2.1 Emphasis on Social Relations

Certainly, the most basic emphasis in social network research is on the importance of social relationships (Reference FreemanFreeman, 2004). Relationships can be conceptualized in terms of pipes through which resources, such as knowledge, flow; and in terms of prisms through which people’s reputations are discerned: you are known by the people you are perceived to be connected to (Reference Kilduff and KrackhardtKilduff & Krackhardt, 1994; Reference Audia and GrevePodolny, 2001). Thus, while brokers can gain advantage from spanning across the gaps in social structure, they also have to be mindful of their reputations in the minds of those whose support they need for the pursuit of initiatives within their organizations (Reference Reay, Goodrick and D’AunnoPodolny & Baron, 1997). Being seen to be connected to quite disparate groups can damage reputation (Reference ZuckermanZuckerman, 1999).

Relationships between organizational actors include positive ties such as friendship (Reference Freeman, Harrison and ZyglidopoulosTasselli & Kilduff, 2018), advice (Reference KrackhardtKrackhardt, 1990), and knowledge exchange (Reference Zietsma, Toubiana, Voronov and RobertsTsai, 2001); but also include negative ties such as hindrance (Reference Clarke, Richter and KilduffClarke, Richter, & Kilduff, 2021) and a preference for avoiding coworkers (Reference Labianca, Brass and GrayLabianca, Brass, & Gray, 1998). Together these positive and negative ties constitute the social capital that is potentially available to an actor, defined as the goodwill inherent in the structure and content of social relations (Reference Adler and KwonAdler & Kwon, 2002: 18). As James Reference ColemanColeman (1988: 108) noted, ties such as friendship and acquaintanceship can be appropriated for other purposes such as getting a job (Reference Fernandez, Castilla and MooreFernandez, Castilla, & Moore, 2000; Reference GranovetterGranovetter, 1973) or facilitating team performance (Reference Clarke, Richter and KilduffClarke et al., 2021). Ties to high status contacts can facilitate career advancement (Reference LinLin, 2001), whereas ties to leaders who are peripheral in their advice networks are detrimental to an actor’s level of influence (Reference Zietsma, Toubiana, Voronov and RobertsSparrowe & Liden, 2005). People who have many negative ties, relative to those who have few, are more likely to harm others and be the target of harm from others (Reference Hoffman and JenningsVenkataramani & Dalal, 2007).

In the modern evolution of network research, the emphasis on social relationships has expanded to include relationship change (Reference Audia and GreveRivera et al., 2010). Researchers have responded to the critiques concerning the neglect of network change (Reference Emirbayer and GoodwinEmirbayer & Goodwin, 1994) with an increasing interest in network dynamics (Reference Chen, Mehra, Tasselli and BorgattiChen et al., 2022). Informal network connections are dynamic in the sense that social relationships shift and change over time as new technology is introduced in organizations (Reference BarleyBarley, 1990; Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010), as new management is appointed (Reference Burt, Ronchi, Wessie and FlapBurt & Ronchi, 1990), as new people are hired (Reference CarleyCarley, 1991), and as people are promoted (Reference Reay, Goodrick and D’AunnoPodolny & Baron, 1997). As networks of relationships change, our chances of becoming happy (Reference Fowler and ChristakisFowler & Christakis, 2008), depressed (Reference Freeman, Harrison and ZyglidopoulosRosenquist, Fowler, & Christakis, 2011), or obese (Reference Christakis and FowlerChristakis & Fowler, 2007) also change.

In their everyday work lives, people have frequent opportunities to expand their networks by meeting new people but, it seems, relatively few people take advantage of social occasions to forge new relationships (Reference Ingram and MorrisIngram & Morris, 2007). Patterns of relationships such as friendship stabilize relatively quickly within a bounded social system (such as a student living group – Reference ZhaoNewcomb, 1961) but under the surface there is likely to be considerable movement. Some actors form stable relations but others “dance between friends throughout the observation period” (Reference Hoffman and JenningsMoody, McFarland, & Bender-deMoll, 2005: 1229).

The amount of churn individuals experience in their personal relations with others may derive in part from differences in underlying personality related to the ease with which individuals manage impressions and social relationships (Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010). Similarly, individuals with a propensity to engage in brokerage (Reference Burt, Jannotta and MahoneyBurt, Jannotta, & Mahoney, 1998) are likely to experience considerable network change given that brokers trade across gaps in social structure (i.e., structural holes – Reference BurtBurt, 1992) and these gaps are subject to rapid decay in competitive organizations (Reference BurtBurt, 2002). Figure 1 illustrates some possibilities of how brokerage opportunities (structural holes) can expand, remain the same, or close over time. Relations between brokers and the unconnected parties to whom they offer a service tend to be fragile in part because of distrust of brokers who benefit from others’ communication difficulties (Reference Reay, Goodrick and D’AunnoStovel, Golub, & Milgrom, 2011).

Figure 1 Brokerage opportunities change or remain the same over time.

2.2 Embeddedness

An influential review declared that “embeddedness in social networks is increasingly seen as a root cause of human achievement, social stratification, and actor behavior” (Reference Audia and GreveRivera et al., 2010: 91). Embeddedness refers to the overlap between social ties and economic ties; or the nesting of social ties within other ties (Reference Kilduff and BrassKilduff & Brass, 2010). People are embedded to the extent that they show a preference for economic transactions with fellow network members (Reference GranovetterGranovetter, 1985). Embeddedness, including reliance on favored contacts for buying and selling, is important to the extent that markets are inefficient (Reference BurtBurt, 1992), but even in markets reputed to be highly efficient people tend to neglect interpersonal relationships at their peril (Reference Abolafia and KilduffAbolafia & Kilduff, 1988; Reference BakerBaker, 1984). Social ties are forged, renewed, and extended through the network rather than through actors outside the network (Reference Audia and GreveUzzi, 1996). Social connections between people that exist at one point in time tend to be repeated in the future (Reference Audia and GreveRivera et al., 2010: 100).

People develop embeddedness when they overlay one type of connection with another type of connection – that is, when they develop multiplex ties in which more than one relationship is involved. Partners in law firms mitigate problems of status competition among their coworkers by developing multiplex ties of advice and friendship (Reference Lazega and PattisonLazega & Pattison, 1999). Leaders of teams develop multiplex ties when they form both advice and friendship links with team members. These ties help leaders improve team performance when team social capital is otherwise impoverished (Reference Clarke, Richter and KilduffClarke et al., 2021). People who are embedded (in terms of having large, dense, and high-quality relationships with colleagues in the workplace) tend to believe that the organization values their contributions and cares about their well-being (Reference Hayton, Carnabuci and EisenbergerHayton, Carnabuci, & Eisenberger, 2012). Individuals who are embedded in dense groups also tend to engage in interpersonal citizenship behaviors in organizations (Reference Chung, Park, Moon and OhChung et al., 2011). But a pair of individuals tends to be less creative to the extent that their dyadic relationship is embedded within a dense network of common third parties – social pressure inhibits creative expression (Reference Lounsbury and GlynnSosa, 2011).

Firm embeddedness refers to social ties among business owners within a community. This type of embeddedness both constrains and enables firm-level outcomes (Reference Reay, Goodrick and D’AunnoUzzi, 1997). On the plus side, social ties in a region create channels for contacts among managers and employees of firms, making it easier for firms to obtain knowledge about opportunities in foreign markets. Business relations embedded in social relations tend to affect outcomes in transitional economies relative to market economies (Reference Luk, Yau, Sin, Tse, Chow and LeeLuk et al., 2008). Managers rely more on relational ties as asset specificity and uncertainty increase (Reference Zhou, Poppo and YangZhou, Poppo, & Yang, 2008). But as ties become denser, there is an increasing likelihood that firms will interact only with local actors rather than pursuing foreign markets (Reference Laursen, Masciarelli and PrencipeLaursen, Masciarelli, & Prencipe, 2012). At the same time, potentially lucrative opportunities for entrepreneurs lie beyond embeddedness within their international communication networks (Reference EllisEllis, 2011).

Embeddedness inhibits opportunism according to social capital theory (Reference GranovetterGranovetter, 1985). But the effects of embeddedness may be culturally contingent. A study of 192 international joint ventures found that, within collectivist versus individualist cultures, embeddedness, in the form of interparty attachments and boundary-spanning ties, was a stronger inhibitor of opportunism (Reference LuoLuo, 2007). A related finding is that, among managers, affect and cognition-based trust are more intertwined in the collectivist culture of China relative to the individualist culture of the USA (Reference Chua, Morris and IngramChua, Morris, & Ingram, 2009). What happens when West meets East? For partnerships between Western-based and Eastern-based firms, commitment to further exchanges predicts export performance and is itself driven by the reciprocal, reinforcing cycle of each partner’s perception of the other’s commitment (Reference O’BrienStyles, Patterson, & Ahmed, 2008).

Because organizations are dependent on each other for resources (Reference Zietsma, Toubiana, Voronov and RobertsPfeffer & Salancik, 1978), they form alliances that help them survive in competitive markets. In knowledge-intensive industries such as biotechnology, firms embedded in collaborative alliances benefit from knowledge exchange that promotes learning and firm expansion (Reference ZhaoPowell et al., 1996). In these networked organizations (Powell, 1990), innovation happens in the interstices between firms rather than from internal research and development (Reference FurnariFurnari, 2014). These cross-unit collaborations provide benefits to multinational organizations to the extent that they organize themselves as collaborative networks (Reference Ghoshal and BartlettGhoshal & Bartlett, 1990).

