The Network is Personal: Introduction to a Special Issue of Social Networks1
Barry Wellman, January 14, 2007 version 2b

1. A Personal View of the Network World

We practice personal network analysis every day. Each of us is the center of our own universe. We know who our friends are, how they are connected to each other, and what kinds of sociability, help, and information they might provide – from the 1 to 5 very strong ties that we feel closest with to the 200 to 2,000 acquaintances whom we barely know (Boissevain, 1974; Wellman 1979; Bernard, et al., 1984; McCarty, et al., 2001; McPherson, Smith-Lovin, and Brashears 2006). We reckon what network capital is available from people who can help us (Sik and Wellman 1998; Kadushin, 2004). We even have some knowledge about the friends of our friends – that Kathy knows Wayne Gretzky. Although we do not have full maps, social networking software is making it easier to link us to friends of friends.

Personal network analysts usually want to know which types of people are in peoples’ network (are they composed mostly of kin or friends?), what kinds of relationships they contain (strong or weak ties; frequent or infrequent contact?), and what kinds of resources flow through different kinds of networks (do kin provide more emotional support than friends?).2 These research interests affect how analysts define populations, approach gathering data, and obtain information. Where whole network analysts typically concentrate on uncovering the structure and composition of one big network, personal network analysts usually study a sample of many smaller personal networks: the worlds according to Muhammad, Dick, and Harriet. Analysts want to identify ties wherever they lead, without being confined to groups, neighborhoods or organizations.3

The personal networks studied are often so small – typically the 2 to 30 of the people most strongly connected to the person at the center – and so numerous that analysts do not use fancy network analytic techniques. Instead of UCINet or Pajek, personal network analysts often use the more prosaic and less structural SPSS (see Müller, Wellman and Marin, 1999). Yet, personal network analysts use the essence of the network paradigm itself: a focus on relationships, wherever situated, rather than assuming that the world is built up on solidary, bounded groups (Wellman 1988).

All of the articles in this issue use the network paradigm to analyze community and social support. This interest in supportive community became more than an academic interest in the last half of 2006 when the media frenzy about “social isolation in America” became a moral panic. Personal network analysts J. Miller McPherson, Lynn Smith-Lovin and Matthew Brashears (2006) used U.S. General Social Survey data to show that the mean number of people with whom adult Americans discuss important matters had shrunk 28 percent in two decades, from a mean of 2.9 (1985) to 2.1 (2004). Suddenly, North American jeremiahs lamented the supposed social isolation of the American public, a panic spurred by the decline in the mean number of confidants, the increase in the percentage of socially isolated people (without any confidents), and apparent introversion of Americans turning more to their spouses and less to friends and relatives. The articles’ authors did more than 100 interviews with newspapers, magazines, television and radio: the journalists and the bloggers almost all believing that the findings revealed an American crisis. For example, Washington Post columnist Sebastian Mallaby (2006) proclaimed that “by some reckoning, social isolation is as big a risk factor for premature death as smoking”. Yet, the panic was overstated – or at least misspecified – as none of the articles in this special issue show social isolation.

2. Central Issues

Although personal network analyses can be wide ranging, certain issues keep cropping up:

1. Reconceptualizing – and documenting – communities as networks rather than as neighborhoods. This has provided fresh light on allegations that community has declined. By only looking in neighborhoods and kinship groups, community researchers had overly focused on questions of social integration and stability rather than on social relationships and mobility. The shift to a network view has provided empirical evidence that communities have not declined – they have just spread out (see Bastani’s, Hennig’s, Grossetti’s and Mok and Wellman’s articles; McCarty, et al., 2001).

2. Showing how social support flourishes in personal communities to provide important network capital through specialized ties. Those that provide emotional support may not provide material aid (Hennig’s and Plickert, Côté & Wellman’s articles; Espinoza, 1999; Lin, 2001).

