Evolutionary roots

Examining the evolutionary roots of the sociotype is necessary for an in-depth comprehension of the new construct. Indeed the crucial novelties of our evolutionary/historical past have revolved around communication matters —e.g. origins of language, emotional expression, group behavior, morals and ethical rules, counting and writing systems, economic and political organizations, knowledge systems, modern media, and so on [20–22]. So fluid and culturally diverse are all the emerging structures of human sociality that, apparently, they defy any precise classification or quantitative specification. However, the presence of a series of significant regularities in the size and structures of social groups, notwithstanding their remarkable variability, suggests the plausibility of a “deep structure” of social bonding for the human species [23–24].

There seems to be an average of social networking, with very ample upper and lower limits, concerning the number and types of bonding relationships that an individual is able to maintain meaningfully. The finding of networking regularities such as the famous “Dunbar’s number” (around 150–200 individual acquaintances) would make evolutionary and anthropological sense [1, 3, 24–27]. These relational findings, integrated within the “social brain hypothesis”, which was originally known as the “Machiavellian intelligence hypothesis” [28], would project an extended clutch on the roots of human sociality, on the origins of language, and on many other traits of social and cultural life. Specifically, the social brain hypothesis has posited that, in primate societies, natural selection has favored larger brains and more complex cognitive capabilities as a means to cope with the challenges of social life [27, 29–32]. Thereafter, due to the overall cortical conformation and brain capacity of our species, vastly enlarged regarding other Anthropoidea, we are able to maintain a really high number of social bonds, meaningfully shared and sustained with the abundant members of our oversized groups. However, like many other brain/mind phenomena, exactly how ‘social bonds’ are made, maintained, differentiated, eroded, restored, finalized, etc. is not sufficiently understood yet—in all probability it is both a universal phenomenon related to crossing some threshold of neuronal complexity and a species-specific phenomenon related to the singularities of the different brains [33].

In the evolution of human societies, language appears as the essential tool for bond-making, although not the only one [29, 34–38]. Human languaging, more often than denotative, becomes cohesive, consensual, identity maker—and above all, a source of mental stimulation. The regular practice of ‘interesting’ conversations induces in our social brain the production of neuropeptides and neurohormones that relieve stress and boost immune system and nervous system’s function [25, 39, 40]. While talking, the specific contents exchanged are often not so important. Rather than the exchange of functional information, it is trivial conversation, gossiping about social acquaintances, what represents the human closest equivalent to grooming exchanges and bond-restoration practices in primate groups [29,36]. Indeed talking may be considered as a new form of grooming in human societies, and comparatively–in energy and physical grounds–it is virtual.

The mental necessity of conversation

Thereafter, the repercussions in human daily life of this new form of virtual grooming can hardly be overstated: talking becomes one of the preferred and most affordable types of mental stimulation. Counting with an appropriate network of people to talk with becomes a necessity for the well-being and mental health of individuals. Having access to, and participating in, amusing conversations becomes an essential ingredient for our social, psychological, and physiological life. The way the different emotions related to social interactions impinge upon language itself and are rearranged within this new channel of expression represents another factor of utmost psychological importance [41–43]. Also, distinguishing several classes of bonds (related to their strength and to their positive or negative valence) turns out to be important in the conversational rewards obtained, as well as in the distinctiveness and degree of surprise of the circulated information, particularly regarding the mentioned preference for the whereabouts of social acquaintances and the importance of the speaker’s own image. The individual fitness within the social group is always at the stake, with relevant gender differences in communication goals and strategies [25, 36, 44, 45].

Within our cognitive dynamics, bonds and conversation are paramount. It can be said both that social bonds claim for their actualization and that linguistic practices claim for their regular realization as well. Both behavioral propensities are reinforcing each other. In fact, the daily conversation/communication budget of each individual has to be apportioned among the different classes of bonds of his/her sociotype so that the talking exercise becomes sufficiently rewarding–provides enough grooming–and that new fitness opportunities may be explored, taking into account the existing diversity of possible encounters and the available communication channels.

