Network analysis represents a very systematic means of examining the overall configuration of relationships within social systems. Social units (e.g., individuals, groups), and relationships, represented at points (nodes) and the communication among the units provides the basis for study. Network analysis provides a form of study used across the academy, including communication, and even the natural sciences, to study specific problems. The wide range of academic disciplines that have found social network analysis appealing has also resulted in a literal Tower of Babel with central concepts like relationships described by many different terms (e.g., edges, arcs). This entry will use terminology traditional to the communication discipline.
Contexts are most often manifested in network studies with the operationalizations of entities. So, if one is interested in an industry nodes might be defined as separate organizations, if one is interested in a team then nodes would represent individuals who are its members. Figuring out how to make a representation of the system, particularly where to draw the boundaries between separate networks, provides a fundamental challenge. Boundaries imply some quality of interaction demonstrably different within the network than communication outside the network, particularly with other networks. The two contrasting approaches involve the realist approach where the researcher adopts the vantage point of the actors (e.g., people with whom one conducts research) and the nominalist approach which would create a conceptual framework (e.g., all members of each academic department—economics, philosophy, communication).
Network analysis examines the relationships between nodes, referred to as “links.” Links represent the communication between nodes that serves as the basic datum. However, in most network analyses, a limited understanding of the function [Page 1634]and quality of communication becomes employed in the analysis. For example, viewing the number of messages between two nodes in a network may be used as a measure of the “strength” of the network, rather than any assessment of the quality or content of the communication.
Links have several properties. Traditionally network analysts have examined their content (e.g., production, innovation, maintenance). Asymmetry is an important property of organizational networks since there are a multitude of differences between organizational members, especially in term of status and the direction of communication. Strength (e.g., importance, frequency) of ties is often examined. The strength of weak ties is perhaps the most well-known concept related to network analysis. A “strong” tie between nodes may exist, even though the number of messages remains relatively small. Reciprocity involves whether each node would characterize the relationship in the same way. Quite often, a linkage lacks reciprocity, one member of the link views the relationship as stronger than the other node. Channels might include written, face-to-face, telephone, or telecommunication networks.
The manner in which these various properties of links are combined can determine the analytical power and depth of any one network analysis. For example, a multivariate network might simultaneously measure weight, frequency, and duration of a link. Multiplexity refers to the nature of overlap, or correspondence, between differing networks (e.g., friendship as opposed to work). The degree of multiplexity has been related to such issues as the intimacy of relationships, temporal stability of relationships, reduction of uncertainty, status, the degree of control of a clique over its members, performance, redundancy of channels, and the diffusion of information.
Measurement issues create some challenges and areas for consideration when examining network analysis. The lack of robustness in the network analysis makes the approach one not favored by scholars.
A combination of data gathering and computer analysis problems sharply limits the size of networks which can be examined. In practice the difficulties associated with the collection of the data creates ceilings on the use of network analysis methods (e.g., observational techniques can be only used with very small n’s). These problems are exacerbated by the difficulties associated with sampling from populations to obtain network data. The advent of web crawlers and other tools primarily developed by physicists interested in network analysis has changed this somewhat, with special tools available to measure massive data sets; big data associated with the Internet.
Recently, human subjects review committees have raised objections to asking respondents to report on behavior involving others who may not consent to the data collection. The problem is that network analysis requires identification of all the respondents. Unlike survey or experimental research, the ability to provide anonymity and confidentiality proves difficult.
Network analysis involves a variety of methodological difficulties in data collection. For example, missing data when analyzing interactions creates serious problems in determining which relationships to analyze. There is also considerable divergence of views as to what is the most important, subjective or objective measurement of networks, which is related to the problem of whether people can accurately self-report their communication linkages. Typically, a network analyst makes a tradeoff between simplicity at the dyadic level to examine complexity at the social system level.
Fortunately, especially in terms of the automated auditing of network data, there have been a number of systematic attempts to come to grips with measurement issues that also addresses some large problems, with small world investigations tackling the whole Internet. For example, examining an entire body of messages over e-mail involving legal actions (E-Discovery, Enron) provides the basis for measurement of network content.
In response to measurement issues, a number of network studies approach using an egocentric approach. The focus is on the radial network which considers the focal network from the point of view of one individual and examines the network of relationships with others. This approach permits more use of more traditional survey research procedures and statistical analysis.
Network analysis permits the employment of a variety of techniques in the examination of various configurations of relationships. Partly because of the focus on individual roles in sociology, social psychology, and in organizational theory, early work in communication network analysis focused on typologies of network roles. Richards’ typology of network roles embedded in his NEGOPY network analysis program led to research that focused on individuals who had limited contact with others and those who were more involved, especially the critical role of liaisons.
