Bridge supervision occurs when manager and boss operate in separate social worlds, a condition increasingly likely as managers work more often from locations outside the office. The concept of bridge supervision was proposed using evidence from managers balkanized into product and geographic silos. Silos facilitate managers segregated from the boss. We here try to replicate support for bridge supervision hypotheses in a cohesive population of HR managers, where bridge supervision would be more difficult. Cohesion does limit the evidence of bridge supervision, but we nevertheless replicate the phenomenon's central hypotheses: bridge supervision is associated with role segregation between manager and boss, and manager performance is unaffected. More, by considering a broader set of network indicators of bridge supervision, we replicate Burt and Wang's focus on just two: mutual contacts, and manager-exclusive density. We close with key features of bridge supervision now replicated, cautions on how easily the phenomenon can be undetected in a cohesive study population, and implications for future research.

We reconcile two conflicting views of the network centralization effect on team performance. In one view, a centralized network is problematic because it limits knowledge transfer, making it harder for team members to discover productive combinations of their know-how and expertise. In the alternative view, the limits on knowledge transfer encourage search and experimentation, leading to the discovery of more valuable ideas. We maintain the two sides are not opposed but reflect two distinct ways centralization can affect a team's shared problem-solving framework. The shared framework in our research is a shared language. We contend that team network centralization affects both how quickly a shared language emerges and the performance implications of the shared language that develops. We analyze the performance of 77 teams working to identify abstract symbols for 15 trials. Teams work under network conditions that vary with respect to centralization. Results indicate that centralized teams take longer to develop a shared language, but centralized teams also create a shared language that is more beneficial for performance. The findings also indicate that the highest performing teams are assigned to networks that combine elements of a centralized and a decentralized network.

Bridge supervision occurs when the connection between manager and boss is a network bridge between separate social worlds. Improved communication technology has facilitated the use of bridge supervision, so manager and boss can easily interact by audio or onscreen as a pair of people disconnected from surrounding colleagues -- but at what cost to the manager, or to effective management? We argue that bridge supervision affects the way in which managers play their role, but not how well the role is played. We find clear support for the argument in a traditional corporate hierarchy. Managers operating under bridge supervision exclude the boss from their work discussions and are conservative in expressing emotion. Behavioral correlates notwithstanding, compensation and good ideas have their familiar association with network brokerage independent of bridge versus embedded supervision. In sum, bridge supervision affects manager style, but not performance. We conclude this paper by discussing the implications of our findings for future research.

We began this work intending to illustrate the network origins of jargon, a signal feature of team learning and the division of labor. In the process, we came to recognize the substantive importance of message timing, which we discuss as the pulse of a network. This paper describes our route to that recognition.

We analyze data from a renovated classic network experiment providing empirical support for three hypotheses. The first, and most familiar from past work, is that teams moving down their learning curve to greater efficiency are prone to shared jargon. As a team moves down its learning curve, language drifts away from day-to-day speech, into jargon. The second and third hypotheses concern network correlates of the drift. With respect to network structure, teams are less likely to converge on jargon when communication is concentrated in one teammate. With respect to pulse, teams are more likely to converge on jargon when communication efforts are numerous and crowded in time.

The two network predictors overlap conceptually. They both involve learning and access to information, but are distinct in their mechanism: Structure provides access. Pulse creates motivation to access. Teammates keeping up with numerous messages concentrated in time have a shared incentive to find shorthand terms (i.e., jargon) that enable faster exchange of information. Network structure predicts team convergence on jargon, but pulse is a stronger predictor. Directions for new research are discussed.

It is well known in economics, law, and sociology that reputation costs in a closed network give insiders a feeling of being protected from bad behavior in their relations with one another. A person accustomed to doing business within a closed network is therefore likely to feel at unusual risk when asked to cooperate beyond the network because of absent reputation-cost security. It follows that business leaders in more closed networks should be less likely to cooperate beyond their network (H1). Success reinforces the status quo. Business leaders successful with a closed network associate their success with the safety of their network, so they should be even less likely to cooperate with a stranger (H2).

We combine network data from a heterogeneous area probability survey of Chinese CEOs, with a behavioral measure of cooperation, to show strong empirical support for the two hypotheses. CEOs in more closed networks are less likely to cooperate beyond their network, especially those running successful businesses: Successful CEOs in closed networks are particularly likely to defect against people beyond their network. The work contributes to a growing literature linking network structure with behavior: here, the closure that facilitates trust and cooperation within a network simultaneously erodes the probability of cooperation beyond the network, thereby reinforcing a social boundary around the network. Taking our results as a baseline, we close sketching new research on personality, homophily, network dynamics, and variation in the meaning of "beyond the network."

Chapter 1. Introduction People you know versus the people they know (social capital in the immediate network: direct access to structural holes, social capital from a neighbor's network: indirect access to structural holes), So what? (business practice, research design, social capital process clues), Overview of the book

PART ONE: ESTABLISHING SECONDHAND BROKERAGE

Chapter 2. Process Clues in Network Spillover Direct Access to Structural Holes (direct network constraint, known returns to direct access, structure as a proxy for process), Indirect Access to Structural Holes (indirect network constraint, research design: network spillover versus network contagion), Possible Returns to Indirect Access (global processes imply strong spillover, local processes imply weak spillover, personal processes imply no spillover, process clues), Summary

Chapter 3. Balkanized Networks Product Launch Network (brokerage opportunities between regions, opportunities within regions, employee returns to brokerage), Supply-Chain Organization (segmented by geography and product, manager returns to brokerage), Summary