The structural and dynamic aspects we are distinguishing in the sociotype (classes of bonds and talking-time budgets) are but two different facets of a unitary social adaptation phenomenon performed by each individual along his/her life. The adaptive sociotype may be conceived as closely following the phenotype’s trajectory along the arch of life—beginnings, development, maturity, and senescence. Analyzing the respective bond structures and conversation-time distributions in these differentiated stages could lead to very interesting comparatives: not only by following the developmental age, but also by taking into account personality, gender, status, professions, cultures, etc.

The sociotype and the growing problem of loneliness

A number of recent studies on social networks, technologically oriented, have tracked vast amounts of interpersonal exchanges [4, 17, 19, 46], but the metrics of the relational structures necessary for personal well-being and mental health have hardly been addressed yet. Hopefully, the progressive delineation of a sociotype construct, susceptible of both theoretical and empirical demarcation, might contribute to a better understanding of the structures and dynamics of human sociality, and might provide some practical help when sociality itself is in crisis, as seems to be happening with the current “epidemics of loneliness” affecting large population tracts and particularly the elderly [47–53].

At a social/economic scale, the diminished relationships and bonding structures of “social capital” would penalize the activities of daily life and would decrease the individual’s well-being [17, 48, 54]. The evidence in fast-developing countries is that economic growth and technological development spurred by the ‘information revolution’ have gone hand-in-hand with an increase in behavioral disorders, family disintegration, social exclusion, and lower social trust [55, 56]. In the 1985 US Census, the average number of confidants was three; in the 2004 Census the average was 2, but the most common figure was zero confidants for almost 25% [57]. The phenomenon is similar in most Western countries, where the ubiquitous presence of Media and of information and communication technologies (ICTs) has dramatically altered life styles. In recent years, there has been a significant transfer of social life and collective entertainment activities to individualized activities such as computer games, Internet, TV watching, and the new online social networks. However, it is unclear the effect that such ICTs pervasiveness and overuse are having in our social relationships and quality of life. In what extent could computers, cell phones, and TVs replace our need of face-to-face relationships? [45, 49, 58–61]. The balance between positive and negative factors is not settled yet. It is at least significant that depression and suicide rates have increased dramatically in the last three decades; and that mental disorders nowadays represent a global disease burden that surpasses cardiovascular diseases and cancer [62–64].

From a psychological and biomedical perspective, in spite of the pervasive loneliness and lack of meaningful relationships in contemporary societies, there is a dearth of adequate indicators gauging the conversational activities of the individual. Actually, none of the existing questionnaires on related topics (e.g. UCLA loneliness scale, MSPSS, SNI, Duke, SELSA, MOS, SSB, Zimet, de Jong, etc.) seems to be centered in the basic, face-to-face relational phenomena focused by the sociotype. Precisely, this is the kind of information that a few of the most recent enquiries are beginning to ask for [65, 66].

The present study

Subsequently, the empirical part of the present study has tried to develop a new psychometric indicator related to the social interactions ‘adaptively’ demanded by the social brain of each individual. By means of a specific survey, different kinds of psychosocial data were gathered around the structure of individuals’ social bonds and on their dynamic update via conversation. According to criteria frequently used in other studies [most important sources have been: 4, 29, 47, 48, 57, 67, 68], we distinguished between three main relational scenarios: the own residence, public spaces, and the workplace (or study). We also distinguished between four types of relationships: family, close friends, acquaintances, and work/study colleagues [in contraposition to the three levels in 67]. This division has oriented the basic structure of the new psychometric indicator. Essentially, the present study–and its related fieldwork–has aimed to develop an accurate and applicable operationalization of a sociotype-inspired psychometric indicator, evaluating its structure, internal consistence, and convergent validity.

