Participants

Data were collected from healthy adults from the UK population attending three science festivals and a museum. Following [23,25], we focus on White participants without a history of mental illness. Participants were required to identify their ethnicity from a standard list of categories: ‘White British’, ‘White (other)’, ‘Indian’, ‘Pakistani’, ‘White Irish’, ‘Mixed’, ‘Black Caribbean’, ‘Black African’, ‘Bangladeshi’, ‘Chinese’, ‘Other Asian’, ‘Black (other)’, and ‘Other’. Participants in the ‘White British’, ‘White (other)’ and ‘White Irish’ were amalgamated for the purpose of this study. Not all participants completed all the survey questions, so the final sample size comprised 398 females (M = 38 years, range = 18–74 years) and 324 males (M = 43 years, range = 18–75 years, N = 315). The associated data are available in S1 Table.

The study was approved by the University of Oxford Central University Research Ethics Committee (CUREC Ref: MS-IDREC-C2-2015-005), and participants completed consent forms before taking part.

Previous papers from the same study have been published on the associations between the social measures and specific gene variants in both the White sample with no history of mental illness used in this current paper [23], and in the remaining non-white and sub-clinical samples [24]. In addition, analysis of data from this study exploring the relationships between digit ratio, genetic variants and social measures has also been published [25]. Whereas the previous papers explored genetic associations, in contrast the current paper investigates associations between the different social measures. These interrelationships between the social measures have not been previously reported and here we present new analyses of the social measures data only (that is, excluding the genetics data), in order to explore the relations between these dimensions of sociality independently of examining the underlying biology.

Procedure

Participants in the study completed a set of questionnaires on mobile devices. Descriptive statistics for all variables are given in S2 Table.

To study disposition, we measured empathy, attachment style and impulsivity. To measure empathy we used the Reading the Mind in the Eyes task (RMET), which was validated on a UK sample consists of 36 photos of the eye region of faces expressing different emotions that are presented to the participant one at a time [28]. The participant is asked to identify the correct emotion being expressed from four options, such as ‘ashamed’, ‘nervous’, suspicious’, and ‘indecisive’. Higher scores indicate greater accuracy of identifying emotions.

In addition, we used the short-form Empathy Quotient (EQ), which was validated in a UK sample [29]. This consists of 22 items, 6 of which are reversed scored, such as ‘I find it difficult to judge if something is rude or polite’ (reversed scored) and ‘I can easily tell if someone wants to enter a conversation’. Participants rate the extent to which they agree or disagree with each statement on a 4-point scale anchored as ‘strongly disagree’, slightly disagree’, slightly agree’ and ‘strongly agree’. Means scores are taken to account for missing responses and higher scores indicate greater empathy. Whereas the RMET scale had a relatively low reliability (Cronbach’s α = 0.585), EQ had a high reliability (Cronbach’s α = 0.902).

To measure the Anxious and Avoidant dimensions of attachment style we modified the short-form Experiences of Close Relationships scale (ECR) to relate to ‘close friendships’ rather than romantic relationships [30]. This measure comprises 12 items (4 reverse scored), 6 for each of the two dimensions. For each item participants provide a score on a 7-point scale from ‘strongly disagree’ to ‘strongly agree’ for statements such as ‘I am nervous when close friends get too close to me’ (avoidant) and ‘My desire to be very close sometimes scares people away’ (anxious). Higher scores indicate higher levels of Anxious or Avoidance attachment. We analysed these as two continuous dimensions of attachment rather than categorising participants into different attachment styles, since categorisation means a loss of power and precision (see [31] for the original ECR measure and discussion). In addition, we used the short-form Barratt Impulsiveness Scale [32] as a self-report measure of Impulsivity, which comprises 15 items (6 reverse-scored) rated on a 4-point scale anchored at ‘Rarely/Never’, ‘Occasionally’, ‘Often’ and ‘Almost always/always’, such as ‘I act on the spur of the moment’ and ‘I save regularly’ (reversed scored). Higher scores indicate greater impulsivity. Both attachment subscales had relatively high reliability (anxious Cronbach’s α = 0.735, avoidant Cronbach’s α = 0.777), as did the Impulsivity scale (Cronbach’s α = 0.795).

To measure attitudes and behavior in relation to sexual relationships, we measured sociosexual orientation, using the revised Sociosexual Orientation Inventory (SOI) [33]. This comprises three sections, with a total of 9 items (three in each section), with higher combined scores indicating that an individual is more promiscuous and willing to participate in short-term sexual relationships. The first section asks how many sexual partners an individual has had (i) only once, (ii) without being in a committed relationship, and (iii) in the past 12 months. Participants are required to pick one of 9 categories demarcated as single values for 0 through 4, then as combined categories ‘5–6’, ‘7–9’, ‘10–19’ and ‘20 or more’. The second section comprises three statements, such as ‘Sex without love is OK’, for which the participant is asked to rate the extent of their agreement on a 9-point scale anchored at ‘strongly disagree’, ‘neutral’ and ‘strongly agree’. One of these statements, ‘I do not want to have sex with a person until I am sure that we will have a long-term, serious relationship’, is reversed-scored. The final section asks participants to rate three questions regarding the frequency of fantasies and sexual arousal on a 9-point scale ranging from ‘never’ to ‘at least once a day’.