Chapter 4. More Connected Networks A Human Resources Organization, HR Returns to Brokerage, Two Divisions in Financial Services (annual network data, regional segregation), Banker Returns to Brokerage, Analyst Returns to Brokerage (industry recognition as a performance metric, who gets elected?), Conclusions (consistent returns to brokerage, negligible returns to secondhand brokerage, probably true in organizations generally)

PART TWO: TESTING THE PERIMETER

Chapter 5. Industry Networks Direct Access to Structural Holes (network data on industry dependencies, industry concentration, baseline effects on industry performance, micro-macro connection), Indirect Access to Structural Holes (expected advantage: maybe, yes, and no, tire cord industry, returns to indirect access), Conclusion (micro-macro consistency, the specific inconsistency, less inconsistent than more extreme, speculation)

Chapter 6. Closure and Stability Network Chaos in Financial Services, Direct and Indirect Embedding (relational, structural, indirect structural, network metrics), Reputation Stability (closure in the aggregate, kinds of closure, no trade off between kinds of closure), Network Stability (closure in the aggregate, kinds of closure, strong indirect compensates for weak direct), Conclusions (no closure, no reputation, closure reinforces status quo by protecting new relations from decay, stability effect of closure spills over from neighbor networks, spillover closure promotes brokerage)

Chapter 7. Mishpoke, Not Inside and Outside Brokerage, Why this Chapter, Network Diagnostics Indicating a Diversity Problem (an instance of women treated as outsiders, broader diagnostic results), Hierarchy Is the Active Ingredient (Karen and Jane, generalizing the example, diagnosis is difficult from inside the network), Strategic Partners and Partner Networks (hierarchy indicates a partner with access to structural holes, Paduka and George, a third step in the network diagnostic), Conclusion (codicil to the broader story, essential feature of contemporary business, exception that proves the rule)

PART THREE: EXPLORING IMPLICATIONS

Chapter 8. Bent Preferences Agency in Networks (assume it away, hold it constant, endogenous agency), Perception in Network Context (marginal evaluation, marginal interpersonal evaluation, bent preferences, network fear hypothesis), Network Defines Peers (connectivity versus structural equivalence, intrepid broker hypothesis, brokers are opinion leaders, brokers display emotion, aside on motivation in teams), Perception Defines the Network (network weights defined, network identity hypothesis, brokers break frame, role equivalence provides frame), Summary

Appendices A. Measuring the Network (population boundary, network survey, selection bias, perceived relations), B. Measuring Access to Structural Holes (bridge counts, constraint, size, density, hierarchy, betweenness, the special case of isolates, indirect network constraint, positional measures), C. Measuring Analyst Accuracy

What is the scope of the brokerage network to be considered in thinking strategically? Given the value of bridging structural holes, is there value to being affiliated with people or organizations that bridge structural holes? The answer is 'no' according to performance associations with manager networks, which creates a problem for consistent network theory across micro to macro levels of analysis. The purpose here is to align manager evidence with corresponding macro evidence on the supplier and customer networks around four-digit manufacturing industries in the 1987 and 1992 benchmark input-output tables. I begin with illustrative evidence on manager networks, to establish a baseline and to explain why direct and indirect access to structural holes can be an advantage. Direct access refers to structural holes in the immediate network of a manager's colleagues, or an industry's suppliers and customers. Indirect access refers to structural holes between friends of friends, in the networks around colleagues, or around suppliers and customers. I then describe the analogous industry network model, introducing the industry data (two years of benchmark performance and network data on detailed American manufacturing industries), and highlighting complementarities between the manager and industry evidence (consistency across levels of analysis, greater variety in manager networks, less endogeneity in the industry networks). Third, I introduce the evidence on industry performance and indirect access to structural holes. In contrast to the manager evidence, about 24% of the industry-structure effect on industry performance can be attributed to structure beyond the industry's own buying and selling, to networks around the industry's suppliers and customers. However, the industry evidence is not qualitatively distinct from the manager evidence so much as it describes a more extreme business environment.

As the network around a set of people closes, it provides a competitive advantage known as social capital. Reputation is the mechanism responsible. Closed networks create a reputation cost for inappropriate behavior. With a reputation cost for inappropriate opinions and behavior, trust is less risky within the network, people are self-aligning to shared goals, and production efficiencies result from donated labor and the speed with which tasks can be completed. Stability is critical to the argument. For closure's reputation mechanism to have its salutary effects, there has to be a credible threat that reputation will survive to affect future relationships. If reputation were to begin anew with each project there would be no reputation cost to proscribed behavior. I draw three conclusions from four years of data on colleague networks around bankers and analysts in a large financial organization: (1) Reputation stability increases quickly with closure. I find that reputation has no stability from one year to the next in networks of colleagues who have little contact with one another. However and this is an intriguing parallel to the social conformity induced by four peers in Asch's (1951) classic laboratory experiment do the same work when you have four mutual contacts with colleagues, and reputation this year is a good predictor of reputation next year. With respect to the people studied here, Coleman (1988:S107) had it exactly right when he said:"Reputation cannot arise in an open structure." (2) Closure's stability effect is concentrated in new relationships. Closure is associated with more positive relations and relations are more robust to decay when embedded in closed networks. However, by the third year of a relationship, closure is less important than the strength of the relationship that has built up between the two people. In other words, closure keeps people in new relations longer than they would stay otherwise, thus protecting new relations from decay. (3) Closure's stabilizing effect operates at a distance from the stabilized network element. Closure among direct contacts, and closure among indirect contacts (friends of friends), make independent and statistically significant contributions to stability. My summary conclusion is that closure creates an endogenous force for the status quo that secures and expands the boundary around a network, protecting new relations until they are self-sustaining, and doing so even for people only indirectly connected at the periphery of the network.

