How New is “New”? Who Gets Added in a Panel Study of Personal Networks?

Abstract

Using longitudinal data from UCNets, we examined newly-listed alters and distinguished between truly new ties who were recently met (typically coworkers and acquaintances) and awakened ties who were previously known to ego (typically extended kin and friends). Half of the newly-listed ties among the younger respondents were truly new, whereas two-thirds among the older respondents were awakened. In both groups, however, most alters were previously listed. These ties mainly included kin, confidants, and advisors, suggesting stability in the network core. Methodological implications for the name-generating process and substantive implications for the understanding of network dynamics and structure are discussed.

Large-scale panel surveys of individuals’ personal, or egocentric, networks that ask respondents to list their network alters have become increasingly common (e.g., Cornwell et al., 2021; Huisman et al., 2011; Mollenhorst et al., 2014; Tulin et al. 2021; Wrzus et al., 2013). Researchers largely accept that this “name-eliciting” or “name-generating” method can provide accurate depictions of a person’s social world, especially when multiple name-generators are employed (see review in Perry et el., 2018). Nonetheless and understandably, methodological concerns about the reliability and validity of this technique have arisen (e.g., Bernard et al., 1990; Eagle and Proeschold-Bell., 2014; Marin, 2004; McCallister and Fischer, 1978; van der Poe1, 1993; White and Watkins, 2000; Yousefi-Nooraie et al., 2019). One of these concerns deals with the question of who respondents tend to name when asked about the people in their social worlds. In this paper, we seek to deepen our understanding of the name-generating process by specifically examining who is newly-listed, or added, over time to respondents’ reports. Taking advantage of the longitudinal nature of the University of California Social Networks Study (UCNets), a research project about personal networks and health that uses an extended name-eliciting survey, we focus on those alters who are mentioned for the first time at the last wave of the panel and ask who they are and how new they really are to the network.

When filling in a name-generator questionnaire a second or subsequent time, respondents may name entirely new associates, that is alters whom they have only recently met who were not known (or not well-known) to them previously; we refer to these alters as truly new ties. However, they may also mention for the first time alters with whom they had a prior connection, alters whom they could or would have named earlier if they had been pressed to be more exhaustive in their listing or been asked different name-eliciting questions. These alters were not named before perhaps because they had been forgotten at the moment, were known in a superficial way, did not play a meaningful role in the respondent’s life earlier, or became dormant due to life circumstances or happenstance. We refer to these not truly new alters as awakened ties.

Distinguishing theoretically and empirically between truly new and awakened ties is important because they might have different substantive implications for network dynamics and consequences for ego. Truly new ties often emerge when people enter new social and institutional settings, such as when they start college, take a new job, move to a different neighborhood, or join a social club. These new environments constitute opportunity structures for social meetings and the formation of ties (Blau and Schwartz 1984; Feld and Carter 1998; Mollenhorst et al. 2014).¹ Because new ties are typically “weak ties,” in the early stage of the development (Burt 2000), and tend to originate in social circles in which ego had not - or had only minimally – been active previously, they have a high potential of linking a person to unfamiliar and non-redundant resources and information, which the literature suggest can be crucial for social and economic mobility (Burt, 1995; Lin, 2001).

Awakened ties, on the other hand, may be likelier to provide a deep well of potential support. Awakened ties refer to ties that went dormant and then got reactivated for a variety of reasons. They often involve old associates whom people trust and to whom they have positive feelings and can thus be relatively easily mobilized in times of need (Levin et al., 2011; McCarthy and Levin, 2019; Quinn, 2013). But awakened ties also include relationships to weak and peripheral alters who have come to assume a greater role in ego’s life and constitute a source of support to ego following the occurrence of some life event or changing circumstances (Killworth et al. 1990; Levine et al. 2011). As Fingerman (2009) argues, ties to alters who were previously not well known to the person “may be dormant and become active when life stages or locations converge” (p. 74).

The distinction between truly new and awakened ties may also provide insight into the processes by which personal networks are formed and change over time. Like archeologists who dig underneath the surface of a site, and drawing on Kilduff et al.’s idea (2006) that networks are built of relationships accumulated over time, layer upon layer, in people’s minds, looking separately at truly new and awakened ties can help reveal the complex and often hidden structure of personal networks. At the methodological level, this distinction can promote our understanding of who is sampled and why samples of alters generated by the egocentric method look the way they do.

In this study, we focus on both the substantive and methodological issues related to naming new ties in the network. Specifically, we examine (1) what the relative shares of truly new and of awakened ties in egos’ networks are; (2) how truly new and awakened ties differ in terms of their roles in networks, their relationships with egos, and in egos’ personal characteristics; and (3) what contexts and circumstances (e.g.., life events) are associated with the presence of truly new rather than awakened ties in ego’s network. Our findings show that new ties of the UCNets panelists were not all the same and varied in degree of newness. Some were brand new, but many others were previously known to ego and form reconnected ties. Truly new ties were more common among the younger participants whereas awakened ties were more common among the older participants. However, most of the alters whom egos named were not new to the survey but rather alters who were listed in previous waves. These were typically kin and people in whom egos confided and from whom they received advice. Overall, these findings point to a tendency toward stability in reported personal networks, at least in the short run.

Background

Which Alters Do Respondents Name in Network Studies? Substantive and Methodological Concerns in the Longitudinal Study of Egocentric Networks

The name-generator technique, it is generally understood, entails respondents offering up subsets of alters based on a mental “sampling” of all their “true” alters, with all the cognitive distortions that recollection processes can entail (e.g., Brashears, 2013; Brashears and Quintane, 2015; Shea et al., 2015; Smith et al., 2020). While name-generators extract alters who roughly fit the criterion specified in the question, for example, close kin, people in whom they confide, or coworkers, there are clearly errors of omission and of inclusion which likely create biased samples.

