The Harvard Breakthrough

The arguments of Mitchell, Barnes and Bott were extremely influential in Britain (see Frankenberg, 1966), but their very success meant that social network analysis came to be identified with the specific ideas of the Manchester anthropologists. That is to say, network analysis was seen to be concerned specifically with informal, interpersonal relations of a 'communal' type, and the method was seen as specifically concerned with the investigation of egocentric networks. As a result, the crucial breakthrough to the study of the global properties of social networks in all fields of social life was not made in Britain.

It was, in fact, at Harvard that this breakthrough occurred. A decade after Homans's initial explorations, a trickle of papers began to appear from Harrison White and his associates which pushed the analysis much further. Soon, the work of students and colleagues of the authors of these publications produced a torrent of papers which firmly established social network analysis as a method of structural analysis.

The key elements in this breakthrough were two parallel mathematical innovations (see the discussion in Berkowitz, 1982). The first of these was the development of algebraic models of groups using set theory to model kinship and other relations in the spirit of L6viStrauss. This led to a re-consideration of the early work in graph theory and in other branches of mathematics and to the attempt to use algebraic methods to conceptualize the concept of 'role' in social structure (White, 1963; Boyd, 1969; Lorrain and White, 1971).

34 Social network analysiy

White's continued explorations of 'block Modelling' (see Chapter 7 below) can be seen as carrying forward the very emphasis on role structure to which Nadel had pointed. The second innovation was the development of multidimensional scaling, a 'scaling' technique for translating relationships into social 'distances' and for mapping them in a social space. Very much in the tradition of Lewin's work on field theory, these developments proved extremely powerful methods of analysis. (For applications in sociology see Laumann, 1966 and Levine, 1972.)

The confluence of these two strands led to the important and influential work of the new Harvard group centred around White (see Mullins, 1973). White had moved to Harvard from Chicago, and his work retained important links with that of Davis and others, who had elaborated on the basic sociometric views throughout the 1960s. The Harvard group developed as mathematically orientated structural analysts, concerned with the modelling of social structures of all kinds. There was no single theoretical focus to their work, the unifying idea being simply that of us,' ng algebraic ideas to model deep and surface structure relations. It was network analysis as a method that united them. The public reception of Granovetter's article (1973) popularized this viewpoint in American sociology and helped to stimulate many other studies. Although it was not a highly technical piece of mathematics - or, perhaps, because of this Granovetter's work was of central importance as a charter statement for popularizing and legitimating the position (see also Granovetter, 1982). Although many researchers continued to work in such areas as the analysis of 'community' structure, others were interested in such phenomena as corporate interlocks and so helped to move network analysis away from its 'residual' focus on purely interpersonal relations. In doing so, they stimulated numerous substantive applications of the techniques. Much of the effort of the Harvard group - no longer based solely at Harvard - was centred on the International Network for Social Network Analysis (INSNA) at Toronto, which acted as a focus for the development of social network analysis under the leadership of Wellman and Berkowitz, both former students of White .21

Two classic studies by Granovetter and by Lee grew out of the earliest discussions of the Harvard school. While they were not explicitly algebraic in their approach, they became important exemplars for others. This was not least because they offered substantive and analytical continuity with earlier sociometric work.

Granovetter's work on Getting A Job (1974) started out from a critical consideration of attempts by labour economists to explain how people find work. In particular, he wanted to explore the ways

Development of social network analysis 35

in which people acquire information about job opportunities through the informal social contacts that they have. His interest was in the kinds of links involved in the transmission of information, whether these were 'strong' or 'weak', and how they were maintained over time. To this end, he selected a sample of male professional, technical and managerial workers in a suburb of Boston who had changed their jobs during the previous five years. Granovetter found that informal, personal contacts were the primary channels through which individuals found out about job opportunities: 56 per cent of his respondents relied on this means, and this was particularly true for information about the higher-paying jobs. These results were not especially striking, being broadly in line with earlier research, and Granovetter set himself the task of identifying those who provided information and the circumstances under which

they passed it.

Granovetter showed that the 'rational' choice of methods for acquiring job information was of little importance. Individuals did not really compare the rewards and costs attached to different sources of information, and there was little active 'search' for jobs. Instead, information was acquired accidentally whenever contacts volunteered the information. The most important people in providing information were work or work-related contacts. They were rarely family or friends, and they tended to be people who were in different occupations from the respondent. The probability that a person would make a job change was dependent on the proportion of work contacts who were in different occupations from him or herself.

To explain these findings, Granovetter drew on an information diffusion model. Those people with job information were assumed to pass this on to a certain proportion of their immediate contacts, who passed it on, in turn, to a certain proportion of their contacts, and so on. Assuming that the information attenuates over time as it passes through subsequent links in the chain, 22 it is possible to track its passage through a social network and to discover the number of people who will acquire the information and their various locations in the network. The acquisition of information, therefore, depends upon, first, the motivation of those with information to pass it on, and, second, the 'strategic' location of a person's contacts in the overall flow of information (Granovetter, 1974: 52).