3. Providing a means to reconcile agency with social network analysis, by studying how people capitalize on their circumstances to maintain their personal networks (Wellman & Frank, 2001; Carrasco, Hogan & Wellman, 2007).

4. Using multilevel analysis to disentangle the effects of ties and the networks in which they are embedded: Is it the nature of dyadic ties (e.g., strong/weak; kin/friend) or of the networks (e.g., large/small; densely/sparsely knit; clustered/integrated) that affects the kind and quantity of resources that flow through these networks. For example, are individual immediate kin more supportive when they are members of networks dominated by immediate kin (van Duijn, van Busschbach & Snijders, 1999; Wellman & Frank, 2001; Mok & Wellman’s article)?

5. Documenting homophily: the similarity of network members to each other (by some criteria) and to the person at the center of the network (Louch, 2000; Agneessens, Waege & Lievens, 2006), and linking homophily to personal network structure. Are ties clumped into discrete clusters or broadly interwoven in a single component (Plickert, Côté & Wellman’s article; McCarty, 2002; Carrasco, Hogan & Wellman, 2007)? Looking beyond consideration of the personal networks themselves, what are the implications of personal network composition and structure to integration and cleavages in larger social systems: how does homophily and dense network structure affect social integration (Laumann, 1966; Ferrand, Mounier & Degenne, 1999)?

6. Tracing changes in personal networks over time, and linking them to the nature of life experiences and of the relationships (Martin & Yeung, 2006; Feld, Suitor & Hoegh, 2007).

7. Recent work has expanded the network approach to thinking about community. For example, analysts have traced (and built tools to facilitate) communities of practice – people obtaining information within organizations – by studying who know whom, who knows what, and who knows who knows what (Swarbrick & Contractor, 2002). Other analysts have disproved the original notion of the Internet as a separate system by demonstrating the integration of online and offline relationships (Wellman & Hogan, 2006; Zhao, 2006).

3. Themes

3.1. Beyond America

This special issue builds on classical personal network research into community and social capital. Although the basic approach to personal network analysis was established in the 1960s and 1970s (Laumann, 1969); Wellman 1979; Fischer, 1982; see also Boissevain’s [1974] observational work), research has continued to build on the original insights and new methods.4 The basis of this special issue are the three papers presented at a personal network session of the 2004 Sunbelt Social Network Conference, held in Portoroz, Slovenia: Michel Grossetti from France; Marina Hennig from Germany; and Diana Mok and myself from Canada. To this, we added two papers from authors who could not be there: Susan Bastani from Iran, and Gabriele Plickert, Rochelle Côté and myself – Canada again. The articles reflect the continuing personal network concern with the “community question” about the interrelationship between large-scale structural phenomena (such as urbanization and technological change) and the composition, structure and content of interpersonal relationships (see also Wellman, 1999). Although the authors did not research or write from the same template, their sensibility is similar, forming a serendipitously found set of personal network analyses in diverse societal contexts.

3.2 Survey Based

All of the articles are based on surveying a fairly large sample of respondents and asking them about their ties with a subset of their network members. (It is almost always too expensive to survey each network member mentioned.) Because survey time and respondent attention are limited, no article tries to analyze entire personal networks. Instead, they focus on socially-close ties.

The researchers all make the standard survey analytic assumption that each person’s network is relatively independent from each other – a reasonable assumption when sampling respondents from large populations. Yet, the analyses do not necessarily assume that the characteristics of network members – and their ties with the person at the center – are independent from each other or from the person at the center. For example, women are more likely to be members of a woman’s network, and networks tend to specialize in having relatively frequent (or infrequent) contact between the respondents and the members of their networks.

3.3 Name Generators

All of the research reported here use name generators to elicit information about network members and their ties to the respondents. They either ask respondents a broad stimulus question – “who are you close with?” – and then ask about details (relationship to respondent, etc.), or they ask specifically about a supportive behavioral practice (“who gives you aid when you are emotionally distressed?”) and then ask about details (relationship to respondent, etc.) The second (support) approach gives direct prompts for areas of interest (see Hennig’s and Grossetti’s articles), but the first (general closeness) approach provides information about important relationships even when they are not supportive (see Plickert, Côté & Wellman and Mok & Wellman’s articles; Marsden & Campbell, 1983; Strait, 2000; Marin, 2004; Marin & Hampton, 2007).