Evaluating executive education is difficult. Lower-level training often has concrete outcomes because the training is intended to make a known process more efficient reduce the number of defective parts, increase the number of units shipped, or reduce the number of customer complaints. The goals of executive education programs are typically less concrete, less about refining what exists than creating what does not yet exist. A typical goal is to improve the informal network of collaborative relations among senior people so they are better aligned with company strategy and better able to quickly detect and pursue market opportunities consistent with the strategy. This paper is about the evaluation of such a program, one grounded in the network structure of social capital. There is abundant cross-sectional evidence on the performance correlates of social capital. Corroborating that evidence, we run a field experiment in which executives educated in the network structure of social capital show performance improvement relative to a control group of untrained, but otherwise equally able peers: Program graduates are 36% to 42% more likely to receive top performance evaluations, 43% to 72% more likely to be promoted (an effect that builds in the two years following the program), and 42% to 74% more likely to be retained by the Company. Active participation matters. The subsequent careers of executives who were quiet spectators in the program cannot be distinguished from the careers of people in the control group, peers who never attended the program.

The social capital of brokerage is evident from the higher compensation, more positive recognition, and broader responsibility given to people who coordinate across the structural holes between groups. This paper is about brokerage among direct versus indirect contacts. Information moved between direct contacts I discuss as direct brokerage, to distinguish it from information moved between friends of friends (people to whom one is only connected indirectly), which I discuss as second-hand brokerage. Analyzing network associations with performance in three study populations, I find that second-hand brokerage has little or no value in a wide variety of circumstances. Brokerage benefits are dramatically concentrated in the immediate network around a person. Why that is so, and conditions under which it is more or less so, are the subjects of this paper. The implication for research design is that brokerage can be measured using designs in which data are limited to the immediate network around an individual. The theory implication is that the social capital of brokerage is a local phenomenon as in the Austrian market metaphor with its emphasis on tacit knowledge about local norms and practice.

This paper is about three points: network bridges are critical to the advantage known as social capital, bridges relative to other kinds of relationships show faster rates of decay over time, and the faster decay in bridges has implications for the stability of social capital. A bridge connects people not otherwise connected; in other words, it spans a structural hole in the surrounding organization. I have four years of data on the social networks of bankers in a large organization. I show that bridge relationships are associated with more positive peer reputations and higher compensation, but bridges decay at an alarming rate. Nine in ten this year are gone next year. I describe factors in the rate of decay, find slower decay in the networks of bankers experienced with bridge relationships, and conclude that social capital accrues to those who already have it. An appendix is included on the kinked decay functions observed in contractual bridge relationships.

This chapter drawn in large part from lengthy review elsewhere of argument and evidence on social capital (see "The network structure of social capital" below) is about current work on the social capital of structural holes. I begin broadly with social capital in metaphor, get more specific with four network mechanisms that define social capital in theory (contagion, prominence, closure, and brokerage across structural holes), then focus on three categories of empirical evidence on the fourth mechanism: evidence of rewards and achievement associated with brokerage, evidence of creativity and learning associated with brokerage, and evidence on the process of bridging structural holes.

To study decay in attachment to an organization, I analyze data on women who obtained an MBA from the University of Chicago's Graduate School of Business (GSB). I measure attachment in terms of network embedding: An alumna is attached to the GSB to the extent that people close to her graduated from the GSB. Behavioral data corroborate the network data in that alumnae measured to be more attached are more likely to have joined an alumni club and made a financial contribution to the school. The hypothesis is that alumnae attachment will decay over time, more slowly when the school is deeply embedded in an alumna's network, more quickly when disruptive events compete for the alumna's time and energy. As expected, attachment declines across the years after graduation (linearly for the first twenty years to about half its initial level), and decay is inhibited when connections with GSB graduates are embedded in stable relations of family, work, or long-term friendship. Decay is remarkably robust to events after graduation (which account for 2% of explained variance in attachment). In other words, an alumna's attachment today was largely determined while she was in school. The results should be of practical value to people who design programs to build personal attachment to organizations, and of theoretical interest to scholars who study such connections

Trust remains an unresolved concern in network models of social capital. The social capital of brokerage depends on trust since the value created by brokers by definition involves new, and so incompletely understood, combinations of previously disconnected ideas, but trust is often argued to require network closure, precisely the condition that brokers rise above. My purpose in this paper is to show how the trust association with network closure is more complex, and decidedly less salutary, than argued in closure models of social capital. Building on earlier work, my argument is framed with respect to two hypotheses describing how closure affects the flow of information in a network. What I will discuss as a bandwidth hypothesis (presumed in closure models of social capital and in related work such as models of reputation in economics) says that network closure enhances information flow. The echo hypothesis (based on the social psychology of selective disclosure in informal conversations) says that closed networks do not enhance information flow so much as they create an echo that reinforces predispositions. Information obtained in casual conversations is more redundant than personal experience but not properly discounted, which creates an erroneous sense of certainty. Interpersonal evaluations are amplifed to positive and negative extremes. Favorable opinion is amplified into trust. Doubt is amplified into distrust. I introduce a baseline model in Section 1 describing an etiology for trust ignoring social context. The bandwidth and echo hypotheses are introduced as contextual extensions of the baseline model in Section 2. In Section 3, I use network data on three study populations to illustrate contradiction between the hypotheses and empirical support for echo over bandwidth. My summary conclusion, in Section 4, is that network closure does not facilitate trust so much as it amplifies predispositions, creating a structural arthritis in which people cannot learn what they do not already know.

