Interpersonal Configurations and Cliques
Theoretical work in the sociometric tradition, I have argued, has involved a considerable effort to uncover ways of decomposing networks into their constituent sub-groups. This search for what have variously been termed as 'cliques', 'clusters', or 'blocks' has also been a feature of the research tradition which developed at Harvard University during the 1930s and 1940s. In this line of work, the investigation of 'informal relations' in large-scale systems led to the empirical discovery that these systems did, in fact, contain cohesive sub-groupings. The task that the researchers then faced, and only partly solved, was to discover techniques which could disclose the sub-group structure of any social system for which relational data were available.
Radcliffe-Brown and, through him, Durkheim were the major influences on this tradition of research. Radcliffe-Brown's ideas had been especially influential among anthropologists in Australia, where he had taught for a number of years. His influence was particularly strong in the work of W. Lloyd Warner, who moved to Harvard in 1929 to join his fellow Australian, the psychologist Elton Mayo. The two men worked together in a series of closely related investigations of factory and community life in America, and they saw these investigations as applications of the structural concerns of Radcliffe-Brown.
Development of social network analysis 17
Mayo had moved from Australia to Harvard in 1926 in order to take on a leading role in the newly developed research programme of its business school. His principal contact with sociological ideas was through the dominating influence of the biologist L. J. Henderson, who actively promoted the work of Pareto among his Harvard colleagues. Henderson held that this was the only appropriate basis for a truly scientific sociology and that it was, furthermore, the only viable political bulwark against revolutionary Marxism. Mayo's psychological concern for individual motivation was complemented by a growing awareness of what Pareto termed the 'non-rational' components of action. Economic action, for Mayo, was not a purely rational form of action, but was structured also by non-rational sentiments such as those of group solidarity. Pareto was also the great exponent of 'elite' theory, and Mayo saw that a managerial elite which recognized this influence of group relations on economic motivation could most successfully control worker behaviour. Warner's contribution to the Harvard research programme, as befitted a trained field worker, showed a greater concern for detailed investigations of the actual patterns of group behaviour which could be found in particular social settings. To Mayo's theoretical and 'applied' concerns, Warner brought an empirical dimension. Despite these differences - or, perhaps, because of them - the work that the two began at Harvard was crucially important in the development of social network analysis. Their careers overlapped there for only six years, but their research proved massively influential. The major projects which they and their colleagues undertook were investigations of the Hawthorne electrical factory in Chicago and a study of the New England community of 'Yankee City'.
The Hawthorne studies have become classics of social investigation, and they need little discussion here (see the useful discussion in Rose, 1975). Briefly, a series of studies of worker efficiency had been undertaken during the 1920s by managers in the Hawthorne works of the Western Electric Company in Chicago. These managers were attempting to discover how alterations in the physical conditions of work (heating, lighting, rest periods and so on) affected productivity, and they discovered, to their considerable surprise, that productivity increased almost regardless of the particular changes that were made. In an attempt to understand these paradoxical results, the managers called on Mayo and his Harvard team for some guidance in restructuring the research programme. Mayo concluded that the crucial factor responsible for increased productivity had been the very fact of participation in the research
project: the workers were pleased that their managers were taking an interest in them, and their sense of involvement and integration into the life of the factory motivated them to greater efforts.
With the advice of Warner, the Hawthorne investigators began an anthropological' study, an observation of work group behaviour in a natural setting in the factory. The scene of their observations was the bank wiring room, and the team approached their research in the factory in the same way that a social anthropologist would carry out fieldwork in a village in an alien society. They recorded all that they could observe of group behaviour, aiming to construct a full anthropological account. The importance of the Hawthorne studies to the development of social network analysis lies in their use of sociograms to report on group structure. Just as the kinship structure of a village community might be illustrated using a genealogical diagram, the Hawthorne team constructed sociograms to illustrate the structure of informal relations within the work group.
The principal report on the Hawthorne studies (Roethlisberger and Dickson, 1939: 500 ff.)-5 includes various sociograms constructed by the research team. They saw these as reflecting the 'informal organization' of the bank wiring room, as opposed to the formal organization which was depicted in the managerial organization chart. Sociograms were constructed to show each of a number of aspects of group behaviour: involvement in games, controversy over the opening of windows, job trading, helping, and friendships and antagonisms. The Hawthorne study was, I believe, the first major investigation to use sociograms to describe the actual relations observed in real situations. In their diagrams, people are represented by circles and their relationships by arrows. The similarity of these diagrams to the sociograms subsequently developed by the group dynamics researchers are obvious, but the researchers give no indication of how they hit upon the idea of such diagrams. There is, for example, no discussion of the evolving work of Moreno. It will be seen from Figure 2.4, however, that the diagrams resemble not only the formal organization charts which were used by managers, but also the electrical wiring diagrams that would have been a very familiar feature of the plant. It must be assumed that the influence of Warner must have encouraged the researchers to adapt conventional anthropological kinship diagrams by drawing on these other influences of the organizational setting.