These sorts of errors have been explored for the classic General Social Survey name-generator asking respondents to name those with whom they discussed “important matters” (e.g., Bearman and Parigi, 2004; Brashears, 2014; Lee and Bearman, 2017; Marin, 2004; Small, 2013, 2017). Across a set of name-generator techniques, including the “important matters” probe, respondents tend to report those who are personally closest to them, such as more intimate and more often seen alters. They do so for several reasons. Respondents are more likely to remember and name alters to whom they feel emotionally connected and who play a central role in their networks, such as family members and close friends (Brewer, 2000; Marin, 2004). Respondents are likely to name alters who are cognitively interconnected, for example, the set of fellow church members, coworkers, or college friends (Brewer, 1993). Respondents also tend to remember more easily alters with whom they regularly engage in meaningful interactions, either positive or negative. While individuals also interact with people whom they do not know well simply because such people are accessible and easily available (Desmond, 2012; Small, 2017; Small and Sukhu, 2016), long-lasting relationships, which are more easily remembered, tend to develop with core network members. The tendency to name core network members is less of a concern to scholars interested in studying people’s inner social circles and perceptions of support and their consequences for individual outcomes, such as health and wellbeing. It may, however, create biases for those wishing to learn about the diversity of networks and their dynamic aspects.

One way to investigate the underlying logic of respondents’ sampling procedure is to leverage panel studies to see which alters remain in, which are dropped from, and which are added to the generated lists between survey administrations. In doing so, however, it is important to keep in mind that the “samples” of alters obtained at each wave also reflect the specific name-generators employed in the study, which determine the criteria of inclusion in the network. Alters who do not participate in the types of exchanges tapped in the name-generators but are active in some other, not measured domain will be less likely to be named (Feld et al. 2007; Marin and Hampton 2007; McCallister and Fischer 1978; van der Poel 1993). This may lead to errors in the interpretation of network change when, for example, alters change their role in the network over time. Alters who are no longer active in a domain tapped in the name-generators may be mistakenly treated as lost. Similarly, those who were previously active in the network but not in the domains measured and showed up in a later wave only because they changed their role in the network, will be wrongly classified as “new” (Marin and Dubash 2021; Mollenhorst et al. 2014). In the latter case, the naming of the alter at the later wave of the survey reflects a change in the content of the relationship rather than the emergence a new relationship with a new associate.

Dropped Ties

Numerous studies show that as many as half of associates are typically dropped from rosters of alters (usually to be replaced), the proportion depending on the specific method used and the time gap between waves (e.g., Badawy et al., 2018; Bidart et al., 2020; Bignami-Van Assche., 2005; Fischer and Offer, 2020; Mollenhorst et al., 2014; van Duijn et al., 1999). Several studies have examined the correlates of alters being dropped, showing that in general, between-ego differences exist but are not as important as between-alter differences. Alters who are relatively peripheral, such as distant kin, known fewer years, less active in ego’s life, less supportive, and less connected with others in ego’s network are more likely to be dropped from one survey to the next (see prior citations, also Cornwell et al., 2014; Goldman and Cornwell, 2018; Marin and Hampton, 2019). A recent study also found that alters who are dissimilar from ego in terms of gender, age, and ethnicity have higher odds of disappearing from the network (Tulin et al., 2021), and that this process of tie decay typically occurs at the early stages of the relationship (see also Burt, 2000). Nevertheless, how much of this culling process reflects substantive changes in respondents’ lives and how much of it reflects a methodological bias remains an issue.

Some researchers have argued, or at least implied, that features of the name-eliciting methodology itself contribute to high rates of dropping names and perhaps to the observed correlations as well. One such feature is that respondents often overlook or forget particular names from one wave to the next; upon being asked, they say that they should have listed the forgotten alter (Fischer and Offer, 2020). Another is respondent fatigue common to this method (Fischer, 2012; Marin and Hampton, 2007; Paik and Sanchagrin, 2013; Valente et al., 2017; Yousefi-Nooraie et al., 2019). Also, interviewer effects are common and sizeable (Brüderl et al., 2013; Eagle and Proeschold-Bell., 2014; Fischer and Bayham, 2019; Herz and Petermann, 2017; Marsden, 2003). Furthermore, the nature of peripheral ties, their being at the hazy border of ego’s network, makes it difficult to obtain reliable counts. Any of these factors, which are part of the cognitive critique, can reduce the reliability of the name-generating technique (Bignami-Van Assche, 2005; van der Zouwen and van Tilburg, 2001).

Added Ties

Concerns also exist regarding ties that are added to the network in longitudinal studies. Attending to added ties can inform us both about the substance of network change and about the methods for capturing it. Fewer studies, however, have systematically examined which alters are likely to be newly-listed and which egos are likely to report on them in follow-up surveys. Although these issues have been overlooked, prior research proposes several explanations.

Social Contexts as Opportunity Structures.

An important step to gathering new alters occurs when ego enters new social and institutional settings. These environments constitute opportunity structures for the formation and development of new relationships (Blau and Schwartz 1984; Fischer 1982; Mollenhorst et al. 2014). First, they determine the potential pool of alters whom a person can encounter. Then they provide opportunities for repeated interactions with these alters and thus shape the ability to develop a relationship with them. Often these interactions develop around shared interests and activities, an idea encapsulated in Feld’s concept of foci (Feld 1982; Feld and Carter 1982).

Life events, both normative and non-normative, can be important triggers to shifts in a person’s social and institutional environment (Wrzus et al. 2013). The convoy model (Antonucci et al. 2010) nicely describes how, in making life-course transitions, taking on new roles, and developing new routines, people recruit new alters while they drop others. For example, Bidart et al. (2020) show how young adults starting school, getting a first job, and moving out of their parents’ homes form new ties in new contexts. Another example is the birth of a child. While ties to friends and more distant associates often weaken and even disappear with the transition to parenthood (Kalmijn 2012), new relationships with other parents in the community and various service providers can be formed (Bost et al. 2012).

Replacing Lost Ties.