We have also aimed to assess the sociotype’s explanatory power to account for loneliness and psychological distress, throughout the relationships with UCLA and GHQ-12 measures, both considered as individual’s well-being metrics likely influenced by the sociotype outcomes; and additionally we have explored possible differences due to age, personality, and gender. Overall, the following working hypotheses have been considered:

1. The sociotype can be measured with appropriate psychometric features as a general unique dimension, and also by means of specific sub-domains; the statistical analysis has to show both a confirmative factor structure and a robust internal consistency for the total scale and for the sub-domains.
2. The sociotype scale should converge with the other constructs related to psychological states of loneliness and distress (UCLA, GHQ-12), demonstrating its validity and relevance by means of a high, significant correlation.
3. The sociotype scale should correlate with the different dimensions of personality, positively with extraversion and negatively with neuroticism and psychoticism (the lie dimension could be more complex, depending on the different domains).
4. The sociotype scale would change across the successive stages of life, with different subscales increasing/decreasing their levels throughout the life course, and expecting an overall decline for the elderly.
5. According to gender, the sociotype sub-scales could show significant differences: traditionally, a stronger social network for males in terms of work/study colleagues would be expected, while for females the stronger networking should appear in terms of family.

Given that the general human need to connect–to which the sociotype refers–seems to be a universal adaptive trait [29], an open exploratory question would concern the relative constancy of the overall sociotype measure. It could be hypothesized as being caught in a similar range of variability for most individuals, although strongly biased in its constitutive sub-domains by gender and age, and by different personal, social and cultural factors and influences—see for instance the “epidemics of loneliness” that affects more intensely the elderly in modern societies. The sociotype means the endless adaptation to the different possibilities of social intercourse around the individual.

Shedding light on all these structural differences and social influences would permit a better characterization of the structure and dynamics of human sociality, guiding the implementation of public programs aimed at strengthening social networking.

Study design

The structure of the study was based on the correlational method with an analytical cross-sectional design. The whole measurements were obtained by means of the self-assessment technique, using a set of questionnaires via Internet, complemented with face-to-face interviews as well.

Participants

Most of the participants accessed the survey via Internet, but around 15% were face-to-face interviews (for the elderly), intentionally trying to cover for the different gender, age, and status characteristics. In all cases, the inclusion criteria were age >18 years, being able to read and write Spanish, and not suffering from severe physical or mental disorders. The final number of participants (n = 1,075) exceeded the validity evaluation criterion [69], resulting in a sample that was psychometrically adequate for the study. The main characteristics of the sample are presented in Section 3.

Procedure and ethics statement

In order to hypothesize the structure of the sociotype model, we started with a qualitative study looking for the main characteristics of social networking and conversational habits in different ages, by means of semi-structured interviews (involving narratives of personal experiences, moods, appreciation of new communication technologies, etc.). See (S2 and S3 Files). This qualitative study involved 45 interviews (conveniently diversified by age, gender and class) [70]. A total of 45 participants were recruited (26 female and 19 male), with a mean age of 60.4 (SD = 22.26); they were intentionally selected to try to cover the referred socio-demographic characteristics. All of these participants signed an informed-consent form approved by the Ethical Committee of Aragón, Spain (CEICA). The interviews were recorded and transcribed by the interviewer herself (Raquel del Moral). These records were subject to content analysis by three researchers (Raquel del Moral, Pedro C. Marijuán, and Jesús Montero-Marín), who independently identified the emerging categories according to which the generality of the topics could be encoded [71]. The broad frameworks of the possible sociotype interactions were tentatively identified. Next, we attempted to determine which conceptual aspects of the sociotype were typical of these frameworks. We empirically defined each of the emerging categories by discussing their ability to capture the different interactions. The appropriate adjustments were made by consensus to ensure that each definition would be comprehensive and exclusive of the others [72].

That initial qualitative experience provided a number of data about how the theoretical notion of the sociotype was felt in different personal circumstances and social domains. Subsequently, a preliminary pilot fieldwork was developed applying a convenience sampling to 165 subjects, all of them students from two education centers, which were interrogated using a very preliminary draft-model [12]. As a result of these works, and in consonance with the above cited literature, we proposed the emergent dimensions of ‘family’, ‘friends’, ‘acquaintances’, and where applicable ‘study or work mates’, as the main factors that a sociotype basic definition should include. Further, a series of 32 selected questions were proposed to conform the initial version of the “Sociotype Questionnaire” (SOCQ), which will be described below in its developmental process and final version.

