" introduced by Fritz Scharpf (Scharpf, 1976;Sombart, 1927).Roads, rivers, channels and circuits are further representations of networked configurations.Although it is difficult to conceive dynamic interaction in images, it can be represented for instance by the collision of billiard balls.
As mentioned above, in network research there is a strong tradition of criticism with regard to metaphorical orientations.Indeed, metaphorical thinking implies serious flaws, but metaphors also made useful epistemic contributions in the evolution of knowledge.Rather close to empirical reality, they can provide fruitful analogies in everyday language and intuitive ideas about complex configurations.For instance, Plato's helmsman metaphor explained in a simple way the superiority of specialised knowledge.Metaphors in this sense are proto-theories.
A first step in the theorisation of metaphors is the explication and specification of fields of meaning.For instance, in network research there are competing network notions and ideas of what networks are.There are at least two rival network concepts with rather different meanings: on the one hand there is a formal network concept based on graph theory, on the other hand there is a network concept embedded in institutional theory which is frequently used in the governance debate.Whereas a network based on graph theory includes all sets of nodes and links, a network based on institutional theory only contains specific configurations, thus only a subset of the graph theoretical concept.For instance, all configurations of nodes and links in Figure 4 are networks in a formal graph-theoretical view, whereas in an institutional perspective only Figure 4-f is seen as a network.Relational configurations in the latter sense are specific governance structures beyond markets and hierarchies (Kenis, 1991;Powell, 2003).However, both network notions are not completely incompatible, since governance structures can be formalised by graph-theoretical models too, but their meanings are different.
If a conceptual system or analytical framework is used to systematically describe and explain a given object of knowledge, then we build a theory about that given object.For some time there was a dominant conviction that a "theory of networks" could not exist, since there are so many types and forms of networks across countless domains and levels of social reality.For instance, how can we compare a mafia-type criminal network to a smart grid in the energy sector?For some time, however, a group of physicists has tried to find comparable network characteristics in all natural and social phenomena -networks of proteins, people, words, or web sites in the internet.Their ultimate goal is to find general network patterns and generalisable "network laws".An example of such a generalisation is the discovery of the law of preferential attachment in so-called "scale-free networks".Scale-free means that the distributional pattern of network links remains more or less unaffected by rescaling.This happens when many nodes are involved in few links, and if few nodes have many links.This pattern has been recognised since Pareto's research on patterns of income distribution.Based on advanced computing power and large data sets, network research could test these preferential attachments in many areas since the 1990s.Newman (2010) published an encompassing textbook covering most of these new streams and directions of network science.
Besides these general network theories claiming general validity in nature and society, there are social and political variants of network theory.A specific type is, for instance, Bruno Latour's actor-network-theory (ANT), claiming descriptive and explanatory power for socio-political and technological configurations (Latour, 1996), or Manuel Castells' theory of the network society (Castells, 2011), or theories on networked socio-technical systems (Mayntz, 1988).Some scholars understand networks also as key-mechanisms of social order in modern societies (Raab, 2009).
Social theories are systems of propositions including more or less "general" social laws and mechanisms.But "general" does not necessarily mean universality across time and space.For instance, some mechanisms only work in specific social domains or institutional contexts.Theories covering such "semi-universal" patterns are so-called middle range theories, whereas grand theories such as rational choice, game theories and versions of systems theory claim comprehensive coverage.Some social theories are mathematised and suggest high precision, whereas others remain vague and often implicit.Their appeal is often to evoke certain images of society that focus on important aspects or trends of social and political development.Castells' theory of the network society is such a case.
The most abstract axiomatic theory in this context, graph theory, can be traced back to Leonard Euler in the 18 th century.Graph theory provides a formalised language for precise descriptions of relationship structures.On this basis, a variety of algorithms can be constructed for the analysis and explanation of all kinds of relationships (Brandes, 2010).However, graph theory is not a theory of real objects (e.g.factual political or technical networks) in the sense that it could refer to general patterns and mechanisms.If graph theorists explain the whole world in the sense of a "theory of everything", their explanations are perplexingly abstract, without carrying much information (Dipert, 1997).
From grand or middle range theories -axiomatic or not -we have to distinguish varieties of implicit knowledge stocks and background knowledge.These include presuppositions, paradigms, and philosophical -isms.All three are located on the left side of the spectrum of precision in Figure 2. Some social scientists see the network perspective as such a new mental entity (Kilduff, 2006;Stegbauer, 2010).
