Linton C. Freeman. The Development of Social Network Analysis: A Study in the Sociology of Science. Vancouver, Canada: Booksurge Publishing, 2004, 205 pp., $15.95 trade paperback; $7.99 nonprintable ebook. Available: http://www.booksurge.com.
Review written by Charles Kadushin, firstname.lastname@example.org, Brandeis University
So you thought you knew who "invented" social network analysis -- Jacob Moreno, right? Not so fast. To begin with, Moreno was not his birth name, it was Jacob Levi and, though he was wildly imaginative, many of the features of "sociometry," the term he indeed coined in 1934 (two years after he published his first network analyses) were probably due to his collaborators. In his carefully researched "history of social network analysis written from a social network perspective," Linton Freeman sets the record straight and turns the lens of social network analysis onto the field itself. The result is a provocative essay in the sociology of knowledge that is a must-read for anyone with the slightest interest in social networks, as well as those interested in the sociology of science.
There is little doubt that social network analysis has now "arrived." Every year since about 1984, there has been a linear growth in the number of substantive areas in which social network analysis has been applied (p. 5). The New York Times celebrated social networks as one the "new ideas" (sic) of 2003 (Gertner, 2003). Many organizations attempt to improve their efficiency through sociometric analyses (Krebs, 2003). Research and development laboratories map major gatekeepers of critical information. There are maps of who works with whom in biotechnology. Epidemiology was founded on the tracing of agents who carried disease, and modern network methods have been applied to the HIV-positive field.
In 2004, Social Science Citation Index recorded approximately 450 hits to "social network or social networks" (and you thought you could keep up with the literature!). On the first results page, ten different journal titles are cited as publication sources. Only by the 53rd citation is a sociology journal mentioned. We don't even get to the Social Networks journal until citation 71. If social network analysis has become popularly and academically almost commonplace, and if the earliest signs of the field were in the mid-19th century, what took it so long to be recognized as a discipline? There are many surprises on this long trajectory, and recounting the history helps Freeman to explain what happened.
Freeman defines social network analysis as having four key features: a structural intuition, systematic collection of relational data, graphic images, and mathematical or computational models. (I would add a fifth feature that is ancillary yet crucial: a study of the flows through the network.) The first four features alone tend to produce a static network, though in Freeman's own work flows are often important. When flows are added, networks become channels through which ideas, values, friendship, esteem, money, sales, disease, or almost anything can travel. The same network structure may pass flows of different kinds, or different structures may better facilitate different flows. The impact of social network analysis and its utility depends in large measure on which flows are studied. The way the different flows capture the popular and academic imagination determine, in part, the place of network analysis. But we are getting ahead of our story.
Most of Freeman's four features have been around in various forms since the 19th century (with major pushes from 1930s through the 1960s) and involved hundreds of major and not so major theorists and investigators other than Moreno. Antecedents of structural intuitions invoke some familiar names: Henry Maine, Ferdinand Tönnies, Emile Durkheim, Herbert Spencer and Charles Cooley "all tried to specify the different kinds of social ties that link individuals in different forms of social collectivities" (p. 15). And for Georg Simmel, "sociology was no more and less than the study of the patterning of interaction" (p. 16). Freeman, however, was surprised to discover that the man who gave sociology its name in the 1840s, August Comte, also believed that the statics of sociology consisted of the interconnections of social actors.
More surprising to contemporary social network analysts may be the fairly widespread systematic data that Freeman discovered had been collected in the 19th and early 20th centuries. In 1810. Pierre Huber, studied patterned interaction among ants and was perhaps the first who systematically studied non-human interaction. In the mid-19th century, in New York state, Lewis Henry Morgan joined a secret society called "The Grand Union of the Iroquois" -- except that he discovered nobody knew anything about the Iroquois. So he wrote an ethnography and eventually, in 1871, published elaborate data on kinship and descent around the world, along with systematic kinship graphs that set the pattern for those used to this very day. In 1834, John Atkinson Hobson studied and presented two-mode charts of corporate overlap in South African finance (p. 18). In fact, he developed hypergraphs long before we called them such. And before Moreno's work became well known, child development studies in the 1920s and 1930s tracked both questionnaire and observational data on child interaction. In 1928, Helen Bott (none other than Elizaboth Bott's mother) even used matrices. In 1923, Elizabeth Hagman showed discrepancies between interview data and observational data on playmates.