The duration of the complete survey was approximately half an hour. Each of the participants was presented with an initial description of the survey (with an informed consent form), which introduced the aims of the study, the advantages/disadvantages of participating, and notification that the data would be processed anonymously (S4 File). An online platform gave support to the completion of the survey and data collection (http://www.surveymonkey.com). A research psychologist administered the questionnaires performed in the face-to-face format (for most elderly participants); afterwards the collected data were dumped on the online platform (S1 Table).

The Ethical Committee of Aragón (CEICA) had previously approved this study (Act: CP13/2014). All the participants provided their informed consent before completing the survey, either by reading the project information and providing verbal consent (face-to-face format), or by explicitly accepting the study conditions (online platform). Given the procedure followed and the kind of generic data requested, the anonymity of the participants in the survey was granted.

Measurements

• Background variables: The survey recorded a set of socio-demographic variables providing a general view of the social circumstances of participants, such as: sex, age, relationships (‘with partner/married’, ‘single’, ‘separate/divorced’, ‘widow/widower’), connivance (‘alone’, ‘partner’, ‘partner and children’, ‘other family’, ‘friends’, ‘residence’), place (‘rural’, ‘urban’), education (‘no studies’, ‘primary’, ‘high school’, ‘university’), employment (‘student’, ‘unemployed’, ‘employed’, ‘retired’), salary (‘<Minimum Wage’, ‘1–2 MW’, ‘2–4 MW’, ‘>4 MW’), social satisfaction (using a Visual Analogical Scale–VAS, from 0 to 100).
• Sociotype Questionnaire (SOCQ): Subjects were asked a set of 32 items, assessing the quality of their relationships with ‘family’, ‘friends’, ‘acquaintances’, and ‘study/work’ mates (8 questions for each one). The first three domains were proposed as subscales of a general sociotype factor, and the fourth one was proposed as an independent scale, to be used when applicable (in this case, 49,5% employed and 11,3% students). The items were developed by a multidisciplinary expert panel (including biologists, psychologists and sociologists), who included the main characteristics of each domain by consensus. The wording of the items was guided by a table of content specifications, enabling their fit, conceptual validity, and representativeness. The number of items was over-dimensioned to select those with the best psychometric properties. In order to counteract the effects of response styles and biases, the survey utilized a forced-choice response format, rating the degree of agreement with each of the statements, some of them in reverse score, using a Likert-type scale with 6 response options, from 0 (never) to 5 (always). See (S1 File).
• General Health Questionnaire (GHQ-12): This is the most extensively used screening instrument to measure psychological distress, being attractive because of its brevity (12 items) [73]. Its psychometric properties have been studied in several countries [74], applying to various types of population, e.g., elderly [75] and urological patients [76]. We used the Spanish validated version [77, 78], with α = 0.76. The correction was conducted assigning values from 0 to 3 to the different possible answers (S5 File).
• The Revised UCLA Loneliness Scale (RULS): This widely used questionnaire consists in a one-dimensional 20-item scale, designed to measure subjective feelings of loneliness and social isolation [79]. It is a revised version of the original UCLA Loneliness Scale [80]. The Spanish validated version herein used counts with adequate psychometric properties, α = 0.94 [81]. Participants rate each item on a Likert-type scale ranged from 1 (never) to 4 (often). See (S6 File).
• Eysenck Personality Questionnaire-Revised (EPQ-R): This questionnaire measures three major dimensions that account for most of the variance in personality [82]. The EPQ-R is an excellent choice to represent the personality domain. This measure has proven useful for numerous applications in human resources, career counseling, clinical settings and biomedical research. A validated Spanish version of this questionnaire was used, with adequate psychometric properties [83]. The EPQ-R scales are: ‘extraversion’ (α = 0.82), ‘neuroticism’ (α = 0.86), ‘psychoticism’ (α = 0.73), and ‘lie’ (α = 0.76). The total number of items forming the Spanish version of EPQ-R is 83, and they are answered assigning ‘yes’ or ‘not’ (S7 File).