This chapter is about two network structures that have been argued to create social capital. The closure argument is that social capital is created by a network of strongly interconnected elements. The structural hole argument is that social capital is created by a network in which people can broker connections between otherwise disconnected segments. I draw from a comprehensive review elsewhere ("The network structure of social capital") to support two points in this chapter: there is replicated empirical evidence on the social capital of structural holes, and the contradiction between network closure and structural holes can be resolved in a more general network model of social capital. Brokerage across structural holes is the source of value added, but closure can be critical to realizing the value buried in structural holes.

This is a position paper on the network structure of social capital. In addition to conclusions about specific aspects of theory and research, my summary points are three: (1) Metaphor versus Mechanism. More than one network mechanism can be cited as responsible the competitive advantage known as social capital. The two mechanisms most often cited are protection within closed networks and brokerage across structural holes, but there are others around which future work will emerge if social capital continues to be such a popular metaphor. My first point is that research and theory will better cumulate across studies if we focus on the network mechanisms responsible for social capital effects rather than trying to integrate across metaphors of social capital loosely tied to distant empirical indicators. (2) Evidence. There is an impressive diversity of empirical evidence showing that social capital is more a function of brokerage across structural holes than closure within a network, but there are contingency factors. Research can be expected to yield wildly inconsistent results across studies that ignore the structure of relations among contacts, content distinctions between kinds of relations, numbers of peers, task uncertainty, or the distinction between insiders and outsiders. (3) Complementarity. The two leading network mechanisms can be brought together in a productive way within a more general model of social capital. Closure can be a significant contingency factor for the value of brokerage. Structural holes are the source of value added, but network closure can be essential to realizing the value buried in the holes. Here is a table of contents: Social Capital Metaphor; Network Mechanisms (Networks affect and replace information, Closure, Structural holes, Social order of disequilibrium); Evidence (Individual and group, Creativity and learning, Process of brokering, Entrepreneurship); Network Dimensions of Social Capital (Network constraint, Size, Density, Hierarchy, Evidence from five study populations), Contingency Factors (Movitation, Network content, Peers and task uncertainty, Network closure, Social capital of outsiders); Conclusions.

Accumulating empirical evidence on American managers shows that social capital effects on performance are a function of the information and control benefits of bridging structural holes the disconnections between nonredundant contacts in a network. Is that network form of social capital unique to Americans? France seemed to us a productive site for comparative research because the image from past research is that French managers are more regulated than Americans; more regulated by bureaucratic authority and more regulated by peer pressure, with both amplified by the greater reliance in France on internal labor markets. People comfortable with knowing their place in a chain of bureaucratic control could be uncomfortable with the negotiated control exercised by network entrepreneurs, so the positive association between structural holes and performance in the United States could be negligible or even reversed for French managers.

We use network and performance data on two study populations of senior managers, one in France and one in the United States, to describe social capital similarities and differences between the populations. The network form of social capital is similar in the two populations: More successful French managers, like Americans, tend to have networks rich in structural holes. The French and American managers make similar distinctions between kinds of relationships. Relations that bridge structural holes are similarly detached from routine work activities for the French and the Americans. The interesting difference is that social capital develops differently in the two populations. The French managers operate with a less porous social boundary around their firm and associate negative emotions with bridge relations. Reinforcing Aix-en-Provence observations on the significance of adult education for Franco-German differences in organization, we find that exposure to peers in other firms via executive education is for our French managers the only factor positively associated with the social capital of bridge relationships.

The tendency for relationships to weaken and disappear I discuss as decay, and functions describing the rate of decay over time I discuss as decay functions. Three conclusions are supported with four years of network data on a study population of bankers and their colleagues in a financial organization. (1) Factors known from cross-sectional evidence to be associated with strong relationships are associated with slow decay; decay is slower in relations between colleagues with a strong prior relationship (inertia), working in the same corporate division (homophily), prominent in the social hierarchy of bankers (status), or connected indirectly through many third parties (embedding). (2) Regardless of slower decay in certain relations, decay has a pattern over time similar to the population ecology "liability of newness" attributed to selection and learning, with the added complication of networks and people aging simultaneously. Decay is a power function of time in which the probability of decay decreases with tie age (years for which a relationship has existed) and node age (years for which a banker has been in the study population). (3) Embedding stability is reponsible for the greater stability of older relationships. The decay-inhibiting effects of age occur where embedding is disrupted but not where embedding is continuous. The third conclusion is interesting in highlighting the first derivative of social structure as a causal variable: embedding has to be measured for its change, rather than level, to see its two distinct effects on relationship decay.

Opinion "leaders" are more precisely opinion "brokers" who carry information across the social boundaries between groups. They are not people at the top of things so much as people at the edge of things, not leaders within groups so much as brokers between groups. The familiar two-step flow of communication is thus a compound of two very different network mechanisms; contagion by cohesion through opinion leaders gets information into a group, contagion by equivalence generates adoptions within the group. Opinion leaders as brokers bear a striking resemblance to network entrepreneurs in social capital research. The complementary content of diffusion and social capital research makes the analogy productive. Diffusion research describes how opinion leaders play their role of brokering information between groups, and social capital research describes the benefits that accrue to brokers.