New ties also appear as a consequence of having dropped previous ties. Findings from the National Social Life, Health, and Aging Project (NSHAP) suggest that older adults add new ties to replace dropped ones (Cornwell et al., 2014; Cornwell and Laumann, 2015, 2018), although in panel surveys this process may be difficult to distinguish from regression to the mean.² This idea resonates with the homeostasis approach, according to which lost and gained ties often offset each other as people tend to replace lost relationships by forming new ones with similar characteristics, thus contributing to stability in terms of network size and properties over time (Wellman et al. 1997; Zettel et al., 2004). There is less evidence concerning how much egos make conscious efforts to substitute lost ties with new ones, with some research on widows suggesting that at least some people do so explicitly (Lamme et al., 1996; Utz et al., 2002). As to the new alters themselves, these studies find that the added associates tend to have weaker relationships with ego than continuously present alters. Striking, however, is the finding reported in Mollenhorst et al. (2014) that most of the newly-named alters were not actually newly met; indeed, respondents had known them an average of 20 years. This highlights the need to distinguish between truly new associates and previously known, reactivated ones.

Dormant and Awakened Ties

Research suggests that dormant ties are pervasive, yet they have received little empirical attention because dormant ties are difficult to capture as most surveys ask about currently active ties or ties that could be activated if needed, thus mainly tapping alters who are present cognitively (Killworth et al., 1990; Tulin et al., 2021). Furthermore, much of the “social capital” literature has treated dormant ties as simply dead connections. Ties that are no longer valuable to the individual, it is argued, become inactive and eventually fade away and disappear from the network (e.g., Burt, 1995; Coleman, 1990). Recent studies, however, suggest that dormant ties are alive and well, even if “asleep,” and can constitute an important source of support. For example, Levin et al. (2011) found that the advice executives in MBA programs received from their dormant ties, with whom they were asked to reconnect as part of the study, was extremely relevant and useful, sometimes even more so than the advice they received from their current associates. The authors conclude by stating that awakened ties are particularly useful and efficient: like peripheral ties they tend to lead to new information and opportunities, but like close ties they often share a history with ego, and this require substantially less investments of time and attention than would new ties.

However, the questions which ties are likely to fall into dormancy, for what reasons, and what leads to awakening, have been overlooked. People stop interacting with others for many different reasons, including disagreements and conflicts. In this case, severed ties have a lower chance of being activated again. However, oftentimes life circumstances, divergent destinies, and simple inertia and happenstance may lead people to lose touch with each other. The potential of falling into dormancy affects almost all social relationships, excluding perhaps those with the closest family members (Levine et al. 2011). If the relationship was based on trust and evokes positive feelings and memories, these ties can be successfully reactivated in the future (Quinn 2013). In this respect, as long as the possibility of reconnection exists, dormant ties cannot be treated as dead (Burt, 1995; Levin et al., 2011; Marin and Hampton 2019) and studying them provides a window into the dynamic nature of networks and their deeper structure (Kilduff et al., 2006). Examining awakened ties can give researchers important information as to which ties are likely to have been dormant, what roles they fulfill in the network, and what leads to their reemergence. Thus, we examine newly-listed alters and ask who these alters are and how “new” they really are.

Data and Methods

UCNets data

This study is based on data from the University of California Social Networks Study (UCNets; see http://ucnets.berkeley.edu/ for publicly available data and documentation), a three-wave longitudinal research project conducted between the years 2015 and 2018 on personal relationships, life events, and well-being. UCNets sampled participants from two distinct age groups (21−30 and 50−70 year-olds) in the greater San Francisco Bay Area. These two understudied age groups were purposefully targeted in order to maximize the number of key transitions and life events respondents would likely experience between waves.

The sampling procedure initially consisted of randomly selecting households within 30 randomly-selected census tracts in the Bay Area, inviting any member of the household in the relevant age groups to participate for pay in the panel. This procedure sufficed for the older sample but not for the younger one.³ The latter was supplemented with an opt-in panel of respondents recruited through previous participants and Facebook. Few differences exist between the young respondents recruited through household solicitations and those recruited in other ways (Lawton and Wilson, 2018). The final sample could not be considered random but it is roughly representative of the population demographically, with the major exception of being skewed toward female and better educated respondents. We use post-stratification sampling weights to correct for these biases as well as for the distributions of race-ethnicity and marital status in all the analyses. We also control for recruitment procedure in all the multivariate analyses of the younger respondents.

In the first wave, most participants were randomly assigned to either a face-to-face interview or to a web survey (all the participants drawn through Facebook filled in the survey online). In successive waves, additional respondents who originally completed the survey in a face-to-face interview were randomly assigned to the online mode. The in-person and online instruments were substantively identical (Fischer and Bayham, 2019) and analyses (not reported) showed that they yielded similar results. The analyses in this study are based on data drawn from respondents who completed all three waves (n = 049, which corresponds to 81% of the original sample). The sociodemographic characteristics of these respondents are presented in Table 1.

Table 1.

Respondent-level characteristics and prevalence of life events (weighted percentages)

	Young cohort %	Old cohort %
Male	49.84	46.98
White	39.28	58.01***
Asian	28.49	20.33**
Latino	22.54	12.27***
Black and other	9.68	9.39
Foreign born	17.49	20.66
Married	28.0	61.0***
Education less than BA	23.18	36.74***
Education BA	58.84	43.58***
Education more than BA	17.97	19.67
Poor health	14.21	16.95
Extroversiona	−.074 (1.018)	.027 (1.001)
Neuroticisma	.063 (.974)	−.021 (.994)
Network size at wave 2a	10.030 (3.569)	10.420 (4.219)
Personal reference or Facebook recruit	51.70	---
Experience of life-event between wave 2 and wave 3
New partnership	9.02	3.33***
Birth of child	9.30	7.40
New job or school	43.50	6.50***
Graduated from school	11.07	---
Death of a close person	23.0	32.0***
Break in friendship	26.0	25.0
End job	5.64	12.33***
School, work, financial prob.	48.95	23.81***
Health problem	12.30	28.20***
Moved near (in Bay area)	20.26	5.02***
Moved far (outside Bay area)	13.31	3.38***

Note: Young cohort refers to 21–30 year-olds and old cohort to 50–70 year-olds at wave 1

^aMean and standard deviation (in parentheses)

^*p < .05.

^**p < .01.