2.3 Structural Patterning

Following on from the discussion of embeddedness, the third leading idea that distinguishes the social network research program concerns structural patterning – the notion that beneath the complexity of social relations are enduring patterns that can be discovered through analysis to show, for example, how actors cluster together or how networks are controlled by a few actors (e.g., Reference Burt, Ronchi, Wessie and FlapBurt & Ronchi, 1990). Some social systems are organized in terms of a cohesive subgroup of core actors and a more peripheral set of actors loosely connected to the core. Where individuals are placed with respect to this core/periphery structure affects their outcomes, including their creativity (Reference Cattani and FerrianiCattani & Ferriani, 2008).

Other social systems can be understood as teams of actors forming and reforming over time. The success of these teams at any point in time depends not just on the accumulation of talent and motivation inherent in the team members (their human capital) but also on the extent of “connectivity and cleavage” (Reference Wellman, Wellman and BerkowitzWellman, 1988: 26) across the whole social system: the success of the team is dependent upon the social structure of the system within which the team operates (Uzzi & Reference O’BrienSpiro, 2005).

Structural analysis reveals the patterns of presence and absence in social networks that indicate clustering, connectivity, and centralization. Block model analysis (e.g., Reference DiMaggioDiMaggio, 1986) and small-world analysis (e.g., Reference Kilduff, Crossland, Tsai and KrackhardtKilduff et al., 2008) are configurational approaches that analyze patterns at the social network level rather than at the level of the individual (Reference Dorogovtsev and MendesDorogovtsev & Mendes, 2003), thereby permitting the study of the whole and the parts of social networks simultaneously (Reference Wellman, Wellman and BerkowitzWellman, 1988). Interest in the effects of structural patterning at different levels of analysis is growing. Individual attitudes, behaviors, and outcomes cannot be fully understood without considering the structuring of organizational contexts in which people are embedded (e.g., Tasselli & Reference Lounsbury and GlynnSancino, 2023), whereas social network structuring and change in organizations cannot be fully understood without considering the psychology of purposive individuals (Reference Muzio, Aulakh and KirkpatrickTasselli, Kilduff, & Menges, 2015).

It is worthwhile emphasizing that the social structure of networks is by no means obvious to those who are members of such networks. Individuals are often mistaken concerning the patterns of relationships that include themselves and their colleagues (Reference Landis, Kilduff, Menges and KilduffLandis et al., 2018). People tend to perceive themselves as more central in their friendship networks than they really are (Reference Kumbasar, Romney and BatchelderKumbasar, Romney, & Batchelder, 1994). They also forget casual attendees at meetings, tending to recall the meetings as attended by the habitual members of their social groups (Reference Freeman, Romney and FreemanFreeman, Romney, & Freeman, 1987). In one memorable example, a chief executive officer (CEO) of a troubled company that was subject to vandalism and bomb threats examined his firm’s social network (gleaned from archival data by researchers) with bafflement. He had perceived his employees as “waves of turtles coming over the hill; hired as they made it to our door” (Reference BurtBurt, 1992: 1). He had not noticed the networks of kin, neighbors, and friends that constituted his personnel. The CEO had no clue about the deep cleavages that existed among his employees. Social network research has the possibility of emancipating people from default structural effects once structure and structural position are understood.

One of the most influential ideas in social network analysis relates to the uncovering of structural features by revealing the extent to which two individuals are structurally equivalent, that is, connected to the same other individuals (Reference Lorrain and WhiteLorrain & White, 1971). Through structural equivalence analysis, classes of equivalently positioned individuals can be detected (Reference Boorman and WhiteBoorman & White, 1976; Reference White, Boorman and BreigerWhite, Boorman, & Breiger, 1976). People who are structurally equivalent in terms of having similar relations to other people in advice and friendship networks tend to have similar views with respect to the organization and the support it offers to employees (Reference Zagenczyk, Scott, Gibney, Murrell and ThatcherZagenczyk et al., 2010). Further, structurally equivalent employees tend to experience similar levels of emotional exhaustion at work even though their exhaustion levels are unrelated to those of their friends and supervisors (Reference Zagenczyk, Powell and ScottZagenczyk, Powell, & Scott 2020).

2.4 Network Outcomes

The fourth leading idea of major importance to contemporary social network research is the emphasis on outcomes. The subfield of social capital research develops the theme that social network connections constrain and facilitate outcomes of importance to individuals and groups (Reference BurtBurt, 2000). Debates rage over the precise meaning of social capital (e.g., Reference Borgatti, Jones and EverettBorgatti, Jones, & Everett, 1998) given that it can be defined as “shared norms or values that promote social cooperation” (Reference FukuyamaFukuyama, 2002: 27) on the one hand and “investment in social relations with expected returns in the marketplace” (Reference LinLin, 2001: 19) on the other. At the individual level, social capital typically refers to the benefits that accrue from individual network connections (Reference Muzio, Aulakh and KirkpatrickTsai & Ghoshal, 1998). The value of social capital to an individual depends on the number of other people occupying the same social network position (Reference BurtBurt, 1997) – in this sense social capital is an arena for competition. The fewer the competitors who are structurally equivalent or in other ways occupy the individual’s place in the social system, the greater the information and control benefits of brokerage across structural holes.

We live in an age in which people accumulate hundreds of friends and acquaintances through social media (Reference ZhaoTong et al., 2008), while at the same time people report having fewer people in whom they can confide than was the case even a decade earlier (Reference McPherson, Smith-Lovin and BrashearsMcPherson, Smith-Lovin, & Brashears, 2006). Social networks are important for survival – people who lack social and community ties are more likely to die than those with more extensive contacts (Reference Berkman and SymeBerkman & Syme, 1979). Yet social and community engagement is declining outside the ranks of affluent young white people (Reference Hoffman and JenningsSander & Putnam, 2010). It might be thought that the massive increase in connectivity since the discovery that people could connect with complete strangers through about five intermediaries (Reference Hoffman and JenningsTravers & Milgram, 1969) would drastically shrink the small world of interpersonal communication. But research suggests that it still takes between about five and seven intermediaries for email users to reach target persons by forwarding messages through acquaintances (Reference Dodds, Muhamad and WattsDodds, Muhamad, & Watts, 2003).

Social capital, irrespective of definitional debates, relates to outcomes of social network positions. Two of the major outcomes of importance to human beings are health and career progress. Health outcomes are clearly related to social capital. For example, longevity in cancer patients is greater for those with larger networks, and this is especially so for younger patients (Reference Muzio, Aulakh and KirkpatrickPinquart & Duberstein, 2010). People with more types of social ties (e.g., spouse, parent, friend, workmate, member of social group) are less susceptible to catching the common cold (Reference Cohen, Doyle, Skoner, Rabin and GwaltneyCohen et al., 1997). Social networks can affect health through a variety of mechanisms including social support, social influence, access to resources, social involvement, and person-to-person contagion (see Reference Audia and GreveSmith & Christakis, 2008 for a review). However, despite the strong and reliable association between the diversity of social networks and longevity and disease risk, there is still little understanding of how interventions might influence key components of the network to improve physical health (Reference Cohen and Janicki-DevertsCohen & Janicki-Deverts, 2009). Complicating matters is evidence that the relationship between health and networks is bidirectional: health behavior also affects social networks. For example, adolescents select friends whose smoking levels are similar to their own. Rather dismayingly, the data show that adolescent smokers are more likely than nonsmokers to be named as friends (Reference Muzio, Aulakh and KirkpatrickSchaefer, Haas, & Bishop, 2012).

Social capital relates not just to health but also to important career issues. Bankers who have strong ties to colleagues from whom they receive important information concerning deals but whose colleagues are only sparsely connected among themselves receive high bonuses (Reference Mizruchi, Stearns and FleischerMizruchi, Stearns, & Fleischer, 2011). People who can potentially act as go-betweens for colleagues who are themselves not connected tend to have higher performance (e.g., Reference 77Mehra, Kilduff and BrassMehra, Kilduff, & Brass, 2001). As a major review of the network structure of social capital makes clear, people who develop large, sparse, nonhierarchical networks that are rich in opportunities to broker connections across structural holes tend to be more creative; they tend also to receive more positive job evaluations, early promotions, and higher earnings (Reference BurtBurt, 2000). In contrast, people whose work-related networks feature densely-connected cliques of friends tend to experience substandard performance in organizations and substandard rewards.

The contrast between brokerage and closure networks is shown in Figure 2, which illustrates a situation in which Jen has taken over Robin’s job. Robin had a relatively closed network, spanning across only one structural hole between Groups 1 and 2. Jen restructured the network and expanded the social capital associated with the job by adding two new clusters of people in addition to the two clusters reached by Robin’s network. Jen, like Robin, only had to manage four ties, but in the reconfigured network Jen bridges across six structural holes between the four groups. Thus, Jen’s network is both more efficient and more effective than Robin’s.

Figure 2 Managing structural holes between groups.

Despite the importance of maintaining a diverse network that provides information and control benefits, the individual is also well advised to build cohesive ties with the “buy-in” network – that small group of people in the organization who have control over the individual’s fate. A lack of cohesiveness among those with fate control impedes the individual’s advancement, whereas the individual’s average closeness to those with fate control has a strong positive effect on mobility (Reference Reay, Goodrick and D’AunnoPodolny & Baron, 1997). Cohesion is also valuable in teams. Meta-analysis shows that the higher the density of ties within a team, the more the team members commit to staying together and achieving their goals (Reference Balkundi and HarrisonBalkundi & Harrison, 2006). But ties external to the team are also crucial. To the extent that the team leader is connected to popular team leaders within the overall organization, the team tends to be more productive (Reference Mehra, Dixon, Brass and RobertsonMehra et al., 2006). More generally, teams that have numerous connections beyond the team to other actors that are themselves disconnected from each other (i.e., nonredundant connections), will have access to a broader diversity of perspectives, skills, and resources, and therefore can be expected to perform well (Reference BurtBurt, 2000: 398).