Given the difficulty of observing interpersonal relations as they develop within an organization, I use iterated prisoner's dilemma games to simulate their development. The goal is to understand how trust could develop as a function of private games, that is, as a function of interaction sequences between two people independent of their relationships with other people. My baseline is Axelrod's results with TIT for TAT showing that cooperation can emerge as the dominant form of interaction even in a society of selfish individuals without central authority. I replicate Axelrod's results, then show that the results only occur in a rare social context, viz., maximum density networks. Where people form less dense networks by withdrawing from unproductive relationships, as is typical in organizations, the competitive advantage shifts from TIT for TAT to abusive strategies. A devious PUSHY strategy wins in moderate to high density networks. A blatantly HOSTILE strategy wins in less dense networks. Abusive players do well in sparse networks because their abuse is lucrative in the initial exchanges of a relationship, before the other person knows to withdraw. Wise players avoiding the abusive players leaves the abusive players free to concentrate on naive players (con men thrive in big cities). The implication is that what keeps abusive players at bay are friends and acquaintances warning managers away from people known to exploit their colleagues. I reinforce the point with illustrative survey data to conclude that private games are not only too dangerous, but also too rare and too slow to be the foundation for trust within organizations. The results are an evidential call for the sociological intuition that trust and distrust cannot be understood independent of the network context in which they are produced.

This chapter is about the tension between two understandings of the role played by social networks in the distribution of information and control, and so resources, within markets and hierarchies. Structural hole theory focuses on the benefits of entrepreneurial opportunity. Network theories of cohesion focus on the benefits of security. They contradict one another on the issue of trust, a contradiction resolved by a network theory of trust and distrust induced by gossip. Distrust is a strategic research site for distinguishing the theories. I present illustrative evidence from words and phrases that senior managers use to explain why they have had so much trouble working with their most difficult colleague. As predicted by the gossip argument, explanations are prone to hostility and character assassination when embedded in strong third-party ties. My summary conclusion from the review and evidence is that the cohesion argument is true, but incomplete, and incomplete in a way that eliminates the social capital contradiction between brokerage and cohesion.

Legitimacy affects returns to social capital. I begin with the network structure of social capital, explaining the information and control benefits of structural holes. The holes in a network are entrepreneurial opportunities to add value, and persons rich in such opportunities are expected to be more successful than their peers. Accumulating empirical research supports the prediction. However, women here pose a puzzle. The entrepreneurial networks linked to early promotion for senior men do not work for women. Solving the gender puzzle is an occasion to see how network models of social capital can be used to identify people not accepted as legitimate members of a population, and to describe how such people get access to social capital by borrowing the network of a strategic partner.

We use survey network and personality profile data to explore the idea that personality varies systematically with structural holes. We draw two conclusions from the analysis: (1) Personality does vary with structural holes. The association is concentrated in a few items, but those few personality items describe three-fourths of the variance in network constraint. (2) The association is consistent with the structural hole argument. People in the least constrained networks claim the personality of an entrepreneurial outsider (versus conforming and obedient insider), in search of authority (versus security), thriving on advocacy and change (versus stability). We summarize with a network entrepreneur personality index that defines a surprisingly accurate probability of the respondent having an entrepreneurial network. We conclude with cautionary evidence from a survey of corporate staff in a large financial organization. Where the personality index is associated with entrepreneurial networks (lower ranks), neither the index nor the networks are associated with manager performance. Where manager performance is significantly linked with entrepreneurial networks (more senior ranks), the personality index is not associated with network structure, and performance is not higher for managers with more entrepreneurial personalities. The personality data are an interesting correlate, but no substitute, for sociometric data.

I present argument and evidence for a structural ecology of social capital that describes how the value of social capital to an individual is contingent on the number of people doing the same work. The information and control benefits of bridging the structural holes Ñ or, disconnections between nonredundant contacts in a network Ñ that constitute social capital are especially valuable to managers with few peers. Such managers do not have the guiding frame of reference for behavior provided by numerous competitors, and the work they do does not have the legitimacy provided by numerous people doing the same kind of work. I use network and performance data on a probability sample of senior managers to show how the value of social capital, high on average for the managers, varies as a power function of the number of people doing the same work.

As a guide to selecting name generators for social capital research, I use network data on a probability sample of heterogeneous senior managers to describe how they sort relations into kinds, and how the kinds vary in contributing to social capital. Managers sort relations on two dimensions of strength Ñ intimacy (especially close versus distant) versus activity (frequent contact with new acquaintances versus rare contact with old friends) Ñ and with respect to two contents Ñ personal discussion (confiding and socializing relations) versus corporate authority (the formal authority of the boss and informal authority of essential buy-in). Comparing name generators for their construct validity as indicators of social capital, I compute network constraint from different kinds of relations, and correlate constraint with early promotion. The correlation is strong for the network of personal relations, zero for the network of authority relations, and strongest for personal and authority relations together. I close with research design recommendations for selecting name generators.

Our argument is that more complex social structures obscure the social frame of reference responsible for contagion. Where it is difficult to answer the question "Who am I?" it is difficult to answer the question "Who is like me?" Ñ which lessens the importance of resolving differences from others' ideas and behaviors. We illustrate the point with evidence on social contagion in business, medicine, and politics.

This paper is about the social capital of people at the top of their organization. I will refer to them as managers. Social capital is integral to their work, so its effects are revealed in interesting detail, especially in contemporary organizations. There is a shift away from bureaucracy, with layers of formal control replaced by fewer layers of negotiated informal control. The shift away from bureaucracy means that managers cannot rely as much on directives from above. They are more than ever the authors of their own work. Firms gain by being able to identify, and adapt more readily to, needed production changes and market shifts. Managers face new costs. Coordination costs once borne by corporate bureaucracy Ñ each person having responsibility for coordination within a limited domain of responsibility Ñ are now borne by individual managers who have responsibility for coordination across broader domains, with a corresponding increase in uncertainty, stress, and potentially disruptive conflict. Enter social capital. The shift away from bureaucracy is a shift to social capital as the medium for coordination within the organization. How social capital provides that coordination, and the resulting benefits to managers who have social capital, is the story to be told here (see Lazega, 1994, for background review of organizational network analysis).

