Closeness, loneliness, support: Core ties and significant ties in personal communities

Social Networks 34 (2012) 206–214

Contents lists available at SciVerse ScienceDirect

Social Networks journal homepage: www.elsevier.com/locate/socnet

Closeness, loneliness, support: Core ties and significant ties in personal communities夽 Jens F Binder a,∗ , Sam G.B Roberts b , Alistair G Sutcliffe c a b c

Nottingham Trent University, UK University of Chester, UK University of Manchester, UK

a r t i c l e

i n f o

Keywords: Core ties Significant ties Personal communities Friendship maintenance Social loneliness Emotional loneliness

a b s t r a c t Past research has introduced further distinctions within the strong ties that form our personal communities. This work aimed at a comparison between core and significant ties in terms of their emotional closeness to ego, the social provisions that are exchanged, the relationship maintenance behaviours reported, and ego’s loneliness. Measures for all these variables were assessed in a survey study. Evidence for a trade-off between the number of ties and their level of intimacy was obtained such that having more core ties was associated with lower levels of intimacy. Distinct signatures for friends from both sets of ties emerged for friendship maintenance behaviours and social provisions exchanged. Further, social and emotional loneliness were differentially related to the number and the intimacy of core and significant ties. © 2011 Elsevier B.V. All rights reserved.

1. Outline Personal networks typically consist of a great variety of ties, differing in their origin and their duration (Grossetti, 2005, 2007). Research has consistently shown that different types of alters (social contacts) serve different purposes for the ego (the network owner) and provide different social provisions (Carbery and Buhrmester, 1998; Granovetter, 1973; Weiss, 1974; Wellman and Wortley, 1990). Further, studies on personal networks have usually followed the notion of an interpretable average network size, be it at the level of wider acquaintance (McCarty et al., 2001) or for core networks (McPherson et al., 2006). These average figures were seen from early on as the result of constraints operating on the ego, the owner of the network (Bernard and Killworth, 1973; Pool and Kochen, 1978). Constraints may consist of limits on time for socialising, limited access to communication technology (Wellman, 1999), or even limits on the amount of social information our brain can process (Hill and Dunbar, 2003; for an overview see Roberts, in press). Thus, although most people would agree that maintaining social ties is a very beneficial thing to do, there might be a limit as to the number of ties that can be maintained.

夽 This research was supported by a joint grant from the ESRC and the EPSRC. ∗ Corresponding author at: School of Social Sciences, Nottingham Trent University, Burton Street, Nottingham NG1 4BU, UK. Tel.: +44 115 8482416; fax: +44 115 8486929. E-mail address: [email protected] (J.F. Binder). 0378-8733/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.socnet.2011.12.001

The present work aims to build on these ideas and to extend them in an investigation of the social ties that are of highest emotional relevance to ego. In particular, we attempt to distinguish between two types of strong social ties, following past research on social networks (Wellman and Wortley, 1990) and recent arguments from evolutionary anthropology (Dunbar, 1998, 2008). We will focus on the size and composition of these sets of ties, and how their size may ultimately be constrained by ego’s finite resources for socialising. Further, we will look into quantitative and potential qualitative differences between both types of alters in terms of the functions they fulfil for the ego and their implications for ego’s emotional well-being. In order to keep alters comparable, we will limit our investigation of differences to friends, that is, non-kin alters. In the following, we will first discuss constraints on network size in more detail. We will then turn towards the question of how to distinguish between different types of alters by drawing on studies of friendships, social support, and loneliness. Finally, to conclude the introduction, we will sum up our areas of investigation.

2. Constraints on network sizes The size of any network will depend on how the network is defined, and a closer look at existing terminologies is necessary here in order to deal with some inconsistencies in reported findings. For practical and theoretical reasons, researchers have often distinguished between core and peripheral networks (McPherson et al., 2006). Importantly, in the context of personal networks,

J.F. Binder et al. / Social Networks 34 (2012) 206–214

this amounts to a distinction between alters at different levels of relevance or emotional closeness to ego. In want of a better term, we will refer to these alters as being part of different sets within the network. Past studies have tended to look either at the overall network size (e.g., McCarty et al., 2001) or at the closest set (e.g., McPherson et al., 2006). It can be argued, however, that sources of crucial, ongoing support to ego do not only reside in a set of a very few intimate alters, but can be found in somewhat less intimate alters as well (Wellman and Wortley, 1990; Plickert et al., 2007). Here, we are particularly interested in what has been called personal communities (Pahl and Spencer, 2003; Spencer and Pahl, 2006; Wellman, 1979), the set of active and relevant ties that includes close friends and family and forms our micro-social world. Several lines of research suggest that personal communities can be usefully divided into two different sets of ties, albeit under different labels. Wellman and Wortley (1990) distinguished between strong ties (‘those you feel closest to outside your home’) with a median size of four and significant ties (‘those who are in touch with you in your daily life and who are significant in your life’) with a median size of seven. Similarly, a recent Pew Social Ties Survey (Boase et al., 2006) separates core from significant ties whereas Wellman et al. (1988) opt for routine and intimate ties. Dunbar and Spoors (1995; Dunbar, 1998) define support ties as ties to significant alters from whom ego would first seek help in times of crisis. This is contrasted with sympathy ties, defined as ties to significant alters that are contacted monthly (Dunbar and Spoors, 1995) or alters whose death would leave you personally devastated (Buys and Larson, 1979). Research by Dunbar and colleagues has shown that both sets taken together average 12–15 alters, with the more exclusive set of support ties containing around five alters (Dunbar and Spoors, 1995). This small number of most intimate ties has also emerged in a number of earlier studies on personal networks (Milardo, 1992). For the present work, we will adopt the recently used labels ‘core’ and ‘significant’ ties for those ties that make up our personal communities. For the sake of simplicity, we will assume that these sets are equivalent to ‘strong’ and ‘significant’ and so forth, although we are well aware that this might be an oversimplification. It remains a matter of debate whether core and significant ties reside in two discrete sets or whether they are organised in different ways. Non-discreteness could in part account for a certain fluctuation in the estimates of tie numbers. As with all sets of network ties, number estimates vary according to the elicitation procedure and the exact definition of these sets. The General Social Survey, for example, defines core ties exclusively as belonging to important discussion partners (McPherson et al., 2006) and arrives at comparatively low figures (but see Fischer, 2009; Spencer and Pahl, 2006 for critiques). Studies using more inclusive definitions (e.g., Boase et al., 2006) unsurprisingly yield higher estimates. Constraints on network sizes have featured prominently in studies by Dunbar and colleagues (Hill and Dunbar, 2003; Roberts et al., 2009). For the most part, however, their research has been concerned with extended personal networks that include core and peripheral ties. As already mentioned, the absolute size of the network is affected by the name generator used and the method of estimation. Asking participants who they know for a number of social categories gives a network size of around 290 (McCarty et al., 2001). In a re-analysis of the same data, Zheng et al. (2006) arrive at a median of 610 for overall network size. In comparison, Dunbar and colleagues have investigated constraints for what they call the active network – including alters that ego is in contact with at least once a year and has a genuine personal relationship with. This active network is thought to have an upper bound of around 150 alters (Hill and Dunbar, 2003). Roberts et al. (2009) found a negative relationship between active network size and mean