These leading ideas (summarized in Table 1) – the emphasis on social relations, embeddedness, structure, and network outcomes – will interweave throughout this Element (Reference Kilduff and BrassKilduff & Brass, 2010).

3.1 Development of Structural-Hole and Weak-Tie Theory

Structural-hole theory, like every generative social theory, has shown vigorous evolution since its earlier articulation in terms of the advantage of disconnected contacts (Reference BurtBurt, 1980). In the earlier articulation the emphasis (borrowing from Reference Lounsbury and GlynnSimmel, 1950) was on the extent to which actors achieved autonomy by occupying positions that had many conflicting group affiliations. Prefiguring the later emphasis on how diverse contacts reduced constraint, the autonomy argument emphasized how “the pattern of relations defining the network position ‘frees’ occupants of the position from constraint by others” (Reference BurtBurt, 1980: 922). In the later development of this argument as it affected interpersonal relations, the emphasis changed from structural positions (occupied by structurally equivalent actors – Reference BurtBurt, 1980) to individual persons, and from freedom from constraint to the contrast between constraint on the one hand and control on the other (Reference BurtBurt, 1992). More recently, the micro-macro dynamic has, following empirical results (Reference BurtBurt, 2007), encompassed ego within the restricted focus of ego's direct contacts, thereby eschewing implications concerning the much wider community (Reference BurtBurt, 2010).

There is also a developing emphasis on differences among individuals in terms of their ability to recognize and take advantage of structural-hole positioning (Reference BurtBurt, 2005: 23). People display consistency across situations in whether they build closed or open social networks, and this consistency is suggestive of individual agency in network construction. Achievement is determined by the individual’s role experience and their role-specific network (Reference 68BurtBurt, 2012). Note, however, that despite this developing emphasis on individualism, structural-hole theory envisages companies benefiting from the activities of individuals who span across structural holes in the social fabric of the organization. These network brokers are “highly mobile relative to the bureaucracy” in providing faster and better solutions (Reference BurtBurt, 1992: 116).

Weak ties are those characterized by infrequent interaction, short history, and limited (emotional) closeness (Reference GranovetterGranovetter, 1973). Weak ties are “ideal vehicles for access and exposure to very different thought worlds – perspectives and approaches that are not only new to the actor but … fundamentally different from each other” (Reference BaerBaer, 2010: 592–593). In the weak-tie approach (Reference GranovetterGranovetter, 1973, Reference Granovetter1983), the emphasis is on bridging to distant clusters rather than on cementing relations with close friends or kin. To break out of the comforting entrapment of one’s close circle of friends and family requires contact with a quite different social circle, contact that is unlikely to derive from a strong tie given that those with whom we maintain strong ties are likely to know the same people we do. It is through weak ties (such as infrequent encounters between two people in the supply chain) that novel opportunities and resources are likely to become available. A key hypothesis in weak-tie theory is: “the stronger the tie between A and B, the larger the proportion of individuals in S to whom they will both be tied” (Reference GranovetterGranovetter, 1973: 1362). This is a theory about how the relationship between two people can affect embeddedness in larger community structures and how community structure can affect the fates both of individuals and of the clusters to which they belong. The micro-macro dynamic in weak-tie theory encompasses not just the individual, the dyad, and the local cluster to which individuals and dyads belong (as in structural-hole theory); it also incorporates the ways in which individuals, dyads, and clusters reciprocally relate to much larger community structures.

Structural-hole theory, among its many other contributions, is valuable for pointing out that bridging ties – whether strong or weak – are key to understanding how individuals achieve advantage in situations in which information represents a scarce resource (Reference BurtBurt, 2005: 18). Structural-hole theory is similar to and builds on weak-tie theory’s discussion of the benefits of diverse information. One of the valuable aspects of Reference BurtBurt’s 1992 explication is the differentiation of these benefits into those of access, timing, and referrals (Reference BurtBurt, 1992: 13–15), benefits encompassed by both weak-tie theory and structural-hole theory. Briefly, in terms of access, some people are better positioned than others to use their networks to screen important news and opportunities. Similarly, some people have personal contacts who provide them with information before others receive it. And in terms of referrals, some people have personal contacts who make sure their names are mentioned at the right time in the right place so that opportunities are made available. All of this is compatible with weak-tie theory, although the emphasis in structural-hole theory is on people “who can speak to your virtues” (Reference BurtBurt, 1992: 15), prefiguring the more recent emphasis on benefits that flow from the immediate set of contacts around the individual rather than from secondary and more distant contacts. Thus, according to a recent treatment of structural-hole theory, accessing structural holes beyond the ego network provides little benefit (Reference BurtBurt, 2007, Reference Burt2010). This more recent development of structural-hole theory differentiates it from weak-tie theory’s emphasis on benefits flowing from afar.

3.2 Juxtaposing Structural-Hole and Weak-Tie Approaches

Differences between the two theories relate to the emphases on control, tie strength, traversing social distance, accuracy of social perception, and micro–macro links.

3.3 Micro–Macro Links: From Juxtaposing to Integrating the Two Theories

If advantage in the structural-hole approach derives from brokerage in ego’s immediate network of alters, then it is not surprising that its theoretical lens centers on local networks and local outcomes (Reference BurtBurt, 2010). Central to weak-tie theory, however, is the exploration of how local, micro relationships lead to global, macro patterns (Reference GranovetterGranovetter, 1973). If weak ties are less prone to triadic closure and, thus, span greater social distance, they are more likely to serve as the crucial informal connections helping to hold together separated business units within an overall collectivity. Formation or deletion of weak ties at the local level can therefore have significant consequences on structural integration or fragmentation at the global level. Local processes such as formation of weak ties, for instance, have been characterized as contributing to the formation of a small world at the global level, where highly clustered groups are connected by short path-lengths (Reference Dodds, Muhamad and WattsDodds et al., 2003; Reference O’BrienRobins, Pattison, & Woolcock, 2005). Thus, a powerful contribution of weak-tie theory is in explaining how local network changes shape global network connectivity, a point less emphasized in structural-hole theory.

Weak-tie theory has always had a double focus: a micro-focus on the strength of the direct tie between the individual and that individual’s contacts within and beyond the workplace, as well as a more macro focus on the structure of ties across the whole community of interests that constitutes the modern firm. It is this double focus that gives the theory much of its distinctiveness – it is one of the few social theories that compellingly relates the activities of individuals to the fates of communities. Extending the theory to the situation of people within an organizational unit who develop strong ties of cohesion among themselves, even in the face of pressures toward globalization, we can say that this internal bonding restricts the opportunities available to each employee of that unit. Relatedly, if each employee exclusively restricts him- or herself to strong-tie attachments, this reduces the resilience of the business unit in the face of unexpected jolts because the business unit will not have allies elsewhere in the wider organization or in the value chain (Reference Krackhardt and SternKrackhardt & Stern, 1988).

The causal mechanisms in weak-tie theory are different from those posited in structural-hole theory. At the individual level, it is access to socially distant resources that provides the individual with advantage in the weak-tie account. At the community level, it is integration through weak ties across fragmented social groups that provides resilience in the face of threats to the social order and persistence of the community. But this is quite different from the active brokerage posited by structural-hole theory that anticipates the individual controlling the flow of information between disconnected alters.

In sharpening the distinctions between the weak-tie approach and the structural-hole perspective, we can ask: how do the classic themes of strategy and serendipity play out in the pursuit of advantage? What, for example, would a strategic approach involve from a weak-tie perspective? One of the advantages of weak ties is that they require lower time commitments relative to strong ties, and, thus, increase the occurrence of serendipitous encounters (for which there is more time, and concerning which there is more likelihood). Thus, one possible weak-tie strategy would be for individuals to develop many weak ties, not knowing which ones might or might not connect to socially distant sources of diverse information and opportunity. This “real options” approach helps hedge against uncertainty concerning how the network and the competitive landscape evolve (Gulati, Reference Gulati, Nohria and ZaheerNohria, & Zaheer, 2000).

We have already discussed the emphasis within the structural-hole approach on the strategic manipulation of ties, but we can say a little more in the way of clarifying theory. There is a consistent emphasis within structural-hole theory on brokers as active agents striving for advantage, as exemplified in this quotation: “[P]eople with networks rich in structural holes are the people who know about, have a hand in, and exercise control over, more rewarding opportunities” (Reference BurtBurt, 2005: 18). But structural-hole theory also incorporates the likelihood that individuals in positions that span across structural holes are at risk of good ideas and rewarding opportunities even if they fail to recognize these ideas and opportunities or choose to pursue them. The emphasis on strategic agency does not rule out the possibility that agency can benefit from serendipity. Given the considerable churn in even the close ties of organizational members (Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010), this “pulsing swirl of mixed, conflicting demands” (Reference BurtBurt, 1992: 33) requires on the part of network brokers an active manufacturing of bridges: “Where structural holes do not exist, they can be manufactured, or their absence can be neutralized” (Reference BurtBurt, 1992: 230).

