Social network analysis is a methodology designed for the study of social relations, and relationalism is a theoretical framework based on the primacy of relations rather than actors. Clearly, what we have here is a match made in heaven! The elective affinities between these two schools of thought are strong and together they present an extremely promising framework for exploring the tangled, dynamic complexities that constitute social life. Yet, the embrace between the two has been somewhat half-hearted on both sides. And indeed, there are conceptual inconsistencies between relationalism and social network analysis that pose challenging theoretical problems with the potential to undermine the coherence of such a combined approach and have led researchers on both sides of the equation to shy away from each other. In the following, I consider some of the issues raised by combining relationalism and social network analysis , the latter of which carries its own theoretical baggage. It is consistent with the theoretical frameworks of both relationalism and social networks to suggest that the process of bridging across distinct actors, and all the inconsistencies and differences that are uncovered in such a process, is what makes relationships generative. The process of addressing or attempting to reconcile those differences produces new, potentially innovative combinations of elements and negotiated settlements. Thus, working through the inconsistencies raised by a relational network analysis may be the one of the more fruitful paths for further development of both relational theory and social networks analysis in the social sciences. Ultimately, I would argue that relationalism can be entirely consistent with social network analysis. It just suggests a certain type of networks analysis , one that is dynamic, open to contingency, and concerned with the cultural, social, and historical context of social structural patterns. Luckily, network science, in and outside of the field of sociology, has been moving quickly toward evolving, rather than static, conceptualizations of social networks and developing new methods that can be fruitfully applied to further the aims of relational research.

While this volume presents many different definitions of relationalism, I want to quickly outline my perspective on what is a vast and contended area of theoretical inquiry before addressing the links to social network analysis . My understanding of relationalism is that its origin lies in the pragmatist philosophical tradition and their reaction to Kant and his influence. Thus, I see Charles Peirce , in particular, but also John Dewey , Jane Addams , George Herbert Mead , and William James as important early thinkers in this tradition. While these thinkers were in many respects forging ahead on a path begun by Kant , they were also working to dismantle central elements of his thought. In particular, they were attacking the notion of the a priori, the idea that consciousness predates existence, and the belief in a strict dichotomy between the material and the ideal, which Kant arguably inherited from Descartes . One of the strategies used by the pragmatists to dismantle the dualism between the material and the ideal was to challenge the idea that consciousness exists prior to material existence—or indeed any such notion of the a priori. Instead of prior to, consciousness for the pragmatists arose from experience, meaning also that consciousness was not distinct from experience, as conceived by Kant , but part of experience. For example, in in his essay “How to Make Our Ideas Clear” (1878), Peirce argued that the beliefs we hold about the world follow from thought, which follows from doubt, which follows from the experience. Following Peirce , reason is produced through experience—which is a direct contradiction to Kant’s position that reason is prior to experience and is necessary as a foundation through which the experience of the world is possible. Mead further developed this line of opposition to Kant by arguing that consciousness, and particularly self-consciousness, is the product of experiencing a social environment (Mead 1934, 186–191) . Mead particularly destabilized the notion of the self as an enduring monolithic entity that encounters the world, and instead produced an image of the identity and self-consciousness as emergent properties that result from grappling with the complex dynamics of the lived world.

This line of critique has been central to the relational movement in social thought. A rallying point has been Mustafa Emirbayer’s call to “reject the notion that one can posit discrete pregiven units such as the individual or society as ultimate starting points of sociological research” (Emirbayer 1997, 287). In this way, “relational sociologists treat social phenomena as processes, constituted by flows of action or interaction, which operate immanently to the life of individuals” (Powell and Dépelteau 2013, 2). One focus of relationalism has therefore been to show that the self —or the modern social theoretical equivalent, the actor—is constructed through the shifting web of relations and the dynamic flux of social environments. Margaret Somers , a central voice of relationalism, has argued that social scientists, and particularly scholars of identity , should be analyzing the social constitution of identity through the analyses of external narratives (Somers 1994, 606) . Similarly, Harrison White has argued that “identities spring up out of efforts at control in turbulent contexts” (White 2008, 1). Indeed, for relationalists, the active force does not reside in the actor at all, instead it is the relations between actors that temporarily coalesce into the units that we recognize as actors, such as individuals, communities, organizations, groups, and nation-states—among other social entities.