207

emotional closeness between ego and alters in their network. From this perspective, time and cognitive constraints appear to result in a trade-off between the number of alters in the network, and the emotional intensity of each relationship in the network. There is conflicting evidence whether constraints of the same nature operate on personal communities. Wellman and Frank (2001) found that strong and supportive ties seemed to operate independently of the overall network. Other studies found that time and cognitive constraints do influence the number of closer ties. Stiller and Dunbar (2007) demonstrated that the number of levels of intentionality an individual can process is related to the number of core ties whereas working memory capacity is related to the number of significant ties. Further, there is evidence that when individuals have less time for social relationships (e.g., entering into a romantic relationship, having dependent children) the number of strong, non-kin ties that they have tends to decrease (McCannell, 1988; Milardo et al., 1993). However, as at the level of extended networks, the constraints may not just influence the size of personal communities but also their composition – for example, more core ties may imply less emotionally close alters as there is a limited amount of time and effort that can be invested in social relationships. Next to the issue of constraints, another area of enquiry in this work concerns the distinction between core and significant ties. Past studies on personal communities have yielded detailed findings on network size and composition (Degenne and Lebeaux, 2005; Fischer, 1982; McPherson et al., 2006), the type of support that these networks offer (Wellman and Wortley, 1990) and how these networks change over time (Suitor and Keeton, 1997; Wellman et al., 1997). However, although the distinction between ‘strong’ and ‘weak’ ties is well known (Granovetter, 1973, 1983) and established as a theoretical concept, differentiation within the close ties of personal communities is little understood.

3. Differences between core ties and significant ties Our understanding of core and significant ties, and the potential differences between them, is strongly influenced by social psychological studies on relationships. We contend that information on functional differences between the two sets of ties can be gained by adopting an exchange perspective on relationships (e.g., Rusbult, 1980) which emphasizes investments in social ties and costs arising from them. Two related traditions emerge from this perspective that are relevant to the present work. Researchers so far have focussed either on maintenance efforts or on the generation and exchange of social provisions. The first approach has been taken in studies on friendships (Hays, 1985; Oswald et al., 2004; Rose and Serafica, 1986). The quality and duration of friendships clearly depends on how much maintenance they receive and of which kind (Hays, 1985, 1989; Oswald et al., 2004). Oswald et al. (2004; see also Oswald and Clark, 2006) have developed and validated a measure of friendship maintenance behaviours that comprises four aspects: positivity, supportiveness, openness, and interaction (i.e., joint activities). These dimensions reliably distinguished between close and casual friends (Oswald et al., 2004), but so far they have not been used to differentiate core from significant ties within personal communities. The second approach goes beyond friendships and addresses exchange processes in social ties in general. Weiss (1974) proposed that different social provisions (e.g., advice, intimacy) are derived from different types of relationships (e.g., parents, romantic partner). This implies that a diverse social network should provide optimal support. The idea has received empirical support (Carbery and Buhrmester, 1998; Plickert et al., 2007; Russell et al., 1984;

208

J.F. Binder et al. / Social Networks 34 (2012) 206–214

Wellman and Wortley, 1990) although provisions are far from being exclusively associated with relationship types. Social networks are a source of highly important support to sustain the ego both materially and emotionally, in times of hardships as well as in every-day life. Individuals with low levels of social support have higher levels of morbidity and mortality, especially from cardiovascular disease (for reviews, see House et al., 1988; Uchino, 2006). In a similar vein, Russell et al. (1984) demonstrated that loneliness was contingent on social network metrics while at the same time related to indicators of anxiety and depression. So it comes as no surprise that research has looked at loneliness as a correlate of malfunctioning support networks. The link between loneliness and network variables such as size, density, and quality of ties has been established by several independent studies (Russell et al., 1980; Sarason et al., 1987; Stokes, 1985). Some researchers have made a distinction between social loneliness and emotional loneliness (see Green et al., 2001 for an overview). Building on the original theorizing by Weiss (1973, 1974), social loneliness is seen as indicative of a lack of social relationships, possibly due to a lack of stimulation and reassurance. Emotional loneliness is seen to occur when there is a lack of intimacy and deep understanding of the kind we get from a romantic relationship, but also from other close confidants. Studies so far have empirically validated both concepts (Russell et al., 1984; Vincenzi and Grabosky, 1989). Social loneliness has consistently been linked to the size of support networks (Green et al., 2001; Russell et al., 1984; Vaux, 1988). In contrast, the same studies have shown that emotional loneliness depends mostly on relationship status, that is, on the presence or absence of a romantic partner.