Further, structural-hole theory advocates the principle of divide and rule, building on the work of Reference MertonMerton (1968: 393–394) and Reference Lounsbury and GlynnSimmel (1950: 185–186). From this perspective, the broker manufactures competition between alters to establish control by creating conflict where it otherwise might not exist: “Make simultaneous, contradictory demands explicit to the people posing them, and ask them to resolve their – now explicit – conflict. Even where it doesn’t exist, competition can be produced by defining issues such that contact demands become contradictory and must be resolved before you can meet their requests …. [I]f the strategy is successful, the pressure on you is alleviated and is replaced with an element of control over the negotiation” (Reference BurtBurt, 1992: 31). The promotion of latent conflict and competition and the guarding against possible collaboration by alters (Reference BurtBurt, 1992: 30–32) differentiate the structural-hole approach from the emphasis on serendipity characteristic of weak-tie theory.

The architecture of the ideal organizational network suggested by both weak-tie theory (Reference GranovetterGranovetter, 1973) and structural-hole theory (Reference BurtBurt, 2005: 12–13) resembles a small world of cohesive clusters (that represent distinctive sources of knowledge, opportunity, and resources) connected by bridges across which knowledge, opportunity, and resources can flow. However, because weak-tie theory emphasizes information from afar, whereas structural-hole theory emphasizes local control of disconnected alters, there may be different implications depending upon which theoretical lens is adopted. To the extent that strong ties are the focus of structural-hole theory, the implications of weak ties may be missed. Weak ties are not means through which ego controls alters, but they may well be pipes through which resources temporary or consistent flow (Reference HansenHansen, 1999). They may be present but unseen in networks exhibiting apparent structural holes, thereby limiting the information benefits to ego and reducing the extent to which control is possible.

An important difference between the two approaches, from a configuration aspect, is that the weak-tie approach posits the possibility of information and resources flowing from afar across unclear boundaries, whereas the structural-hole approach seems to require a well-bounded social network within which the network broker can operate. Thus, structural-hole theory has found a natural home in the analysis of social networks within bounded business units (Reference BurtBurt, 2007), whereas weak-tie theory has been applied to situations such as labor markets where the boundaries of opportunity from the individual’s perspective are unclear (e.g., Reference Freeman, Harrison and ZyglidopoulosMontgomery, 1992). In the boundaryless modern workplace, it may be possible for an updated weak-tie theory to help us understand how careers develop and resources flow as vertical, horizontal, external, and geographical organizational boundaries are minimized, and as people pursue advantage both within and across current employers (Reference Arthur and RousseauArthur & Rousseau, 1996).

In conclusion, both weak-tie and structural-hole approaches are relevant and generative theoretical frameworks for research on social networks. For weak-tie theory, a recent experiment (Reference Lounsbury and GlynnRajkumar et al., 2022) on 20 million people over a 5-year period, during which 600,000 new jobs were created, showed support for the causal claim that weak ties increase job transmission, but suggested that, after a point, there were diminishing marginal returns to tie weakness. And weak ties were particularly generative of job opportunities in digital industries. Structural-hole theory contributes to a burgeoning research program on different types of brokerage (e.g., Reference Freeman, Harrison and ZyglidopoulosNicolaou & Kilduff, 2022), and the antecedents, mechanisms, and outcomes of brokerage positions and processes (Reference Kwon, Rondi, Levin, De Massis and BrassKwon et al., 2020).

4.1 What Kind of Research Do You Propose?

Because the study of organizational social networks is a fertile research arena, in which new studies contribute to the sophistication of existing theoretical approaches in answering an evolving set of questions, understanding the basic assumptions of your research approach is important. As with any research initiative, a social network research endeavor can proceed according to different underlying assumptions. Being clear about the kind of research you are engaged in can facilitate progress and reduce misunderstanding among members of the research team. There are four basic types of research endeavor, organized according to questions concerning epistemology and ontology. The four approaches are illustrated in Figure 3, which poses two questions as you begin your research journey: Are you trying to move closer to the truth about the world? And do you believe that the theory you are working with represents reality? In this section, we present the rationale behind the four approaches. Then, we discuss basic methodological choices that researchers interested in social network analysis must face.

Figure 3 Four different approaches to social network research.

4.2 Research Design

4.2.2 Ego-Network Design

Different from the whole-network approach, ego-network (or egocentric) design involves identifying each individual ego’s direct contacts and the connections among those contacts (Reference O’BrienWasserman & Faust, 1994). People directly connected to ego are called alters. Under this design, information concerning alters’ characteristics, ego–alter relations, and alter–alter relations can be supplied by ego (e.g., Reference BurtBurt, 2004) or collected unobtrusively from employment records, for example (Reference Burt, Ronchi, Wessie and FlapBurt & Ronchi, 1990). This data collection approach helps answer questions concerning the network ties or networking behavior of egos in a social context.

Ego-network research, indeed, focuses on the local social networks surrounding egos, rather than the full set of relations among all egos and alters. Thus, ego-network research primarily focuses on outcomes related to ego rather than structural features of whole networks. For example, controlling for the size of ego’s network and for the extent to which ego’s network features a rival who connects to many of ego’s contacts, we can calculate the extent to which ego’s alters are directly connected to each other with network constraint, a measure of social network brokerage that predicts speed of promotions and other advantageous outcomes (Reference BurtBurt, 1992). Figures 4a and 4b illustrate the ego networks for Avery and Carol. In these two ego networks, we see that Avery is a friend of Chris, Carol, and Emily, who are also friends of each other. By contrast, in Carol’s network, some people, such as Jack and Emily, are not friends. Thus, Carol has a brokerage role whereas Avery does not.

Figure 4a Avery’s ego network.

Figure 4b Carol’s ego network.

Ego-network research typically offers little insight regarding global patterns of connections (Reference Freeman, Harrison and ZyglidopoulosPerry, Pescosolido, & Borgatti, 2018) unless the researcher aggregates the different ego networks into a whole network (e.g., Reference BurtBurt, 2004). Ego-network research is particularly useful for accessing network data on individuals located in relatively large organizations. A whole-network approach would burden each respondent with the necessity of recalling connections among hundreds or even thousands of alters. To achieve a balance between the completeness and the quality of data, researchers often limit the number of alters to be listed by egos (e.g., Reference Brands and MehraBrands & Mehra, 2019). Missing data in an ego-network design is less problematic than in a whole-network design.

As mentioned already, ego networks can be combined to construct a whole network (Reference Weeks, Scott, Borgatti, Radda and SchensulWeeks et al., 2002). Even if ego-network data are collected anonymously, alters listed by different egos may be identified and matched based on the alters’ attributes, such as demographic information, with the aid of software packages like SPIDER (Semi-automated Processing of Interconnected Dyads using Entity Resolution; Reference Young and HopkinsYoung & Hopkins, 2015). But researchers should be cautious about possible errors in the process of identifying alters (Reference Freeman, Harrison and ZyglidopoulosPerry et al., 2018). And there are ethical and legal constraints on the identification of individuals with which researchers must comply.

4.3 Sampling and Bounding Networks

The identification of social network boundaries is a critical step in network research (see Reference Agneessens and LabiancaAgneessens & Labianca, 2022 for a discussion). Sometimes, a group has an easily observed boundary, such as an organizational department (e.g., Reference Kumbasar, Romney and BatchelderKumbasar et al., 1994). In other cases, boundary specification requires compiling a list of the members of the population, collecting all the direct and possibly indirect ties of interest to the researcher (e.g., Reference ZhaoPowell et al., 1996), and establishing the period over which the data will be collected. For example, a study of the spread of poison pills through the US intercorporate network used the Fortune 500 list of companies as the initial boundary set but had to exclude forty-two companies that featured missing data and thirty-two firms that were not publicly traded (Reference DavisDavis, 1991). The time interval was fixed as between 1984 and 1989. The network measure of interest was a board interlock. These data on ties between companies had to be checked against standard directories even though initial data collection used computerized routines.

Similarly, with an ego-network design, participants and their alters need to be identified. One source of ego-network data is the US 1985 General Social Survey (GSS), a national probability sample of 1,395 adults. To investigate the extent to which, under job threat, status affects network recall, researchers reduced this sample to 806 people through the elimination of data from the nonemployed and the exclusion from the sample of respondents for whom other necessary data were missing (Smith, Reference Muzio, Aulakh and KirkpatrickMenon, & Thompson, 2012). The GSS ego-network data were collected using the following name-generator question and follow-up probing:

In other cases, snowball sampling can help establish the boundary of the network beyond the initial sample of people identified by the researcher. The process involves collecting information on the contacts of the original sample members and continuing to collect information on the contacts of the contacts until few new names are added to the sample (Reference Reay, Goodrick and D’AunnoScott, 2000: 61). This process provides reasonable estimates of dyads and triads in the larger population of interest (Reference FrankFrank, 1978, Reference Frank1979).

4.4 Data Collection

4.5 Visualization Using Graphs

Imagine that your research project requires you to represent the friendship network among employees within a research and development (R&D) department and to understand the connections between this network and the innovation activities of the firm. You have already thought through the theoretical approach that will best help you approach your research question. Now, it is time to think about methods and visualization. How do you go about doing this? One intuitive way to represent any set of relationships among people is to draw a graph. Graphs have long been used for the visualization of social network relationships (e.g., Reference Roethlisberger and DicksonRoethlisberger & Dickson, 1939), but they also capture the data necessary for systematic analysis. The theory of graphs provides systematic vocabulary and mathematical operations to describe, denote, and quantify network structural features (Reference Harary, Norman and CartwrightHarary, Norman, & Cartwright, 1965). In this section, we illustrate basic graph theoretical concepts.