^***p < .001 (two-tailed tests) for differences by age group

Measures

One of the major advantages of UCNets is its use of an extensive egocentric network approach to collect comprehensive information about people’s social relationships. At each wave, respondents were asked seven name-generating questions encompassing different social activities and exchanges: (1) the people with whom they socialized, (2) the people whom they confided in, (3) the people who gave or would give them advice, (4) practical help, (5) assistance during emergencies, (6) those to whom they provided help, and (7) the people they found sometimes difficult or demanding. In the first stage, respondents were asked to name the people with whom they engaged in these interactions (up to nine names in the first question and up to six names in all the other questions). In the next stage, respondents provided information about each of the named alters, specifying the role alter played in the network, their sociodemographic characteristics, and relationship with ego.

An alter was considered a network member in a particular wave if that alter was mentioned in at least one of the seven name-generators at that wave. This study focuses on all the alters named at wave 3 (n = 4,080 and 5,861 in the young and old cohorts, respectively)⁴ and it distinguishes between previously-listed ties, that is alters who were listed in previous waves (i.e., in wave 1 or wave 2 or both), and newly-listed ties, alters who appeared for the first time at wave 3. As Table 2 shows (see upper panel), respondents reported on average about 3 new names at wave 3 which they did not mentioned previously, with a slightly higher number in the young cohort (3.42 among the younger respondents and 2.91 among the older respondents.) These numbers are substantial, representing about one-third of the network in the younger age group and slightly more than one-fourth of the network in the older age group.

Table 2.

Mean number of previously- and newly-listed ties at wave 3, overall and by type of tie (weighted)

	Young cohort	Old cohort
Aggregated person-level
Mean (SD)
Previously-listed	6.88 (2.94)	7.45 (3.01)
Newly-listed	3.42 (2.32)	2.91*** (2.35)
Truly new	1.83 (1.79)	.85*** (1.28)
Awakened	1.56 (1.66)	1.98*** (1.98)
N Respondents	382	558
Alter-level
% Newly-listed ties (out of total number of ties)	34.7	29.0***
% Type of newly-listed tie (out of all newly-listed)
Truly new	53.3	29.1***
Awakened	45.4	67.9***
Othera	1.2	3.0***
Total	100	100
N Ties	4,080	5,861

Note: Young cohort refers to 21–30 year-olds and old cohort to 50–70 year-olds at wave 1

^aThe “other” category includes responses that did not fit the “new” or “awakened” categories in the open-ended follow-up question.

^*p < .05.

^**p < .01.

^***p < .001 (two-tailed tests) for differences by age group

However, as we noted in the introduction, newly-listed ties may not necessarily be totally new associates whom respondents had not known at all or not known well previously. Other reasons may explain why an alter who is not truly new would be mentioned for the first time at wave 3. For example, the respondent might have forgotten to mention that alter in the previous waves, or that the alter is someone with whom the respondent had lost touch until recently or someone whom they may have known in a literal sense but had not been significantly involved with before and therefore not listed previously. UCNets provides additional information that allowed us to consider these possibilities and differentiate between truly new and awakened ties. For each alter named, respondents were asked whether they had met that alter since the prior survey, roughly a year. Thus, newly-listed alters who were met between waves 2 and 3 were coded as new. For those newly-listed alters who had not been recently met, respondents were asked the additional follow-up question “why do you think you listed [alter] for the first time today?” The response categories were (1) “did mention but gave a different name,” which we recoded into previously-listed; (2) “did know before but forgot to list,” which we treated as awakened; (3) “did know before but now plays a larger role in life,” which we also treated as awakened⁵; and (4) “really just got to know,” which we coded as new. A fifth option was the open-ended category “some other reason” (n = 281). We carefully read the text the respondents who checked this option provided and based on its content recategorized the answers into either truly new (e.g., “he is a recent coworker of mine,” “new client,” and “I just started seeing him”); awakened (e.g., “we have had more contact since I moved back East,” “we drifted apart but now are friends again,” and “we got more involved after I found out he also has cancer”); or left them as “other” (e.g., “her dog fell in love with Leo (my dog)” and “not sure”).

At the alter-level, using information drawn from the name interpreters, we categorized alters based on the role they played in the network of the respondent. The role relationship variables include immediate kin (spouses, parents, adult children, and siblings), extended kin (other relatives), and for the nonkin: romantic partners, friends, roommates, neighbors, workmates, schoolmates, churchmates, and acquaintances. The variables for types of interaction between alter and ego are measured with a series of dummies that refer to how the respondent is engaged with the alter. They are based on the name-generators we described above (socialize, confide, advice, practical help, emergency help, providing support to alter, and alter is difficult). We control for two additional alter-level characteristics. Emotional closeness refers to whether the respondent reported feeling especially close to the alter and physical distance from alter is measured with the variable indicating whether the alter lives over one hour away from the respondent.

At the respondent-level, we examine several life events that are associated with a change in social context. Some events pertain to entering new social and institutional environments which can provide the opportunity to create new ties: new partnership (i.e., getting married or entering a new romantic relationship), new job or school, birth of a child (in the young cohort) or a grandchild (in the old cohort), and graduating from school (in the young cohort only). Other life events may increase the need or wish to reconnect with old associates: the death of a close person, major break in friendship or other (non-romantic) relationship, ending a job (becoming unemployed or retiring), experiencing school, work or financial problems, and having a health problem. In addition, we examine residential moves, which include moved near (in the Bay Area) and moved far (outside the Bay Area).⁶ All the life event variables are dummies indicating whether the respondent reported experiencing the event between wave 2 and wave 3.

Finally, we control for several background characteristics and personal traits, including gender, race and ethnicity, marital status, education level, self-reported health, and extroversion and neuroticism from the brief Big Five inventory. In the young cohort, we also control for whether the respondent was recruited to the study through Facebook or personal reference. The descriptive statistics for all the respondent-level variables are presented in Table 1.

Findings

How Prevalent Are Truly New and Awakened Ties?

The distribution of the newly-listed ties reveals interesting results (see lower panel of Table 2). Slightly more than half of all the newly-listed ties in the young cohort were truly new ties whereas the vast majority (about two-thirds) of apparently new ties in the old cohort were awakened ties. This discrepancy by age group reflects the numerous opportunities to form new ties that younger people face as they enter adulthood, a life stage characterized by high mobility and the transition into new settings (Bidart and Lavenu, 2005; Bidart et al., 2020), and the accumulation of numerous associates over the life course among older adults (Antonucci et al., 2010; Quinn, 2013). The percentage of “other” newly-listed ties (i.e., alters who could not be categorized as either truly new or awakened) is very small and we therefore excluded them from the multivariate analyses reported below.