4. Study aims Based on the reasoning outlined above, we can identify three main areas of investigation for the present work, all based on comparing core and significant ties. Such a comparison is, we believe, useful even if the difference between the two should merely be one of quantity rather than quality. First, constraints on the size of personal communities may be reflected in a negative relationship between the number of alters in each set of ties and the degree of emotional closeness between ego and those alters, as research has shown at the level of the extended network (Roberts et al., 2009). This rests on the assumption that social time is limited and the frequency of contact between ego and alter is closely related to the emotional intensity of the relationship (Hill and Dunbar, 2003; Roberts and Dunbar, 2010). Individuals may either invest lots of time and have frequent communication with a small number of people (resulting in a small network but with high emotional closeness) or less time and less frequent contact with a larger number of people (resulting in a large network but with lower emotional closeness). However, as stated above, it could also be that strong ties are subject to different processes as implied by Wellman and Frank (2001). Second, for both maintenance behaviours and social provisions we aim to compare the pattern of relationship activities for core and significant ties. In this, we are following several lines of research that strongly suggest differentiation within personal communities (e.g., Dunbar, 1998; Milardo, 1992; Wellman and Wortley, 1990). In particular, we are interested in the question whether core ties simply provide everything at a higher level than significant ties, or whether both types of ties show distinct signatures. Third, no attempt has been made so far to look at the relationship between the concepts of social and emotional loneliness on one hand and core and significant ties on the other. Given the different functions assigned to these ties, we expect them to be differentially related to the types of loneliness. It is plausible to assume that social

loneliness should be negatively related to the size of one’s personal community. However, it is less clear whether this holds to the same degree for core and significant ties. Regarding emotional loneliness, one possibility is that a negative relationship only holds for core ties, due to the high levels of intimacy associated with them. In the following, we will present preliminary evidence to address these issues. Our analyses rest on a university-based sample which has a number of implications. Universities as organisations typically contain a larger proportion of (young) students and a smaller proportion of staff, all of them in tendency with a higher socio-economic background as compared to the general population. As networks develop over the life time (Suitor and Keeton, 1997; Wellman et al., 1997) and the exchange of social provisions depends on life stages (Carbery and Buhrmester, 1998), we are cautious to take levels of mean scores and even relationships among variables as representative of a wider part of society. Our main goal, however, the comparison between different types of ties, can be attempted if we keep in mind the nature of the sample and take the findings as first evidence. 5. Method 5.1. Sample A total of 303 participants, 70% female, 30% male, were recruited at a university based at a large city in England. Students (72%) as well as staff (28%) were recruited in a proportion in line with the organisation’s overall composition. This procedure yielded a sample with a broad age range (18–59 years) and a mean age of 26.5 years (SD = 9.1). The vast majority (88%) reported being locally based. Participants were asked whether they considered themselves to be in a relationship with a romantic partner which was affirmed by 60%. Further analyses focusing on the sample composition are reported below. 5.2. Procedure and measures Participants responded to an online survey. This method has been validated extensively against other techniques for obtaining self-report measures (e.g., Naus et al., 2009) and has been used by Nettle (2007) for eliciting core and significant ties. Care was taken that each participant was able to access the survey only once. First, core and significant ties were assessed with name generators that followed closely definitions used in prior studies (Dunbar and Spoors, 1995; Buys and Larson, 1979; Nettle, 2007). After providing background information, participants were first asked to give an initial estimate of the number of core ties using the definition “people you would seek advice, support or help from in times of severe emotional or financial crisis”. They were then told to produce a list of all the people fulfilling this definition. For each alter listed, respondents were asked how emotionally close they were to the alter on a scale from 1 to 10, with 10 indicating maximum closeness. This measure has been used successfully in previous studies (Hill and Dunbar, 2003; Roberts et al., 2009). Further, they reported on the alter’s gender and whether there was any relation by birth or marriage to the participant. The procedure was the same when eliciting significant ties. Participants were explicitly told to consider people in addition to the ones they had listed before. The definition used was “people whose death would leave you personally devastated”.1 The order in which

1 Of course, there are numerous implications the death of another can have. The demise of a business partner or a spiritual leader could have devastating consequences, for example. In the context of this study, however, it was obvious to respondents, by the additional information assessed, that only personally known

J.F. Binder et al. / Social Networks 34 (2012) 206–214

core and significant ties were assessed and the imposed exclusivity of groups follows the notion that core ties should display highest levels of intimacy, yet may also fit the definition for significant ties (Dunbar and Spoors, 1995). In the next part of the survey, participants were instructed to choose one friend from among the core ties and one friend from among the significant ties, ideally a same-sex friend representative of the non-kin contacts listed. Detailed measures for these two friendships were then obtained in the form of maintenance behaviours and social provisions. Friendship maintenance behaviours were assessed using the scales developed by Oswald et al. (2004). Their instrument consists of 20 items designed to capture four dimensions of maintenance: positivity (e.g., making each other laugh; expressing thanks), supportiveness (e.g., support during a difficult time; accepting the other for who they are), openness (e.g., having intellectually stimulating conversations; sharing of private thoughts), and joint activity (e.g., making an effort to spend time together; visiting each other’s homes). Several studies (Oswald and Clark, 2006; Oswald et al., 2004) have demonstrated the reliability and validity of the scales. For each dimension, item ratings were averaged such that higher mean values indicated higher frequencies of behaviours on a scale from 1 to 5. For the chosen alter from the list of core ties (‘core friendship’) internal consistency was satisfactory for supportiveness (Cronbach’s ˛ = .90), openness (˛ = .79), and joint activity (˛ = .84). For positivity, one item (not returning each other’s messages) was removed from the scale in order to improve reliability from ˛ = .60 to ˛ = .73. Internal consistencies for the friend from the significant ties (‘significant friendship’) were highly similar. Social provisions were measured using an adapted version of the Social Provisions Questionnaire (SPQ) by Carbery and Buhrmester (1998). The SPQ asks for the degree of exchange for each social provision postulated by Weiss (1974). One provision, nurturance of others, was removed from the list because it did not seem to apply well to the context of friendships. This left eight provisions: companionship, affection, guidance/advice, reassurance of worth, intimate disclosure, instrumental aid, emotional support, reliable alliance. Rating scales again ranged from 1 to 5 with higher values indicating a higher degree of exchange. The order in which core friendship and significant friendship were presented in this part of the survey was varied on a random basis resulting in two different versions of the survey. We checked for order effects using t-tests on all friendship variables revealed no differences between the versions, and order of presentation was therefore dropped from all analyses. Finally, social and emotional loneliness were assessed following the procedure by Russell et al. (1984). Participants were asked to read through two short descriptions of loneliness and then to rate how intensely they had been experiencing both kinds of loneliness on scales from 1 to 5 with higher numbers indicating more intensity. Social loneliness was described as follows: “A possible type of loneliness involves not belonging to a group or social network. While this may be a set of friends who engage in social activities together, it can be any group that provides a feeling of belonging based on shared concerns, work or other activities.” For emotional loneliness the description ran “Another possible type of loneliness is the lack of an intense, relatively enduring relationship with one other person. While this relationship is often romantic, it can be any one-to-one relationship that provides feelings of affection and security.” Both measures have been validated by Russell et al. (1984) using the UCLA loneliness scale

others could be listed. Further, any strong dependency on financial or spiritual offers would also be picked up by the first name generator.