Enter Moreno, née Levi. Freeman's account adds spice to an already fantastical character. Moreno had dark side: "self-centered, self-serving ... admitted hearing voices, he sometimes thought he was God, and he was convinced that others were always stealing his ideas" (p. 31). Though to gain a full appreciation of his bizarre side, there is nothing like reading Who Shall Survive, available in the original 1934 edition for about $175 (what to give your favorite network scholar) or the even more bizarre 1953 edition that -- oddly -- costs about the same. "For the most part Moreno seemed to be unfocused but, when he was involved with a woman who could serve as a muse, he succeeded in concentrating and was able to write" (p. 34). George Homans thought Helen Jennings' book on leadership a far better one than Moreno's classic. Nonetheless, by the mid-1930s, Moreno became "something of a social science celebrity" (p. 36). He started Sociometry, and got a virtual who's who in American social science to join his enterprise: Paul Lazarsfeld did some mathematical modeling of the probability of choice -- but Allport, Boaz, Bogardus, and Bruner (just to stop at the Bs) were also associated. And don't forget Newcomb's classic study that is still analyzed and the network data distributed with UCINET. Many of Moreno's studies and graphs also continue to be cited: Who Shall Survive still has just under 500 citations in Science Citation Index, and the sociometric studies that he did with Jennings at the Hudson School for Girls are models for a successful network intervention.
So why did Moreno's "invention" of sociometry seem to peter out by the 1940s? Freeman credits the other side of Moreno's personality for this -- voices, playing God, a series of marriages to beautiful women who encouraged him but turned his attention more and more to psychotherapy. (From my observations, Moreno was not the only famous social scientist who married a series of women, though his may have been more beautiful -- though, in truth, I never saw them.) "His commitment to mysticism, his bombastic personality, and his megalomania ... were too much for the regular members of the academic community to bear"(p. 42).
One of the most interesting chapters in the relatively unknown history of network studies was the Harvard group in the late 1920s and 1930s. Most of us are familiar with Homans' gloss on the Bank Wiring Room, also enshrined in UCINET's distributed data sets. Homans came to this interest back in the 1930s, when he reviewed the Western Electric material along with Elton Mayo, T. N. Whitehead, Fritz Roethlisberger, and Lloyd Warner. It was Warner who had suggested to Mayo that he focus on social structure and patterned interaction. Warner's own famous Yankee City studies had studied interpersonal networks and "produced 'literally tons' of empirical data"( p. 46). Warner organized the "Deep South" project, again enshrined in current network analyses. Not coincidentally, this led to the involvement of St. Clair Drake, who later became undergraduate Linton Freeman's charismatic mentor.
In the early 1930s, functionalist Henderson had organized a seminar around the sociology of Pareto that attracted the major sociologists then at Harvard (such as Parsons, Merton and Kingsley Davis), as well as such stars as Crane Brinton and Bernard DeVoto, along with Mayo's business school crowd. George Homans, "a recent Harvard graduate, a young aspiring poet" was hired as the seminar administrator (p. 55). Homans became a Harvard Junior Fellow, and in 1939 became an instructor at Harvard. In this day of degree credentialing it is worth noting that Homans never had a Ph.D. This lack obviously did not prevent him from developing theories of interaction and propositions about them, matters that he had started work on as early as the mid 1930s. Anthropologist Conrad Arensberg, who introduced Moreno to members of this circle, as well as William Foote Whyte (author of the classic Street Corner Society), were also Junior Fellows. And there were others, such as Eliot Chapple, who are now well known to structural analysis. Though charts and data abounded, Chapple and Arensberg advocated more formal mathematical analyses (even drawing upon electrical network circuit theory ) and tried to apply them to more open systems.