Statistical analysis

Means (SD) and frequencies (percentages) were calculated on the socio-demographic data. From the proposed items of the sociotype questionnaire, we selected those with the best discrimination coefficient (item-rest coefficient) in their respective domain, taking into account the criterion of ≥0.30 from the Classic Test Theory point of view [84]. To analyze the factor structure, we randomly split the sample into two halves: the first sub-sample (n1 = 538) for the initial Exploratory Factor Analysis (EFA), and the second sub-sample (n2 = 537) for the Confirmatory Factor Analysis (CFA). Mardia’s coefficients [85] were estimated to evaluate items’ distribution in multivariate terms. Polychoric correlation matrices, especially developed for the analysis of relationships between polytomous categorical variables, were calculated; KMO index as a measure of sampling adequacy, and Barlett’s test of sphericity to check if there is enough redundancy between the items to be summarized with a few number of factors, were verified [86], ensuring beforehand that the determinant of the matrices were greater than 0.00001 in order to discard possible problems of multi-collinearity [87].

We used parallel analysis [88] to identify the number of factors, replacing the raw data by optimal implementation based on minimum rank factor analysis after generating 500 random correlation matrices [89]. A factor is significant if the associated eigenvalue is bigger than that corresponding to a 95th percentile of the eigenvalues derived from the random dataset. This method is the best solution to decide the number-of-factors-to-retain [90, 91]). The unweighted least squares (ULS) was the method used for factor extraction in the EFA, in view of its demonstrated robustness, especially when working with polychoric matrices [92]. The rotation method was Promax (k = 4.00), given the correlated solution expected, using raw varimax as clever rotation start. To select the items to be included in each factor, we used the criterion of loadings w >0.5 [93], and we used the Item Response Theory (IRT) parameterization by the multidimensional normal-ogive graded response model, which is derived from the assumption of normally distributed measurement error [94], with an ≥0.65 as criterion to interpret the pattern of item discriminations. The percentage of explained variance in each item by means of communality values (h2) was calculated. We tested the appropriateness of fit by using the goodness of fit index (GFI) and the root mean square of standardized residuals (RMSR), which are explained bellow. From the proposed items, we selected those with the best discrimination coefficient in their respective domain, taking into account the Classic Test Theory point of view [84], and using the criterion of item-rest correlations ≥0.30 in the corresponding domain.

We examined the absolute and incremental fit of the emergent SOCQ model by confirmatory factor analysis, applying unweighted least squares, and using the GFI, the adjusted goodness-of-fit index (AGFI), the root mean square of standardized residuals (RMSR), the normed-fit-index (NFI), and Bollen’s relative-fit-index (RFI). GFI and AGFI refer to explained variance and values >0.90 are acceptable [95]. RMSR is the standardized difference between the observed and the predicted covariance, indicating good fit values <0.08 [96]. NFI measures the proportional reduction in the adjustment function when going from null to the proposed model and is considered acceptable when >0.90 [97]. RFI takes into account the discrepancy for the model evaluated and for the baseline model and it is very good close to 1 [98]. Standardized factor saturations (λ and γ), from an analytical point of view of the models, were also considered.

We examined the internal consistency of the scales using congeneric, tau-equivalent and parallel models of reliability [99]. The congeneric model is the least restrictive and assumes that each individual item measures the same latent variable, with possibly different scales, degrees of precision, and magnitude of error. The tau-equivalent model implies that individual items measure the same latent variable on the same scale and with the same degree of precision, but with possibly different degrees of error. The parallel model is the most restrictive model; it assumes that all items must measure the same latent variable on the same scale, with the same degree of precision and with the same amount of error. In order to reach parsimony, we chose the more restrictive model that fit good enough with the data, applying the ULS method [100]. The reliability value was calculated by squaring the implied correlation between the composite latent true variable and the composite observed variable, to arrive at the percentage of the total observed variance that was accounted for by the true variable [100]. Item-rest and mean item-rest correlations were also calculated to assess the degree of relationship among the finally selected items.

