SNA primarily deals with two levels of analysis, node-level and network-or graph-level. Node-level analysis examines nodes within the structure of a network to determine a particular node’s importance relative to other nodes. A key concept in node-level analysis is the idea of centrality, which is affected by a node’s location in a network.
Broadly, centrality is a measure of node’s structural importance or contribution to a network (Borgatti et al., 2013). There are hundreds of different types of centrality that measure a node’s importance in a variety of ways, however, there are four commonly used measures: degree centrality, closeness centrality, betweeness centrality, and eigenvector centrality.
Degree centrality is simply a measure of how many ties a node has—in other words, a count or sum of the number of links into a particular node. Closeness centrality
is a measure of how close a node is to all other nodes within a network in terms of geodesic distance. A node with high closeness centrality is usually in a position to hasten the flow of information through a network because it is close to a high number of other nodes. Conversely, an actor with low closeness centrality may be either disconnected or on the periphery of a network. Betweeness centrality is a measure of how much a node lies on the shortest path between all other nodes. Betweeness can be interpreted to mean a node’s potential for controlling flow through a network—that is, it denotes a gatekeeping or brokerage role, which means such a node could control or distort information flow (Borgatti et al., 2013). Eigenvector centrality is a measure of how well a node is connected to other well-connected nodes. An interpretation of eigenvector centrality is that nodes with high scores are connected to influential nodes with lots of connections—
expressing the idea of, “It’s not what you know but who you know.”
The various measures of centrality may indicate different things in different contexts. For example, while an actor with high degree centrality may appear to be very popular and well-connected, that actor may also experience a high level of constraint (to be discussed later) because of the obligation to reciprocate, and thus may be less productive than another actor with a lower degree centrality score. Therefore, it is imperative to have an idea of not only the meaning of a centrality score, but also what it may mean in a given context.
As noted, SNA emphasizes a node’s placement in the overall network structure, but it also recognizes that individual nodes may have unique nonrelational characteristics.
These nonrelational characteristics of individual nodes are called attributes, and can represent things such as gender, race, ethnicity, age, or years of education (Everton, 2012). Noting the attributes of nodes can be useful in development and identification of subgroups.
A final node-level concept worth mentioning is equivalence, which implies some type of interchangeability, sameness, or equal value between entities. Perhaps the simplest form of equivalence is structural equivalence, which exists when two nodes are connected to exactly the same nodes. Put another way, structurally equivalent actors have exactly the same relationships to all other actors—i.e., “identical ties to and from
identical other actors” (Everton, 2012; Wasserman & Faust, 1995, p. 468). This definition of equivalence is its most basic and strict form, and nodes that follow this strict definition will have the same centrality and other node-level measurement scores.
Other measures of equivalence exist as well, such as automorphic equivalence and regular equivalence, that are not as strict in definition, but potentially just as useful.
Broadly, automorphic equivalence means that two actors are structurally similar if they occupy indistinguishable positions in a network. For example, squad leaders in a military unit would be considered automorphically equivalent if they were in charge of the same number of Marines (although not necessarily the same Marines) and were supervised by the same number of people (although not necessarily the same people) in identical structural positions (Everton, 2012). With regular equivalence, “actors do not need to have identical ties to identical other actors nor do they occupy indistinguishable positions in a network” (Everton, 2012, p. 292). In other words, actors must have identical ties to and from regularly equivalent actors, where regularly equivalent actors do not have to be connected to the same actors, but must be connected to actors in the same classes (Everton, 2012). This concept should be familiar to those in the Marine Corps, considering that different MOSs have similarly placed Marines in different units, who do similar functions, and are evaluated by similar people. Indeed, much of the Marine Corps’ personnel evaluation system is based on the concept of equivalence.
Equivalence is relatively straightforward in a hierarchical organization such as the military, where redundancies are expected and necessary. In non-hierarchical and dispersed organizations, equivalence may be rarer and identifying it has implications that affect how to confront such a network. For example, if two actors are structurally equivalent, the removal of one actor does not completely sever a connection between disparate groups of actors, because another actor exists with the same contacts and relationships. The key point is that while various forms of equivalence have different meaning, all of them may be useful in identifying larger network structure and subgroups or subnetworks (to be discussed later).