209

(Russell et al., 1980) and in subsequent studies (e.g., Green et al., 2001). 6. Results After reporting on preliminary checks on the sample and describing the basic characteristics of the sets of core and significant ties, we go on to report findings in the order in which we stated our areas of investigation. First, we relate the number of core and significant ties to tie strength within each set. Second, we move on to a more detailed comparison of two alters, one from each set, in terms of friendship maintenance behaviours and the exchange of social provisions. Finally, the relationship between social and emotional loneliness on one side and characteristics of core and significant ties on the other is investigated. 6.1. Preliminary analyses and set characteristics Since online surveys provide less control over the situation in which participants give their responses, we performed additional checks for participant self-selection and motivational biases. Participants who submitted complete information (n = 200) were compared to those who submitted incomplete information (n = 103). Significant differences emerged for the reported number of core ties (t(300) = 3.95; p < .001) and significant ties (t(257) = 4.17; p < .001). However, the two subsamples did not differ in terms of student/staff ratio, location, gender distribution, age or proportion of romantic relationships, thus making systematic biases due to self-selection seem unlikely. We therefore inspected all incomplete responses more closely. Participants with incomplete responses had left the study either before entering any names for significant ties (n = 44) or during the elicitation of this set (n = 59). With the elicitation of ties forming the first part of the survey, this means that these participants had not provided any information on social provisions, maintenance behaviours, or loneliness in the following part. It is therefore plausible to assume that incomplete responses are due to time constraints at the time of responding. On this basis, we decided to carry out all further analyses on the subsample of n = 200 who filled in all parts of the survey. Restricting the sample changed average numbers of core and significant ties only to a small degree (M = 5.34 vs. M = 5.78 for core ties, M = 9.47 vs. M = 10.32 for significant ties).2 Basic characteristics of core and significant tie sets are summarized in Table 1. Set sizes were roughly in line with the ones found previously (Dunbar and Spoors, 1995; Milardo, 1992; Wellman and Wortley, 1990): M = 5.78 (Median = 5) for core ties and M = 10.32 (Median = 10) for significant ties. Mean emotional closeness was higher for core than for significant ties (M = 8.60 vs. M = 7.29; t(199) = 13.43; p < .001). We also examined issues of kinship and relationship status. There was no difference between tie sets for the proportion of kin. However, being in a romantic relationship was associated with having a higher proportion of kin among core ties: M = .46 vs. M = .37; t(197) = 3.85; p < .001. Likewise, mean emotional closeness was increased among core ties for those with romantic partners: M = 8.73 vs. M = 8.42; t(197) = 2.40; p = .02. Yet, the number of core ties remained the same (without partner M = 6.02; with

2 Additionally, all analyses were performed for staff and students separately. Staff were older than students (35.7 vs. 23.1; t(299) = 13.7; p < .001) and were more likely to be in a romantic relationship (76% vs. 54%; t(299) = 3.75; p < .001), thereby increasing sample variability as intended. However, staff and students did not differ in the sizes for support and sympathy group they reported. In the main analyses, the magnitude and direction of effects obtained were essentially the same although statistical significance was sometimes affected, probably due to the different sample sizes. Results are presented here for staff and students taken together. Further information can be obtained from the corresponding author upon request.

210

J.F. Binder et al. / Social Networks 34 (2012) 206–214

Table 1 Summary of characteristics of core ties and significant ties. Core ties Set size Mean Median Kin proportion Emotional closeness Females Set size Proportion of female ties Males Set size Proportion of female ties

Table 2 Friendship maintenance behaviours for core friendship and significant friendship. Significant ties

5.78 (2.66) 5 .41 (.26) 8.60 (0.92)

10.32 (5.99) 10 .40 (.30) 7.29 (1.41)

5.98 (2.50) .64 (.19)

11.29 (5.91) .57 (.20)

5.27 (3.00) .48 (.26)

7.75 (5.35) .46 (.27)

Note. Standard deviations in parentheses. All other values are mean scores unless otherwise indicated.

partner M = 5.61; p = .28). Relationship status did not affect any characteristic of the significant ties set. 6.2. Set sizes and intimacy of ties We made use of hierarchical linear modelling (HLM) to investigate the relationship between set sizes and mean emotional closeness of ties. HLM takes account of the nested structure of the data (closeness ratings for each tie nested within core or significant set) and allows to make full use of the sample size (Bryk and Raudenbush, 1992; Van Duijn et al., 1999; Wellman and Frank, 2001). For each set, a two-level model was specified. On the first level (network ties), closeness was predicted from the random intercept. On the second level (respondent), the random intercept was then predicted from the sizes of the core and significant sets. First, we applied the above model to closeness among core ties. While number of core ties was negatively related to closeness of core ties ( = −.10; p < .001), the number of significant ties showed a positive relationship with closeness of core ties ( = .04; p < .001). We then used closeness among significant ties as the outcome variable. Similarly to before, number of core ties was negatively related to closeness of significant ties (ˇ = −.10; p < .01). The number of significant ties, however, was not significant. Background variables had no impact on the effects in both analyses. Including age, gender or relationship status as additional predictors left effects virtually unchanged. Given the discrepancy between findings for core and significant ties, we investigated this matter in further detail by entering the ratio of set sizes (i.e., number of core ties divided by number of significant ties) as an additional predictor at level two of the model. This was based on the idea of a possible trade-off between set sizes, in which case they would need to be considered in relation to each other in the analyses. The ratio had a mean of 0.86 (SD = 1.08; Median = 0.56) reflecting the fact that the core ties set was generally smaller than the significant ties set. For closeness of core ties as outcome variable, results remained unchanged, and ratio was not significant. In contrast, for closeness of significant ties as outcome variable, ratio was negatively related to closeness ( = −.37; p = .01). In other words, the more core ties in comparison to significant ties, the less intimate significant ties become. At the same time, the number of core ties no longer played a significant role. To provide a further illustration, we dichotomised the number of core and significant ties, via median splits, and used these as two factors in an ANOVA on closeness of significant ties. In line with the regression findings, a significant interaction emerged: F(1, 148) = 8.72; 2 = .06; p < .01. The condition of fewer significant ties in combination with more core ties showed the lowest closeness ratings (M = 6.65) as compared to all other combinations.

