Social network analysis is a methodological and theoretical approach for studying relationships among people, organisations, groups, places, objects or other entities by representing those relationships as a network of connected nodes and ties. A node represents an actor or unit of analysis, while a tie represents a relationship or connection between nodes. Depending on the research question, ties may represent friendship, communication, collaboration, kinship, trade, information exchange, advice, conflict, co membership or many other relationships.
The central idea is that social outcomes can depend not only on attributes of individual actors but also on their positions within patterns of relationships. Social network analysis therefore shifts attention from isolated individuals to structures of connection. It can be used to investigate questions such as who occupies central positions, which groups are densely connected, how information travels, whether networks contain distinct communities and how relationships affect opportunities, influence and access to resources.
Social network analysis has roots in several intellectual traditions, including sociometry, graph theory, anthropology, sociology and mathematical approaches to relational structure. Early sociometric work examined interpersonal relationships within groups, while later researchers developed increasingly formal concepts for describing network structure. Graph theory supplied mathematical representations of vertices and edges, while sociological research developed concepts for interpreting patterns of connection in social terms.
The growth of computers and digital communication greatly expanded the scale of network analysis because researchers can now process large relational datasets derived from surveys, organisational records, communication systems and online platforms. Contemporary network research ranges from small classroom friendship networks to international trade networks and large digital communication systems. Despite technological changes, the central principle remains relational: social structure can be examined through the pattern, direction, strength and distribution of connections among entities.
Networks can be represented as directed or undirected, weighted or unweighted, one mode or two mode, and static or dynamic. A directed network records the direction of a relationship, such as who follows whom or who sends information to whom, while an undirected network records relationships that are treated as reciprocal, such as certain forms of friendship or co membership. Weighted networks assign values to ties, such as communication frequency or transaction volume.
Bipartite or two mode networks connect two different categories of entities, such as people and organisations. Network data can therefore be structured in many ways, and the representation chosen affects the analysis. Researchers also distinguish between complete networks, in which relationships within a defined population are studied, and egocentric networks, in which the relationships surrounding particular focal individuals are examined.
Several measures are central to social network analysis. Degree centrality concerns the number of direct connections associated with a node, although the meaning differs between directed and undirected networks. Betweenness centrality measures the extent to which a node lies on paths connecting other nodes and can indicate a potential brokerage position. Closeness centrality captures how near a node is to other nodes according to a specified distance measure. Eigenvector based measures consider connections to other well connected nodes.
Density describes the proportion of possible ties that are actually present in a network, while clustering measures can identify tendencies for connected actors to have connections with one another. Components identify disconnected portions of a network, and community detection methods attempt to identify densely interconnected groups. These measures are mathematical descriptions rather than automatic explanations of power or importance, so their interpretation must depend on the type of relationship and substantive context being studied.
Network structure can influence the movement of information, resources, opportunities and social influence. Individuals occupying brokerage positions may connect otherwise separated groups and potentially gain access to diverse information, while dense networks can facilitate trust, repeated interaction and rapid diffusion within groups. Network research has therefore been used to examine employment opportunities, innovation, organisational collaboration, political mobilisation, disease transmission, migration, criminal associations, scientific collaboration and online communication.
A famous sociological insight associated with network thinking is that weak relationships can sometimes connect individuals to social circles containing information unavailable within their immediate close-knit group. Network analysis also distinguishes structural position from individual characteristics: two people with similar education, income or occupation can have different opportunities because they occupy different relational positions. This does not mean networks mechanically determine behaviour, because actors can create, maintain, avoid and transform relationships.
Data collection is one of the most difficult aspects of network analysis because researchers must define what counts as a relationship and determine whose relationships are included. Surveys may ask respondents to identify friends, collaborators or sources of advice, while digital records can provide traces of communication, transactions or online connections. Administrative records, organisational membership data and archival documents can also provide network information.
Researchers must consider missing ties, inaccurate recall, changes over time and differences in willingness to report relationships. Ethical concerns can be particularly serious because network data can reveal information about people who never directly consented to participate, including relationships between identifiable individuals. Anonymisation is challenging because network structure itself can sometimes make individuals identifiable. Researchers therefore need careful governance, secure data handling and clear decisions about whether the network boundary is justified and ethically defensible.
Social network analysis differs from simply producing a list of relationships because it provides formal tools for analysing relational structure. Conventional statistical methods often treat individual observations as the primary units, whereas network data frequently violate assumptions of independence because one person's relationship is inherently connected to another person's relationship. Network models therefore require specialised statistical and computational techniques. These include exponential random graph models, stochastic actor oriented models, network regression approaches and longitudinal network methods, among others.
Researchers may also combine network analysis with qualitative interviews, ethnography or conventional statistical data to understand why particular ties exist and how structural patterns relate to social outcomes. Network measures should never be interpreted without attention to sampling, measurement choices and the substantive meaning of ties, because a highly connected node in a communication network does not necessarily have the same social significance as a highly connected node in a friendship or organisational authority network.
Social network analysis is especially valuable because it makes relational structure visible and measurable. It can reveal clusters, bridges, isolated actors, concentrations of connectivity and pathways through which information or resources may move. At the same time, network analysis has limitations. Correlation between network position and an outcome does not by itself establish causation, because people may choose relationships partly because they already share characteristics.
Network boundaries can be arbitrary, relationships can be difficult to measure, and digital trace data may capture platform behaviour rather than complete social relationships. Centrality measures can also be misleading if researchers treat mathematical prominence as synonymous with social power. The strongest network research therefore combines careful conceptualisation, appropriate mathematical methods, substantive knowledge and attention to context. Its distinctive contribution is the systematic study of social relations as structures that both reflect and potentially shape social behaviour.