It is here that we can begin to see the enormous potential that social network analysis holds for pursuing a relational research agenda. This agenda may be causal or descriptive. If social entities are the product of relations, then a means for studying the interactions, transactions , exchange, predispositions, and affectual relations is necessary for understanding how the social objects that populate are world are produced. A relationalist ontology could also make an even more far-reaching claim. If relations between social actors are responsible for the actions, intents, and dispositions of the actors, then social scientists interested in outcomes would be best served by focusing their attention on where the actions that are generative of actors lie in the relationship. In this sense, actors are just a temporary lens for perceiving, but also frequently obscuring, what is the real causal nexus: the intersecting pattern of relations. If relationalists are correct, for example, this implies that causal arguments based in individual units of analysis such as individuals, organizations, or nations are misleading. One might then suggest that researchers could instead use relations as units of analysis; however, given the relational framework , this also does not quite go far enough. The problem here is that a method that compares units, even relations, as if they are independent is also problematic. A relationalist ontology suggests that units are not independent—indeed, they are mutually constitutive. There is even less reason to imagine that relationships are independent than actors, and we should not compare relations as though they are independent in order to conduct, for example, a statistical analysis of predictors of tie formation. And even using more advanced techniques for controlling for the interdependencies between relations or actors falls short of carrying through the theoretical commitments of relational sociology, which instead calls for a type of analysis that is not controlling for relational interlinkages but instead is explicitly focused on analyzing the interlinkages and their effects. The analysis of interlinkages is of course one of the distinctive contributions of social networks to methods in the social sciences. And this analysis may also be performed in a descriptive mode that allows relationalists to directly observe the changing pattern of relationships that for many do not just produce but in fact constitute social life.

It must first be acknowledged that social network research captures an even more diverse field than relational sociology. Many people think of social networks as a method, but the field of social networks is better understood as a research area that encompasses different but related methodological approaches, at least two different theoretical orientations, and a focus on a particular empirical object: the social network. A social network is a set of relations that link actors. The relations and the actors they connect can be represented as a matrix, a list of relations, or as a network visualization. Figure 13.1 represents a social network of information transferred between captains of East India Company ships in an overseas trade network of the seventeenth century (Erikson and Samila 2016).

Fig. 13.1
figure 1

Network of captain-to-captain ties

Full size image

In this network, the nodes represent captains and the arcs linking the nodes represent observable instances in which information about ports was transferred between captains. The arcs are directed, meaning that they represent the direction of the flow of information between captains. Whereas in many network visualizations arrows express the direction of the tie, in this visualization, which was created using Gephi, the curvature of the line represents the direction, where flow goes clockwise.

In some ways, this is a quintessential social network. It captures informal and transient ties between actors that would be difficult to map out or analyze with another method. In other ways, this network is unusual. It captures transactions rather than relationships. Meaning more precisely, the network ties indicate a transfer of information between actors, not a fixed relationship in which one actor identifies the other as friend, partner, confidante, or suchlike. Those transactions may be indicative of more enduring relationships between the captains, but this is an empirically undetermined question. Informal relations between individuals are the bread and butter of social networks and a core substantive concern in social network research. They have been central to social network research since Mark Granovetter’s pioneering work on the importance of weak ties (1973). Social network analysis , however, is more than capable of handling strong ties , weak ties , transactions , dispositions, kinship ties, and many other types of permanent, semi-permanent, and fleeting ties between alters. Network analysis is also able to analyze different types of networks at the same time using multiplex methods.

Networks, however, do not have to link individuals. Extremely interesting social network research has been produced on dolphins, wolves, and tree networks, among others, and a large proportion of research in the field has been devoted to networks of organizations. The restriction that does apply is that the actors and ties must be social in nature. Thus, the analysis of computer networks, which is a very developed field , does not fit the bill. This does not mean, however, that very useful techniques cannot be pulled from the field of computer science and applied to social networks. And indeed, much of the progress of in the last decade of social network research has come from borrowing and applying the sophisticated methods developed in chemistry, computer science, physics, and epidemiology.