But the circle fell apart: Warner, Gardner, Davis, Drake and Whyte left for the University of Chicago, and Arensberg (who had some influence on me) went to Columbia. In part, there were some intellectual differences on whether or not one could or should make advances in what Homans called, in derogatory fashion, "micromeasurement." In perhaps greater measure, academic politics were at work. As a result, Harvard in the 1930s is rarely credited with having been an incubator of network analysis. I would add that a tension became apparent that has been an underlying motif in the development of network analysis: Developing mathematical rigor in studies of networks of limited size and within what Russ Bernard calls "network in a 'box'" -- that is, within the walls of a school, classroom, organization or the equivalent. Not that small closed systems are easy to analyze, but it is much more difficult to be rigorous about large open system social networks with fuzzy boundaries. The issue was cast at Harvard in part as an interest in qualitative field work versus quantitative measurement, but underlying it was the movement of the majority of the Harvard circle, especially the sociologists and anthropologists, towards an interest in large scale social systems. They simply could not see why graphing relations in bounded groups could contribute anything to an understanding of social systems. In part, it was a question of what flows could and should be studied: informal unnamed relationships in closed systems, or formal instituted role relations and obligations between named institutionalized statuses such as mother, or student. The flow of prestige in small groups did not seem applicable to the flow of social ranks and norms in large systems.
Then follows, almost up to the 1970s, what Freeman calls the "Dark Ages," in which various circles and research teams in different universities all worked on network analysis and theory. Each did very important work, but it resulted in no integrated cumulative effort. Some of us who have been around for a while might resent being assigned to the Dark Ages, but that is another matter -- and Freeman includes himself. In part, the disparate efforts in this period were not seen as linked because the term "network," which might have been applied as a blanket term to all the work, was absent. Much as Molière's bourgeois gentilhomme was startled to learn that he was speaking in prose, so most of the intellectuals Freeman recounts as having worked in the Dark Ages did not see their endeavors as being studies in social networks, and thus overlooked the commonalty between them.
For example, Kurt Lewin formed a stellar group at MIT that included just about everyone who became prominent in post-World War II social psychology and fundamentally defined the field. But his theory was called "field theory" and his shop was the "Research Center for Group Dynamics." They were interested in, among other things, the processes of communication and influence. Festinger, Schachter and Back produced a book called Social Pressure in Informal Groups (Festinger, Schacter and Back, 1950), but only the subtitle ("A Study of Human Factors in Housing") revealed that the major effort was a study of sociometry. After Lewin's death and the breakup of the Center, Alex Bavelas created the "Group Networks Laboratory" in which he, with the aid of mathematician Duncan Luce (one of my teachers at Columbia) and others worked on their famous experimental studies of communications in groups of five. But again, this group broke up as early as 1950. Festinger and Cartwright moved to the University of Michigan, and Festinger then went on to Minnesota -- and the term "network" was lost. In the meantime, mathematician Frank Harary was recruited at Michigan, bringing graph theory to what eventually became the network world. But their classic, Structural Models (Harary, Norman and Cartwrwright, 1965), only gets around to talking about networks in the last chapter, with one example of hypothetical data about population movements. Although there are a few examples of sociometric analyses earlier in the book (one from Cartwright's housing study), they are not associated with the idea of network.
There were other centers and starts. In the 1940s, a group at Michigan State worked on the sociometry of influence in rural sociology. This did not create an enduring school, but the studies of rural transmission of innovation eventually influenced Everett Rogers, who had 'discovered' sociometry in the course of his dissertation, in 1955. Later Rogers moved to Michigan State and was responsible for training some of the current major figures in the network field. Elihu Katz, of the Columbia University's personal influence school (including his mentor Paul Lazarsfeld, James Coleman, and Herbert Menzel) made use of the rural sociology tradition developing the "two step" -- later "N step" -- flow of communication and influence, also in the 1950s. The Columbia group influenced me (and Ron Burt) -- but Katz was surprised when I told him that he was one of the fathers of the network field, because the word "network" was originally as foreign to him as it was to me when I first started my work on social circles. Oddly, Robert Merton's theories of status and role sets were almost pure network theories but neither he nor his student Rose Coser, who has offered the most rigorous specification of these theories (Coser, 1975), used the term. I once argued with Coser that network methods and theories really could clarify her work. She remained dubious, much as the Harvard large social system school could not see the utility of network analysis.