By asking about specific network members, the articles elicit more detail (and possible overstatements), then when respondents are asked to estimate the size of their networks (see the discussion by Carrasco, Hogan & Wellman, 2007). Other ways of understanding personal networks are also available that do not ask respondents to name specific network members: Resource generators ask if respondents can get different types of resources: “Do you know anyone who can provide you with emotional aid?” (Van der Gaag & Snijders, 2005). Position generators inquire about whether respondents know someone in a number of different categories (usually occupations): “Do you know a carpenter? A physician?” (Fu, Lin & Hsung, 2001; Erickson 2004; Flap & Völker, 2008). Position generators provide useful estimates of a personal network’s diversity and its access to network capital.

As using name generators takes a long time, size estimates, resource generators, and position generators are often useful in surveys, and they correlate with other social phenomena. Another technique, contact diaries (Fu, 2007), enumerates network members in a method that harks back to Boissevain’s (1974) following people around. Such diaries may identify frequently-contacted trivial ties, but not identify important, infrequently-contacted ties, such as a parent living abroad.

3.4 Multiple Types of Ties

All of the articles use statistical techniques to study multiple types of ties: asking not just about social closeness, but differentiating among different types of social support, such as emotional aid and large-scale material aid. The ties themselves are mixtures of the specialized – giving only one type of support or sociability – or multiplex – using multiple means of communication to exchange several types of support (see Hennig’s article).

The articles pay attention to variations in the composition of personal networks. For example, Mok and Wellman analyze if variations in residential distance of socially-close network members affect contact and supportiveness. Bastani finds that kinship (“family”) is the dominant organizing principle of Iranian personal networks. All of the networks exhibit general tendencies to homophily: liking those who are like you. Age, gender and socioeconomic similarities are widespread, less so among kin.

All of the articles report that the average personal network has between 10 and 27 socially-close ties, a much higher number than the 2.1 discussion partners reported in the “social isolation” article (McPherson, Smith-Lovin & Brashears, 2006). This suggests that social isolation is not endemic, once the analytic focus is broadened to include multiple types of close ties.

One analytic absence is that the articles scarcely deal with the structure of these personal networks: they do not analyze patterns of which network members are connected to each other. Although obtaining this information can be time-consuming, it can yield powerful results. For example, one recent study finds that networks with large, diversified components tend to be more socially active (Carrasco, Hogan & Wellman, 2007).

4. What the Articles Say

Susan Bastani’s “Family Comes First: Men’s and Women’s Personal Networks in Tehran” presents a picture which may be a revelation to western readers who learn only about theocratic and nuclear issues in Iran. Based on interviews with both wives and husbands, she shows that middle-class networks in Tehran are similar to western networks in their size. Yet their composition is different: immediate kin predominate, and with them, densely-knit networks. These networks are gender homogeneous: kin have some male-female ties but friends are of the same sex. Tehrani women are not hidden in the household, but, live active mobile lives, with a complex division of networking between husbands and wives.

Marina Hennig analyzes the nature of families (with children) in contemporary urban Germany. Her “Re-Evaluating the Community Question from a German Perspective” focuses on testing theory about the nature of personal community. Using complex indicators of social support, she finds heterogeneous networks of kin, neighbors, friends, and acquaintances, with each role type tending to provide different kinds of support. Kinship is the most important. Hennig suggests that “it may be the low percentage of kin in North America that requires more explanation than the high percentages found in Europe and Iran” (page ###). Several types of structures are apparent, especially densely-knit network cores (often composed of kin) surrounded by more sparsely-knit peripheries (often friends, neighbors and workmates: see also Wellman & Wortley, 1990).