We used the SOCQ dimensions as independent variables in multivariate linear regression models, in order to assess the contribution of the sociotype construct to explain ‘loneliness’ and ‘psychological distress’, controlling the possible influence of the personality traits. Previously, we evaluated the degree of association regarding all the constructs, by means of Spearman’s R coefficients. Standardized beta coefficients were used to assess the individual contribution of each variable, and the Wald test was used to evaluate their significance. Adjusted multiple determination coefficients (R2y.123) were also calculated to observe their grouped explanatory power, and their significance was assessed by means of analysis of variance [101]. Partial correlation coefficients (Ry3.12) −to indicate the correlation between two variables when the effect of the other variables included in the equation are removed− and semi-partial correlation coefficients (Ry(3.12)) −the square of which shows the increase in the coefficient of determination after including a specific variable in a model, partialising the influence of the other included variables− were also calculated. The basic assumptions of the regression models were evaluated by using the K-S test over the conditional distribution of the residuals to ensure they were normally distributed, by the Durbin-Watson test to rule out possible autocorrelations in the error terms (adequate with a roughly value = 2.00), and by the tolerance values (1- the squared multiple correlation of a given regressor with the remaining), to discard co-linearity problems [101].

Student's t-test for independent measurements were used to contrast possible differences in the SOCQ dimensions by sex, and the one-way ANOVA in the case of age (groups: ‘18–30 years’, ‘31–45 years’, ‘46–65 years’ and ‘>65 years’). The basic assumptions of both contrasts (independence, normality and heterocedasticity) were revised. All the tests used were bilateral (α<0.05). Packages SPSS v19, FACTOR v10, and AMOS v20 were used to conduct the statistical analysis.

Socio-demographic characteristics of participants

A total of 1,075 participants completed the study. All of them were Spanish (with diverse regional backgrounds), 66.8% females and 33.2% males, between the ages of 18–95 years (Mean = 49.79; SD = 21.47), 52.3% of them with partner or married and 25.6% singles, 87.8% living in an urban context, 58% with university studies, 49.7% employed, and 28.1% retired. The main socio-demographic characteristics of all the participants are shown in Table 1.

Download:

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

Table 1. Characteristics of the study participants (n = 1,075).

https://doi.org/10.1371/journal.pone.0189568.t001

Exploratory Factor Analysis (EFA)

Table 2 shows the features of the 16 general SOCQ items finally selected according to the method previously described based on the Classical Theory of Tests (n₁ = 538) [Mardia’s = 37.66 (p<0.001); KMO = 0.80; Bartlett's = 2,927.40 (p<0.001)]. Parallel analysis identified three factors, explaining 69.8% of the variance. F₁ presented topics associated with ‘friends’, F₂ with ‘family’, and F₃ with ‘acquaintances’. The Schmid-Leiman second order factor solution presented the values of F₁ = 0.55, F₂ = 0.67 and F₃ = 0.54. The model presented appropriate fit (GFI = 0.99; RMSR = 0.04). Table 2 also shows the characteristics of the SOCQ at work/studies items (n = 328) [Mardia’s = 9.24 (p<0.001); KMO = 0.82; Bartlett's = 350.70 (p<0.001)]. Parallel analysis identified one factor, explaining 69.9% of the variance. F₁ presented topics related to ‘mates at work/studies’. The model presented appropriate fit (GFI = 1.00; RMSR = 0.03).

Download:

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

Table 2. Psychometric features of the SOCQ by using Exploratory Factor Analysis^*.

https://doi.org/10.1371/journal.pone.0189568.t002

Confirmatory Factor Analysis (CFA)

The characteristics of the general SOCQ matrix (n₂ = 537), were: Mardia’s = 33.28 (p<0.001); KMO = 0.84; Bartlett's = 3,140.90 (p<0.001). Fig 1 shows the general SOCQ structure using CFA from an analytical and standardized point of view. The three first order factors turned out to be highly influenced by a general second order factor (G), with loadings over F₁ = 0.51, F₂ = 0.94, and F₃ = 0.53, and explaining 73.7% of the variance. The item loadings with regard to their respective latent factor were high (from 0.52 to 0.82). The general SOCQ structure presented adequate fit indices with no using correlations between the error terms (GFI = 0.99; RSMR = 0.05; AGFI = 0.98; NFI = 0.98; RFI = 0.97). The characteristics of the SOCQ at work or studies matrix (n = 328) were: Mardia’s = 11.29 (p<0.001); KMO = 0.82; Bartlett's = 1,003.78 (p<0.001). Fig 1 also shows the SOCQ at work/studies structure by using CFA. The only one factor explained 71.7% of the variance, with loadings from 0.73 to 0.83. The SOCQ at work/studies structure presented adequate fit with no using correlations between the error terms (GFI = 0.99; RSMR = 0.02; AGFI = 0.99; NFI = 0.99; RFI = 0.99).