In a graph, nodes (or points or vertices) represent actors in a social network, for example Avery and Chris in Figure 5a. Ties (or lines or edges) between two nodes represent social relations, for example friendship, in this case. Two nodes are adjacent if they are directly linked by a tie, and the number of adjacent nodes is called the degree of a node. For example, in Figure 5a Jack is adjacent to Chris, Carol, and Michael. Jack has three connections, hence a network degree of three.

Ties within a specific graph represent a single type of relationship. For example, a friendship network is represented in Figure 5a, whereas Figure 5b represents a task communication network. Some types of social relation, such as talks with, tend to be reciprocated, whereas other types of social relation, such as gives advice, are directed from one person to another person without reciprocation. An example of a directed tie is represented in Figure 5c in which the one-way arrow from Avery to Jack shows the flow of advice. If two persons, such as Avery and Emily, advise each other, the tie is represented by a double-headed arrow.

Figure 5a Friendship network.

Figure 5b Task communication network.

Figure 5c Advice-giving network.

Some actors have no direct ties between them but can still connect via others. For example, in the communication network in Figure 5b, Avery and Jack are not connected, but information can still flow between Avery and Jack via Emily. Thus, Avery–Emily–Jack forms a path, namely, a sequence of nodes without revisiting. This path is the shortest one between Avery and Jack and is defined as the geodesic distance (or distance) between them.

The information in graphs is captured mathematically in adjacency matrices. For example, Figure 6 contains the data represented in Figure 5c. In an adjacency matrix, nodes are represented by rows and columns. The tie from node i to node j is indicated by the entry in row i and column j. For example, Avery’s advice-giving to Jack is indicated by “1” in Row 1, Column 4 in Figure 6. The matrix diagonal is filled with zeros by convention unless people’s ties to themselves are well-defined.

Figure 6 Matrix of binary advice-giving relationships.

The graphs and matrices capture the presence or absence of relationships between people either in binary terms – zeros and ones – or in more nuanced terms to indicate the relative strength of relationships. For example, in the matrix depicted in Figure 7, a higher number indicates a stronger tie in terms of higher frequency of advice-giving. In a graph, there are also multiple ways of visualizing the features of social connections, such as adding value on top of lines, or adjusting tie width according to tie strength, as shown in Figure 8.

Figure 7 Matrix of valued advice-giving relationships.

Figure 8 Advice-giving network with values indicated by line thickness.

4.6 Data Analysis

4.6.1 Characterizing Networks: Centrality

Social network analyses allow us to understand the structural characteristics of whole networks and the structural features of each individual’s network position. In this section, we focus on one of the most important node-level properties in social network analyses: centrality. Centrality incorporates a set of concepts that indicate different aspects of individuals’ structural importance in social networks (Reference BorgattiBorgatti 2005; Reference Borgatti and EverettBorgatti & Everett, 2006). In undirected networks, the widely used centrality measures include degree centrality, eigenvector centrality, betweenness centrality, and closeness centrality. We introduce each of these four measures next and summarize their equations and interpretations in Table 2.

Table 2 Summary of centrality measures.

Measure	Equation	Interpretation
Degree centrality	$d_i={\sum }_j{x}_{ij}$ , where $d_i$ is the degree centrality of actor i, and $x_{ij}$ is the value in row i and column j of the adjacency matrix.	Popularity and exposure to flows through the network; direct opportunity to influence or be influenced
Betweenness centrality	$b_i={\sum }_{h<j}\frac{g_{hij}}{g_{hj}}$ , where $b_i$ is the betweenness centrality of actor i, ghij is the number of geodesic paths that link h and j via i, and ghj is the total number of geodesic paths that link h and j.	Control over things flowing through the network (gatekeeping; brokering)
Closeness centrality (normalized)	$c_i=\left(n-1\right)/[{\sum }_{j}d\left(i,j\right)]$ , where $c_i$ is the closeness centrality of actor i, n is the number of nodes in the network, and d(i, j) is the geodesic distance from i to j.	The speed at which an actor receives things flowing through the network
Eigenvector centrality	Let X be the adjacency matrix of a network, λ the largest eigenvalue of X, and e the eigenvector: X $e$ = $\lambda e$ , thus, $e$ = $\frac{1}{\lambda }Xe$ . Then the ith component of e gives the eigenvector centrality of actor i: $e_i=\frac{1}{\lambda }{\sum }_j{x}_{ij}{e}_j$ , where $e_i$ is the eigenvector centrality of actor i, and $x_{ij}$ is the value in row i and column j of the adjacency matrix.	The well-connectedness of each actor, often interpreted as status

Degree centrality (Reference FreemanFreeman, 1979; Reference Muzio, Aulakh and KirkpatrickWang et al., 2014; Reference Reay, Goodrick and D’AunnoO’Mahony & Ferraro, 2007) can be thought of as an actor’s popularity in that it comprises a count of each actor’s ties. For example, in the idea-sharing network illustrated in Figure 9, Emily shares ideas with three persons, that is, Avery, Chris, and Jack. Thus, Emily’s degree centrality in this network equals three. Degree centrality indicates the extent to which an actor is visible in the network and the extent to which an actor is exposed to emotional support, work-related advice, gossip, disease, good ideas, and other influences (Reference Borgatti, Everett and JohnsonBorgatti et al., 2018).

Figure 9 Idea-sharing network.

Betweenness centrality (Reference FreemanFreeman, 1977; e.g., Reference Freeman, Harrison and ZyglidopoulosTasselli & Kilduff, 2018) captures how often a given actor occupies a network position that falls on the shortest path between two other actors. This measure of centrality is interpreted as an actor’s potential to play the role of a gatekeeper who can control flows through networks (Reference BrassBrass, 1984). For example, actors with high betweenness centrality can filter or distort information flowing via them, and can separate or liaise between disconnected alters (Reference Borgatti, Everett and JohnsonBorgatti, Everett, & Johnson, 2013). As such, betweenness centrality can be used as a measure of brokerage if the size of each individual’s network is controlled for (e.g., Reference Muzio, Aulakh and KirkpatrickOh & Kilduff, 2008). For egocentric research, the appropriate measure is ego betweenness, which is specific to the set of actors directly connected to ego. Based on the equation in Table 2, Emily’s betweenness centrality in the team’s idea-sharing network is seven, whereas Carol’s is four. Thus, compared with Carol, Emily has more control over the ideas flowing within this team.

Closeness centrality (Reference FreemanFreeman, 1979; e.g., Reference Hoffman and JenningsPerry-Smith, 2006; Reference Zietsma, Toubiana, Voronov and RobertsTsai, 2001) is often normalized in use and captures the distance between one actor and other actors in a network. It equals the reverse of the sum of geodesic distances between an actor and others. This value is then multiplied by n-1 for normalization, with n representing the number of nodes in a network. A higher closeness centrality indicates that an actor can reach other actors quickly via a smaller number of links (Reference BorgattiBorgatti, 2005). Based on the equation in Table 2, Chris’s closeness centrality in the team’s idea-sharing network is 0.42, whereas Jack’s is 0.63. Thus, Jack is likely to hear new ideas shared in this network more quickly than Chris. Note that in networks that include two completely disconnected actors, the distance between them is not well-defined and closeness centrality cannot be calculated directly. A few options are available to address this issue, such as recoding the distance as the number of nodes (Reference FreemanFreeman, 1979) or setting the reverse distance as zero (Reference Valente and ForemanValente & Foreman, 1998).

Eigenvector centrality captures the idea that some network contacts are more important than others. It is similar to degree centrality in terms of indicating actors’ exposure to flows through the network, but it has been described as a type of “turbo-charged” degree centrality (Reference Borgatti, Everett and JohnsonBorgatti et al., 2013). Unlike degree centrality, which considers each actor to be connected equally with the focal actor, the eigenvector measure assigns a weight to each actor being directly linked to the focal actor (Reference BonacichBonacich, 1972; e.g., Reference Jensen and WangJensen & Wang, 2018; Reference Shipilov, Greve and RowleyShipilov, Greve, & Rowley, 2010). Specifically, a person’s eigenvector centrality is a weighted sum of eigenvector centralities of this person’s adjacent contacts. The intuition behind this measure is that the centrality of an actor, for example Avery, depends not only on how many actors Avery is connected to but also on whether Avery knows influential others who are also central in the network (Reference BonacichBonacich, 2007).

Eigenvector centrality is often interpreted as a person’s status in a network. For example, in the idea-sharing network in Figure 9, Avery is connected to Emily, and Michael is connected to Carol. Thus, Avery’s and Michael’s degree centralities are both equal to one. But Avery’s single contact, Emily, is a prominent member of the network in that she has three connections. Thus, Avery has higher eigenvector centrality than Michael because of her link to the well-connected Emily.

Measuring centrality is more complicated in directed networks where relationships between two actors may be asymmetric. Whereas betweenness centrality can be used in directed networks, the other three measures require adjustments. The basic principle of these adjustments is to consider the network as having two versions, with each version representing one direction of the network. For example, for degree centrality, in the directed advice-giving network shown in Figure 5c, we could calculate Carol’s indegree centrality, which reflects how much advice she receives from others, and outdegree centrality, which reflects how much advice she gives to others (Reference O’BrienWasserman & Faust, 1994). A similar approach applies to eigenvalue centrality and closeness centrality (Reference Borgatti, Everett and JohnsonBorgatti et al., 2018).

4.6.2 Network Structures

Network research is often referred to as structural analysis (e.g., Reference Kilduff and KrackhardtKilduff & Krackhardt, 1994) in deference to its emphasis on discerning and analyzing structural features of the social world. In this section, we discuss some of the basic structures that feature in social networks.