More recently, Burt has articulated the concept of structural holes. Burt argues that market-oriented competitive behavior comprises of how individuals gain access to the “holes” in networks. The absence of linkages between groups provides opportunities for brokerage since actors can pursue their autonomous interests, free of the constraints imposed by cohesive groupings.
Network analysis indices, or mathematical expressions of linkage patterns, are very sophisticated means of attacking levels of analysis problems. Indices associated with pathways primarily deal with how easily a message can flow from one node to another node in a network. They are intimately related to matrix manipulation and graph theory. Another way of conceptualizing this problem is in terms of the small world studies originated by Milgram.
Individual positioning indices (e.g., anchorage and integrativeness) try to mathematically capture an individual’s location within the relationships displayed within the network. The most commonly examined of these indices tries to reveal how central an individual is in a network. Freeman distinguished three types of centrality. Degree or local centrality refers to the number of immediate contacts an individual has, while closeness or global centrality refers to number of ties needed to reach directly or indirectly all others in a network. Betweenness centrality refers to strategic location as the shortest distance between two points in the network, whether an actor stands between two nodes. So, brokers become go-betweens by serving as the node that sends messages from one grouping in a network to another. A broker has a centrality in the network making the person possible to facilitate, impede, or bias the transmission of messages from different groups.
Perhaps the greatest level of development in network indices comes in the area of the relative connectiveness of social aggregates, groups/cliques, within larger social systems. Essentially the issue of connectiveness refers to whether or not all of the possible linkages in an aggregate are being utilized. This has important implications for processes like attitude formation in groups and a group’s relative cohesiveness.
Because of its focus on relationships, network analysis does not mesh well with traditional statistical analytic frames, such as analysis of variance and regression. (This is especially problematic for the discipline of communication, which at its root assumes dependence of actors.) The recent explosion of interest in network analysis became possible as the number of computer application with differing and often unique capabilities grew that focused on sophisticated visualizations of networks.
Despite the development of comprehensive software packages (e.g., UCINET) network analysts are not quite in the same position as are statistical analysts in the social sciences generally, where researchers rely on the companies that sell SPSS, SAS, and similar software packages to test software, market it to users (paying attention to factors which enhance marketability such as ease of learning and use), correct “bugs,” and diffuse new applications.
The awareness of social networks is, quite literally, an important survival tool for individuals which has resulted in much popular interest in networking. Network analysis is a practicable method for examining the overall configurations of relationships in a large social system, which can also provide an elegant description of them. Network analysis can describe and analyze complex organizational arrangements using a more holistic analytic perspective by examining specific and direct information on the pattern of an individual’s linkages. The impact of the change to a network moves from the individual to a more conceptual focus on relationships as the [Page 1636]unit of analysis. Such a focus permits the derivation of other measures from the aggregation of these individual linkages, including clique identification, roles, and metrics (e.g., connectedness), permitting the aggregation of data using the potential of many different levels of analysis (interpersonal, group, organization, and community) across various domains such as health communication and diffusion of innovation.
J. David Johnson
See also Health Communication; Human Subjects, Treatment of; Internet Research, Privacy of Participants; Measurement Levels; Methodology, Selection of; Organizational Communication; Privacy of Participants; Reliability of Measurement; Sampling, Methodological Issues in; Small Group Communication; Social Network Systems; Social Networks, Online
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). UCINET for windows: Software for social network analysis. Harvard, MA: Analytic Technologies.
Burt, R. S. (1992). Structural holes: The social structure of competition. Cambridge, MA: Harvard University Press.
Freeman, L. C. (1977). A set of measures of centrality based on betweenness. Sociometry, 40, 35–41. doi:10.2307/3033543
Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology, 78, 1360–1380. Retrieved from http://www.jstor.org/stable/2776392
Johnson, J. D. (2009). Managing knowledge networks. Cambridge, UK: Cambridge University Press.
Monge, P. R., & Contractor, N. S. (1987). Communication networks: Measurement techniques. In C. H. Tardy (Ed.), A handbook for the study of human communication (pp. 107–138). Norwood, NJ: ABLEX.
Monge, P. R., & Contractor, N. S. (2003). Theories of communication networks. New York, NY: Oxford University Press.
Richards, W. D. (1985). Data, models, and assumptions in network analysis. In R. D. McPhee & P. K. Tompkins (Eds.), Organizational communication: Traditional themes and new directions (pp. 109–128). Beverly Hills, CA: Sage.