Behaviour

Core friend

Significant friend

Effect size d

Openness Supportiveness Joint activities Positivity

4.07 (0.75) 4.26 (0.85) 3.66 (0.93) 4.17 (0.71)

3.62 (0.83) 3.80 (0.90) 3.31 (0.88) 3.96 (0.74)

.57 .52 .38 .29

Note. Standard deviations in parentheses. Effect sizes were computed using pooled standard deviations.

6.3. Friendship maintenance behaviours and social provisions In order to compare the pattern of activities within relationships from both sets of ties, we focussed on the reported frequencies of friendship maintenance behaviours and the exchange of social provisions for the two friends sampled from both sets. To this end, we used repeated-measures ANOVAs with type of friend and behaviours or provisions as within-subjects factors. Mean scores for the analyses are presented in Tables 2 and 3. For the frequency of maintenance behaviours, main effects emerged for type of friend (F(1, 187) = 72.49; 2 = .28; p < .001) and maintenance dimension (F(3, 561) = 72.12; 2 = .28; p < .001). Overall, behaviours were more frequently engaged in with the core friend (M = 4.04) than with the significant friend (M = 3.67). In addition, an interaction between type of friend and maintenance dimension indicated that the difference between core friend and significant friend varied across dimensions: F(3, 561) = 6.18; 2 = .03; p < .001. In follow-up t-tests, we found significant differences on each maintenance dimension; but in terms of effect sizes, the gap between friends was most pronounced for openness, followed by supportiveness, joint activity, and positivity. We found a similar pattern of results regarding the exchange of social provisions. Again, a main effect for type of friend indicated a stronger exchange of provisions with the core friend (M = 4.22) than with the significant friend (M = 3.75; F(1, 176) = 72.00; 2 = .29; p < .001). Also, the degree of exchange varied across provisions: F(7, 1232) = 38.89; 2 = .18; p < .001. Levels were highest for guidance and advice (M = 4.26) and lowest for instrumental aid (M = 3.56). Importantly, type of friend and provisions interacted (F(7, 1232) = 2.36; 2 = .01; p = .02) such that differences between friends, measured by effect sizes, were biggest for reliable alliance and emotional support and smallest for affection and companionship. In both sets of findings, then, differences between core and significant friend were most pronounced for behaviours and provisions that are usually associated with stronger, deeper, and longer lasting friendships. We further compared emotional closeness for core and significant friendships and found that they differed significantly from each other: Mcore = 8.66; Msignificant = 7.75; t(188) = 7.17; p < .001. This mirrors the mean difference found for the full sets of ties reported above.

Table 3 Exchange of social provisions for core friendship and significant friendship. Social provision

Core friend

Significant friend

Effect size d

Reliable alliance Emotional support Instrumental aid Intimate disclosure Reassurance of worth Guidance/advice Affection Companionship

4.44 (0.84) 4.44 (0.88) 3.84 (1.08) 4.03 (1.11) 4.32 (0.97) 4.44 (0.75) 3.96 (1.13) 4.32 (0.91)

3.93 (0.90) 3.92 (0.97) 3.27 (1.06) 3.46 (1.17) 3.85 (0.97) 4.08 (0.88) 3.50 (1.17) 3.98 (0.90)

.59 .56 .54 .50 .48 .44 .40 .38

Note. Standard deviations in parentheses. Effect sizes were computed using pooled standard deviations.

J.F. Binder et al. / Social Networks 34 (2012) 206–214

211

Table 4 Summary of regression analyses for the prediction of social and emotional loneliness from the number of ties. Predictor

Criterion Social loneliness

Number of kin, core ties Number of kin, significant ties Number of friends, core ties Number of friends, significant ties

Emotional loneliness

b

SE(b)

ˇ

b

SE(b)

ˇ

−.025 −.068 −.011 −.052

.050 .021 .039 .017

−.04 −.24** −.02 −.23**

−.082 −.040 .162 −.084

.059 .025 .046 .020

−.09 −.13 .27*** −.32***

Note. R2 for social loneliness = .09 (p < .01). R2 = for emotional loneliness = .13 (p < .001). ** p < .01. *** p < .001.

6.4. Social and emotional loneliness In order to compare relationships between core and significant ties on one side and different forms of loneliness on the other, we ran a set of regression analyses. In particular, emotional and social loneliness were predicted from the number of kin and the number of friends in the sets of core and significant ties. We distinguished between friends and kin as they may have differing effects on social and emotional loneliness. Results are summarized in Table 4. Social loneliness was negatively related to the number of significant ties, but not to the number of core ties. With increasing numbers of kin (ˇ = −.23) and of friends (ˇ = −.24) among significant ties, loneliness decreased. The pattern was more complex for emotional loneliness. Here, the number of core friends was positively related to loneliness (ˇ = .27) while significant friends again showed a negative relationship (ˇ = −.32). Given that earlier on an association between number of ties and closeness to these ties had emerged, we also used closeness to alters rather than their number as predictors of loneliness. The results of these analyses are summarized in Table 5. No significant effects were found for social loneliness. For emotional loneliness, increasing closeness to core friends was negatively related to loneliness (ˇ = −.29) whereas closeness to significant friends showed a positive relationship (ˇ = .26). Closeness to kin did not play a role. In sum, both forms of loneliness are differentially affected by core and significant ties. While social loneliness was plausibly related to the number of significant ties, the larger of both sets, results were more complex for emotional loneliness. We will turn to this issue in detail in the discussion.