One of the discoveries that surprised Freeman most was the pivotal work done at MIT in the late 1950s, sparked by political scientists Karl Deutsch and Ithiel Pool. The famous "underground" paper of 1958 that Pool wrote with Manfred Kochen about the probabilities involved in the "small world" was eventually published in the first issue of Social Networks (Pool and Kochen, 1978).
Freeman traces his own intellectual network heritage as beginning with lectures of St. Clair Drake (who had worked on the "Deep South" study) and, later, his reading of Bavelas and Leavitt's studies, as well as Bales and Moreno. At Syracuse, he and his colleague Morry Sunshine "failed at that time to make the connection among these several projects" (p. 111). Freeman's own work at Syracuse University in the 1960s is noted as a new approach to community studies, in which he and colleagues tracked who interacted with whom over which issues. I cited that work in my own "Power, Influence and Social Circles: A New Methodology for Studying Opinion-Makers" (Kadushin, 1968), but in studying Yugoslav elites I did not at the time use the term "network" and I do not believe that Freeman did in his studies either. Both of us found that the so-called "bias" of decision-making studies of elites (one got a different vision of who counted depending on who was asked) was in fact the very essence of the data, provided one systematically tracked who said what about whom and who did what with whom -- in short, analyzing what we would now call the social network of influence.
An important contributor starting in the 1960s was Edward O. Laumann who, like the Columbia group and Freeman in his early work at Syracuse, was essentially concerned with large systems and was mainly a survey researcher. Working first at Michigan and later at Chicago, he used the famous Detroit Area Study to develop early studies of large scale ego networks. He incorporated network ideas in studies of stratification and political elites and many of his former students became leaders in network studies. He is now, of course, famous for his surveys of sexual behavior, which is indeed another form of network behavior and is now of great interest to epidemiologists of AIDS.
The United States was hardly the only beacon of light during Freeman's Dark Ages. The group at Manchester University and the London School of Economics in the 1950s were well known as a fountain of network research and theory. Gluckman, Barnes, Mitchell and Nadel all used the term 'network,' which may account for their clear place in the current network field. The anthropologists also did important field work and interesting analyses. In Freeman's view, Radcliff-Brown influenced them all and "anticipated exactly the developments that took place almost forty years later" (p. 103). Curiously, Freeman does not mention Boissevain (1974) whose Friends of Friends remains my favorite introduction to the network field. Boissevain, who was closely related to the Manchester school, strongly attacks structural functionalists, however -- Radcliff-Brown in particular -- for failing to track actual relationships and the behaviors of "real people" and, rather, explaining behavior in terms of the "dominant values that support the system" (Boissevain, p. 12). This may not be a fair reading of structural functionalism, but the thrust of Boissevain's argument is that network data and theory provide a more substantiated view of what goes on in society.
Levi-Straus, in France, was not only a major influence on structuralism generally, but produced network graphs and worked with a leading French mathematician, André Weil. In the 1960s, Claude Flament wrote an important introduction to graph theory. In Amsterdam, Robert Mokken and his students worked on elite studies and interlocking directorates, and had important influence on network studies of elites. Bolstered by better training in mathematics typical of Dutch schools, the Dutch eventually became world leaders in mathematical statistical developments in network theory and analysis.
Yet Freeman feels that even by the end of the 1960s, with all these developments both in the United States and Europe, "no version of network analysis was yet universally recognized as providing a general paradigm for social research" (p. 121).
Starting in the 1970s, apparently something different happened. My account of this differs somewhat from Freeman's, in part because he modestly downplays his own role. There were several interrelated developments. First, two major training centers developed, one on the West Coast at UC Irvine, which Freeman headed -- and which benefited from the previous network interests of mathematical anthropologist Doug White and attracted another key mathematical anthropologist, Kim Romney. The other developed at Harvard under the aegis of Harrison White, who basically did it all by himself. Both must have been brilliant teachers because they inspired the core of network theorists and analysts of the last 30 years. True to the principles of networks, while their written work was important, their major contribution was the network of students who interpreted their ideas.