Where the purpose of social network analysis was originally a method through which to describe social structure itself, the intent has evolved over time to encompass describing structures, understanding the role of relations in social processes , and understanding the role of the structure of relations in social processes . These are distinct because relationships can be analyzed without reference to the larger structure within which they sit. The field has persuasively established that (1) the existence of relationships has a distinct impact on many social processes of interest to social sciences, and (2) the structure of those relationships also has a distinct and unique impact on social processes of interest. To these ends, the methods of social network research have included clustering analysis, blockmodeling, visualization techniques, community detection algorithms, mathematical modeling, computational modeling, triadic analysis, exponential random graph modeling, and more recently, directed acyclic graph analysis, as well as a host of other more general statistical and qualitative methods of research.

These methods are diverse and come not only from different traditions but also from entirely distinct disciplines. Despite this diversity, social network analysis should not be thought of as merely a grab bag of methods. There are many strong social network theorists and two distinct theoretical traditions (Erikson 2013) . One is relationalism. The other, formalism, has arguably had an even more powerful, though perhaps less clearly articulated, impact on social network analysis . Formalism can be dated back to the early work of Georg Simmel , one of the primogenitors of social network research. Georg Simmel’s work as an urban scholar is widely known and has a long history in sociology. Less well known to date has been Simmel’s legacy as a neo-Kantian philosopher and the impact of this philosophical heritage on his approach to social forms. In his early work, Simmel was explicitly attempting to expand Kant’s work into the realm of the social sciences. Where Kant asked “How is Nature Possible?” Simmel wondered, “How is Society Possible?” (Simmel 2009, 40). For Kant , the answer was that there were certain essential structures of reason—such as conceptions of time, space, and causality —that are prior to our experience of the world and, indeed, make that experience possible. Simmel posited in his essay that social forms make society possible.

Social forms, for Simmel , are “conditions which reside a priori in the elements themselves, through which they combine in reality, into the synthesis, society” (Simmel 1972, 8). Simmel further elaborated on the connection between social forms and Kant’s critical faculties in writing, “the sociological apriorities are likely to have the same twofold significance as those which make nature possible. On the one hand, they more or less completely determine the actual process of sociation as functions or energies of psychological processes. On the other hand, they are the ideational, logical presuppositions for the perfect society” (1972, 9). Thus, these social forms are not produced by social experience, but instead they give form to social experience. Simmel’s social forms include, among others, superordination, subordination, competition, the division of labor, parties, and the act of representing others (i.e. political representation or principal–agent relations).

The idea of social forms is appealing for social network researchers because it provides a powerful and relatively clear way of conceiving of how social network configurations affect social outcomes. Simmel himself provided many early examples of the impact of purely formal structures of relations between individuals, that is, social network patterns. The most famous example is of the triad. As Simmel pointed out, the shift from a dyad , a relationship between two people, to a triad, a relationship between three people, fundamentally changes both the nature of relationships between those actors and the potential for what kinds of patterns of social organization may occur. The dyad contains two possibilities: connection and disconnection. The linked dyad is inherently an intimate connection, since each dyadic actor has only one other link; so, they are tightly bound to their partner. The addition of one new individual, which is the transformation from a dyad to a triad, alters the essential nature of the relationships between individuals. In a fully linked triad, one in which all actors have relationships with all other actors in the triad, the intimacy of the relationship decreases and a sense of belonging to a group is created. These changes occur because actors are not entirely dependent upon one other individual in order to retain a relationship, intimacy decreases, and the existence of the group can withstand the loss of one individual. This condition has the larger, powerful implication that the group has a life that exists outside of the impact of any one actor.

Most interesting, however, is the possibility for social organization and strategy that emerge with the creation of a triad. Actors now have the structural possibility of forming ties with two disconnected individuals, giving them a strategic advantage over both—they can act as a broker between the two, gaining resources through the act of linking and transferring otherwise isolated individuals, and play the two off against each other by, for example, threatening to form an exclusive tie with either one. Simmel referred to this as the strategy of the laughing third, tertius gaudens.

This aspect of Simmel’s thought has been interpreted as part of his larger emphasis on how the number of actors involved in a social process can fundamentally alter its essential nature. Any network researcher, however, will immediately perceive that it is not the number but the way in which the number allows for new structural configurations between the actors that is doing the active work in these examples. It is the alternation of ties with the absence of ties that produces a context for strategic action and organizational hierarchy within even this extremely small group structure.

Social network research has embraced this insight and developed a formal means of evaluating the group structure of triadic configurations with the triadic census.