A dyad consists of two actors between whom there is either a tie or an absence of a tie (Reference O’BrienWasserman & Faust, 1994). Figure 10 shows the three possible dyadic states: null, that is, no tie; asymmetric, that is, a one-way tie; and reciprocated, that is, a two-way tie. An asymmetric positive one-way relationship such as friendship or resource provision may provoke an impetus for the relationship to be reciprocated or disbanded to restore balance (Reference HeiderHeider, 1958; Reference Wellman, Wellman and BerkowitzWellman, 1988). Across a social network, the extent of reciprocity can be measured as the proportion of reciprocated ties relative to all ties. For example, in the advice network displayed in Figure 5c, only the tie between Avery and Emily is reciprocated, meaning that overall reciprocity is 11 percent. Low reciprocity in a network may indicate status hierarchy, with some high-status actors receiving many nominations that they do not reciprocate.

Figure 10 Three states of a dyad.

Triads. Long considered the building blocks of informal networks (Reference Holland, Leinhardt and LeinhardtHolland & Leinhardt, 1977), triads involve three actors and the presence or absence of ties among them. Triads, relative to dyads, provide for the possibility of alliances (two against one), brokerage (one brokering between the other two), and the formation of majorities and minorities (Reference Lounsbury and GlynnSimmel, 1950).

Balance theory (Reference HeiderHeider, 1958) alerts us to the importance of triadic transitivity in positive relationships such as friendship. Ignoring cases of vacuous transitivity involving, for example, three isolates, transitivity refers to whether the triad is complete. For example, if Ann regards both Bill and Colin as her friends, then the triad is transitive if Bill and Colin are friends with each other. The transitivity principle is important in weak-tie theory, where a friendship group of three people in which the link between two is missing is labeled the forbidden triad, forbidden because of the assumption of strong pressure on two people who have a mutual friend to become friends (Reference GranovetterGranovetter, 1973).

The members of a dyad embedded within a triad (i.e., Simmelian dyads – Reference Krackhardt, Kramer and NealeKrackhardt, 1998, Reference Krackhardt1999) are constrained in their attitudes and behaviors because of their connections to a third party according to theory (Reference Lounsbury and GlynnSimmel, 1950) and empirical research (Reference Krackhardt, Kramer and NealeKrackhardt, 1998; Reference 75Krackhardt and KilduffKrackhardt & Kilduff, 2002). But if a person is embedded in more than one such triad, the person may gain new ideas and access to resources by virtue of being a “multiple insider” – someone who benefits from the cohesion available within closed groups and the nonredundancy that comes from being able to move across such groups (Reference Freeman, Harrison and ZyglidopoulosVedres & Stark, 2010).

A clique refers to a complete network in which every actor is directly connected to every other actor and has no common link to anyone outside the clique (Reference Luce and PerryLuce & Perry, 1949). For example, Figure 11 illustrates the interaction network among fourteen participants and four instructors at a National Science Foundation summer camp in 1992 (Reference Borgatti, Everett and FreemanBorgatti, Everett, & Freeman, 1999). There are ten cliques in the camp. Cliques may emerge based on shared demographic characteristics such as gender and ethnicity (e.g., Reference Mehra, Kilduff and BrassMehra et al., 1998). People who bridge two or more cliques are Simmelian brokers (Reference KrackhardtKrackhardt, 1999) who may face paralyzing pressures to conform (Reference KrackhardtKrackhardt, 1999: 206) or who may find themselves liberated to pursue innovative activities (Reference Freeman, Harrison and ZyglidopoulosVedres & Stark, 2010) based on how well their dispositions are matched to the brokerage challenge of moving between different closed groups (Reference Freeman, Harrison and ZyglidopoulosTasselli & Kilduff, 2018).

Figure 11 Interaction networks and cliques in the National Science Foundation summer camp.

Centralization refers to the extent to which a network is centralized around one or a few actors. It is measured as a ratio of actors’ centrality scores, most typically their degree centrality scores (Reference FreemanFreeman, 1979). The nominator is the sum of the difference between the most central node’s centrality and every other node’s centrality, whereas the denominator is the maximum possible centralization score for that network. The maximum centralization occurs in a star network (one actor connected to all others with no other connections between actors) as illustrated in Figure 12. In the help network shown in Figure 13, the most central person is Carol, whose degree centrality is five. Thus, the nominator = 16: (5−1) + (5−2) + (5−2) + (5–2) + (5−2) + (5−5), whereas the denominator = 20: (5−1) + (5−1) + (5−1) + (5−1) + (5−1) + (5−5), and the centralization score = 16/20 = 0.8. By contrast, the camp social network in Figure 11 has no centralized actors so the centralization score is 7 percent.

Figure 12 A star network with six nodes.

Figure 13 A help network with six nodes.

As an important network structural feature, network centralization affects both individual and organizational outcomes. For example, in a longitudinal study that examined how social networks influenced the effectiveness of enterprise system implementation, researchers found that centralized structures made the implementation more likely to fail (Reference Lounsbury and GlynnSasidharan et al., 2012). But at the individual level, employees with high indegree centrality reported implementation success even when they worked in centralized units. Thus, individual centrality and network centralization jointly affected people’s self-perceptions of success.

Networks can be assessed as to the extent to which they exhibit a core/periphery structure, which, in the extreme, features core members connected to everyone and periphery members connected only to core members and not to other members of the periphery (Reference Borgatti and EverettBorgatti & Everett, 2000). In Figure 14, Emily, Chris, and Avery are the core actors, whereas the other three are peripheral actors.

Figure 14 A network with ideal core/periphery structure: Graph and network matrix.

There are two ways of identifying core/periphery structures in social networks (Reference Borgatti, Everett and JohnsonBorgatti et al., 2018). The discrete method involves an optimization process: Actors are assigned to be either core or peripheral such that the correlation between the data matrix and the ideal matrix is maximized, indicated by a measure of fit. This optimization makes sure that the partition of core and peripheral actors maximizes the core–core ties and minimizes the periphery–periphery ties. Although this method matches the essential features of a core/periphery structure, it provides an oversimplified description of network structure.

The continuous method provides a more comprehensive picture of the extent of core/periphery structuring. This method generates a node-level coreness value by modelling the existence or strength of ties between a pair of actors i and j as a function of the coreness of each actor. If we use x_ij to denote the entry of row i and column j in a network matrix A^*, and use c_i to denote the coreness of actors i, this method sets x_ij equal to c_ic_j. If both actors have high coreness, they are connected; and if both actors have low coreness, they are not connected. Then a least-squared procedure is implemented to identify coreness scores for each actor to minimize the Euclidean distance between the real data matrix A and the matrix A^*. For example, Figure 15 illustrates the interaction network among workers in negotiation for higher wages in a tailor shop in Zambia (Kapferer 1972) and the list of people with the highest and lowest coreness scores generated based on the continuous method.

Figure 15 Time 1 interaction networks of Zambian workers (Kapferer, 1972).

In organizations, core/periphery positions expose people to trade-offs between resource support and fresh ideas. Core people in organizations are likely to obtain the resources and legitimacy that are essential to achieving success (e.g., Reference Hargadon, Thompson and ChoiHargadon, 2006). But peripheral people are theorized by some researchers as likely to generate new ideas because they face less pressure to conform to field norms (Reference Lounsbury and GlynnPerry-Smith & Shalley, 2003). The creativity of peripheral actors was summarized by polymath Reference O’BrienMichael Polanyi (1963: 1013) as follows: “I would never have conceived my theory, let alone have made a great effort to verify it, if I had been more familiar with major developments in physics that were taking place.” An alternative view is that higher creative performance is found among people who occupy an intermediate core/periphery position in their organizations (Reference Cattani and FerrianiCattani & Ferriani, 2008).

Small-world networks are characterized by high local clustering and short average paths (Reference Watts and StrogatzWatts & Strogatz, 1998). Local clustering means that actors in the network tend to be well connected within several distinct clusters, and short average path-length means that any actor in the network can reach any other actor via a small number of intermediate actors. Figure 16 illustrates an example of a small-world network. In the classic small-world research conducted by Travers and Milgram (1969), researchers invited 396 people in Nebraska and Boston to mail a folder directly to a Boston-area stockbroker if they personally knew this individual, or to mail the folder to a personal acquaintance who might know the individual. In this study, sixty-four folders were successfully delivered to the stockbroker. The mean number of intermediaries for these delivery chains was 5.2. In similar follow-up research, 540 white people in Los Angeles were invited to generate acquaintance chains to either a white- or a black-target person in New York. It turned out that 33 percent of persons completed the white-target chains and 13 percent completed the black-target chains (Reference Korte and MilgramKorte & Milgram 1970).

Figure 16 A small-world network.

Small-world networks are features of the real world, but also feature in people’s perceptions of their networks. A cognitive social network research effort showed that, given the difficulties in organizing and tracking even small social networks in organizations, people used small world principles in their perceptions of friendship networks. People perceived their work colleagues as interacting in dense clusters, with connections across the clusters between the most popular people within the clusters. The actual networks showed no features of small-worldness whereas the perceived networks exhibited small-world features of clustering and connectivity (Reference Kilduff, Crossland, Tsai and KrackhardtKilduff et al., 2008).