Then the field got "organized" and began to use what turned out to be the magical name ("network") in a fairly uniform way. The organization part came from the initiative of Barry Wellman, a Harrison White student, who was then and is still now at Toronto, who started an organization, INSNA (International Network for Social Network Analysis), which for many years was distinguished by having no formal organization, no bylaws, and no financial oversight. There was also a newsletter, Connections, which Barry edited. His practical efforts in creating community were matched by his theoretical and empirical studies of both geographically-centered communities and virtual distant communities, thus reaffirming the famous Kurt Lewin aphorism, "There is nothing so practical as a good theory." The organization grew in part from these centers, but also from a computer network called EIES. It was a precursor of now familiar technology: It combined some of the features developed much later by various forms of "groupware" together with an email system, a bulletin board and a list server, which Freeman used as the foundation of an electronic network of networks. While Freeman was still at Bucknell, he obtained an NSF grant to create an electronic network of social networkers. It was open to a group of about 40 people involved in the network field who were invited to a kickoff conference at Bucknell in January 1978. This system began to knit the field together. I personally had no patience with the slow (300 baud) text editor because the university DEC20 that I used from my office had a full screen editor. I quickly gave up. Others stayed with it and I believe gained greatly from it. In 1978, Freeman also started the journal Social Networks, which helped to define and formalize the field as an interdisciplinary venture.
In 1981, in Florida, Russ Bernard and Al Wolfe started the "Sunbelt Network" conferences, meeting every February in a warm place. It capitalized on EIES, and Barry Wellman was coopted: in part, the Sunbelt would serve as the annual meeting of INSNA and bring the field together, and abstracts of the papers were to be published in Connections. Perhaps just as important, Barry got away from cold Canada for a few days. As Freeman details in the book, there had been previous meetings of persons involved in the network field that brought together diverse persons who, with hindsight, we can now see became the major figures of the field. But these meetings were always by invitation only. The result was that some people felt themselves to be insiders and others felt left out. In contrast, the genius of INSNA was that anyone could belong; everyone was invited to the "Sunbelt" and anyone who wanted to could give a paper. Now, as a sign of the maturity of the field, INSNA and the Sunbelt have become institutionalized, for better or for worse. INSNA is a formal organization and the Sunbelt actually rejects some papers. Connections has a budget and referees papers.
Freeman notes the good feeling and lack of overt or nasty competition in the network field as it evolved. He attributes this civility to the interest in mathematical theory and precise definitions that it required, thus limiting any haggling over vague social science concepts. I agree about the unusual character of relations among network scholars, but my limited acquaintance with physicists suggests that mathematics may not be the answer. And certainly large egos are not absent in either physics or social networks. But social systems have a strong tendency to reproduce their origins and advance in either benevolent or malevolent spirals. Social networks was lucky in its organizational start as an arena for informal exchanges among people who felt they were pioneers and who were trying to solve problems. In this latter respect, a concern for algorithms that could make sense of messy network data may have helped, because participants in the field were always looking for something they could use on their recalcitrant data.
White and Freeman, while sharing a passion for formal theory and a genius for teaching, were otherwise quite different in their interests. White was more interested than Freeman in large scale social systems, yet the Harvard developed computer program for "blockmodelling" and the algebra of blocks was readily applicable only to small systems. The irony is that both of their interests in formal mathematics led to the meticulous investigation of small systems. I have no count of the number of times that the Sampson Monastery data with its eighteen members was analyzed according to different algorithms, but there are over 40 hits on Google. Social scientists interested in large scale social systems had to look elsewhere. The survey analysis of ego networks, trivial mathematically and not involving real whole networks, was one way to go that initially attracted Laumann and later Wellman, Fischer, and Burt. Those interested in corporate overlap and national elites had to use clustering methods that were centered around chopping large systems into manageable components using maximum likelihood methods and algorithms borrowed from (of all places) library science. The recent advent of fast, inexpensive,large scale parallel computing made it possible, in theory, to chop extremely large social systems into analyzable parts -- but for various reasons these methods and algorithms have not made much of a dent on the field.