Figure 13.2 presents the sixteen possible configurations in a directed graph of three actors. The circles represent actors and the arrows represent directed relationships between actors. Thus, resources or affect may pass from one person to the other in a reciprocated or unreciprocated flow. As is evident, just these simple conditions produce a relatively rich array of social possibilities. And researchers have associated different triadic configurations with different local relational types. For example, transitivity in social networks is a property in which friends like their friend’s friends. It is linked to a theory of cognitive dissonance, when individuals try to minimize the disagreements within the set of their close acquaintances (Festinger 1957; Davis 1979). Transitivity is present in and linked to the triadic patterns 030T, 120D, 120U, and 300. Triads 120C and 201 contain transitive patterns but also have intransitive relational patterns within the network of three. Calculating triadic censuses with this theory in mind then can lead to conclusions about relational patterns, the strength of friendships, and the cognitive factors that weigh on relational decisions—quite a lot to extract from the formal properties of a network. And the triadic census is only the start of the different formal relational patterns that can be identified with the tools of network analysis .

Fig. 13.2
figure 2

Directed triadic configurations

Full size image

For Simmel , however, the triadic configurations are not just markers of different relational types—they give form to relations. In doing so they create subjective states for the actors participating in those social relations. For example, the restless shifting and transient interactions of people in metropolitan areas create intelligent, rational, blasé, and calculative individuals. The objective conditions of the number of relationships experienced by individuals create these subjective states and characteristics. This relationship is, in many ways, a researcher’s dream come true, because observable external states can be used to reveal the hidden subjective states of individuals. And this idea—that structure shapes content and the social structure of relationships determines the fundamental characteristics of a society—was embraced by early social network theorists.

This idea also, at first glance, appears to fit entirely with a strong program of relationalism. Here we have a concrete means by which relations determine actors, their subjectivities, and their actions. If the goal of relationalism is to dissolve actors into relations, then this is a potentially excellent path to follow. Yet, the problem is the very concreteness of the relations and their patterns. Relationalism is not only about dissolving actors. Relational theorists have been most explicit that they are critiquing the very notion of the a priori as well as the idea of a strict dichotomy between the material and ideal (Emirbayer 1997, 287; Somers 1994, 605, 621, 628, 2008, 205, 1998, 751; White 2008, 1; Powell and Dépelteau 2013). Fixed and determinate social objects are anathema to the creative and dynamic flux that makes up a relational ontology . And the idea that social forms exist in a different space than the interactions they make possible is also problematic. These aspects of relationalism render a wholehearted embrace of much formal social network analysis a challenging and potentially off-putting project for relational sociology.

In theorizing social forms, Simmel explicitly transposed the ideal a priori of Kant into the realm of the social. Within his framework, these should be understood a concrete, crystalline structures that have a determinative impact on social life. Simmel took a different line in his turn to vitalism, but this later work had much less of an impact on research in social network analysis . The social forms are a social a priori, and network analysts have understood them in that sense. Thus, social network researchers have produced methods for identifying different relational configurations , with the idea that they will have the same impact across contexts. A prominent example of this is the research on structural holes. Structural holes are in some ways a generalization of the brokerage advantage of Simmel’s laughing third. The structural hole concept is based on a measure of relational dependence, which weighs the dependency of individuals on their alters against the connectedness of their alters’ connections to each other (Burt 1995). In practice, the measure is similar to but more complex than ego network density—in other words, the number of ties within one individuals circle of acquaintances relative to the possible number of ties within the actor’s circle of friends. A similar idea is captured by the clustering coefficient (Watts and Strogatz 1998).

Researchers using the structural holes concept put a high value on low ego density because occupying this type of structural position places individuals in an advantaged location from which they can control the flow of resources and information between others in their network—or even possibly piece bits of existing information together into new ways, thus serving as a source of innovation (Burt 2004). While the universality of the impact of structural holes has been challenged, the impetus behind the research was to reveal the impact of structure by demonstrating the consistent “vision advantage” that accrues to individuals who occupy positions characterized by low local density. The success of this research program, which has been very generative, rests largely on the strength of the universality of the link between structural holes and innovation , which is understood to work across contexts. The concept of structural holes is not an isolated incidence in network analysis ; brokerage positions, centrality, density, weak ties , and strong ties have all been attributed with generalizable properties throughout the literature. Indeed, the powerful appeal of network research often lies in the idea that it is possible to pull abstract configurations out of messy contingent circumstances in order to get to the heart of what social processes are driving outcomes. The problem of course is that instead of essentializing individuals and falsely treating them as fixed entities, the strong formalism of social network analysis can be guilty of universalizing relational patterns as fixed, pregiven entities. Here we have a direct contradiction with the central tenets of relationalism. And this contradiction, I would argue, is the core of the tension that has arisen between social network research and a relational sociology.