To understand the extent to which a network displays small-world properties, we calculate the small-world quotient (Reference Watts and StrogatzWatts & Strogatz, 1998) as illustrated in Table 3. The quotient is made up of two criteria: the extent to which the network shows much higher clustering than a random network of the same size, and the average path-length. The clustering coefficient measures the average interconnectedness of ego’s alters in a network. Take a friendship network, for example: the clustering coefficient equals the extent to which ego’s friends are also friends of each other, averaged across all egos in the network (Reference ZhaoWatts, 1999). The path-length between two actors in a network equals the smallest number of ties that an actor needs to traverse to reach the other actor (Reference Watts and StrogatzWatts & Strogatz, 1998). Taking the average of all individual path-lengths between all connected individual actors generates the average path-length in a network. The clustering coefficient and average path-length values are then adjusted to account for the properties of a random network of the same size and density, because dense networks exhibit more clustering and shorter path-lengths. In a random network with n nodes and k average ties per node, the expected clustering coefficient is k/n, and the expected path-length is ln(n)/ln(k) (Reference Dorogovtsev and MendesDorogovtsev & Mendes, 2003). The actual clustering coefficient and the actual path-length are divided by the corresponding expected values, generating a clustering coefficient ratio and a path-length ratio. The ratio of these two ratios then produces the small-world quotient (e.g., Reference Kilduff, Crossland, Tsai and KrackhardtKilduff et al., 2008).

Table 3 Formula for small-world quotient.

Variable	Formula
Clustering coefficient (CC)	$\frac{{\sum }_{i=1}^n{C}_i}{n}$ , where $C_i=\frac{A_i}{k_i\left(k_i-1\right)}$ and Ai is the actual number of ties between node i’s ki adjacent nodes
Expected network clustering coefficient (CCexpected)	k/n, where n is the number of nodes in a network and k is the average number of ties per node
Clustering coefficient ratio (CCratio)	CC/CCexpected
Path-length (PL)	$\frac{2}{n\left(n-1\right)}{\sum }_{i=1}^n{\sum }_{j=1}^n{L}_{min}\left(i,j\right),$ where Lmin is the minimum path-length connecting node i and node j
Expected network path-length (PLexpected)	ln(n)/ln(k), where n is the number of nodes in a network and k is the average number of ties per node
Path-length ratio (PLratio)	PL/PLexpected
Small-world quotient (SW)	CCratio/PLratio

4.6.3 Duality of Network Structure: Two-Mode Networks

A two-mode network captures the intersection of persons within groups and of groups within individuals (Reference BreigerBreiger, 1974). For example, the Southern Women data set (Reference Davis, Gardner and GardnerDavis, Gardner, & Gardner, 1941) features eighteen women’s attendance at fourteen events, with data collected from newspaper records. Figure 17 shows the network matrix, with the rows representing the women and the columns representing the events. An entry of “1” means that the person attended the event, and “0” means nonattendance.

Figure 17 Two-mode women–event dataset.

Two-mode network data can be converted to one-mode data to infer relationships among a single set of entities (Reference Borgatti, Everett and JohnsonBorgatti et al., 2018). In the women–event data set example, we can construct a women–women network based on how many events two women both attended. Figure 18 shows the valued data matrix (with the first two letters representing the corresponding person in the column headings) and the graph (with the co-membership values reflected by the thickness of the lines). We can also construct an event–event network based on how many members are shared by two events, as shown in Figure 19.

Figure 18 Converted women–women network matrix and graph.

Figure 19 Converted event–event network matrix.

Note that the women–women and event–event matrices are not independent of each other but involve duality: The tie that links two persons is a set of events forming the intersection of the events’ attendance (Reference Zietsma, Toubiana, Voronov and RobertsSimmel, 1955; Reference BreigerBreiger, 1974).

Two-mode affiliation data are used in organizational research on topics such as interlocking directorates (e.g., Reference Davis and GreveDavis & Greve, 1997) and project collaborations (e.g., Reference Cattani and FerrianiCattani & Ferriani, 2008; Uzzi & Reference O’BrienSpiro, 2005). Both actors and the places where they interact can be depicted in the same representation using correspondence analysis (e.g., Reference Kilduff and BrassKilduff & Brass, 2010).

5.1 Agency and Structure: The Eternal Tension

Social network research currently strains to incorporate both the big data revolution involving networks with millions of nodes (e.g., Reference Lee, Kim, Ahn and JeongLee et al., 2010) and a new emphasis on purposeful action and the pursuit of advantage by individuals (e.g., Reference BurtBurt, 1992; Reference Hoffman and JenningsTasselli & Kilduff, 2021). With big data, there is excitement over the possibility of examining the properties of very large networks as they evolve and change. Researchers in this tradition examine common features across heterogeneous networks including biological networks, cocitation networks, and the World Wide Web (e.g., Reference Dorogovtsev and MendesDorogovtsev & Mendes, 2003). The emphasis is on examining the ways in which clusters develop, the processes by which highly connected actors develop even more connections (e.g., Reference NewmanNewman, 2002), and the extent to which networks exhibit resilience to attack (e.g., Reference Lounsbury and GlynnMoore & Westley, 2011). Key questions in this structural configuration research include: How do structural features such as large components featuring millions of connected nodes affect flows across the network? And does the network resemble a small world (i.e., a network characterized by a high degree of clustering together with short paths between any two actors – Uzzi & Reference O’BrienSpiro, 2005)?

On the side of individual agency, the publication of Reference BurtBurt’s (1992) Structural Holes book, with its description of the tertius gaudens broker who spans across gaps in the social structure, brought a new focus on the ways in which people use social connections for advantage. In this world of competitive action between individuals, agency is ever-present: “The tertius plays conflicting demands and preferences against one another and builds value from their disunion” (Reference BurtBurt, 1992: 34).

A contrasting agentic approach is represented by the tertius iungens broker, the third who joins others to create collaboration among those who might not otherwise engage in work projects (Reference Hoffman and JenningsObstfeld, 2005). The tertius iungens research emphasis is on the process of bringing people into collaborative endeavors rather than on the structure of advantage that was emphasized in some prior brokerage research (Reference Lounsbury and GlynnObstfeld, Borgatti, & Davis, 2014).

The big data and agentic approaches to social networks pull against each other. The big data emphasis draws on the structural foundations of social network research to capture the lineaments of giant webs of interconnections, whereas the research on individual agency sees the structure of social networks as representing opportunities for individual actors to gain advantage through systems of relationships that can be forged, bridged, and broken (Reference Burt, Kilduff and TasselliBurt, Kilduff, & Tasselli, 2013).

Thus, structure and agency represent two ways of looking at social networks. In examining the structure of a network, the focus is on the overall pattern of ties at different levels of analysis. At the individual ego-network level, structure concerns whether the people to whom the individual is tied, that is, the alters, are themselves connected to each other (e.g., Reference Muzio, Aulakh and KirkpatrickOh & Kilduff, 2008). At the level of whole networks, structure is examined in terms of whether there is evidence of a core/periphery structure (e.g., Reference Cattani and FerrianiCattani & Ferriani, 2008), or whether networks exhibit small-worldness (e.g., Reference Kilduff, Crossland, Tsai and KrackhardtKilduff et al., 2008).

These structural emphases neglect the question of whether and how agentic actors create, reproduce, and transform social structures in their own interests (Reference Emirbayer and MischeEmirbayer & Mische, 1998). Change in social networks from an agentic perspective is prompted by motivated people in pursuit of outcomes that are important at the collective (e.g., Reference Hoffman and JenningsObstfeld, 2005) or at the individual level where social ties can be regarded as investments “in social relations with expected returns in the market-place” (Reference LinLin, 2001: 19).

The tension between agency and structure involves a dualism between individuality, representative of the “push” factor of motivation from within, and social networks, representative of the “pull” factor of structures of opportunity from without. Models of social action that incorporate both the motivations and capabilities of individuals and the constraints and opportunities provided by network structures are available (e.g., Reference Muzio, Aulakh and KirkpatrickTasselli et al., 2015) but are difficult to incorporate within a single study.

5.2 Network Volatility

Social network research has tended to privilege stability rather than change, with just 11 percent of social network papers published in the last two decades explicitly assessing network dynamics (for a recent review, see Reference Chen, Mehra, Tasselli and BorgattiChen et al., 2022). The argument for more research on dynamics is that network volatility is intrinsic in network research. Social networks are complex adaptive systems constituted both by established structures of relationships and by evolving patterns of expectations and perceptions (Reference Kilduff, Tsai and HankeKilduff et al., 2006). But many relationships, such as friendship, tend to be relatively stable, as one study of MBA students’ friendships observed: “Over the time period studied there was no significant change in homophily among the racial groups’ networks, despite the explicit promotion of diversity in recruitment of students, formation of heterogeneous classes and teams, and active support by the MBA program administrators” (Reference Mollica, Gray and TrevinoMollica, Gray, & Trevino, 2003: 123).

Early work by Barnard argued that people in social contexts attract and repel each other like “components in a magnetic field” (Reference BarnardBarnard, 1938: 75). The idea of repulsion and attraction, in terms of both homophily and propinquity, was also inherent in most foundational work on the dynamics of network relations. Reference ZhaoNewcomb (1961) observed the “acquaintance process” of initial strangers – “seventeen men who were transferring from other institutions of higher learning to the University of Michigan” (Reference ZhaoNewcomb, 1961: 2) – during the period of development and stabilization of their relationships. According to Newcomb’s analysis, reciprocated relationships among strangers tended to stabilize over a period of about three weeks. But a closer examination of those results suggested a different view, showing that reciprocity often tended to fluctuate and some individuals “danced between friends” over the entire observation period of fifteen weeks (Reference Hoffman and JenningsMoody et al., 2005: 1229).