Michel Grossetti asks, “Are French Networks Different?” in Toulouse from those studied by Fischer (1982) a generation ago in northern California. He finds more residential proximity than Fischer, Toronto (Mok and Wellman) or Tehran (Bastani), and higher levels of social support than in California. The French report having large networks, with a mean of 27 members, but as Grossetti notes, differences in the wording of name generators affect comparative analyses of personal network size and composition. Otherwise, the overall message is of similarity with Fischer’s study.

The two articles that conclude this issue use data from the second East York (Toronto) study. Mok and Wellman use multilevel analysis and distance modeling to ask how much contact and support declined with distance. The decline is neither linear nor smooth, with substantial decreases in contact at about 50 (face-to-face) and 100 miles (both face-to-face and phone). Distance also affects obtaining some forms of social support: not only borrowing small household items and small amounts of money, but also getting major emotional support. Their data were collected in 1978 before the advent of the Internet when contact was by expressways, airplanes, and telephones. It is a baseline for contemporary studies of the Internet age.

Plickert, Côté and Wellman use these same Toronto data to investigate the circumstances under which network members reciprocally exchange support. Their basic finding is relational: the key predictor of reciprocity is the one-way provision of support. Giving and getting are closely associated, especially for exchanges of the same kind of thing – such as a hug for a hug. By contrast, multilevel analysis does not show much effect of either personal network structure or the personal characteristics of either the respondents or the members of their network.

5. Inside Personal Networks

Similarity and variety is often in the eye of the beholder, but to this reader/editor, the articles show more similarities than differences. None of the networks are hyper-local; none of the networks are role homogeneous; none are very sparsely knit or densely knit. And most importantly, all of the networks exist: the respondents – or their families – are not floating free in unsociable and supportive space, contrary to the fears of Beck (1986/1992), Bellah, et al. (1996), and 2006’s social isolationist pundits. Grossetti remarks, “I did not expect to discover such marked convergences between the personal networks of the Toulousains in 2001 and the Californians in 1977-1978” [p. ###]. He further argues that “network structures have hardly changed at all since the 1970s, in a variety of semi-industrialized, industrialized and post-industrialized milieus…. Despite the evolution in communication methods and in our way of life, the part of social life constituted by social networks has remained remarkably stable” [p. ###].

Taken together, the articles depict a world suffused with differentiated sociability. People have many ties, with different types of ties providing various kinds of supportive resources. Kin dominate personal networks in Iran and Germany, while North Americans and the French have ties to mixtures of kin, friends and (to a lesser extent) neighbors, workmates and fellow members of organizations. Within these foci, people often have some choice, connecting with network members who are like them and whom they like. The more voluntary the focus (e.g., friends as compared to kin), the more homophilous the tie.

In the twentieth century, the household has been the key operator of personal networks, with wives arranging sociable get-togethers and in-laws exchanging support as much as blood relatives. This was true in Toronto a generation ago (Wellman and Wortley, 1989, 1990), and in contemporary Iran and Germany. Yet, in some societies, there may be a turn away from the household to the individual as the basic personal networking unit. Instead of stay-at-home housewives, family cars, and household phones, there are trends to women working outside of the home, individual mobile phones, personal internet logons and individual automobile use. Although there is little social isolation, the village, organization, group or the household may have diminished roles as the key organizers of personal communities.

The emerging picture is of “networked individuals” operating connecting somewhat autonomously out of “networked households” (Wellman, 2001; Kennedy & Wellman, 2007). But how do such network individuals operate? Fortunately, at least two groups are working now to develop efficient software for analyzing the structure and dynamics of large batches of personal networks.


6. References

Fischer, C., 1982. To Dwell Among Friends. Berkeley: University of California Press.

Marsden, P., Campbell, K. E., 1984. Measuring tie strength. Social Forces 63, 482-501.

Wellman, B. (Ed.), 1999. Networks in the Global Village. Boulder, CO: Westview Press.