Download:

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

Fig 1. Analytical perspective of the SOCQ by using confirmatory factor analysis.

https://doi.org/10.1371/journal.pone.0189568.g001

Reliability models of the SOCQ

Table 3 shows the reliability models tested for the SOCQ. All the scales and sub-scales fitted better with the congeneric model, being the estimates obtained of R = 0.81 for the ‘general SOCQ’, R = 0.81 for ‘family’, R = 0.90 for ‘friends’, R = 0.71 for ‘acquaintances’, R = 0.87 for ‘work/studies’. The average of item-rest values for the ‘general SOCQ’ was 0.52, being of 0.64 for the ‘family’ sub-scale, of 0.77 for the ‘friends’ sub-scale, of 0.45 for the ‘acquaintances’ sub-scale, and of 0.72 for the ‘work/studies’ sub-scale.

Download:

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

Table 3. Fix indices for the reliability models of the SOCQ.

https://doi.org/10.1371/journal.pone.0189568.t003

Explanatory power of the SOCQ regarding loneliness and psychological distress

The SOCQ factors showed important associations with the other constructs (Table 4).

Download:

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

Table 4. Relationships of the SOCQ dimensions with the other constructs.

https://doi.org/10.1371/journal.pone.0189568.t004

The explanatory power of the regression models was very high (Table 5). ‘Loneliness’ was explained (R² = 0.62; p<0.001) by ‘family’ (Beta = -0.24; p<0.001), ‘friends’ (Beta = -0.29; p<0.001), ‘acquaintances’ (Beta = -0.18; p<0.001), ‘extraversion’ (Beta = -0.17; p<0.001), ‘neuroticism’ (Beta = 0.29; p<0.001) and ‘psychoticism’ (Beta = 0.05; p = 0.040). ‘Psychological distress’ was explained (R² = 0.42; p<0.001) by ‘family’ (Beta = -0.08; p = 0.005), ‘friends’ (Beta = -0.09; p = 0.009), ‘acquaintances’ (Beta = -0.08; p = 0.010), ‘work/studies’ (Beta = -0.15; p<0.001), ‘extraversion’ (Beta = -0.10; p = 0.005), ‘neuroticism’ (Beta = 0.45; p<0.001). It was possible to accept the basic assumptions needed to go ahead with the regression, with tolerance values from 0.55 to 0.85.

Download:

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

Table 5. Regression models for the SOCQ with regard to loneliness and psychological distress.

https://doi.org/10.1371/journal.pone.0189568.t005

Differences in SOCQ according to sex and age

As we can see (Table 6), males presented higher scores in ‘work/studies’ than females [(Mn = 10.28; SD = 6.90) vs. (Mn = 8.99; SD = 7.35); p = 0.005]; while females did it in ‘family’ [(Mn = 17.01; SD = 3.18) vs. (Mn = 17.46; SD = 3.10); p = 0.027]. In terms of age, the SOCQ-general, ‘friends’ and ‘work/studies’ showed a decreasing trend (p<0.001), ‘acquaintances’ showed an increasing trend (p<0.001), and ‘family’ did not show differences among age groups (p = 0.333). ‘Social satisfaction’, measured by the VAS, did not show significant changes either by sex (p = 0.217) or by age (p = 0.262). The general assumptions of independent groups to develop the analyses were fulfilled.

Download:

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

Table 6. SOCQ differences according to sex and age.

https://doi.org/10.1371/journal.pone.0189568.t006