Scientific worldviews such as "X-isms" (Bunge, 1979) are more elaborated than metaphors, even if they are more imprecise and inexplicit.X-isms often include "ideological" orientations and emphasise the general superiority of certain interpretations and analytical perspectives.A holist, for example, holds the view that individual and social action is not an analytical entity as such, but is determined by macro-structures -such as class divisions or functional differentiation structures (ontological X-ism), and that only explanatory models based on these structures can provide satisfactory explanations (methodological holism).In contrast, individualism derives macro-level social phenomena exclusively from the action and interaction behaviour of individuals.Not all differ in a clear-cut way.For instance, structuralism and relationalism overlap in some areas.Some recent research programs combine different X-isms.For instance, Coleman and Burt provide combinations of individualism and structuralism (Burt, 1982;Coleman, 1990).

Relationalism and the Network Paradigm
Relationalism is a new social ontology aiming to explain the world primarily or exclusively on the basis of relations.It is a contrasting alternative to substantialism or essentialism.Together with structuralism it is probably the most influential background philosophy for network research.In the 1970s, some theory-based network analysts used structural-functionalist systems theory as foundational theory.For instance, Laumann and Pappi applied Parsons AGIL scheme in their empirical network research on local political networks (Laumann, 1976).
In the 1980s, Wellman and Berkowitz presented the network approach as structural analysis as such and emphasised the close links between network analysis and variants of structuralism (Wellman, 1988).Particularly influential at that time was French structuralism, with its focus on deep structures, but also variants of American structuralism gained popularity in sociology.Harrison White, a prominent figure of the so-called "Harvard Structuralists" (Scott, 2000) became, at least since the publication of "Identity and Control", one of the most important spokespersons of the "new structuralism" or "relationalism" (Mützel, 2010;White, 1992).It was particularly this perspective that influenced the relationalist manifesto, where a new ontology was formulated exclusively based on relations (Emirbayer, 1997).
In this respect relationalism appears as a special form of structuralism which does not include all kinds of structures (such as distribution structures, infrastructures, control structures, symbol structures), but focuses exclusively on relations.A relational structure is a structure that can be represented by links between points.In mathematics, this structure is also called a graph -a set of nodes and edges.In this context the relationalist world view stresses the primacy of relationships over the related.
Promoters of modern relationalism present long lists of renowned theorists and philosophers supporting this peculiar worldview.Emirbayer quotes Marx, Simmel, Cassirer, Dewey, Bentley, Elias, Foucault, Bourdieu, and Luhmann as relationalists.In particular, Karl Marx, he conceives as a "profoundly relational thinker" (Emirbayer, 1997, p. 297) and bases this view essentially on two Marx quotes: one proposition in "Grundrisse" states that "society does not consist of individuals, but expresses the sum of interrelations, the relations within which these individuals stand".Another statement in Marx's 6 th Thesis on Feuerbach is: "[b]ut the essence of man is no abstraction inherent in each single individual.In reality, it is the ensemble of the social relations." A further key witness for Emirbayer is Cassirer, a student of Heidegger.His book "Concept of substance and function concept" is quoted with the paragraph: Things "are not assumed as independent existences present anterior to any relation, but ... gain their whole being... first in and with the relations which are predicated of them.Such 'thing' are terms of relations, and as such can never be given in isolation but only in an ideal community with each other" (Cassirer, 2000, p. 36).
One of the most stylised arguments is taken from Harrison White's "Identity and Control" where it is stated: "[t]here is no tidy atom and no embracing world, only complex striations, long strings reptating as in a polymer goo" (White, 1992, p. 4).Curiously, even Luhmann is assigned to this new philosophy of networks since his system theory, in fact, reduces social relations exclusively to relations of communication (Luhmann, 2007).
However, the idea that the world only consists of relations, and that there are relations without relata is rather mysterious.While it is conceivable in the cultural sphere that meanings arise as nodes in semantic networks, so to speak "from nowhere", this rela-tions-without-relata-idea is hardly generalisable.Perhaps this is the reason why in particular the sociological and cultural studies versions of relationalism put their focus on the cultural level of analysis (Fuhse, 2010).But this is only one of the various streams of this new X-ism.
Emphasis on social relations has also been stressed by the historical sociology version of relationalism in which Elias, Mann and Tilly are seen as forerunners (Hobden, 2001).In this perspective, Daniel Nexon has placed the process-based nature of social development moves in the forefront: "Relationalists [...] adopt a scientific ontology that privileges processes over sub stances.A focus on any form of social transaction necessarily gives priority to pro cesses rather than 'things', since transactions are dynamics exchanges rather than entities.But relationalism implies a stronger claim about the fundamental priority of processes" (Nexon, 2009).
Andrew Abbott delineates this view from the mechanism approach, which usually would require an agent or the intentional actions of an individual (Abbott, 2007).