Other famous foundational studies combined ethnographic inquiry and a network approach in investigating the evolution of network structures over time, including the “karate club” study conducted by Reference ZacharyZachary (1977), which analyzed the structure of relationships in a karate club before and after its split into two different clubs. Due to the ideological conflict between the club president and the club instructor over both the price of karate lessons and the type of karate being practiced, the club was differentiated before the split in two highly centralized blocks around the two actors. Over time, these opposite pressures led the club to divide into two distinct clubs following the two leaders. Through the use of block modelling and sociograms, Zachary’s analysis suggests that a network aimed at achieving a goal will, in the presence of goal conflict, tend to form two groups differentiated on the basis of different goals.

A later approach to volatility emphasized the role of “shocks” in providing opportunities for actors to restructure their ties. External shocks can include technological change (Reference BarleyBarley, 1990; Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010) and industrial action (Reference MeyerMeyer, 1982), whereas internal shocks can include the potential distortive effects of new management (Reference Burt, Ronchi, Wessie and FlapBurt & Ronchi, 1990). Shocks can increase the degree of uncertainty experienced by individual actors, with efforts to reduce uncertainty resulting in a change of communication patterns between individuals and groups (e.g., Reference BarleyBarley, 1990).

More recent studies have addressed theoretical issues underlying patterns of volatility and stability inherent in network dynamics, with emphasis given to temporal antecedents and consequences of brokerage dynamics (e.g., Reference Burt and MerluzziBurt & Merluzzi, 2016). In a study using four years of data on the social networks of bankers in a large organization, interpersonal bridges relative to other kinds of relationships showed faster rates of decay over time (Reference BurtBurt, 2002), with nine out of ten bridges vanishing in the average period of a year. A more detailed analysis of these results, however, showed that decay varied according to an actor’s experience in managing structural holes: slower decay was found in the networks of bankers experienced with bridge relationships, suggesting that social capital tends to accrue to those who already have it.

Dispositional forces, such as the individual’s personality, contribute to the churning of interpersonal connections such that “individuals help (re)create the social network structures they inhabit” (Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010: 639). The social structures that individuals forge tend to maintain an overall inertia over time, even as the connections underlying those structures are shaped by the happenstance of individual choices and external events (Reference Hoffman and JenningsMoody et al., 2005).

Overall, volatility and stability in network connections represent a duality characteristic of “boundedly rational actors creating and re-creating the social structures within which their opportunities and constraints evolve” (Reference Kilduff, Tsai and HankeKilduff et al., 2006: 1038). We lack evidence, however, on the advantages and disadvantages associated with the different degrees of volatility (and stability) that individuals experience in their networks of connections. Volatility can enhance opportunities for network advantage (Reference BurtBurt, 2002). But volatility can also imperil individuals’ career prospects in cases where the legitimacy necessary for the adding and cutting of relationships is absent (Reference BurtBurt, 1992).

The research on network dynamics employs its own vocabulary and methods, as summarized in the glossary provided in Table 4.

5.3 Boundaries of Social Networks

Social network research has long focused on the advantages, constraints, and actions of focal individuals. Neglected in this focus have been the alters to whom individuals are connected. An alter-focused research endeavor has begun, however, with examination of ways in which alters affect ego’s centrality (e.g., Reference Audia and GreveTasselli, Neray, & Lomi, 2023) or brokerage (e.g., Reference Kleinbaum, Jordan and AudiaKleinbaum et al., 2015). But the question arises of where to draw the boundary delineating possible influence on the individual. Should the boundary be restricted to individuals’ direct contacts, or should it be extended to include indirect contacts of individuals’ actors (i.e., the direct contacts of their contacts)? Research is conflicted on this question. Economic outcomes in organizations may be affected only by the set of direct contacts around the individual (Reference BurtBurt, 2007), but this remains true only if the indirect ties of high-ranking colleagues are neglected (Reference Galunic, Ertug and GargiuloGalunic et al., 2012).

The question of whether to include indirect connections in network research is relevant because longitudinal studies conducted in a variety of settings find significant effects linking direct and indirect ties to propensity to suffer from obesity, smoking cessation (Reference Christakis and FowlerChristakis & Fowler, 2007), happiness in the workplace (Reference Fowler and ChristakisFowler & Christakis, 2008), and social status and prestige (Reference Lin, Ensel and VaughnLin et al., 1981). As yet unanswered are the effects of direct and indirect ties in relationships that are negative (Reference Labianca and BrassLabianca & Brass, 2006), conflictual (Reference Klein, Lim, Saltz and MayerKlein et al., 2004), or emotional (Reference Menges and KilduffMenges & Kilduff, 2015). To what extent are people affected by experiences, cognitions, and feelings that are several connections removed?

These questions have particular urgency for the growing area of social network interventions designed to improve intraorganizational functioning by identifying coalitions and analyzing intergroup relations (Reference O’BrienNelson, 1988), and at the macro level to improve health outcomes, disseminate innovations, and make organizations more effective (Reference ZhaoValente, 2012). Social network research has matured sufficiently for its ideas and technologies to be substantially useful in practice.

5.4 Personality and Networks

Is network change driven by individual action, structural embeddedness, or the coevolution of individual characteristics and behaviors and the properties of network interactions (Reference Muzio, Aulakh and KirkpatrickTasselli et al., 2015)? Currently, we lack research that examines the dynamic patterns connecting individual identity and network configuration. Although personality variables have been treated by micro-foundational network research as independent (and immutable) predictors of network change (e.g., Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010), future work can investigate the possibility that the individual’s personality itself changes as he or she experiences changing network positions. New research shows that relationship experiences, such as friendship and kinship, affect personality development (Reference Muzio, Aulakh and KirkpatrickMund & Neyer, 2014). Previously, the discovery that personality in the form of self-monitoring influenced the occupation of advantageous network positions (e.g., Reference 77Mehra, Kilduff and BrassMehra et al., 2001; Reference Zietsma, Toubiana, Voronov and RobertsSasovova et al., 2010) challenged the structural hegemony of network research (Reference MayhewMayhew, 1980). Now, the challenge is to build on the volume of work that shows the mutability of personality (Reference Zietsma, Toubiana, Voronov and RobertsTasselli, Kilduff, & Landis, 2018) to gauge the extent to which individuals change their dispositions in response to social network opportunities and constraints.

6.1 Brokerage as Individual Advantage and Community Contribution

An individual view of agency (e.g., Reference LinLin, 2001) emphasizes individual achievement through network connections, whereas the embeddedness tradition (e.g., Reference GranovetterGranovetter, 1973. Reference Granovetter1985) focuses on social structure constraints and facilitation. These views compete on whether people or networks fuel social action. This tension is both reflected in and intrinsic to social network theory, but also gives opportunities for integration of competing views. Structural-hole theory (Reference BurtBurt, 1992, Reference Burt1997), for example, depicts brokerage as a highly agentic activity in which people negotiate between the “pulsing swirl of mixed, conflicting demands” for their own advantage (Reference BurtBurt, 1992: 33). Brokers are seen as network entrepreneurs who enable change (Reference Burt, Jannotta and MahoneyBurt et al., 1998). At the same time, structural-hole theory suggests that achievement accrues to those who provide value to the community of interacting participants by supplying good ideas (Reference BurtBurt, 2004), mentoring junior colleagues (Reference BurtBurt, 1992), and performing distinctive work (Reference BurtBurt, 1997). There is increasing interest in taking alter-centered approaches rather than continuing to focus on advantages accruing to the broker (Reference BrassBrass, 2022). Thus, there is an opportunity to bridge between the individual advantage and the embeddedness approaches. Indeed, recent work indicates that successful social network brokers are those who engage in “punctuated brokerage,” a pattern of interaction that features intermittent brokering with periods in which the broker retreats within a closed, rather than an open, network structure (Reference Burt and MerluzziBurt & Merluzzi, 2016).

6.2 Network Cognition: From Bias to Opportunity

Implicit in the work on network cognition is the assumption that accurate perceptions of networks are advantageous in terms of future interactions and organizational outcomes (e.g., Reference BrandsBrands, 2013: S93). Network accuracy is seen as helping individuals scan the map of their social world in search of opportunities and advantage (e.g., Reference KrackhardtKrackhardt, 1990). The alternative possibility is that misalignments in network perceptions are leading indicators of network change. Rather than seeking to correct individuals' mistaken network perceptions, therefore, as prescribed in prior research and practitioner advice (e.g., Krackhardt & Hanson, 1993), individuals can be made aware of the possibility that environments can be enacted through purposeful efforts, so that actual relationships can catch up with perceptions. Network misalignment, therefore, could be recategorized as a form of cognitive social capital that has the potential to be converted into actual social capital.

6.3 Past Ties

Research on ties that span temporal configurations suggests that it is not only current social capital that helps facilitate individuals’ outcomes. Dormant ties (originated in the past and reestablished in the present) – that is, “former ties, now out of touch” – are repositories of vital and accessible help (Reference Levin, Walter and MurnighanLevin, Walter, & Murnighan, 2011: 923). Given this, we envisage research aimed at analyzing the effects on present interactions of the legacy of past relationships. People forge and change ties in the present, but their actions may be embedded in networks of past ties, also referred to as “ghost ties” (Reference Kilduff, Tsai and HankeKilduff et al., 2006). Research on network memory shows the temporal effects of structural holes and closure on performance (e.g., Reference Hoffman and JenningsSoda, Usai, & Zaheer, 2004). Future research is needed to investigate the often-hidden influence of prior social network ties, both positive and negative, on current networking patterns.