Social ties: the basic units of information and observation for empirical network research
Social network research deals with social ties, actual and perceived, between two actual people. An individual has many social ties of different sorts (parent, co-worker, friend etc.). We identify a social tie by naming the connected two people and the type of social tie, or ties, they have.
The FROM label is ambiguous. When we are dealing with interview or self-report data SNA assumes that we have information FROM A, with respect TO their relation(s) with B. I will call this data point a directed tie, abbreviated to di-tie. With relations such as friendship we would not want to know if B considers A a friend. Is this a reciprocated, bi-directional relationship? The example gives B’s response and we know this is a bi-directional social tie.
When we do not have direct information from B we may ask A if she considers that B would consider her (A) a friend. To capture this different type of data we must add the informant name as an additional identifier for the tie.
This type of question elicits an informant’s perceptions of the relations around them. SNA designates this as cognitive data.
The specification of the informant node ID also clarifies situations where there is directionality in the questions (Who do you go to for advice? Who comes to you for advice?) and situations where the informant has observed relations among people around them (Are X and Y friends?)
This data records that both A and B say that X and Y friends.
Notice that the question: Are X and Y friends? only implies that X and Y either have a mutual bi-directional (or ‘undirected’) relationship or do not. There is no possibility for the informant to say that A is a friend of B but B does not consider that A is a friend.
Most network scientists assume that network relations are always bi-directional (undirected). They are used to data that is objective observations of associations. Because of this they will assume that any FROM-TO pairing implies the reverse directional tie and design software to create the reciprocated di-ties automatically. Within SNA software however the user specifies when this ‘symmetrisation’ should be done.
SNA datasets seldom specify an INFORMANT node as it is needed only in very specific circumstances, the short FROM-TO format is usually sufficient. The differences between directed and undirected data have to be watched more carefully however.
Linking attribute data (node data) to tie data
A researcher will nearly always have information about the people used to identify ties. Such attribute data is a standard, case-by-variable data table. To link to the tie data the case IDs in this node data list must match, exactly, the node labels in the tie data list. Matching of the actual content of these ID fields creates a relational database structure with the node IDs as the key field.
With this relational data structure in place we can integrate the node data into our network diagrams. We can also inspect node data from the two nodes that identify a tie to see if the tie has formed between people who are alike, or not alike, on any personal attribute.
Social networks, culture and social structure
The following quote describes the
… we do not observe a ‘culture’, since that word denotes, not any concrete reality but an abstraction. But direct observation does reveal to us that … human beings are connected by a complex network of social relations. I use the term ‘social structure’ to denote this network of actually existing relations. (Radcliffe-Brown, 1940 cited in Prell 2012: 30)
Radcliffe-Brown was writing at a time when anthropologists such as Margaret Mead and Ruth Benedict were describing whole societies in terms of their cultures understood and described, holistically, through the observer’s empathic observations. He is rejecting this approach.
The quote suggests an empiricist approach involving the cumulative observation of specific social relations – a compilation of tie data. Social structure is just the fact of connectivity. It is like the ‘structure’ that occurs when we drop a clutch of ‘pick-up sticks’, a jumble rather than a structure. We can even imagine that this unstructured ‘jumble’ of connections can extend across whole populations. This is the appeal of small world theory and the idea of ‘six degrees of separation’.
Even a jumble will have some sort of structure such as denser, more jumbled areas, less dense bits and the odd isolate. With social relations such differentiation may lead to informal groupings or communities. When such groupings become formalized and institutionalised they have publicly visible boundaries (who is a member, who is not) and even a collective, corporate public identity.
SNA deals with the fact of grouping with 2-mode data formats. The two ‘modes’ are the lists of node IDs for the persons, on the one hand, and, on the other, the collective or corporate entities or subgroups. Inductive approaches of subgroup detection look for groupings embedded in the tie data. 2-mode data analysis works directly with publicly visible membership listings.
I designate these strategies of investigation and analysis as network topography. We examine them more closely on Day 3 and will return to the concepts of social structure and culture at that time.