We study how the production of trust changes as relations aggregate into social structures. The simplest social context is an isolated dyad -- a pair of people disconnected from others. Their games are private. The more usual social context for trust is an embedded dyad -- a pair of people surrounded by variably close friends, foes, strangers, and acquaintances. The two people play their games in public; a public composed of the third parties surrounding them. We argue that third-party gossip serves to reinforce existing relations; making ego and alter more certain of their trust (or distrust) in one another. We make two trust predictions; direct connection affects level, indirect connection affects intensity. Analyzing network data on a probability sample of diverse senior managers, we show that the trust predictions are correct and learn a great deal about how third parties have their effect. Our principle conclusions are three: (1) Consistent with a repeated games explanation of trust, trust is most likely within strong relations and distrust is most likely within weak relations. (2) Consistent with the gossip argument, indirect connections significantly increase the likelihood of trust within strong relations at the same time that they significantly increase the likelihood of distrust within weak relations. (3) Consistent with the gossip argument, third parties mutually tied to ego and alter are associated with increased trust while third parties more tied to ego than alter are associated with decreased trust. The kinds of indirect connection interact in different ways to produce the observed aggregate positive and negative third-party effects on trust.

Two network contagion models are used to describe corporate contributions officer evaluations of nonprofit organizations seeking philanthropic donations. Contagion by cohesion predicts that behavioral communication between contributions officers results in them sharing the same evaluation. Contagion by structural equivalence predicts that symbolic communication via role playing between officers similarly positioned in the interorganization network of contributions officers results in similar evaluations. We find strong evidence of contagion, robust over differences in the evaluated nonprofit organizations and differences between officers. The evidence is overwhelmingly of contagion by structural equivalence.

Network analysis provides a useful guide for collapsing ostensibly non-network data into analytical categories. I illustrate the point here using a familiar variable, years of age. Viewed structurally, age is a network pattern characteristic of being a specific number of years old. So viewed, years of age can be collapsed into socially distinct age categories where each category is a status in the social structure of age in a study population. For illustration, I describe the structure of relations defining age statuses in the American population. Each status is a unique pattern of relations with kin of specific ages, spouses of specific ages, and friends and coworkers of specific ages. In the mid 1980s, Americans were distributed across nine age statuses; I children (ages 1-18), II students (19-24), III young adults (25-30), IV twilight youth (31-36), V middle-age adults (37-46), VI older adults (47-52), VII senior adults (53-60), VIII retiring adults (61-66), and IX the elderly (over 66). The most severe changes in 1985 were happening to Americans in their late 40s born at the beginning of World War II and in transition from age status V to status VI. When observed in 1985, they were in the process of replacing their parents with their children as important discussion partners and learning to live with much greater age heterogeneity in their other contacts, both in their marriages and their friends and coworkers beyond the family. Women were about to leave their prominent position in heterosexual society defined by age status V and men were about to enter a menopausal period characteristic of status VI.

Two network contagion models are used to describe corporate contributions officer evaluations of nonprofit organizations seeking philanthropic donations. Contagion by cohesion predicts that behavioral communication between contributions officers results in them sharing the same evaluation. Contagion by structural equivalence predicts that symbolic communication via role playing between officers similarly positioned in the interorganization network of contributions officers results in similar evaluations. We find strong evidence of contagion, robust over differences in the evaluated nonprofit organizations and differences between officers. The evidence is overwhelmingly of contagion by structural equivalence.

Social scientists who study organizations, whether they study as theorists, researchers, or consultants have long been aware of the importance of informal control mechanisms for an organization's functioning. The primary hurdle separating that awareness from more effective implementation, however, has been methodology. With a case study of a severely troubled firm, we illustrate how developments in network analysis can be useful in diagosis and management. Using readily available corporate personnel records, we describe the social structure of interpersonal relations within and beyond the firm over a thirty year period. We show how control and leadership shifted over the thirty years, and how conflict within the firm is grounded in the changing structure of relations.

Hummell and Sodeur (1987) propose a practical solution to detecting role equivalence in social network data. The solution is very fast, equally applicable to symmetric and asymmetric relations, involves no iterative computing, and is now readily available as one of the equivalence options in STRUCTURE. Unfortunately, their paper is only available in German in a book published for their colleagues in Germany. The purpose of this brief note is to give their extremely useful idea wider exposure.

This chapter was an exploratory effort to apply models of network form to understand network content (as introduced in the below 1985 article with Thomas Schott). The section on age status in this chapter was expanded into the above 1991 article on the network structure of age in America. The goals for this chapter were twofold: how can we see the way relations are understood in a study population (versus understandings assumed by the network analyst), and is what analytical traction is available from understanding dimensions of network content, substitutable relationship labels within content domains, and content ambiguity within a study population. The study population is the General Social Survey (GSS) national probability sample of Americans in 1985. The chapter was written as a complement to Peter Marsden's (1987) article on the form of the GSS American discussion networks.