6.5. Moderation of loneliness effects by relationship status Levels of self-reported loneliness usually depend on presence or absence of a romantic relationship. We wanted to know whether the effects related to loneliness reported so far were dependent on respondents’ relationship status. We therefore used relationship status as a moderator variable (Aiken and West, 1991; Baron and Kenny, 1986) in each regression described in the previous

section. For each predictor variable an interaction term was generated by centering the variable on its mean score and then multiplying it with relationship status (coded as 1 and 0 for respondents with and without a partner, respectively). All interaction terms, together with the other predictors and relationship status, were then entered in a regression. For the prediction of social loneliness, relationship status interacted with the number of core kin (ˇ = .22; p < .05). Separate regressions for respondents with and without a partner showed the following. Without a partner, overall prediction was markedly improved (R2 = .24 as opposed to R2 = .09). The number of core kin showed a negative relationship with social loneliness (ˇ = −.21; p < .05) as did the number of significant kin (ˇ = −.28; p < .05). Friends, on the other hand, no longer played a significant role. With a partner, overall predictive power was reduced (R2 = .05), and the only predictor to remain significant was the number of significant friends (ˇ = −.23; p < .05). When we predicted social loneliness from emotional closeness, relationship status interacted with closeness to significant friends: ˇ = .39; p < .05. Again, separate regressions were then carried out showing that, as before, overall prediction was improved in the group without a partner (R2 = .21) whereas in the group with a partner it was quite low (R2 = .03). Without a partner, the closeness to significant friends was negatively related to loneliness (ˇ = −.46; p < .05). No other significant effects emerged. Taken together, these findings suggest that having a romantic partner weakens the association between network characteristics and social loneliness. In contrast to social loneliness, no moderation effects were found for the prediction of emotional loneliness from number of ties, and a visual inspection confirmed that effects remained essentially unchanged in both subgroups. The same held regarding the prediction of emotional loneliness from closeness. 7. Discussion 7.1. Summary of findings This study was designed to gain a better understanding of the commonalities and differences between core and significant ties

Table 5 Summary of regression analyses for the prediction of social and emotional loneliness from the intimacy of tie sets. Predictor

Criterion Social loneliness

Closeness to kin, core ties Closeness to kin, significant ties Closeness to friends, core ties Closeness to friends, significant ties

Emotional loneliness

b

SE(b)

ˇ

b

SE(b)

ˇ

−.014 .046 −.128 −.092

.119 .073 .107 .103

−.01 .07 −.14 −.10

−.202 .107 −.346 .279

.141 .086 .126 .122

−.14 .14 −.29** .26*

Note. R2 for social loneliness = .04 (ns). R2 for emotional loneliness = .12 (p < .01). * p < .05. ** p < .01.

212

J.F. Binder et al. / Social Networks 34 (2012) 206–214

within personal communities. We found the number of ties in line with previous findings in this area (Dunbar and Spoors, 1995; Milardo, 1992; Wellman and Wortley, 1990). As stated previously, however, we are cautious to read too much into mean scores due to the nature of our convenience sample. Our main study aim was to investigate how these sets of ties would compare: in terms of constraints on their number, the social exchanges they are part of, and their relationship with social and emotional loneliness. In all these areas, differences between core and significant ties emerged. Although our distinction between ties simply rests on a particular use of name generators, we still find it remarkable that changes in emotional closeness are reflected so clearly in several other variables. Regarding constraints on network size, the number of core ties was negatively related to the emotional closeness of the alters in this set. Thus, even at the level of the very closest social relationships, there seems to be a trade-off between quantity and quality – egos with more core ties were, on average, less close to these people than egos with less core ties. Fewer and closer core ties could imply better support as previous research suggests. Core ties have been found to be emotionally intense and at the same time to involve frequent interaction (Roberts and Dunbar, 2010) and the exchange of emotional, financial and material support (Plickert et al., 2007). Wellman and Frank (2001) found that those with fewer intimate ties were more likely to receive support from each alter, as compared to those with more intimate ties. In comparison, for significant ties, there was no direct relationship between the number of ties and emotional closeness. Instead, we found a relationship with the ratio of set sizes – the more core ties in comparison to significant ties, the lower the emotional closeness to significant alters. Thus it seems that having a comparatively large number of core ties also has implications for emotional closeness among significant ties: egos who invest heavily in their core ties may have less time and energy to invest in their significant ties. This issue clearly needs further empirical consideration before more definite conclusions can be drawn. Regarding the exchange signatures of friends in both sets, we found that differences between core and significant friends were most pronounced for behaviours and provisions that are usually associated with stronger and deeper friendships – openness, supportiveness, emotional support, reliable alliance and guidance/advice. Both friends fall clearly into the small group of closest network ties, the personal community, varying on the emotional closeness scale by an average of around 1 point on a 10 point scale. Our findings indicate that even such small changes in closeness are accompanied by changes in the functions that non-kin ties serve, in terms of the level of support and advice they provide. Further, our findings demonstrate the validity of the scale of emotional closeness, which is a simple measure of tie strength used in a number of studies (Hill and Dunbar, 2003; Roberts et al., 2009), and which by now has now been shown to correlate both with communication frequency (Hill and Dunbar, 2003) and maintenance behaviours between individuals. Our findings also show that different facets of loneliness are differently related to core and significant ties: in their associations with set sizes, with levels of intimacy, and how these associations are affected by relationship status. Social loneliness was unrelated to the number of core ties, but decreased with increasing numbers of significant ties. This suggests that having more significant ties provides a general feeling of social connectedness – you have a relatively large circle of friends and family you can socialise with and call on for support if needs be. Having a romantic partner seemed to act as a buffer against this mechanism such that the association between number of ties and level of loneliness became markedly weaker. It needs to be noted, though, that relationship status is confounded with age and student status in our sample. Staff, who