"By contrast, the relational view presupposes an act, usually a thing that was done, and a scene, usually a set of connections in social time and social space that create the concentric and crosscutting loci for action.Moreover, the relational view makes the scene the dominant term of the pair.For the relational account defines an act as a making of relations within a scene" (Abbott 2007, p. 9).
The relational approach would explicitly seek a process-related understanding of social affairs, in which all social entities -the results of action, actors and relationships -are endogenised (Abbott, 2007, p. 19).
While the above quotation from Marx implies a multidimensional concept of structure, in which conditions may include both material conditions and relationships, the type of relationalism that was sketched in the last paragraphs implies in contrast a radicalisation in the sense that all structures and relations are reduced to interaction structures.Within the spectrum of quantitative and graph theory-based network research, relationalism thus is focusing on a very specific type of relation with the implication that other social structures and levels do not play important roles in the constitution of society.But, as we will see in the next section, interaction is only a special type of relationship.

The Diversity of Social Relations
There is a wide range of contents and formal characteristics in social relations.Many relational structures in modern societies overlap to such a degree that they form a "relation box" which can be represented by a stack of matrices, in which each matrix represents one type of social relation (Winship, 1988).Both Weber and Simmel, the classics of German sociology, conceive multiple social relationships as struts and pillars in the constitution of society.In the chapter on social relations of his main work Economy and Society, he listed a number of concrete relations that are effective in society which include fighting, hostility, sexual love, friendship, filial piety, market exchange, contract compliance, competition, class community and a national sense of community (Weber, 1980).It is well known that, in Simmel's perspective, the intersection of social circles plays a fundamental role in the constitution and integration of society (Simmel, 1910).He was one of the first scholars to think of society as a network, as a collection of multiple relations (Simmel, 1910, p. 391).Influenced by this formal perspective, Leopold von Wiese even proposed a sociological system in the 1930s that was based almost purely on the analysis of social relations.However, this perspective only received limited recognition in Germany and remained almost completely unknown abroad (Wiese, 1933).
Social studies centring on the analysis of relations only gained currency in social network analysis since the 1970s, and in this context a great variety of relationships have been studied during recent decades.Various textbooks provide overviews on the great variety of relationships that are studied in this analytical perspective (Knoke, 1982;Knoke, 2008;Wasserman, 1994).One of the most recent typologies of relations has been presented by a research group around Borgatti which differentiates at the level of basic types between similarity relations, social relationships, flows and interactions (Borgatti, 2009).
Based on the bibliography above, Leifeld (2007) presented interesting data on all different types of relationships that were explored in the above-mentioned quantitative policy network studies.In most studies, data on information exchange, contact, cooperation and influence reputation were collected.Less studied are relations exchange of resources, alliances, common interest, conflict, competition and trust.Relationships on common beliefs were the least investigated at that time.Recently, however, these relations have moved under the heading "discourse network analysis" increasingly into the focus of public policy analysis (Janning, 2009;Leifeld, 2013;Schneider, 2013).
It is important to emphasise that different types of relations are often based on different functional mechanisms and causal structures.For instance, the interaction of armed groups is based on a completely different process than the generation of trust between persons.Even interactions may follow quite diverse logics, if we consider the differences between coercion, competition and bargaining (Scharpf, 2000).Energy flows differ completely from communication flows, the transmission of information.Even within communication flows there may be different logics at work when it is based on "open channels" in contrast to "closed conduits" (Owen- Smith, 2004).And information infrastructures are not the same as information flows.Pathways or channels are usually opportunity structures, not the information itself.For instance, membership in working bodies and policy-making committees provide opportunities for interactions, but these meeting bodies are not interactions themselves (Brandes, 2009;Leifeld, 2012).

The Flaws of Radical Relationalism
In light of the diversity of relations emphasised above, it is highly problematic to conceive only one type of relations -interactions -to be the tissue of social life.And it is even more problematic to conceptualise relations without relata as was suggested by Cassirer and White.However, at this point it should be mentioned that in the paragraph that was cited by Emirbayer, Cassirer does not make an ontological statement about the world as such, but refers only to symbolic structures in the area of number theory.Never-theless, Hartmann called such a perspective in which relations exist without relata a "pure relationalism," that ultimately would boil down to nonsense (Hartmann, 1964).