Important differences in our images of economic networks result from the seemingly innocuous choice between measuring transactions as proportional variables rather than measuring them as marginal variables. Market boundaries defined by proportional transactions emphasize differences between specialized markets, production markets with a single principal supplier or consumer market. Boundaries defined by marginal transactions emphasize differences between diversified markets, production markets defined by unique transaction patterns with multiple supplier and consumer markets. Compared to the results reported in Burt (1988b) with proportional measures of transaction strength, the results reported here with marginal measures offer a substantively richer map of market boundaries to guide organization research by offering clearer distinctions between kinds of market environments in which organizations operate. At the same time, the results obtained with marginal transaction measures corroborate the conclusion that market boundaries were by and large stable during the 1960s and 1970s. In sum, proportional transactions are well suited to their traditional use in economic input-output models tracing the flow of resources through a network. Marginal transactions are the more useful measure for sociological studies of market boundaries for organizational analysis because they more clearly reveal variation in the resource flow patterns that define structurally equivalent (substitutable) production activities as a market.

Much of the evidence of coordination between corporations and their markets comes from cross-sectional studies conducted within portions of the American economy during the past two decades. We know, especially for manufacturing during the late 1960s, that certain structural qualities of markets predict profits and the organization of large firms. But this evidence is open to an uncomfortable empirical question: To what extent did the social-structural qualities determining resource dependence in American markets change during the 1960s and 1970s so as to limit the generalizability of cross-sectional evidence? The analysis here shows that markets were dramatically stable in the social structure of production relations known to predict the structure of large firms. Relying principally on Department of Commerce data, the article traces the American economy through the 1960s and 1970s in terms of 77 broadly defined markets, describing the stability of market boundaries and patterns of transactions with suppliers and customers, the enduring profit inequalities generated by the social structure of the markets, and the constant sources of market constraint to be managed by firms designed to operate within each market. The implications are that organizational research with cross-sectional can be generalized (within specified limits) to other periods of time, organizations can be selected for study from a stable sampling frame of corporate markets, and organizational behavior can be studied over time for its success or failure as an adaptation to known market constraints.

I discuss and illustrate the extent to which different relation measures and pattern similarity measures can be expected to generate different structural equivalence results. Measures of network relations and pattern similarity are reviewed to establish clear comparisons between structural equivalence measures. Using Monte Carlo sociometric choice data drawn from four strategically designed study populations, alternative relation and pattern similarity measures are combined in a factorial design generating six measures of structural equivalence within each study population. I draw three conclusions: (1) There is significant reliability across alternative measures. (2) The reliability increases with the clarity of boundaries between statuses in a study population. (3) The noticeable differences between structural equivalence measures that exist under conditions at all weaker than strong equivalence are principally a function of how relations are measured rather than how relation pattern similarities are measured. I draw two inferences for applied network analysis: (1) Structural equivalence should be computed from path distance measures of network relations (however normalized) rather than being computed directly from binary choice data. (2) Renewed methodological attention should shift from how we measure pattern similarity to how we measure relationships.

Two classes of network models are used to reanalyze a sociological classic often cited as evidence of social contagion in the diffusion of technological innovation: Medical Innovation. Debate between the cohesion and structural equivalence models poses the following question for study: Did the physicians resolve the uncertainty of adopting the new drug through conversations with colleagues (cohesion) or through their perception of the action proper for an occupant of their position in the social structure of colleagues (structural equivalence)? The alternative models are defined, compared, and tested. I draw four conclusions: (a) Contagion was not the dominant factor driving tetracyclineÕs diffusion. Where there is evidence of contagion, there is evidence of personal preferences at work. (b) Where contagion occurred, its effect was through structural equivalence, not cohesion. (c) Regardless of contagion, adoption was strongly determined by a physicianÕs personal preferences, but these preferences did not dampen or enhance contagion. (d) There is no evidence of a physicianÕs network position influencing his adoption when contagion is properly specified in terms of structural equivalence. The ostensible prestige effect is spurious, resulting from biases created when cohesion is used to model contagion. In short, the product of reanalyzing the Medical Innovation data with recent developments in network theory is clearer, stronger evidence of social contagion and redefinition of the social structural conditions responsible for contagion.

Data obtained with the GSS ersatz network density item are compared to density data obtained with the more traditional, more costly, GSS sociometric network items. The inexpensive ersatz density data are not independent of network density, but they are almost completely unreliable. The full range of possible densities occurs at each level of ersatz density and only 1 to 2 percent of variation in network density can be described with ersatz density. Hypotheses operationalized with the ersatz density variable specified as a predictor will be biased toward the null hypothesis. Given this GSS experiment, the reliability of conclusions from studies replacing sociometric network items with inexpensive items purporting to measure network structure should be interpreted with caution.

The people identified as important discussion partners in the GSS network data were cited in order of strength of relationship with respondent; the first cited person having the strongest relntion, the second having the next strongest. and so on. On average, the third citation is a turning point. There is a steep, linear decline in relationship strength ncross the first people cited as discussion partners and a slower, but continuing decline, across the fourth and fifth people cited. Order effects on closeness and contact frequency are described in the comext of network size and relation content. There is a kinship bias only in deciding who to name first: spouses tended to be the first discussion partner cited and other kin tended not to be. There is a sex homophily bias across all respondents - people of one's own sex were cited as discussion partners before members of the opposite sex - but it emerged differently for men and women. Women, especially married women, expressed sex bias in the people with whom they spent time while men expressed sex bias in the people with whom they felt close. Men claimed closer relations with women thnn men but in fact listed their important discussion partners in descending order of closeness and began the list with the names of other men. Finally, there is evidence of a co-worker bias in discussion relations beyond the family; respondents tended to mention co-workers as daily contacts but late in their list of important discussion partners. With the exception of the spouse bias, all evidence of contenl bias is markedly wenker than the consistent tendency for respondents IO list discussion relations in descending order of closeness and contact frequency.