The tension is not a submerged or hypothetical problem. It has been articulated on a number of occasions in various ways, as criticism of the atheoretical nature of network analysis (Burt 1980; Emirbayer and Goodwin 1994; Granovetter 1979; Mitchell 1979; Rogers 1987), but also more recently in a focused critique of the neglect of problems of culture and meaning (Fuhse 2009; Fuhse and Mützel 2011; Mische 2011; Pachucki and Breiger 2010; McLean 2016) . Fortunately, there are paths forward for pursuing a relational social network analysis .

Two important issues that immediately stand out and have long served as rallying cries for relational theorists that are also invested in the tremendous potential network analysis holds for advancing a relational perspective are the twin concerns of context and content. Following Simmel , a lot of formal network analysis has tried to find structural constants. Simmel suggested this approach when he wrote, “In sociology, the object abstracted from reality may be examined in regard to laws entirely inhering in the objective nature of the elements. These laws must be sharply distinguished from any spatio-temporal realization” (1972, 28–9). Early social network analysis and structural theory was mainly focused on establishing a universal principle that patterns of social networks consistently mattered. This principle is clearly not inconsistent with relationalism. The danger arises when there is an attempt to link a specific network pattern with a consistent outcome—that is, a universal law. To take a more established example, Ivan Chase’s work on dominance hierarchies in animal behavior was very influential for early network researchers because he demonstrated that dominance hierarchies emerge from and are based on interactional patterns—not the attributes of those interacting (1980). So far, this is consistent with relationalism. The contingencies of interactions are centrally responsible for things that we perceive as durable structures. However, Chase’s work was also received in a way that lent itself to the assumption that social interactions universally lead to dominance hierarchies. There was certainly a strong emphasis on status-ranking in social network research in the closing decades of the twentieth century. Some form of social organization is necessary to group life, but cooperative relations are an alternative to dominance hierarchies that are also observed among animals. Yet, cooperative relations are less likely to emerge under common laboratory conditions in which animals compete for resources, are forcibly grouped and regrouped, and lack the freedom to exit from the setting (Estevez et al. 2007). Thus, conditions in many experiments can lead to situations in which dominance relations are more likely than cooperation, which is an important fact that may be ignored if the context of the laboratory is taken for granted.

Relationalism explicitly recognizes the importance of context and consistently works to bring it into the analysis—the underlying assumption being that relationships are not fixed and isolated determinative forms but instead fit into a more complex set of larger circumstance that shape the impact of as well as the structure of relations. The real problem is to draw boundaries on the relevant context. Historical network research provides an excellent example of successful relational work by embedding network structures into specific socio-economic and cultural circumstances, such as moments of imperial and national development and global trade settings (Adams 1996; Barkey 2008; Bearman 1993; Erikson 2013; Gould 1995; Somers 1993) . For example, network research on state development has revealed consistent support for the importance of associational networks in the state formation process. There is evidence that associational ties were crucial to state development in the Meiji Empire (Ikegami 2005), the British Empire (Olson 1992), and Switzerland (Wimmer 2011). These case studies are in part so impressive because they not only show that associational ties were important but also detail which groups were linked and how relations between these specific groups contributed to the unique trajectory of state formation in their respective areas. Thus, the emphasis on the importance of the networks in the works is not easily separated from a detailed investigation of the empirical framework.

We can see in work showing that informal ties between congressmen in the early history of the United States depended upon the widespread custom of bipartisan co-residence at boarding houses in Washington DC (Parigi and Bergemann 2016) that crucial contextual details about local culture play incredibly important roles in stories such as these. In more contemporary circumstances, it has been demonstrated that both the structure and outcome of network ties are powerfully affected by the resource levels of a community (Desmond 2012). Differences in organizational task and topic studied in classroom settings alter network effects (McFarland 2001). And larger social and cultural orientations to legal institutions impact network structure and function (Kirk and Papachristos 2011).