Next, we examine what in the nature of the tie and alter’s relationship with ego distinguishes the truly new from the awakened ties. Table 3, top panel, shows the percentages for the previously-listed and newly-listed ties by the role the alter played in the network and their social interaction with ego. Overall, very few kin were truly new ties. In the young cohort, about 7 percent of immediate kin and almost 8 percent of extended kin were truly new. These numbers were even smaller in the old cohort. An examination of some of the open-ended answers in the survey (results not shown) indicate that new kin tended to be recent additions to the family, such as new spouses, in-laws, children, and grandchildren. The percentage of truly new ties among immediate kin was particularly small (less than one percent) in the older age group. This is consistent with the finding that at later stages in life people are more likely to lose than gain new family members (Antonucci et al., 2010). By contrast, in both age groups, about one-fourth of the extended kin were awakened ties (29 percent among younger respondents and 26 percent among the older respondents), suggesting that relationships with more distant relatives may be less stable or less in the front of people’s minds (the percentage of awakened ties was substantially higher among extended kin than among immediate kin).

Table 3.

Percentage of previously-listed, truly new, and awakened alters by role relationship and type of exchange at wave 3 (weighted)

	Young cohort				Old cohort
	N	Previously-listed	Truly new	Awakened	N	Previously-listed	Truly new	Awakened
Role relationship
Kin
Immediate kin	895	88.4	6.8	4.8	1,23	90.9	.1***	8.3***
Extended kin	263	63.0	7.8	29.2	716	67.2	5.3**	25.8*
Nonkin
Romantic partner	133	84.2	12.5	3.1	67	81.2	12.4	6.4
Friend	1,834	59.1	20.7	20.0	2,928	67.9***	7.9***	23.8**
Roommate	219	56.9	34.5	8.3	61	63.5	33.6	2.4*
Neighbor	83	42.1	33.8	24.0	409	63.6***	14.0**	21.2
Workmate	311	37.0	43.2	19.5	404	55.7***	19.5***	24.2
Schoolmate	243	57.8	26.6	15.5	54	57.9	9.2**	30.1*
Churchmate	120	51.8	21.4	26.7	266	57.9	15.4	26.6
Acquaintance	116	22.7	49.7	27.1	204	26.5	40.9	31.3
Type of exchange
Socialize	1,996	65.8	19.3	14.89	3,477	75.7***	6.77**	17.12*
Confide	1,173	82.4	8.01	9.37	1,862	87.0***	4.02***	8.67
Advice	1,083	85.4	5.77	8.68	1,609	88.2*	3.01***	7.80
Practical help	713	65.4	18.58	15.86	885	75.9***	8.22***	15.75
Emergency help	1,176	79.1	10.70	10.17	1,860	86.2***	3.68***	9.66
Provide support	1,361	76.0	14.12	9.91	2,427	79.1*	6.40**	14.11**
Difficult relationship	431	69.6	17.67	10.51	572	77.8**	6.20***	14.00

Note: Young cohort refers to 21–30 year-olds and old cohort to 50–70 year-olds at wave 1

^*p < .05.

^**p < .01.

^***p < .001 (two-tailed tests) for differences between age groups

With respect to nonkin, the results show that a relatively large proportion of acquaintances (about 50 and 40 percent in the young and old cohorts, respectively) and roommates (more than 30 percent in both age groups) were new associates. Reflecting the progress through college and, for some, to post-graduate programs that many young adults experience, the percentage of new friends and new schoolmates was substantially higher in the younger than older age group. Schoolmates, by contrast, were more likely to be awakened ties among the older respondents. Similarly, neighbors and workmates were more likely to be truly new ties in the young cohort but awakened ties in the old one. In general, these findings highlight variations in social life associated with the different stages in the life-course of younger and older adults. They, in fact, confirm more impressionistic accounts of life-stage differences, that young people are building networks and older people are largely relying on networks they had built.

Turning next to types of interaction with ego, the lower panel in Table 3 shows that in the young cohort almost one-fifth of the alters with whom respondents socialized, from whom they received practical help, and whom they considered difficult were truly new ties. In the old cohort, the percentage of truly new ties in all types of interactions was very small (less than 10 percent). Older respondents reported more awakened ties than truly-new ties, especially for socializing (17 percent), practical help (almost 16 percent), and the provision of support to alter (14 percent). By and large, these results indicate that the types of interactions examined in this study tended to occur with alters who were continuously active - or, at least, consistently reported to be - in the network. This is particularly the case for alters in whom respondents confided and from whom they received or would receive advice and emergency help.

Origins of Truly New and Awakened Ties

Respondents were asked an additional series of questions about a subsample of the names⁷ they mentioned. One was “Which one of the following ways best describes how you first met [name]? Choose one.” Although it provides information about fewer alters, answers to this question allowed us to learn more about who the newly-listed ties were and what their origin in the network was. The results presented in Table 4 echo our previous findings by showing that certain settings serve as opportunity contexts for meeting new associates. In the younger cohort, these include school (20 percent), work (27 percent), and through friends (17 percent), and in the older cohort, work (19 percent) and through religious and other organizations and groups (20 percent).

Table 4.