Enter the physicists, capitalizing on some old network theory ideas such as "small world" but armed with excellent mathematical and computational skills and, perhaps even more important, a tradition of making simple models of complex phenomena. This work, led by a wave of newcomers to social networks such as Watts, Newman, and Barabasi, jumpstarted a new literature that was, at its beginning, unaware of most work in the existing network field. Freeman shows that citation using the concept "small world" produces two nearly distinct communities of researchers: the "old" network experts and the "new" physical scientist modelers (who benefit, as Freeman points out, from different publishing patterns and norms in the physical as opposed to social sciences). But these new arrivals, who now offer tools for the possible exploration of very large systems, are becoming ever integrated into the social network small world -- though their initial use of classic social science flows and classic ideas of social structure were limited. I suppose they may be considered well integrated when easy-to-use computer programs are developed that will analyze huge data sets without requiring a physicist's mathematical sophistication to be applied to traditional concerns of social system theory.
Which brings me to an odd and not necessarily triumphalist conclusion. Why did social networks eventually take off as a field in the 1970s? It was the combination of important centers of influential teachers, the network organization of the network field (the mirror on the mirror), the universal use of the name "network" -- and also the development of several relatively easy-to-use computer programs that operated on desktop computers, which managed network data and which encapsulated the mathematical insights that Freeman found so crucial to the transformation of the field from one of 'network as metaphor' to 'network as a mathematical expression.' These programs, for better and only sometimes for worse, allowed social scientists with minimal formal skills to employ hard-nosed analytic techniques to an enormous variety of applications, much as computers have enabled complex statistics to be applied to data sets whose analysts barely understand what they are doing. We may shake our heads at such apparently thoughtless work, but much of it has advanced social science and given us new understandings. What has resulted, however, is that the social network field has often been regarded by outsiders as a set of tools rather than as a set of integrated, interesting concepts and propositions. There remains a large world for grounded social network ideas to conquer.
 Though Freeman admits that Homans was inspired by Pareto's association of sentiments with interaction, Freeman, a confirmed anti-psychological reductionist also believes that Pareto's sociology is "made up almost entirely by psychobabble" (p. 54), a reading not shared by many serious structural thinkers (e.g. Zetterberg in European Proponents of Sociology Prior To World War I, 1968 and http://www.zetterberg.org/Books/b93e_Soc/b93eCh4.htm). Freeman also observes that, despite his being a psychiatrist, Moreno contributed to structuralism at a time "while mainstream sociology had become psychologistic" (p. 36). Anyone who has struggled through Parson's Structure of Social Action might not read mainstream sociology as "psychologistic."
 Classical network theory in the physics of electrical circuits deals with flows through nodes, or what are called "ports." Notable are Kirchoff's laws that deal with conservation of energy in circuits between ports. These ideas are seductive, but it turns out that electrical circuits are far too orderly for direct application to social networks.
Boissevain, J. 1974. Friends of Friends: Networks, Manipulators and Coalitions (London: Basil Blackwell).
Coser, R.L. 1975. "The Complexity of Roles as a Seedbed of Individual Autonomy." In: Coser L. A. (Ed.), The Idea of Social Structure: Papers in Honor of Robert K. Merton (New York: Harcourt Brace Jovanovich), pp. 237-262.
Festinger, L., Schacter, S., Back, K. 1950. Social Pressures in Informal Groups: A Study of Human Factors in Housing (Stanford, CA: Stanford University Press).
Gertner, J. 2003, December 14. "The 3rd Annual Year in Ideas," New York Times Magazine (Late Edition, Final, Section 6): p. 92.
Harary, F., Norman, R.Z., Cartwrwright, D. 1965. Structural Models: An Introduction to the Theory of Directed Graphs (New York: Wiley).
Kadushin, C. 1968. "Power, Influence and Social Circles: A New Methodology for Studying Opinion-Makers," American Sociological Review 33: pp. 685-699.
Krebs, V. E., 2003. Orgnet.com. Available: http://www.orgnet.com/.
Pool, I. D. S., Kochen, M. 1978. "Contacts and Influence," Social Networks 1 (1): pp. 5-51.