In drawing the sociograms of the bank wiring group, certain general conventions were followed, but these were artistic rather than sociological conventions. The precise location of each circle on the page was decided by the artist, the principal constraint being
Development of social network analysis 19
Figure 2.4 A Hawthorne sociogram
simply that any sub-groups identified by the observers should be drawn as close to one another as possible. Apart from this, purely artistic principles of clarity and simplicity governed the design: the number of lines which cross one another, for example, should be as small as possible, and the lines should not vary too much in length. The sub-groups identified by the researchers - they called them 'cliques' - were those which the workers themselves recognized as important elements of their situation. Much as any anthropologist might use 'native' categories and concepts as pointers to the structural features of group life, the workers' own terms were taken as indicators of the existence of 'cliques'. 'The group in front' and 'the group in back' were identified from observations of group behaviour and from group vocabulary as the two sub-groups within the bank wiring group. There was no attempt to use the sociograms themselves to identify sociometrically defined 'cliques" the socially perceived sub-groups were simply mapped on to the sociograms. Having plotted group structure in this way, however, the researchers made little further use of them. They appear to have lacked any theoretical understanding of how social networks might shape the behaviour of individuals.
Warner, meanwhile, had begun a study of the small New England city of Newburyport, to which he gave the pseudonym 'Yankee City'. His fieldwork was carried out between 1930 and 1935, and the research was conceived as a full-blown anthropological study of a modern, urban community. As such, it combined observation with the use of interviews and historical documents. The end of the main phase of fieldwork, however, coincided with a growing antagonism between Warner and Mayo, and Warner left Harvard for Chicago University, where his mentor, Radcliffe-Brown, was already a visiting professor. Warner and Radcliffe-Brown had two years together at Chicago, a period when the analysis of the fieldwork material from Yankee City would have been at its most intense.
20 Social network analysis
Warner spent the rest of his career at Chicago, and it was from there that he supervised and sponsored a number of related studies, most importantly that of 'Old City' in the American Deep South.
Warner's own early work had used the methods and ideas of Durkheim and Radcliffe-Brown in the traditional manner to study an Australian tribe, and it was through his contact with Mayo that he first formulated the idea of applying anthropological methods to the study of a modern urban community. Warner had originally intended to study the district of Chicago in which the Hawthorne works were located, but the work of the Chicago school of sociologists forced him to conclude that the district was 'disorganized' and so would not be amenable to anthropological investigation (Park et al., 1925). Warner felt that only in New England and in parts of the southern United States would he find the kind of established and integrated communities that he wished to study.
Warner's work shows a rich variety of theoretical influences. While the influence of Radcliffe-Brown was uppermost, he allied this with an organismic, systems model of society which, undoubtedly, shows the influence of Henderson's interpretation of Pareto. This led Warner to emphasize such factors as stability, cohesion and integration in the structuring of communities. But he also drew on Simmel's ideas of reciprocal relations and of the influence of numbers on group life. It was, I have suggested, Simmel (1908) who pioneered the analysis of dyads and triads as the building blocks of social life. Following the terminology of Simmel and other German sociologists, also adopted by Moreno, Warner talked of social 'configurations', holding that the social organization of a community consists of a web of relations through which people interact with one another.
The social configuration which comprises a modern community, argued Warner, consists of various sub-groups, such as the family, the church, classes and associations. Alongside these is also to be found the sub-group that he terms the 'clique': an informal association of people among whom there is a degree of group feeling and intimacy and in which certain group norms of behaviour have been established (Warner and Lunt, 1941:32). A clique is 'an intimate non-kin group, membership in which may vary in numbers from two to thirty or more people' (Warner and Lunt, 1941: 110).8 For Warner, therefore, the clique has the same social significance in community studies as the informal group had in the Hawthorne factory studies. The concept describes a particular configuration of informal interpersonal relations.
Development of social network analysis 21
The Yankee City researchers claimed that a large number of these cliques could be identified in the city. The major cliques were the groups that many Yankee City respondents referred to by such terms as 'our crowd', 'our circle', and so on. Having discovered the existence of these cliques from the comments made by those they studied, Warner and his associates claimed that they were second in importance only to the family in placing people in society. People are integrated into communities through 'informal' and 'personal' relations of family and clique membership, not simply through the 'formal' relations of the economy and political system. Any person may be a member of several different cliques, and 'such overlapping in clique membership spreads out into a network of interrelations which integrate almost the entire population of a community in a single vast system of clique relations' (Warner and Lunt, 1941: 1 1 1). This is undoubtedly one of the earliest, if not the earliest use of network terminology to describe the structuring of whole societies into sub-groups.