Contexts of where truly new and awakened ties were first met (weighted)

	Young cohort		Old cohort
	Truly new	Awakened	Truly new	Awakened
Same family (by birth or marriage)	5.3%	15.6%	9.1%	24.2%
Grew up in same neighborhood	0	1.2%	0	.4%
Met in high school, college, grad school or post-graduate program	20.3%	30.6%	.7%	4.6%
Met as neighbors	6.8%	0	7.6%	6.8%
Met at work	26.6%	15.0%	18.9%	13.9%
Met at religious (e.g., church) or other group or organization	7.2%	11.6%	18.2%	20.3%
Met through friends	17.4%	8.1%	9.1%	12.8%
Met through family member (e.g., spouse, child)	2.4%	2.3%	10.6%	6.8%
Met online	6.3%	6.4%	6.1%	0
Other	7.7%	9.2%	19.7%	10.3%
Total	100%	100%	100%	100%
N	207	173	132	281

Notes: based on UCNets’ subsample of alters. Young cohort refers to 21–30 year-olds and old cohort to 50–70 year-olds at wave 1

While these contexts are important for the creation of new associates, they also appear to contribute to enlarging the pool of dormant ties that can be awakened. Thirty percent of the awakened ties in the young cohort were first met at school and twenty percent of the older cohort’s awakened ties were first met at religious and other organizations and groups. Hence, these organizational settings provide the opportunity to create meaningful ties, which even if they then become dormant for a long time can be successfully activated at some point in the future. Not surprisingly, family members were also a major source of awakened ties in the older cohort. Respondents in this age group mentioned that the origin of nearly one-fourth of their awakened ties was the family.

Who Is Likely to Be a Truly New or an Awakened Tie? Results from Multilevel Analyses

We use multilevel modeling to simultaneously examine the alter- and ego-level characteristics associated with the (log) likelihood of an alter being a truly new or an awakened tie, while controlling for the respondents’ background characteristics. Since our dependent variable includes three categories (i.e., whether a tie is truly new, awakened, or previously-listed), we assess a multinomial logit regression in which truly new ties and awakened ties were compared to previously-listed ties, the omitted category. Table 5 presents the odd ratios of these analyses. In both age groups, immediate kin and extended kin were far less likely to be truly new ties compared to nonkin. Among the younger respondents, immediate kin also had lower odds - by almost 90 percent - of being awakened ties. Stated differently, family members, particularly immediate kin, tended to be continuously-listed members of the network.

Table 5.

Multilevel multinomial regression results predicting the log-likelihood an alter would be named a truly new or awakened tie versus a previously-listed tie, odds ratios (weighted)

	Young cohort		Old cohort
	Truly new OR	Awakened OR	Truly new OR	Awakened OR
Intercept	.912	.753	.781	1.380
Alter-level variables
Role relationship
Kin
Spouse and immediate kin	.199***	.115***	.019***	.192***
Extended kin	.128***	.955	.177***	.784
Nonkin
Romantic partner	1.169	.405	2.076	.756
Friend	.458**	1.002	.251***	.750
Roommate	1.483	.398	2.139	.076***
Neighbor	1.590	1.547	1.014	.736
Workmate	3.537***	1.476	1.607	.929
Schoolmate	1.475	.898	.735	1.708
Churchmate	.780	1.240	1.546	1.524
Acquaintance	3.117**	2.768*	3.002***	1.237
Type of exchange
Socialize	1.032	.541**	.637**	.457***
Confide	.458***	.599**	.542**	.449***
Advise	.355***	.703+	.556**	.523***
Recent practical help	1.435	.968	1.222	.965
Emergency help	.647	.848	.514**	.510***
Provide support to alter	.876	.567***	.729	.667**
Difficult alter	.820	.578	.386***	.382***
Alter-level characteristics
Emotionally close	.486**	.603*	.416***	.504***
Live over one hour away	.637*	.963	.874	1.333*
Ego-level variables
Life events between wave 2 and wave 3
New partnership	1.311	1.439	1.114	1.253
Birth of child	4.885***	1.240	1.599+	1.047
New job or school	.891	.965	1.591	.763
Graduated from school	1.023	1.321	---	---
Death of a close person	1.522	1.299	1.305	1.127
Break in friendship	1.031	.974	1.271	1.482**
End job	.916	.832	1.045	1.476*
School, work, financial prob.	1.353	1.405*	.991	1.075
Health problem	1.002	1.007	1.453*	1.092
Move near	1.748**	.922	1.560	.576*
Move far	2.965***	1.761**	4.600***	.702
Sociodemographic variables
Male	1.071	1.149	.778	1.212
Asian	.702	.975	1.407	1.192
Latino	1.131	.678	1.382	1.170
Black and other	.675	1.399	1.212	1.580*
Foreign born	1.018	.877	1.528	.762
Married	.973	.983	.715*	.812*
Education less than BA	.705	1.111	1.182	1.132
Education BA	.820	.904	1.200	1.138
Poor health	.880	.654	.823	.861
Extroversion	1.056	1.067	1.006	1.021
Neuroticism	1.077	.913	1.006	.907
Network size (wave 2)	.948*	.961	.867	.958**
Personal reference or Facebook recruit	1.168	1.249	---	---
Variance components
Between-ego variance	.394	.158	.549	.183
Chi-square (intercept)	566.170***	486.748***	678.371***	695.583***
N alters	4,059		5,764
N respondents	382		558

Note: Young cohort refers to 21–30 year-olds and old cohort to 50–70 year-olds at wave 1

⁺p < .06.

^*p < .05.

^**p < .01.

^***p < .001 (two-tailed tests)

Similarly, alters named as friends were significantly less likely to be truly new ties compared to alters not named as friends, by more than 50 percent in the young cohort and some 75 percent in the old cohort. This finding is not surprising considering that many friendship relationships take time to develop. Additional analyses we conducted with the subsample (results not shown) corroborate this idea by showing that in this study friends were known to ego for some 15 years on average (a number very close to that for awakened ties,17 years). By contrast, acquaintances were about three times likelier to be truly new ties. In this younger cohort, workmates were especially likely to be new associates. The odds of alters who were described as workmates being truly new ties was more than triple that of alters not described as workmates.

Turning to the results for types of exchanges, we found that in both age groups alters who served as confidants and advisors had a significantly lower likelihood of being either truly new or awakened ties. In other words, alters engaged in these exchanges actively persist in the network, perhaps because emotional support and being a source of advice involve a relatively high degree of intimacy and thus tend to characterize interactions with core network members (for similar findings see Marin and Hampton 2019). However, the distinctiveness of confidants and advisors remained significant even after controlling for emotional closeness. No other type of interaction was significantly related to the likelihood of being a truly new tie among respondents in the young cohort. In the old cohort, however, alters with whom respondents socialized and those who were considered a source of help during emergencies were less likely to be listed as truly new ties.

