Accepted Manuscript Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences Baoshan Zhang, Qianyun Gao, Marjolein Fokkema, Valeria Alterman, Qian Liu PII: DOI: Reference:
S0049-089X(15)00093-9 http://dx.doi.org/10.1016/j.ssresearch.2015.05.003 YSSRE 1790
To appear in:
Social Science Research
Received Date: Revised Date: Accepted Date:
21 July 2014 3 April 2015 10 May 2015
Please cite this article as: Zhang, B., Gao, Q., Fokkema, M., Alterman, V., Liu, Q., Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences, Social Science Research (2015), doi: http://dx.doi.org/10.1016/j.ssresearch.2015.05.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
title page
Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences Baoshan Zhanga · Qianyun Gaoa · Marjolein Fokkemab · Valeria Altermanc · Qian Liua a
School of Psychology, Shaanxi Normal University, China
b
Department of Clinical Psychology and EMGO Institute, VU University Amsterdam, Netherlands
c
Department of Management, University of Florida, America
Q. Gao Corresponding author School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
phone numbers: 8615529055201 B. Zhang School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
M. Fokkema Department of Clinical Psychology and EMGO Institute, VU University Amsterdam, Amsterdam, Netherlands, 1081HV e-mail:
[email protected]
V. Alterman Department of Management, University of Florida, PO BOX 117165, Gainesville, FL 32611 e-mail:
[email protected]
Q. Liu School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences Loneliness is a common unpleasant experience resulting from deficiencies in a person’s social relations in either quality or quantity (Perlman & Peplau, 1984), which is prevalent during adolescence (Goossens, 2006; Hawthorne, 2008). It has been well established that adolescent individuals have to cope with a number of poignant developmental challenges (Laursen & Hartl, 2013), of which the two most important are establishing intimate social relationships and developing concepts of the self (Sippola & Bukowski, 1999). But these two developmental tasks are conflicting, and the contradiction between separating the self as unique from others and increasing need for wanting intimacy with others gives rise to a high risk for loneliness (Sippola & Bukowski, 1999). As a consequence, almost 80 percent of adolescents constantly feel lonely (Hawkley & Cacioppo, 2010), and numerous studies have suggested that the prevalence of loneliness reaches a peak in early adolescence (Heinrich & Gullone, 2006; Roekel, Scholte, Verhagen, Goossens, & Engels, 2010). Loneliness has negative influences on adolescents’ mental health and their behaviors (Heinrich & Gullone, 2006). The sharp increase in loneliness during adolescence is associated with lower global life satisfaction (Kapıkıran, 2013), poorer academic performance (Benner, 2011), poorer adjustment (Larson, 1999), higher school dropout rates (Asher & Paquette, 2003), poorer health practices (Stickley, Koyanagi, Koposov, Schwab-Stone, & Ruchkin, 2014), lower self-esteem (Vanhalst, Goossens, Luyckx, Scholte, & Engels, 2013), fewer health-promoting behaviors (Lasgaard, Goossens, Bramsen, Trillingsgaard, & Elklit, 2011) and more severe internet addiction (Huan, Ang, & Chye, 2014; Özdemir, Kuzucu, & Ak, 2014). Additionally, serious loneliness could precipitate a greater crisis for adolescents (Harris, Qualter, & Robinson, 2013; Hawkley & Cacioppo, 2010; Qualter et al., 2013), such as depression (Ladd & Ettekal, 2013), anxiety (Lasgaard et al., 2011), social phobia (Schinka, Dulmen, Mata, Bossarte, & Swahn, 2013) as well as suicide (Dulmen & Goossens, 1
2013), resulting in a greater burden to family and society (Heinrich & Gullone, 2006), as well as to social resources (e.g., greater use of the health care system (Geller, Janson, McGovern, & Valdini, 1999)). As a consequence, chronic loneliness has a vital influence on adolescents’ emotional (Qualter, Brown, Munn, & Rotenberg, 2010; Schinka, VanDulmen, Bossarte, & Swahn, 2012; Vanhalst et al., 2012), mental and physical well-being (Harris et al., 2013; Vanhalst et al., 2013), and on their behavior (Schinka et al., 2013). Therefore, it is very important to further study adolescents’ loneliness and its related protective factors. Previous studies have demonstrated that social support and interpersonal relationships are two important predictors of adolescents’ loneliness (Bowker & Spencer, 2010; Eldeleklioğlu, 2008; Merz & Jak, 2013). For example, there is clear evidence that social support is negatively correlated with loneliness (Kong & You, 2013), and decreased social support is strongly linked to increased feelings of loneliness in adolescence (Hudson, Elek, & Campbell-Grossman, 2000; Kapıkıran, 2013). In addition, the negative correlation between interpersonal relationships and loneliness is also quite robust (Cheng & Furnham, 2002), and research has suggested that poor-quality interpersonal relationships are closely linked to loneliness (Vanhalst, Luyckx, & Goossens, 2014). Furthermore, the predictors mentioned above may be manifested differently