The Yankee City reports used various diagrams to model such things as class structure and family organization, and it is hardly surprising that the researchers also constructed clique diagrams. To represent the social structure that they described, they drew cliques as a series of intersecting circles in a Venn diagram (Warner and Lunt, 1941: 113), but they do not advance to any formal, structural analyses of these diagrams. In the second volume of the Yankee City report, however, there was an attempt to undertake what would now be termed a 'positional analysis' (Warner and Lunt, 1942: 52, Figure X). They present a series of matrices which show the numbers of people occupying each of a number of structurally defined positions. Figure 2.5 shows the format of one of these diagrams. Having identified six classes and 31 types of clique in Yankee City, Warner and Lunt cross-classified class and clique membership in a data matrix. Each type of clique was defined by the predominant class composition of its overall membership, and the cells of the matrix show the numbers of people in each class who were members of each of the 31 types of clique.9 From among the large number of possible combinations - 6 times 31, or 186 they argue that only 73 'positions' actually occurred. All the remaining cells in the matrix were empty. By constructing similar matrices for class against each of a number of other social groupings (types of formal association, types of family etc.,) they were able to combine the matrices together, stacking them one on top of another, and they identified 89 structural positions in the overall ' combined networks The particular procedure that they employed
22 Social network analysis
Figure 2.5 A matrix of cliques
was rather cumbersome, and it is unnecessary to go further into its outmoded operation, but the Yankee City work remains interesting for its attempt to pioneer such methods of formal structural analysis.
Colleagues of Warner began an investigation of 'Old City', in the southern United States, during 1936, and in this research they further explored the idea of the 'clique' (Davis et al., 1941). In looking at 'coloured society' in Old City, they follow Warner's method of seeing cliques as intersecting circles, mapping the overlapping memberships of the most active cliques in a 'space' defined by class and age (Davis et al., 1941: 213, Figure 12). They referred to 'social space' and its 'two dimensions', but there is no explicit mention of any of the work of Lewin on topological field models. The major innovation of this study was its attempt to explore the internal structure of cliques. The researchers argued that a clique could be seen as comprising three 'layers': a 'core' of those who participate together most often and most intimately, a 'primary circle' of those who participate jointly with core members on some occasions but never as a group by themselves, and a ,secondary circle' of those who participate only infrequently and so are 'almost non-members'. On the basis of their investigation of 60 cliques, using similar techniques to those of the Yankee City researchers, they suggested a number of structural hypotheses
Development of social network analysis 23
about the connections between cliques. They argue, for example, that peripheral, lower class members of a clique might be able to contact higher class members of another clique only through the higher class core members of their own clique.
The ideas that emerged in the Hawthorne, Yankee City and Old City research developed in parallel with those of the sociometric tradition of small group research, but there is no evidence that the leading figures in the two traditions were even aware of one another's work during the 1930s and 1940s. In the work of George Homans, however, there occurred the first important intersection of these two strands of research. Homans, a faculty member in the Harvard sociology department, was dissatisfied with the 'grand theory' of Harvard colleagues such as Parsons, which he felt operated at a much too abstract level of analysis. Homans felt that social theory had to be built up from the foundations of a firm understanding of small-scale social interaction. To this end, he began, during the late 1940s, to try to synthesize the mass of smallgroup research that had been undertaken in the United States. He aimed at nothing less than a theoretical synthesis of this work, drawing on the experimental work of the social psychologists and the observational work of sociologists and anthropologists. His theoretical synthesis centred around the idea that human activities bring people into interaction with one another, that these interactions vary in their 'frequency', 'duration', and 'direction'," and that interaction is the basis on which 'sentiments' develop among people. Homans saw Moreno's sociometry as providing a methodological framework for applying this theory to particular social situations. To illustrate his ideas, he re-examined a number of earlier studies.