There are also specific methodological advances that have been designed in order to allow network researchers to incorporate context. I do not recommend these methods as a substitute for attention to culture and history—but I do recommend that relational sociologists take advantage of new methods of multilevel network analysis . Much the same as multilevel statistical analysis, multilevel network analysis allows for social networks between individuals to be embedded within networks of links between higher levels of social organization , civic action groups, organizations, or even states, among others. Multilevel modeling recently applied to the networks of elite French cancer researchers has shown that productivity and productive strategies are simultaneously linked to position in a network, position in an organization, and the organization’s position in the larger inter-organizational network (Lazega et al. 2008). Emannuelle Lazega and Tom Snijders have published an entire volume devoted to the description of various techniques of multilevel network modeling, which all hold significant potential for researchers attempting to include context in their work—as would be consistent with a relationalist perspective (Lazega and Snijders 2015).

Relationalists—and others—have also long bemoaned the tendency in social network analysis to treat all ties as exactly the same: binary objects with two states, on or off (Zuckerman 2008). This tendency again has roots in Simmel’s legacy, as he hoped to bracket out structural from cultural forms with the intention of identifying social configurations that produced consistent outcomes across different cultural circumstances—a little like clear pipes that could be filled with different colored liquids. The configuration of the pipes does the important work, even if the contents produce superficial differences in their appearance. But even in Simmel , we can see a tension between the effect of the social form and the affective contents of that form. For example, tertius gaudens, the laughing third that profits from the disconnect between friends has a structural equivalent in tertius iungens, the one who joins. The tertius iungens attempts to bridge parties in order to facilitate the flow of information between them and is also associated with the innovation that results from bridging disconnected areas of knowledge. The tertius iungens, however, is attempting to bring parties closer together, in the way that a mediator may provide a non-partisan link across warring parties (Obstfeld 2005). The form is the same, but the intent of the parties and effect of relationships are different—with consequential impacts on the nature of the relationship and very possibly the rate and quality of innovation that results from such partnerships. In this case, the contents of the ties—rather than their form—shape outcomes in distinctive and important ways, suggesting that form is not so easily divided from contents.

The sensitivity of relationalists to the contents of ties clearly improves analysis in such circumstances. Social network research has shown the central relevance of relational contents in settings that range from Renaissance political patronage systems to Brazilian youth activist meetings (McLean 2007; Mische 2009). This orientation is consistent with the core tenets of relationalism because it dissolves what they see as an artificial boundary between an abstract notion of structure (e.g. social forms) and the messy and meaning-leaden realities of everyday interactions. For relationalists, it is not just that the contents of ties should be studied as well as the structure of ties—it is that the contents are the ties, creating and sustaining relationships and through their intricacies and biases also shaping the structure of ties and the outcomes they produce. This means that, for example, a friendship does not call forth a certain set of interactions. Instead, a repeated exchange of acts of kindness, consideration, and respect make up that thing which we refer to—in an overly reified manner—as a friendship.

Here again there is a specific branch of network methodology that is still advancing but can provide researchers with concrete methods for dealing with distinct relational types. Networks with multiple relational types are called multiplex networks. Blockmodeling was for many years the primary technique for dealing with multiplexity, and early papers such as John Padgett and Christopher Ansell’s work on the rise of the Medici have long since established that the layering of different types of ties can be crucial to understanding important historical processes. This perspective and an array of analytical strategies are also on display in John Padgett and Walter Powell’s more recent book The Emergence of Markets and Organizations (2012). Multiplexity has been incorporated into advanced techniques for community detection (Mucha et al. 2010), triadic analysis (Cozzo et al. 2015), and a host of other sophisticated techniques by researchers working across disciplines.

Beyond these established calls for more contents and context, I would also suggest that relationalists should focus more on interactions than relationships when conducting social network analysis . This raises the question of how we conceptualize and operationalize ties. Discerning what constitutes a tie in social network research is a much more difficult and involved task than is recognized by people outside of the field . In the past, ties were often constructed from questionnaires asking individuals about their relationships. But we have a choice of conceptualizing ties as fixed psychological dispositions to others (such as a marriage, a friendship, or a family tie) or as interactions (such as sending an email, loaning some money, sharing a room, or engaging in some light conversation). What is at stake is an underlying presumption as to whether the relationships shape the interactions or the interactions produce and shape the relationships. Given the strong turn away from fixed a priori elements in relationalism, I would argue that it is important to embrace a conceptualization of ties based in the messy realities of everyday interactions. And there is clear support of this position in the relational literature. For example, Jan Fuhse emphasizes the actions taken by individuals in his description of how patterns of social activity emerge over time: “social structures result from this doing rather than from individual dispositions and attributes” (Fuhse 2009, 187) . Indeed, relationalists may think of social structures as merely a different way of referring to the cumulation of past actions, not as something that is in fact meaningfully different or which stands apart from these empirical phenomena.