depending on gender. Therefore, another important issue is the patterns of the correlations between interpersonal relationships, social support and loneliness may be different in different sex groups. Previous studies have shown that the focus of interpersonal relationships is different for girls and boys: girls tend to establish dyadic relationships and value intimacy, self-disclosure, and emotional support, while boys prefer to develop relationships with larger groups, involving companionship and joint activity (Gabriel & Gardner, 1999; Maccoby, 1990). As for social support and loneliness, there are also significant differences between boys and girls. For example, Salomon and Strobel (1997) found that girls looked for social support more readily and got more emotional support than boys. In 2
addition, Stokes and Levin (1986) found that different standards were used to evaluate the feeling of loneliness in different sex groups: boys tend to use group-oriented criteria, whereas girls tend to value the quality of dyadic relationships. Although prior studies have established associations between the quality of interpersonal relationships, social support and loneliness (e.g., Bowker & Spencer, 2010; Eldeleklioğlu, 2008; Merz & Jak, 2013), few studies have focused on the individual effects of different types of interpersonal relationships on social support and on loneliness, the role of social support in the association between the quality of different types of interpersonal relationships and loneliness, as well as the topic of whether the associations between the quality of various types of interpersonal relationships, social support and loneliness are consistent across gender. It is important to explore the above patterns of relationships for the following two reasons: firstly, if we know the possible factors that mediate the association between interpersonal relationships and loneliness for boys and girls separately, we can target these factors in future interventions when reducing loneliness in different sex groups; secondly, it is not clear how different types of interpersonal relationships influence loneliness in adolescence and whether there exists any gender differences in the relationship between loneliness and its related protective factors. Therefore, examining this possible mediation mechanism in different groups could greatly contribute to the body of knowledge in this area. Based on the above arguments, the main emphasis of the current study is placed on examining the associations between the quality of various types of interpersonal relationships in school settings (i.e., opposite-sex relationships, same-sex relationships and teacher-student relationships), social support and loneliness. The purposes of the present study are as follows: Firstly, we will explore the mediating effect of social support on the association between the quality of different types of interpersonal relationships and loneliness. Secondly, we will examine the gender difference in 3
the patterns of the associations between the quality of various types of interpersonal relationships, social support and loneliness. Finally, we will examine whether the individual effects of different types of interpersonal relationships on social support and loneliness are consistent across gender. In the following paragraphs, we first introduce our theoretical framework, from which we derived hypotheses regarding the associations between the quality of interpersonal relationships, social support, loneliness, and gender. 1 Theoretical Framework and Hypothesis Development 1.1 The Quality of Interpersonal Relationships and Loneliness Interpersonal relationships can be defined as the patterns of emotion, thought and behavior when individuals interact with others (Plutchik, 1997). Prior studies have demonstrated the bi-directional association between the quality of interpersonal relationships and loneliness (Stensland, Thoresen, Wentzel-Larsen, Zwart, & Dyb, 2014; Woodhouse, Dykas, & Cassidy, 2012). On one hand, the quality of interpersonal relationships is strongly associated with loneliness during adolescence (Cheng & Furnham, 2002; Woodhouse et al., 2012). For example, there is evidence that self-disclosure (an important indicator of interpersonal relationships) is negatively correlated with loneliness (Stokes & Levin, 1986), and within the context of negative interpersonal relationships, researchers have found that rejected students show higher levels of loneliness (Crick & Ladd, 1993; Woodhouse et al., 2012). On the other hand, prior research has also demonstrated that loneliness predicts poorer social interactions (Hawkley, Burleson, Berntson, & Cacioppo, 2003). In summary, the quality of interpersonal relationships is a significant predictor of loneliness, and a higher level of loneliness will result in deficiencies in social relationships during adolescence. Although both the quality of interpersonal relationships and loneliness have influences on adolescent students (Giordano, 2003;Heinrich & Gullone, 2006), they do so from quite different aspects: the quality of 4