Additionally, most types of exchanges (except recent practical help) were associated with lower odds that the alter would be an awakened tie. Alters perceived to be able to help in emergencies were less likely to be awakened ties, but in the older cohort only. In general, these results, which were quite similar in the two age groups, suggest that respondents mainly engaged - or reported engaging - socially with people who were continuously active members of their network.

Interestingly, among older respondents, alters perceived to be difficult seem to remain in the network over time. Difficult alters were less likely to be either truly new or awakened ties than to be listed in previous waves. This finding could be interpreted as contradicting socioemotional selectivity theory, according to which people tend to forsake less meaningful and rewarding relationships as they age (Charles and Carstensen 2010), but they may also reflect the higher level of dependency of older adults and the often ambivalent relationships they have with close network members (e.g., Fingerman et al. 2004).

Looking at the alter-level control variables shows that alters to whom respondents felt especially close were significantly less likely to be either truly new or awakened ties, supporting the idea that meaningful emotional ties tend to persist in the network. This idea is further corroborated by our finding (not shown) that emotionally close alters were known on average for a longer time (almost 27 years) than alters who were not considered emotionally close (16 years). Alters who lived over one hour away from the respondent were less likely to be truly new among younger respondents, highlighting the role of physical proximity in young people’s formation of novel connections even in the digital age (Mok et al. 2010; Small 2017). Among the older respondents, distant alters had significantly higher odds of being awakened ties.

Next, we next turn to an examination of the life event variables, starting with events that usually involve entering new social and institutional contexts.⁸ The only significant association is found for the birth of a child. The odds of young respondents who have had a child between waves 2 and 3 to report truly new ties in their network was almost five times higher than that of respondents who did not have a child. The birth of a new child can be treated as an important change in context, which often gives parents the opportunity to meet new people, such as other parents and various service providers (Bost et al., 2002). A similar association but of a much lower magnitude and marginally significant is observed for the birth of a grandchild in the older cohort. In the older cohort, having had a health problem since the previous wave is also found related to higher odds of reporting truly new ties. These new associates may be health providers and other professionals providing care, assistance, and advice to ego who are recruited to the network following the onset of the condition and its diagnosis (Perry and Pescosolido 2012). Interestingly, though, having had a health problem is not significantly related to the likelihood of awakening dormant ties.

We then examine the life events expected to increase the need or wish to reconnect with old associates. The results indicate that in the young cohort, having experienced work, school, or financial problems since the previous wave is associated with higher odds of reporting awakened ties. These types of problems are typical of the transition to adulthood in contemporary society and may reflect the time it takes for many young people to establish themselves in the labor market and the high level of uncertainty they face in those stages of the life-course (Swartz 2009).

Other significant associations were observed in the older cohort and pertain to the loss of a relationship and the heightened social needs it may entail. Older respondents who experienced the death of a close person since the previous wave had higher odds of reporting awakened ties. Similarly, those who experienced a major break in a friendship or other (non-romantic) relationship reported an increased likelihood of having awakened ties. Perhaps such a loss leads people to fall back on former relationships and prompts their reconnection to them. McCarty and Levin (2019) indicate that active and dormant ties are often substitutable. They found that people tended to report fewer dormant ties when they had many active ties in their networks and argue that “when active ties are scarce, dormant ties may represent a greater source of social identity and viable opportunity for productive interactions in the future” (1440).

We also found an important effect for residential moves. Respondents in the younger cohort who moved out of the region reported an increased likelihood of having truly new, as well as awakened, ties in their network. Older respondents who moved out of the region were more than four times likelier to have truly new ties than respondents who stayed put. These findings corroborate the idea that residential mobility, especially when it occurs at a larger distance, entails a change of context that grants novel occasions to form new social connections. Moving but staying in the same area is a less drastic change because it allows people to remain active with their locally-based networks. Correspondingly, our analyses reveal that among the older respondents moving near was associated with a reduced likelihood of reporting awakened ties. But among the younger respondents, it was related to an elevated likelihood of having truly new ties.

Finally, Table 5 shows that the respondents’ sociodemographic and personal trait variables did not matter much. Most of the coefficients were not statistically significant, except for the lower odds of married respondents and higher odds of Blacks in the older cohort to report awakened ties. Network size at wave 2 was associated with a reduced likelihood of having truly new ties in the young cohort and awakened ties in the old cohort. This finding may reflect the lower need of people with larger networks to add new associates, but it could also capture a ceiling effect on the number of named alters or fatigue effects. Overall, and consistent with prior research (e.g., Marin and Hampton 2019), the results of the multilevel multinomial logit models indicate that most of the variance was at the alter-rather than the respondent-level, suggesting perhaps that forming new ties and reconnecting to old ones may not be about ego as much as about the circumstances of the ties and their characteristics.

Discussion and Conclusion

In this study, we took advantage of the longitudinal egocentric network data in UCNets to examine newly-listed, or added, ties. We focused on those alters who were mentioned for the first time at the later wave of UCNets and asked who they were and how new they really were to the network. We also identified alters whom respondents had known but not named at the time of previous waves and asked why they listed them only later. This investigation generated important insight into the “sampling” process involved in the name-generating procedure at the methodological level, and into the sources of change, as well as continuity, in networks at the substantive level.

We found that similarly to dropped ties (Fischer and Offer 2020), newly-listed ties are not all the same; they vary particularly in degree of newness. Some, a minority, are brand new, but many others were previously known to ego; they are revitalized or newly-expanded ties. In the young cohort, about half of the newly-listed alters were truly new associates. These mainly included recently met coworkers, neighbors, roommates, and acquaintances. Consistent with the structure of opportunity approach (Feld, 1982; Feld and Carter 1982; Fischer, 1982; Mollenhorst et al., 2014), these types of relationships tend to emerge in new settings that ego has entered, such as college and workplace. New ties may also form following life events that shift a person’s social environment, such as becoming a parent in this study. It is possible that some of these new ties will develop over time into more intimate relationships and move closer to the network core while others will go into dormancy and become part of one’s pool of potential helpers in the future.