It was at this point in his argument that Granovetter introduced his now-famous argument on 'the strength of weak ties'. The importance of strong ties is well understood. Those to whom a person is closest (family and close friends, workmates etc.) have many overlapping contacts with one another. They all tend to know

36 Social network analysis

and to interact with one another in numerous situations and so there is a tendency for them to possess the same knowledge about job opportunities. Information which reaches any one of them is more than likely to reach them all. Conversely, they are less likely to be the sources of new information from more distant parts of the network. The information received is likely to be 'stale' information already received from someone else. It is through the relatively weak ties of less frequent contacts and of people in different work situations that new and different information is likely to become available. What this means is that 'acquaintances are more likely to pass job information than close friends' (Granovetter, 1974: 54). In almost all cases studied by Granovetter, information came directly from an employer or one of the employer's direct contacts - there was, typically, a maximum of one intermediary. Links through more than two intermediaries were very rare. It was the short, weak chains of connection that were of greatest significance in the receipt of useful job information.

A comparable, and slightly earlier study was Lee's work on The Search for an Abortionist (1969). In a situation where abortion was illegal, Lee wanted to discover how women acquired information about the opportunities for such terminations. Doctors who undertake illegal terminations cannot advertise and must often operate from hotel rooms rather than from clinics. Those who seek an abortion must, therefore, try to obtain information from those of their friends and acquaintances who may have had some experience with abortion in the past, as these people are likely to have that information or to be able to put them in contact with others who can help.

To study this process, Lee contacted abortionists and women who had had recent experience of an abortion. In constructing her sample she was, interestingly, having to use information search techniques which were similar to those used by the women themselves. Like Granovetter, she used a mixture of interviews and questionnaires to gather her data. Having explored various aspects of their life and social background and their attitudes towards conception and abortion, Lee turned to an examination of their search for an abortionist. The search involved the making of informed guesses about who might be able to help, either by providing the name of an abortionist or mentioning a further contact who might help. Lee found that women approached an average of 5.8 people before successfully contacting an abortionist the actual numbers of contacts ranging from 1 to 31. A number of the contacts, of course, were 'dead ends', and the 'successful chains' varied in length from one to seven steps, the average length being

Development of social network analysis 37

2.8. Over three-quarters of the successful chains involved two or fewer intermediaries (Lee, 1969: Chapter 5). Contacts tended not to be relatives or those in authority (employers, teachers etc.), and the most important channels were female friends of the same age.

Both Granovetter and Lee explored network processes through the use of simple frequency tabulations, making only qualitative comments on the structure of the network relations that they discovered. Indeed, Lee argued that it is extremely difficult to trace the structure of overlapping personal networks in large-scale systems. Their studies were, however, important as outgrowths of and contributors to the systematic and analytical development of social network analysis. Their studies showed the power of even the most basic of social network methods, and they suggested an immense power for the more rigorous techniques being developed by their Harvard colleagues.

The power of social network analysis has become apparent in its use as an orientating idea and specific body of methods. But the application of formal mathematical ideas to the study of social networks has encouraged some writers to suggest that social network analysis offers the basis for a new theory of social structure. Barnes and Harary (1983), for example, have argued that it is possible to advance from the use of formal concepts to the use of formal theory. They argue that the promise of social network analysis can be realized only if researchers move beyond the use of formal concepts for purely descriptive purposes (see also Granovetter, 1979). Mathematics consists of theorems which specify the determinate logical links between formal concepts. Barnes and Harary argue that if the formal concepts prove to be useful ways of organizing relational data, then the theorems too should be applicable to those data. The application of theorems drawn from formal mathematics, then, 'reveals real world implications of the model that might otherwise have not been noticed or utilized by the designer of the model' (Barnes and Harary, 1983: 239).

Some have gone further, and have suggested that developments in social network analysis already point the way to novel frameworks of sociological theory, or to the re-assertion of earlier theories. Particularly influential, for example, have been advocates of an exchange theoretical perspective on social networks (Emerson, 1962, 1964; Cook, 1977, 1982), which is associated with wider 'transactionalist' approaches (Bailey, 1969; Boissevain, 1974) and rational choice theories (Lin, 1982). (See also the discussion in Banck, 1973 and van Poucke, 1979.) Whether social network

38 Social network analysis

analysis will, in the long run, point to the predominance of a particular theoretical framework is not a matter that will detain me in this book. It is undoubtedly the case that social network analysis embodies a particular theoretical orientation towards the structure of the social world and that it is, therefore, linked with structural theories of action. But it seems unlikely that any one substantive theory should be regarded as embodying the essence of social network analysis. The point of view that I shall elaborate in this book is that social network analysis is a particular set of methods and not a specific body of theory.

Return to Table of Contents