In practice, this approach to conceptualizing ties has two implications for research design. The first is that it greatly behooves relational network researchers to gather behavioral data on actual interactions. This means moving away from surveys and name generators and moving toward archival or observational data. There have always been tremendous amounts of data on interactions sitting in the historical record. Family histories, financial records, credit arrangements, and citation patterns are all both abundant and important records of interactions and exchanges between individuals in many different cultures and contexts, and make up only a small number of examples of the kinds of interactional data that can be produced from inventive use of archives and commercial and state documentation. These types of records have been supplemented with an almost overwhelming bounty of data on online interactions. And researchers have not neglected the opportunity of putting new mobile technologies to work in tracking physical interactions in real-time. All of these data, and indeed the very nature of interactions, which are fleeting and temporally situated, suggest the second implication, which is that relationalists should embrace a dynamic approach to network analysis . This, of course, is more easily said than done as it is much simpler to identify a structural pattern in a stable network than to pull regularities out of a complex and shifting mass of interdependencies. On top of which, dynamic networks raise difficult questions about the nature of ties that many people grapple with on a regular basis and that have indeed plagued humanity for some time—such as how do you know when a friendship is over? Or when does love begin?

Despite the complexities, there has been significant work in advancing techniques to manage continuously evolving networks. Ronald Breiger, Kathleen Carley, and Philippa Pattison produced an edited volume that contains both helpful reviews and introductions to new techniques and advances (2003). The spread of sentiment over time has been modeled by James Fowler and Nicholas Christakis (2008), visualization techniques for dynamic networks have been explored by Jim Moody, Daniel McFarland, and Skye Bender-Moll (2005), while Gueorgi Kossinets and Duncan Watts have addressed quantitative approaches to evolving network structures (2006). Computational modeling of networks also holds tremendous potential for piecing out the impact of and regularities in interactional patterns as they unfold over time. And these works are but a small sample of the progress that has been made in this area over the last fifteen years or so.

Implicit in these suggestions is, I hope, the larger recommendation that relational researchers do not give up on using the most sophisticated tools of network analysis . This is an area where relationalist network analysis has failed to live up to its full potential. While a significant literature of relational work has employed advanced techniques of network modeling, there is a tendency for relationalists to avoid deploying the full battery of network analytical techniques or to treat networks as more of an analytical concept than a methodological tool. I would encourage relational researchers to embrace the new techniques in network science. Indeed, the influx of dynamic data has led to a convergence with relational approaches, as researchers are very much knee-deep in vast reams of temporally shifting, layered, and contingent interactional data . Thus, embracing the advance guard of these techniques can only help a relationalist agenda.

Finally, I want to make a strong plea for relationalists to keep the importance of structure and pattern in mind. There is a tendency in relationalism, as I have argued elsewhere (Erikson 2017), to focus energies on dissolving the actor into its constituent parts of relations. While I certainly do not object to this goal, I fear that too much focus on the individual will shift attention away from one of the truly innovative and important contributions that have been made by network theorists over time (which is of course also one that underscores the central importance of actors and their intentions in social processes ): that the patterns of ties, or the network structure of relations, between actors has an independent effect on social outcomes that we as researchers need to take into account. If relationalists focus too much on how relations construct individual actors, we miss the bigger social outcomes that take place at the level of the collectivity and that very much rely on relations and their structure. This includes everything from neighborhood segregation, political polarization, market volatility, and crime waves to world trade patterns and disease outbreaks. Social networks affect these important outcomes. To properly understand how they do so, we need to avoid treating them as static, reified objects, an approach that does not reflect the reality of networks as they exist in time. Relational theory gives us a framework from which to approach this task. It has the potential not just to benefit from the field of social networks but to help advance the field as well