show a higher percentage of relationships, may also have developed different strategies to keep social loneliness low. For emotional loneliness, having more core ties unexpectedly implied higher levels of loneliness whereas having more significant ties was indicative of lower levels. One way of resolving this puzzle is to include the role of emotional closeness in an explanation. As stated before, more core ties imply lower levels of closeness. At the same time, lower closeness levels among core ties imply more emotional loneliness. A bigger core tie set, then, on average contains ties that are less suited to counter emotional loneliness because they are further removed from the ideal remedy, namely a romantic partner or confidant. The pattern of findings is reversed for significant ties. It may be that having more significant ties containing less close friends – indicative of less emotional loneliness – simply means that alters in the core set are chosen with care and contain few, very close contacts which in turn reduces emotional loneliness. At present, these processes must remain speculation. 7.2. Limitations and caveats There are several limitations to this research as well as some open questions which need to be noted. The first concerns the nature of our sample. A convenience sample such as the one used here does not readily allow for a generalization of effects. For this, a large, representative random sample would be preferable. Whether differences between core and significant ties that we found will hold up to a more general and rigorous test remains to be seen. A similar caveat can be raised regarding the use of online surveys. While our analyses make strong biases due to self-selection or participant motivation seem unlikely within our sample, clearly more research using different methods to validate our approach and to replicate the findings is desirable (for recent developments in network elicitation, see Hogan et al., 2007). Another question is to what degree the differences between core and significant ties are qualitative or quantitative in nature. Some theories have postulated fundamental functional, and therefore qualitative, differences (Dunbar, 1998). However, statistical differences between two groups are not clear indicators of deeprunning qualitative differences and might just as well be due to an underlying continuum. One possible solution to this problem is to look for dissociations between core and significant ties using multiple variables. In other words, we would need to identify variables that should be related only to core ties and variables only related to significant ties. The more such variables are used successfully for distinction, the more we can be confident about qualitative differences. This logic is implicit in our investigation of social and emotional loneliness, but at this point asserting qualitative differences would be a foregone conclusion. Another area of future research concerns longitudinal studies on how the relationships among core and significant ties change over time. As this was a cross-sectional study, causal conclusions about the associations between various factors have to be tentative. This is most apparent when the relationship between loneliness and network layers is considered. While we have predominantly focused on possible effects of set sizes on loneliness, we cannot, and do not want to, rule out effects of initial loneliness on changes in the network sets. The same holds for the question of constraining factors: does intimacy constrain network size or vice versa? Longitudinal studies allow for determining how states at time 1 affect states at time 2, thus yielding more confidence in causal statements. 7.3. Outlook In this study we demonstrated that not all ‘strong ties’ are the same, and that in fact there are important differences within strong ties. Although several studies on social networks have

J.F. Binder et al. / Social Networks 34 (2012) 206–214

differentiated within the strong ties of personal communities (e.g., Boase et al., 2006; Wellman and Wortley, 1990), there has been a tendency in other research areas to regard these communities as uniform and assess it simply in terms of numbers. For example, the negative relationship between strength of social support and morbidity and mortality has been widely replicated (Uchino, 2006). However, level of social support is often measured as a single, uni-dimensional concept, without any consideration of either the strength of individual ties, or the distinction between core and significant ties (Hammer, 1983; Lin et al., 1999). Including simple, quick-to-complete measures of both tie sets, and the emotional closeness of ties therein, may be a way of determining more precisely the mechanisms that are behind the association between social support and health outcomes. Our findings are also relevant to recent work in evolutionary anthropology. Functional differences between ties and constraints on network size are central tenets of the Social Brain Hypothesis (SBH) developed by Dunbar (1998). SBH links network size to brain evolution and posits that primates have unusually large brains for their body size due to the complexity of primate social relationships. SBH further predicts constraints on the number of social relationships humans can maintain at any given level of intimacy (Hill and Dunbar, 2003; Dunbar, 2008), just as for other primates (see Roberts (in press) for detailed discussions of constraints on personal networks). In our study, we have mapped our definitions of core and significant ties onto the two most intimate sets of network ties recognised by SBH (Dunbar and Spoors, 1995). The fact that we found differences between ties in the way their number is possibly constrained, in the way they are maintained and marked by social exchanges, and in the way they are differentially related to loneliness, can also further our understanding of the nature and maintenance of relationships from an evolutionary perspective. References Aiken, L.S., West, S.G., 1991. Multiple Regression: Testing and Interpreting Interactions. Sage, Thousand Oaks, CA. Baron, R.M., Kenny, D.A., 1986. The moderator–mediator distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology 51, 1173–1182. Bernard, H.R., Killworth, P.D., 1973. On the social structure of an ocean-going research vessel and other important things. Social Science Research 2, 145–184. Boase, J., Horrigan, J.B., Wellman, B., Rainie, L., 2006. The Strength of Internet Ties. Pew Internet & American Life Project, Washington, DC. Bryk, A.S., Raudenbush, S.W., 1992. Hierarchical Linear Models: Applications and Data Analysis Methods. Sage, London. Buys, C.J., Larson, K.L., 1979. Human sympathy groups. Psychological Reports 45, 547–553. Carbery, J., Buhrmester, D., 1998. Friendship and need fulfillment during three phases of young adulthood. Journal of Social and Personal Relationships 15, 393–409. Degenne, A., Lebeaux, M.-O., 2005. The dynamics of personal networks at the time of entry into adult life. Social Networks 27, 337–358. Dunbar, R.I.M., 1998. The social brain hypothesis. Evolutionary Anthropology 6, 178–190. Dunbar, R.I.M., 2008. Cognitive constraints on the structure and dynamics of social networks. Group Dynamics: Theory, Research, and Practice 12, 7–16. Dunbar, R.I.M., Spoors, M., 1995. Social networks, support cliques, and kinship. Human Nature 6, 273–290. Fischer, C.S., 1982. To Dwell among Friends: Personal Networks in the Town and City. University of California Press, Berkeley, CA. Fischer, C.S., 2009. The 2004 GSS finding of shrunken social networks: an artifact? American Sociological Review 74, 657–669. Granovetter, M.S., 1973. The strength of weak ties. American Journal of Sociology 78, 1360–1380. Granovetter, M.S., 1983. The strength of weak ties: a network theory revisited. Sociological Theory 1, 201–233. Green, L.R., Richardson, D.S., Lago, T., Schatten-Jones, E.C., 2001. Network correlates of social and emotional loneliness in young and older adults. Personality and Social Psychology Bulletin 27, 281–288. Grossetti, M., 2005. Where do social relations come from? A study of personal networks in the Toulouse area of France. Social Networks 27, 289–300. Grossetti, M., 2007. Are French networks different? Social Networks 29, 391–404. Hammer, M., 1983. Core and extended social networks in relation to health and illness. Social Science and Medicine 17, 405–411.