One section of the Old City report has achieved considerable fame - among network analysts at least - because of its re-analysis by Homans. In this section, Davis and his colleagues used matrix methods to look at the involvement of 18 women in 14 social events (Davis et al., 1941: Chapter 7)." Homans took these data, presented them in matrix form, and set out one of the first published statements of the method of 'matrix re-arrangement' in social network analysis (see also Festinger, 1949). The Old City matrix shows 18 rows (women) and 14 columns (events), with an 'x' entry placed in a cell to represent the participation of a particular woman at a specific event. The raw matrix, argued Homans, was not necessarily arranged in any significant order - the columns, for example, were simply arranged in the date order of the events. For this reason, the crosses appear to be distributed at random across
24 Social network analysis
the matrix. A re-arrangement of the rows and columns of the matrix, bringing together the events in which particular women predominate, would, he believed uncover important structural features of the clique. He described his method as follows:
we put in the center the columns representing events ... at which a large number of women were present, and we put toward the edges the columns representing the events ... at which only a few women were present. As far as the lines [rows] are concerned, we put together toward the top or bottom the lines representing those women that participated most often together in social events. A great deal of reshuffling may have to be done before any pattern appears. (Homans, 1951: 83)
Homans argued that this 'reshuffling' must go on until the distribution of the crosses in the cells shows a clear pattern, and he produced a re-arranged matrix in which there were clear signs of a division into two 'cliques' among the women: there were two distinct clumps of crosses in the re-arranged matrix. Homans's method is analogous to what has subsequently come to be called 'block modelling', but he made no use of any formal mathematical methods. In fact, his re-arrangement seems to be simply a trial-anderror process which continued until he was able to spot an apparently significant pattern.
Figure 2.6 shows a simplified version of the re-analysis undertaken by Homans. The matrices show artificial data for the participation of eight people in eight events. In matrix (i), the 'X' entries are scattered evenly across the whole matrix, but a re-arrangement of the rows and columns into the order shown in matrix (ii) brings out a structural opposition between two distinct sub-groups: Ann, Chris, Ed and Gill participate together in events 1, 3, 5 and 7, while Beth, Don, Flo and Hal participate jointly in events 2, 4, 6 and 8. There are two separate sets of people and two specific categories of events. It can be appreciated that re-arrangement by trial-and-error would not be such an easy task, even for such a small matrix, when the data are not so tightly structured as in this artificial example. The real data on 18 women and 14 events would have taken a considerable amount of time to analyses There is, furthermore, no certainty that the final results produced by Homans would be the same as those produced by any other researcher, as there are no criteria by which a 'correct' result can be identified. It is for these reasons that later attempts at this kind of analysis have involved a search for programmable algorithms, so that computers can reliably undertake the task of re-arrangement.
To illustrate his position further, Homans re-analysed the Hawthorne data on the bank wiring room. Using the sociograms
Development of social network analysis 25
Figure 2.6 Matrix re-arrangement
constructed by the observers, he looked at the cliques which Roethlisberger and Dickson had identified (Homans, 1951: 66-70). But Homans retained these original clique identifications, and did not attempt a sociometric investigation of clique structure along the lines of his analysis of the Old City data. He does imply, however, albeit without elaboration, that the matrix re-arrangement method had been used by the original Hawthorne researchers (Homans, 1951: 84). 13
The theoretical framework that Homans constructed to explain group behaviour was an elaboration of the model of the early smallgroup researchers, in which the group is understood as a system within an environment. He divides the structure of any group into
26 Social network analysis
an 'internal system', which expresses the sentiments that arise through the interactions of its members, and an 'external system' through which group activities are related to the problem of environmental adaptation. 14 The environment itself consists of the physical, technical and social contexts of group behaviour. Homans's main concern was with the internal system, which he saw as a more scientific concept than that of the 'informal organization' which it designates. His interest, therefore, was in the scientific elaboration of the insights of research on informal organization by translating these insights into propositions about the structure of 'internal systems'.
To this end, he set up a number of hypotheses about the internal system, starting from the assumption that people who interact frequently with one another will tend to like one another and that, as the frequency of their interaction increases, so the degree of their liking for one another will increase. If there are frequent interactions in the external system, because of such environmental constraints as the demands imposed by supervisors and managers, then the members of the work group will tend to develop sentiments of liking and will engage in further interactions with one another, unrelated to the needs of the external system. It is in this way, he argues, that the internal system gets elaborated into complex social configurations divisible into cliques. 15
Despite the power of Homans's theoretical synthesis of sociometric and anthropological research, there were few major advances that were directly inspired by his work. Homans himself became increasingly concerned to explore the explanation of social behaviour using behaviourist and rational choice models, and he became identified with the framework of 'exchange theory' (Homans, 1961). Robert Bales, a colleague of Homans, carried out some interesting small-group research (Bales, 1950), but he did not use a sociometric approach to his work and became increasingly linked with Parsonian structural functionalism (Parsons et al., 1953). The work of many who had contributed to the development of the idea of balance returned to exclusively psychologistic concerns, and the influential text of Festinger (1957) became an important charter statement in directing these researchers back into the social psychology of perception. The area of group dynamics all but stagnated, with most advances being in the purely mathematical problems of balance, cliques and clusters. While these mathematical explorations were to prove important and fertile sources for the advances later made by Harrison White, they had little impact on the shape of social research during the 1950s and 1960s.
Return to Table of Contents