interpersonal relationships influence adolescents from an external aspect derived from their social interpersonal environment, whereas loneliness impacts them from an internal aspect derived from their inner experiences (Heinrich & Gullone, 2006). The dissatisfaction with the quality and quantity of their external social relationships results in the inner feeling of loneliness (Perlman & Peplau, 1984). Therefore, the current study focuses on exploring the effect of external interpersonal relationship on internal loneliness, concerning the protective factors to the experience of loneliness. The interactions with family members, teachers, opposite-sex peers, same-sex peers and romantic partners form the main interpersonal relationships during adolescence (Galanaki & Vassilopoulou, 2007; Giordano, 2003). Although romantic relationships and peer relationships are closely correlated for adolescents in high school (Connolly, Craig, Goldberg, & Pepler, 1999), previous research has suggested that these two types of interpersonal relationships are significantly different: romantic relationships are characterized with passion, commitment and exclusivity, while peer friendships are characterized with affiliation and based on similarity and shared perspectives (Connolly et al., 1999; Giordano, 2003). In addition, while most adolescents experience peer-relationships during high school (Mortimer & Call, 2001), not all adolescents experience romantic relationships (Carver, Joyner, & Udry, 2003). As a consequence, peer relationships influence adolescent individuals from a more universal aspect (Mortimer & Call, 2001), and thus, although the quality of romantic relationships is also a predictor of loneliness (Moore & Leung, 2002), most studies did not take romantic relationships into consideration when exploring the effect of interpersonal relationships on adolescents (e.g.,Dong, Liu, & Ding, 2012). For most adolescents, school life is the most important part of their lives, and it significantly influences their psychological well-being, in ways such as general social-emotional health (Eccles et al., 1993) and self-esteem (Shiu, 2001). Based on the argument stated above, interpersonal relationships in school consist 5
mainly of opposite-sex peers relationships, same-sex peers relationships and teacher-student relationships (Dong et al., 2012; Wo, Lin, Ma, & Li, 2001). These three relationships are vital in adolescents’ mental development (Giordano, 2003), and play an important role in adolescent academic performance (Dong et al., 2012), emotion distress (Kenny, Dooley, & Fitzgerald, 2013) and risk behaviors (Rudasill, Reio, Stipanovic, & Taylor, 2010). Based on the strong association between general interpersonal relationships and loneliness, as well as the predicting effect of general interpersonal relationships on loneliness (e.g. Heinrich & Gullone, 2006), we expect that the qualities of different types of interpersonal relationships in school are significant predictors of loneliness in adolescence. Hypothesis 1: The qualities of opposite-sex relationships, same-sex relationships and teacher-student relationships will be three significant predictors of loneliness during adolescence. That is, increased qualities of these three relationships will be linked with decreased feelings of loneliness. 1.2 Mediating Effects of Perceived Social Support Social support can be characterized as “the perceived or actual instrumental and/or expressive provisions supplied by the community, social networks, and confiding partners” (Lin, 1986), and may come from family, friends, or a significant other (Zimet, Dahlem, Zimet, & Farley, 1988). It has been suggested that the quality of interpersonal relationships could influence social support. For example, a large body of research has revealed that adolescents with good quality of interpersonal relationships obtained more support from their teachers and peers, whereas adolescents with poor quality of relationships were less likely to get that support from their relationships (Gumpel & Ish-Shalom, 2003; Malecki & Demaray, 2003). Additionally, prior research has shown that social support is also strongly correlated with loneliness. A meta-analytic study focusing on loneliness has demonstrated that social support is an important predictor for loneliness in adolescence (Mahon, Yarcheski, Yarcheski, Cannella, & Hanks, 2006). More specifically, social support from parents, classmates, 6