The idea of structural balance is used to suggest quantitative intervals between relationship strength response categories in the GSS network data. In contrast to an assumption of equal intervals between the categories, the intervals appear quite unequal. Relations with "less close" discussion partners are about 0.17 the strength of relations with "especially close" discussion partners. The middle category of relations between discussion partners appear to be little more than acquaintance relations; about 0.2 of the distance between "total strangers" and people who are "especially close."

Distinctions among kinds of relations (friendship, advice, intimacy, and so on) are typically ad hoc in empirical research. These ad hoc distinctions among relation contents increase the likelihood of equivocal research conclusions. We develop three ideas indicating how standard, well-known, network models of relationship form can be used to clarify relationship content. (a) We begin with an idea for recovering the semantic context in which a relation content occurs. This context is cast as a network of tendencies for contents to be confused for one another and the form of the network dissected with network models of relation form holds insights into the ways in which relation contents are understood in a study population. (b) The network concept of structural equivalence is used to define content domains composed of specific relation contents that are substitutable for one another in described relationships. (c) The network concept of network prominence is used to define the ambiguity of contents in described relationships. The proposed perspective is analogous to a linguistic componential analysis of relation content.

This is an argument for obtaining network data in the General Social Survey (GSS). The proposal requires a discussion of how and why at least minimal network data ought to be obtained in a probability sample survey of attitudes and behaviors. I begin with general concerns; briefly describing the proposal, available experience with the proposed items in large probability samples, how the proposed items are different from existing GSS items, kinds of variables that the proposed items would generate, and kinds of research questions that could be addressed if the proposed items were included in the GSS. I then address focused questions likely to arise in deliberations over the proposal; explaining how much interview time the proposed items are expected to require, why one rather than multiple name generators are proposed, why recording five alters is proposed, why intimacy is proposed as the name generator criterion content, why a short form is proposed for obtaining formal data, how priorities among name interpreter attribute items were established, how the proposed items elicit data on the strength and content of relationships, and how the proposed data might be coded for easy access by GSS users.

Corporate philanthropy is analyzed as a cooptive relation, akin to advertising, directed at persons collectively as a consumer sector of the American economy. The strength of this cooptive relation is predicted from a network definition of the extent to which corporations in an economic sector have a market incentive to institutionalize their relations with people as consumers. As predicted, the proportion of corporate net income donated to charity covaries with the extent to which firms in a sector are dependent on consumption by people and able to do something about eliminating uncertainty in the demand for their product. In fact, the specified structural effect of the market on the rate of corporate giving is stronger than the income and tax incentive effects typically specified in a microeconomic model. Methodologically, the discussion illustrates a strategy by which network analysis is often used to inform analyses of individuals: social context constraints on an actor are captured in a network model of context, then specified as parameters in a microeconomic decision model.

A method is described for interviewing a random sample of persons drawn from a large population so as to describe role-sets defining statuses in the population social structure. The key to the method is a connection between the concept of an actorÕs network position in social structure and combinations of attributes that define statuses in the social structure. With data obtained in a survey interview with a randomly selected respondent, it is possible to describe the relational pattern defining his Òersatz network positionÓ in the population social structure. Given ersatz network positions for a representative sample, it is possible to test hypotheses concerning status/role-sets stratifying the population.

Moving away from description of directorate ties as a cooptive device, we test a theory explicitly predicting cooptive uses of corporate directorates from the structure of the market in which firms operate. The theory is based on a network model of structural autonomy. It takes as exogenous information the sales and purchase transactions between establishments in sectors of the economy, locates those sectors most constraining pricing discretion within each sector, then predicts where establishments should be connected by interorganizational relations as directorate ties (establishments connected through corporate boards by ownership, direct interlocking, or indirect financial interlocking) in the 1967 American economy, we find the theoryÕs predictions to be accurate. Each of the three types of directorate ties tends to occur where there is market constraint, and tends not to occur in the absence of constraint. Further, the three types of ties are coordinated as multiplex directorate ties. Where establishments in one sector constrain those in another, there is a strong tendency for all three types of directorate ties to exist between the two sectors. Where there is no such constraint, all three tend to be absent. Support is weaker for intrasector in comparison to intersector cooptation. Whatever the cooptive intent of the directorate ties described, they are patterned as if they were intended to coopt market constraints on corporate pricing discretion.

My purpose here is to define and illustrate a concept of structural autonomy based on recent developments in network analysis. The concept is stated in terms of the pattern of relations defining a network position. It incorporates aspects of oligopoly from economics and group-affiliation from sociology. Eight hypotheses are derived from the proposed concept. The hypotheses concern the effects on autonomy from the pattern of relations defining a network position, the places in social structure where cooptive relations should appear (as well as places where they should not), and the increase in autonomy that can be expected from effective cooptation. Numerical illustration is provided. As a useful research site, firms in manufacturing industries in the 1967 American economy are treated as structurally equivalent actors, and total industry profits take to be a result of relative autonomy across industries. As expected, the industries with high structural autonomy tend to have high profits, and firms tend to merge with other firms so as to coopt constraints on industry autonomy.

The existence of an actor as a set of asymmetric relations to and from eveyr actor in a network is specified as the position of the actor in the network. Conditions of strong versus weak equivalence of actor positions are defined. Network structure is characterized in terms of structurally nonequivalent, jointly occupied, network positions. Social distances from network positions are specified as unobserved variables in structural equation models to extend the analysis into the etiology and consequences of network structure.