In the old cohort, by contrast, newly-listed ties were overwhelmingly connections to associates who were previously known to ego. This discrepancy reflects the different life stages of each group and their implications for social life: the higher mobility and frequent transitions into new institutional and social contexts associated with early adulthood (Bidart and Lavenu, 2005; Bidart et al., 2020) and the greater stability of middle-aged people, as well as their accumulation of connections over the life course (Antonucci et al., 2010; Quinn, 2013).Young adults are generally in the process of building their networks of active and (future) dormant ties whereas older people, although they continue to form new relationships as our findings clearly show, tend to rely on the networks they have built over decades.

Nevertheless, in the young and old cohorts alike, the vast majority of the alters the respondents named in the third wave of UCNets were not new but rather had been mentioned in the first wave, the second wave, or both. This finding suggests that when respondents fill in the name-generating questions asking about engagement in various social exchanges and activities, they tend to name core network members. Those alters continuously present in the network were most likely kin – especially immediate kin − and friends, as well as alters in whom the respondent confided, from whom they received advice, and (if older) with whom they socialized.

Not surprisingly, the percentage of previously-listed ties was significantly higher in the old age group, pointing to greater continuity in the social lives of middle-aged people compared to their younger counterparts. Resonating with the socioemotional selectivity thesis (Charles and Carstensen, 2010), this finding may also point to the tendency of older people to interact with a more restricted group of alters who over the years have become increasingly meaningful to ego.

The overall large share of persistently active ties in the networks of UCNets’ participants – 65 percent of the alters in the young cohort and 71 percent in the old cohort had appeared in wave one or two − highlights their relatively stable nature, at least in the short run. This trend could be explained by the types of social activities captured in the name-generators employed in UCNets, mainly the exchange of support, and their bias toward the network core (Burt, 1984; Perry et al., 2018; Small, 2017). It may also result from the relatively short duration of UCNets, three to four years between the first and last wave. Mollenhorst et al. (2014), for example, reported much lower stability in their study of personal networks in the Netherlands over a seven-year period. They found that new alters (i.e., alters who were mentioned only in the second wave in 2007 but not in the first wave in 2000) constituted on average about three-fourths of the network alters. Clearly, a longer time frame provides more opportunities for major changes to occur at both the alter and ego levels.

When interpreting these results, however, one should bear in mind that they refer to ego’s perceptions of who is active in the network based on the specific name-generators tapped in UCNets. These perceptions are not necessarily accurate. As noted in the introduction, network members who engage with ego in domains not measured in the study have a very low likelihood of being reported in the network (Feld et al. 2009; Fischer and McCallister 1978; Marin and Hampton 2007; van der Poel 1993). But beyond this bias, some ties may not behave in ways consistent with one’s expectations. In other words, a gap might exist between alters perceived to be supportive and those who will actually provide support when the need arises (Norris and Kaniasty 1996). This distinction implies that networks may be less stable than what this study suggests, specifically that the number of truly new ties may be underreported. To illustrate this point, consider for example an emergency situation, in response to which ego may turn to new associates for help because they are highly accessible (e.g., Desmond 2012; Small et al. 2016). But ego may not report these new alters when asked about a hypothetical emergency situation because in the absence of an actual need ego is likely to think about the core members of her or his network.

Although newly-listed ties were a minority in this study, their examination revealed important findings related to the dynamic nature of social relationships. As noted above, our results suggest that a large share of newly-listed ties consist, in fact, not of truly new alters but rather of alters with whom ego was previously – or is still currently - engaged. Some of these alters were not mentioned in the prior waves simply because they were forgotten, but others refer to dormant ties that got awakened at the third wave. A larger proportion of extended kin were awakened rather than truly new ties. In the old cohort, the percentage of awakened ties was also relatively high among friends, neighbors, workmates, schoolmates, and churchmates. These findings reflect the larger pool of past social connections whom older respondent can mobilize and from whom they can draw support if the need arises. It is not surprising that older respondents reported more awakened ties than their younger counterparts considering that the possibility of reconnection likely occurs with relationships that have lapsed among people “who have achieved the necessary longevity and life stage at which it is feasible to reestablish some form of connection” (Quinn 2013, p. 399).

By revealing which ties are likely to be awakened and for whom, this study provided a glimpse into people’s past and the contexts in which they were once active, thus allowing researchers to learn about the layered structure of networks (Kilduff et al. 2006). Our findings suggest that dormant ties are very much alive in people’s minds and can be easily reactivated. They also shed new light on the events likely to trigger the reappearance of dormant ties in the network. We found that in the older cohort, having experienced a major relationship break was significantly associated with reporting awakened ties. Considering the limits that people experience in the number of ties they can simultaneously maintain active (Killworth et al., 1990), it could be that a relationship break created the need, as well as more room, both cognitively and practically, to reactivate dormant ties (see McCarthy and Levin, 2019). These findings have implications for the understanding of the recall and listing process in egocentric research. They suggest that the elicitation process in most network surveys, which are not based on a qualitative intensive questioning of ego’s total of connections, (e.g., Bidart et al. 2020), is triggered by recent activity and thus likely underestimates their latent sources of support.

While this study extends previous research on dormant ties in important ways, it leaves many questions unanswered. For example, how long were these connections dormant before they got awakened and how likely they are to fall back into dormancy? Levin et al. (2011) propose that many social ties are like a pendulum; they move in and out of dormancy. Future research would greatly benefit from exploring these issues in a more comprehensive way while linking them to changes in the lives of both ego and alter. Our study, by distinguishing between truly new and awakened ties, is a first step in this direction. It gave us the opportunity to uncover parts of the deeper layers of personal networks and suggests that there is more complexity to them than meets the eye.

Highlights.

• This study examines which alters were mentioned for the first time in the later wave of UCNets.

• Only one-third of all alters were newly-listed, suggesting stability in the network core.

• Most of the newly-listed ties in the young cohort were truly new ties whereas most of the newly-listed ties in the old cohort were awakened ties.

• Alters who were not new mainly included kin, confidants, and advisors.

• Multilevel models estimate the alter and ego characteristics associated with the likelihood of a tie being truly new or awakened.