Following the elicitation of any particular relationship type, researchers are generally interested in additional information beyond just the presence or absence of those ties. Therefore, researchers generally follow each name generator with a set of name interpreter questions. Name interpreters are primarily used to gather additional information about ego’s ties and/or alters. This combination of name generator and name interpreters accomplishes the general aims of an ego network design, the sampling aspects of which will be elaborated further below.
Up to this point, I’ve mainly been talking about ties as though the only option is for them to be present or absent. Dichotomous ties definitely dominated social network research for a long time. However, ties can also have values or signs, a possibility increasingly incorporated into social network data collection and analyses (Krivitsky 2012). A valued tie allows for strength or frequency of the reported relationship. The most cited network finding is the importance of weak ties for locating novel information (Granovetter 1973). This differentiates weak (acquaintance) ties from strong (close) ties. Strength of relationships can be categorical (like in the strength of weak ties) or ordinal. A common ordinal tie strength are those that ask for respondents to provide rank-ordered nominations, where the closest relationships are named first and weaker relationships follow (Bearman, Jones, and Udry 1997). For social interactions, the values of ties can indicate the frequency of those interactions (Marcum and Schaefer 2019). While valued ties may be seen as part of the generation process for identifying relationships, it is often saved for the name interpreter follow-ups. Signed ties similarly move beyond dichotomous relationships to capture the potential for relationships to have both positive and negative valences. For example, if studying “like” ties, a signed graph can allow for the possibility that the absence of liking is not the same as a dislike tie (Doreian 2017).
|Name 1||Name 2||Name 3||…|
|How do you know [NAME]?||1. family||1. family||1. family|
|(check all that apply)||2. friend||2. friend||2. friend|
|3. co-worker||3. co-worker||3. co-worker|
|4. neighbor||4. neighbor||4. neighbor|
|5. other||5. other||5. other|
|About how often did||1. daily||1. daily||1. daily|
|you see [NAME]||2. weekly||2. weekly||2. weekly|
|in the past year?||3. monthly||3. monthly||3. monthly|
|4. < monthly||4. < monthly||4. < monthly|
|5. never||5. never||5. never|
|How long have you known [NAME]?||__yrs __mos||__yrs __mos||__yrs __mos|
|What is [NAME]’s gender?||1. Male||1. Male||1. Male|
|2. Female||2. Female||2. Female|
|About how old is [NAME]?||____yrs||____yrs||____yrs|
Valued and signed ties provide specific examples of a more general tendency for name interpreter questions to provide the capacity to contextualize and enrich the information about each tie elicited by a name generator. Let’s assume your research project used an adapted version of the “important matters” name generator described above, where you asked respondents to name those they’d trust to talk to about highly personal topics. A first set of followup name interpreters you’d likely want to include can address the general nature of the relationship ego has with the named alter. Following the examples in Table Table 2.1, these could address things like how they primarily know each alter (e.g., are they a family member, co-worker, neighbor, friend, etc.), or how long the respondent has known them. These questions could also aim to gather additional information about the nature of the tie that is the focus of the name generator. For example, you could ask how close they feel to the person (some may only trust those they are especially close to; others may prefer to confide in strangers Small (2018)), or how frequently they interact with the named person (just because they name them as a potential confidant does not imply anything about the frequency or recency of those interactions).
Relational name interpreters can also ask about other relationships ego has with that same alter. For example whether they have given money to or received money from those named as important matter discussion partners. It’s important to note that this third category should not be thought of as an additional name generator, since it is only being asked within those who responded positively to an initial name generator. Name interpreters that ask about other relationship types do not provide the opportunity for a full elicitation of those with whom ego has that relationship independently of the initial name generator.31 As an example, the Add Health name generator asks about friends, then one of the name interpreters asks if for each nomination whether they have talked to that friend about a problem within the last week. This interpreter question tells us which friends they talked to about problems, but does not provide a full enumeration of those who they talked with about problems (e.g., if that list would include non-friends). As such, any time the research foci necessitates capturing different sets of relationships, each should be included as its own name generator.
In addition to gathering additional information about the relationship involving the ego-alter pairs, a second class of name interpreters generally seeks to gather additional information about the named alters themselves. If we return to the notion of asking network questions in the context of a general population survey, these name interpreters—asking about alters’ characteristics—may be the only opportunity the study has to gather any information about those alters. Among the most commonly observed patterns in network studies is homophily—the tendency of relationships to form between pairs of nodes who are similar (McPherson, Smith-Lovin, and Cook 2001)—which could not be assessed without information about the identified alters’ characteristics. As such, most survey approaches to gathering network data (especially ego network designs) will include a set of name interpreters asking about alters’ characteristics. These often focus on basic sociodemographics (gender, age, race, etc.), and other information that one could reasonably be expected to know about their network partners. Alter characteristics allow researchers to then investigate questions about the composition of ego’s network (e.g., “What proportion of ego’s alters have more than a college education?”), in addition to questions about homophily on such characteristics (e.g., “What proportion of ties are among pairs whose ages are less than five years apart?”).
A third class of name interpreters then frequently also asks a respondent to provide estimates of the relationships among their alters. As can be seen in Figure 2.2, these questions are typically asked in an undirected fashion (e.g., “Do person A and person B know each other?”),32 and captures general features of those relationships that are likely to be known to others (i.e., it makes less sense to ask respondents about relationships that their alters may be engaged in privately). From this type of question, we can estimate structural features of an ego network. For example, Figure 2.3 portrays the hypothetical ego network corresponding to the nominated alters from Figure 2.1.33 Some researchers new to networks may wonder why we even bother asking a respondent about relationships that they are not part of. As with the other name interpreters above, this is often more about gathering additional contextual information about ego’s relationships of interest to the study, than it is about providing precise estimates of those additional ties. For example, a common question in networks is about the presence and theoretical processes underpinning local network closure (Coleman 1988). If a person names 4 alters, all of whom are isolated from one another, the expectations deriving from those ties (e.g., the ability to socially sanction ego) would differ substantially than if those four alters are, in-turn, connected to one another.34 E.g., the ties arising from a dense friend group, seen at the bottom left of Figure 2.3, may be able to exert more social control over ego than the less dense group of family connections located at the top of the figure.
Everything described above is readily adaptable into a social survey that is sampled at the individual level, and seeks to capture characteristics of those individuals. In that case, network features become like any other individual trait that can be gathered within the survey, and can generally be analyzed with the same methods (and corresponding assumptions) to other individual-level data (Perry, Pescosolido, and Borgatti 2018). What differs here is merely that the types of information emerging from eliciting relational data allows the researcher to also examine some questions based in relational theories.
However, while I was describing these approaches, you may have logged some objections about better ways to gather some of the information elicited from a respondent about their partners (and their relationships). For example, instead of relying on a lone respondent to provide socio-demographic characteristics of their own and their friends, why not ask the friends about their own characteristics? In the context of a general population survey that is sampled at the individual level, however, it is important to remember that it would be highly unlikely that any of the respondents’ named alters would also be included in the sample.35 This inclusion/exclusion of alters from available data raises questions best answered not by measurement decisions, but in strategies for sampling networks. Before turning to describing the three primary sampling strategies employed in social networks research—including several that are not subject to the types of constraints raised here—we turn to an alternative measurement strategy that focuses on the characteristics within, rather than the specific members of, one’s personal networks.