"There are relations of relations, in which relata are themselves configurations of relations.And because relations are most likely to be rationally comprehensible and expressible in the structure of reality, there is a tendency of relationalism to resolve everything that exists in relations.In this way one gets pure relationalism in which the successive stages of a relationship exist without a base point of relatedness, i.e. without final relata.The world is then one big spider web of relationships, without the entities that are related.This immense nonsense is contrasted with the counter term of the relation, the substrate.A relation presupposes a relatum.Relata in this sense are the substrates of the relation" [transl.by VS] (Hartmann, 1964, p. 214). 1f course, relations can have an immense impact on their relata and may even cause them to transform or even to merge.This is, for instance, the central theme in the complexity theory advanced by Stuart Kauffman (1996).Self-organisation of the world emerged out of chemical reaction networks.The network as a metabolic graph is a kind of opportunity structure for catalysis and reaction of different substrates, creating new entities and organisational levels.But even here initial substrates form the starting points.
How problematic it is to conceptualise relations without relata can be demonstrated by the emergence of the modern nation-state.Since its formation, it has been involved in an increasingly dense web of international relations.Inspired by sociological relationalism, Jackson and Nexon accused current political science theories in International Relations to misapprehend modern state-formation in an essentialist perspective (Jackson, 1999).In contrast, the relational view would give a more convincing explanation of the rise of the modern nation-state, where a specific constellation of interaction would have produced the modern nation-state as a new form of organisation.
It is interesting that Nexon modified his perspective about 10 years later.In a detailed reconstruction of the historical development of European nation-states after the Reformation in the 16 th Century he discussed a whole range of relations and relata -cultural, economic and political relations, as well as rulers, elites and the public (Nexon, 2011).His approach ultimately tries to explain how transnational links were the relational infrastructure of the spread of Reformation and Counter-Reformation.These relations generated identity relationships within and between the pre-state segments that ultimately led to new border relations between different segments.But even here, each time-specific relationship was connected to specific nodes, and the latter did not emerge from the former.Complex relationships exerted transformative effects on their relata which again pro-duced new relata and new relations.In addition to the rulers there were local intermediary powers, elites and populations as diverse nodes in a complex transnational network that transformed the prior composite political regime into territorial nation-states.The core idea of radical relationalism that relations exist independently of their relata is thus ultimately untenable.Relations always need relata as a basis for interconnection (Bunge, 2000).
The last mentioned version, including relation without relata, might be called "pure" or "radical relationalism".This extreme position has to be delineated from a moderate version of relationalism in which the entities exist on their own right.Relata precede relations or at least they occur at the same time.The emergence of a relationship then causes changes of state within entities, i.e. relations are changing their relata.This can happen directly, for instance by information exchange causing preference changes within actors, or indirectly, for example, by similarities between relata (homophily) which trigger willingness to share and exchange information.In a further step, these information flows can in turn be transformative for the relata in question, for instance by learning processes.It is also conceivable that an existing relationship, such as an institutional infrastructure, facilitates the development of these relations.Multiplex networks in which several relations exist in a parallel way may invoke different causal chains to act and interact simultaneously.In the analysis of such networks it is important to stress the diversity of such mechanisms.

Conclusion
If relationalism refers to an ontology in which relations alone count in reality and nodes or "relata" are only structural effects without any self-efficacy, it is ultimately a similar form of reductionism as we have encountered in individualism, holism or institutionalism.
Human behaviour and social processes, however, cannot purely be explained by the self-efficacy of individuals and organisations, yet only through the institutional control systems or relational structures in which they are embedded.Ultimately, it is difficult -if not logically impossible -to imagine a pre-existing relation without relata.Much more convincing is the idea of relational configurations in which an altered macro constellation produces new relata, dissolves old ones or transforms existing relata into new entities.In the above mentioned example a certain macro-political constellation of international relations creates new political entities such as nation-states.These new units never arise ex nihilo, but are in most cases transformations or "recombines" of pre-existing traditional political structures.This is a central idea of the theory of evolution.Yet, such transformations can be quite radical and can result in new emergent entities.These, however, are never created from scratch, but are always modifications of things that already exist.
Another conclusion of these considerations is that one should not talk in an abstract manner about "relationships" or "interactions".There is a huge range of relations that are associated with completely different mechanisms.How these diverse relations interact both in nature and society is still a largely unexplored research area.Society is multiplex or "pluri-relational" and there are many different relations between heterogeneous relata in society at the same time.There are also, as shown above, many levels of social structure, which cannot be described in relational structures alone.All this is still a great challenge for theory building.
Relationalism as a background philosophy, as a form of "catch-up theorising" of network studies in general and policy network research in particular is therefore of limited use.Relations in public policy are ubiquitous and heterogeneous.It is hard to imagine a universal theory of relations, since their mechanisms are often completely different.Theories always have to be based on specific mechanisms and operative logics of things and processes.

Figure 2 .
Figure 2. The epistemic space of science (own source)

195Fisher
Figure 3. Network Metaphors (own charts and pictures)