If the participant was in a sexual relationship, we also used the Relationship Assessment Scale (RAS) [34,35] to assess relationship satisfaction, with higher scores indicating greater satisfaction. The original measure is made up of 7 questions, 2 of which are reverse scored, which participants are asked to answer on a 3-point scale with anchors that vary between items. For example, ‘how well does your partner meet your needs?’ is answered from a choice of ‘poorly’, ‘average’ or ‘extremely well’. However, to maintain consistency with the form of the other measures used in the survey, we rephrased each question as a statement, such as ‘my partner meets all my needs’ that was answered on a 5-point scale ranging from ‘strongly’ disagree’ to ‘strongly agree’. Both the SOI and RAS scales had high reliability (Cronbach’s α = 0.850 and 0.910 respectively).

To explore social relationships in the wider social network, we measured participants’ Support Social Network Size by asking them to record their relationship to the individuals they would turn to for help and support during times of difficulty and distress, and totalling the number of individuals listed ([7,36] following [37,38]). In addition, we measured how integrated or close participants felt to their local community using a modified version of the Inclusion of Other in Self scale, using the label ‘community’ rather than ‘other’ [39]. This measure consists of a sequence of seven diagrams, each of which comprises two circles, which become increasingly overlapped as the scale moves from 1 to 7. The labels of these two circles were modified to ‘self’ versus ‘community’ (rather than ‘other’).

Analyses

Descriptive statistics are given in S2 Table separately for females and males. We conducted principle component analyses (PCA) in IBM SPSS Statistics for mac, Version 24 (IBM Corp., Armonk, NY, USA). We excluded RAS scores to maximize sample size: 135 females and 95 males did not report being in a romantic relationship at the time of the survey and so did not complete the RAS. We also excluded RMET because initial examination of the correlation matrix indicated very low correlations between this variable and the others (none were above r = 0.25). Data for the remaining seven variables yielded sample sizes of 398 females and 324 males. A sample of 300 is generally considered acceptable for PCA [40]. KMO scores were >0.6 and Bartlett’s tests were significant (p<0.0001) for both males and females. Oblique (direct oblimin) rotation of the factors was used. Initially factors with eigenvalues >1 were extracted but examination of scree plots led to forced extraction of three factors (see Results).

To examine the structural relationships between the variables, we conducted path analyses. We ran multiple linear regressions predicting each variable in turn from all the remaining variables, and used the standardized ß values to identify significant relationships between pairs of variables. It should be noted that all these relationships were reciprocal. To explore potential sex differences in how different facets of sociality interlink, we conducted these analyses separately for male and female participants (see S3 & S4 Tables for the partial relationships between the variables from multiple linear regressions in females and males respectively).

PCA factors

Females.

Based on eigenvalues greater than 1, two factors were extracted for females, which together explained 47% of the variance. The first factor included a negative loading of avoidant attachment (-0.760), and positive loadings of EQ (0.718), IOS (0.639) and network size (0.492). The second factor showed positive loadings of impulsivity (0.733), anxious attachment (0.613) and SOI (0.569). However, a scree plot suggested that three factors could be extracted and together these explained 61% of the variance (Table 1). For females the first factor included EQ, avoidant attachment and IOS, whereas the second factor included impulsivity and SOI. The third factor had positive loadings for anxious attachment and network size, but both these variables also loaded, albeit to a lesser extent, onto the first factor. Given that a loading below 0.4 is not considered meaningful [40], the loading of anxious attachment in particular on Factor 1 should probably be ignored.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. PCA factor weightings for females and males.

The strongest loadings for each variable are highlighted in grey.

https://doi.org/10.1371/journal.pone.0216210.t001

Path analysis

Fig 1 shows the path analyses for females and males. In males, there is a particularly clear distinction between sociosexual orientation, impulsivity and anxious attachment on the one hand and, on the other, engagement in the wider social network (support network size and feelings of inclusion in the local community), empathy and avoidant attachment. Although these two clusters are also evident for females, the path analysis suggests that these clusters are more integrated in females. The links between the two clusters in females are via avoidant attachment and empathy. Moreover, in females, but not males, network size shows a significant positive relationship with anxious attachment. In contrast, in males, support network size is significantly positively associated with IOS, but this association was not found for females.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Path analyses between dispositional and social variables in females (top) and males (bottom). The line weights of the arrows represent p-values of p<0.05, p<0.01 and p<0.001: the thicker the line the lower the p-value. The two PCA clusters are shown in different shades of grey. EQ: Empathy Quotient; IOS: Inclusion of Other in Self scale; SOI: Sociosexual Orientation Inventory; Avoidant/Anxious: scores on dimensions of Attachment measured by the Experience in Close Relationships scale.

https://doi.org/10.1371/journal.pone.0216210.g001

Strengths and limitations

This study builds on past work by looking at different facets of the social world and associated individual dispositional characteristics in parallel, in order to look at the interplay between them, rather than narrowly focusing on one or two at time. The findings have important implications for understanding individual differences in dispositional traits and how they translate into social behaviour, as well as practical applications in terms of interventions for loneliness. Feeling disconnected can happen both with respect to romantic relationships and to the wider social network, and our results suggest that interventions need to specifically target one or the other because they are relatively distinct domains. However, more generic interventions might be more successful in women, because changes may be more likely to ripple between these different types of relationship if common underlying dispositional traits, such as the ability to empathise, are targeted.

Although this study brings greater insight into how different facets of human sociality and individual dispositions interact, future work should aim to incorporate more objective measures rather than self-report, as well as a greater number of different dimensions, such as relationship quality and other types of relationships such as with kin and colleagues, in order to more fully decipher the complex interplay of human social worlds. Moreover, the sample was limited to UK-based participants who reported being ethnically White and no history of mental illness, so these findings cannot be generalised to other demographic samples. For instance, there may be important cultural differences in how males and females manage their social worlds. In addition, there may have been confounding variables that we did not take into account when asking people about their current social worlds, such as significant life events.