213

Hays, R.B., 1985. A longitudinal study of friendship development. Journal of Personality and Social Psychology 39, 460–470. Hays, R.B., 1989. The day-to-day functioning of close versus casual friendships. Journal of Social and Personal Relationships 6, 21–37. Hill, R.A., Dunbar, R.I.M., 2003. Social network size in humans. Human Nature 14, 53–72. Hogan, B., Carrasco, J., Wellman, B., 2007. Visualizing personal networks: working with participant-aided sociograms. Field Methods 19, 116–144. House, J.S., Landis, K.R., Umberson, D., 1988. Social relationships and health. Science 241, 540–545. Lin, N., Ye, X.L., Ensel, W.M., 1999. Social support and depressed mood: a structural analysis. Journal of Health and Social Behavior 40, 344–359. McCannell, K., 1988. Social networks and the transition to motherhood. In: Milardo, R. (Ed.), Families and Social Networks. Sage, Newbury Park, CA. McCarty, C., Killworth, P.D., Bernard, H.R., Johnsen, E.C., Shelley, G.A., 2001. Comparing two methods for estimating network size. Human Organization 60, 28–39. McPherson, M., Smith-Lovin, L., Brashears, M.E., 2006. Social isolation in America: changes in core discussion networks over two decades. American Sociological Review 71, 353–375. Milardo, R.M., 1992. Comparative methods for delineating social networks. Journal of Social and Personal Relationships 9, 447–461. Milardo, R.M., Johnson, M.P., Huston, T.L., 1993. Developing close relationships: changing patterns of interaction between pair members and social networks. Journal of Personality and Social Psychology 44, 964–976. Naus, M.J., Philipp, L.M., Samsi, M., 2009. From paper to pixels: a comparison of paper and computer formats in psychological assessment. Computers in Human Behavior 25, 1–7. Nettle, D., 2007. Empathizing and systemizing: what are they, and what do they contribute to our understanding of psychological sex differences? British Journal of Psychology 98, 237–255. Oswald, D.L., Clark, E.M., 2006. How do friendship maintenance behaviors and problem-solving styles function at the individual and dyadic levels? Personal Relationships 13, 333–348. Oswald, D.L., Clark, E.M., Kelly, C.M., 2004. Friendship maintenance: an analysis of individual and dyad behaviors. Journal of Social and Clinical Psychology 23, 413–441. Pahl, R., Spencer, L.,2003. Personal communities: not simply families of ‘fate’ or ‘choice’. In: Working Papers of the Institute for Social and Economic Research, Paper 2003-4. University of Essex, Colchester, UK. Plickert, G., Côté, R.R., Wellman, B., 2007. It’s not who you know, it’s how you know them: who exchanges what with whom? Social Networks 29, 405–429. Pool, I.D., Kochen, M., 1978. Contacts and influence. Social Networks 1, 5–51. Roberts, S.G.B. Constraints on social networks. Proceedings of the British Academy, in press. Roberts, S.G.B., Dunbar, R.I.M., 2010. Communication in social networks: effects of kinship, network size, and emotional closeness. Personal Relationships 18, 439–452. Roberts, S.G.B., Dunbar, R.I.M., Pollet, T.V., Kuppens, T., 2009. Exploring variation in active network size: constraints and ego characteristics. Social Networks 31, 138–146. Rose, S., Serafica, F., 1986. Keeping and ending casual, close and best friendships. Journal of Social and Personal Relationships 3, 267–277. Rusbult, C.E., 1980. Satisfaction and commitment in friendships. Representative Research in Social Psychology 11, 96–105. Russell, D., Peplau, L.A., Cutrona, C.E., 1980. The revised UCLA loneliness scale: concurrent and discriminant validity evidence. Journal of Personality and Social Psychology 39, 472–480. Russell, D., Cutrona, C.E., Rose, J., Yurko, K., 1984. Social and emotional loneliness: an examination of Weiss’ typology of loneliness. Journal of Personality and Social Psychology 46, 1313–1321. Sarason, B.R., Shearin, E.N., Pierce, G.R., Sarason, I.G., 1987. Interrelation of social support measures: theoretical and practical implications. Journal of Personality and Social Psychology 52, 813–832. Spencer, L., Pahl, R., 2006. Rethinking Friendship: Hidden Solidarities Today. Princeton University Press, Princeton. Stiller, J., Dunbar, R.I.M., 2007. Perspective-taking and memory capacity predict social network size. Social Networks 27, 93–104. Stokes, J.P., 1985. The relation of social network and individual difference variables to loneliness. Journal of Personality and Social Psychology 48, 981–990. Suitor, J., Keeton, S., 1997. Once a friend, always a friend? Effects of homophily on women’s support networks across a decade. Social Networks 19, 51–62. Uchino, B.N., 2006. Social support and health: a review of physiological processes potentially underlying links to disease outcomes. Journal of Behavioral Medicine 29, 377–387. Van Duijn, M.A.J., Van Busschbach, J.T., Snijders, T.A.B., 1999. Multilevel analysis of personal networks as dependent variables. Social Networks 21, 187–209. Vaux, A., 1988. Social and emotional loneliness: the role of social and personality characteristics. Personality and Social Psychology Bulletin 14, 722–734. Vincenzi, H., Grabosky, F., 1989. Measuring the emotional/social aspects of loneliness and isolation. In: Hojat, M., Crandall, R. (Eds.), Loneliness: Theory, Research, and Applications. Sage, Newbury Park, CA. Weiss, R.S., 1973. Loneliness: The Experience of Emotional and Social Isolation. MIT Press, Cambridge, MA. Weiss, R.S., 1974. The provisions of social relationships. In: Rubin, Z. (Ed.), Doing unto Others. Prentice-Hall, Englewood Cliffs, NJ.

214

J.F. Binder et al. / Social Networks 34 (2012) 206–214

Wellman, B., 1979. The community question: the intimate networks of East Yorkers. American Journal of Sociology 84, 1201–1231. Wellman, B., 1999. The network community. In: Wellman, B. (Ed.), Networks in the Global Village. Boulder, CO, Westview Press, pp. 1–47. Wellman, B., Carrington, P.J., Hall, A., 1988. Networks as personal communities. In: Wellman, B., Berkowitz, S. (Eds.), Social Structures: A Network Approach. Cambridge, UK, CUP. Wellman, B., Frank, K., 2001. Network capital in a multilevel world: getting support from personal communities. In: Lin, N., Cook, K., Burt, R. (Eds.), Social Capital: Theory and Research. Piscataway, NJ, Aldine Transaction.

Wellman, B., Wong, R.Y., Tindall, D., Nazer, N., 1997. A decade of network change: turnover, persistence and stability in personal communities. Social Networks 19, 27–50. Wellman, B., Wortley, S., 1990. Different strokes from different folks. Community ties and social support. American Journal of Sociology 96, 558–588. Zheng, T., Salganik, M.J., Gelman, A., 2006. How many people do you know in prison? Using overdispersion in count data to estimate social structure in networks. Journal of the American Statistical Association 101, 409–423.