and close friends is negatively correlated with the risk of loneliness (Lasgaard, Nielsen, Eriksen, & Goossens, 2010). Similarly, it has been found that enhanced social support could reduce the level of loneliness (Masi, Chen, Hawkley, & Cacioppo, 2011). Based on the above arguments, it is likely that social support would be a mediator in the association between the quality of interpersonal relationships and loneliness. Hypothesis 2: Perceived social support will be a mediator in the association between the qualities of different types of interpersonal relationships and loneliness in the school context. That is, better qualities of opposite-sex relationships, same-sex relationships and teacher-student relationships in school will be associated with more social support, and the increasing support will result in reduced feelings of loneliness. 1.3 Gender Differences in the Associations between Qualities of Interpersonal Relationships, Social Support and Loneliness Gender differences have been found in the levels of various types of interpersonal relationships. For example, girls experience better relationships with their peers in general (Cheng & Chan, 2004; Kenny et al., 2013), and report more positive qualities in same-sex relationships (Blyth & Foster-Clark, 1987) and teacher-student relationships (Salomon & Strobel, 1997). Boys, on the other hand, report more positive qualities in their opposite-sex relationships (Blyth & Foster-Clark, 1987). Furthermore, prior studies also revealed that different types of interpersonal relationships play different roles for boys and girls. For example, adolescent girls obtain more mental support from their peers, such as greater levels of intimacy and emotional support, whereas boys obtain more instrumental support (Buhrmester & Furman, 1987; Frey & Röthlisberger, 1996; Furman & Buhrmester, 1992; Salomon & Strobel, 1997). Based on these arguments, boys and girls will not only show different levels of the different types of interpersonal relationships, but these relationships will also provide different support for them. Thus, it is possible that the patterns of associations between different types of interpersonal relationships and loneliness should be different across gender groups. 7
Hypothesis 3: Gender differences will exist in the patterns of associations between the quality of interpersonal relationships, social support and loneliness among adolescent students. 1.4 Individual Effects of Different Types of Interpersonal Relationships on Loneliness Different types of interpersonal relationships offer different resources to adolescents (Weiss, 1974). Positive teacher-student relationships provide students with the scaffold and instruction to develop key social, behavioral and learning competencies (Brophy, 1985; Pianta, 1999), while peer-relationships (same-sex and opposite-sex relationships) serve as significant sources of intimacy (Ladd, 2005; Way & Greene, 2006), providing an emotional comfort to buffer developmental challenges in adolescence (Mortimer & Call, 2001). In addition, adolescents tend to establish relationships with their same-sex peers more easily and faster than their opposite-sex peers, and thus same-sex relationships are prominent in peer-relationships (Bukowski, Sippola, & Hoza, 1999; Hartup & Stevens, 1999). Furthermore, teachers often change yearly for adolescents, and consequently, it is common for teachers and students to distant each other before they can establish a stable relationship (Galanaki & Vassilopoulou, 2007). Due to the fact that teacher-students relationships tend to be more unstable than same-sex or opposite-sex relationships, and same-sex relationships are more prominent than opposite-sex relationships, it is likely that these three relationships would contribute differently to loneliness. Hypothesis 4: The qualities of different types of interpersonal relationships will show different associations with loneliness in adolescence. That is, same-sex relationships will contribute more to loneliness than opposite-sex relationships or teacher-student relationships, and opposite-sex relationships will contribute more to adolescents’ loneliness than teacher-student relationships.
8
opposite-sex
a
relationship
social support
b same-sex
g
relationship e c teacher-student
f
d
loneliness
relationship
Figure 1. Interrelationships between different types of interpersonal relationships, social support and loneliness. 1.5 Model Specifications To test our hypotheses, we formulated the structural equation model presented in Figure 1. This model specifies that the quality of opposite-sex relationships, same-sex relationships, and teacher-student relationships have direct effects on social support and loneliness. Social support served as an endogenous variable with a direct effect on loneliness. The three interpersonal relationships in school (same-sex relationships, opposite-sex relationships and teacher-student relationships), which served as exogenous variables in the model, were allowed to correlate with each other and also have an indirect effect on loneliness through social support. 2 Method 2.1 Participants Participants were 10th and 11th grade students recruited from four mixed-sex high schools in three cities from mainland China. The expected average age of high school students in China is 17 years (range from 16 to 18 years) according to the official guideline. For all participants, written participant informed assent and parental informed consent for participation were obtained. The current study was also approved by the university's Institutional Review Board and the Committee on Ethics of Research in Humans at each authors’ institution. The self-reports were administered by researchers, and small gifts such as a pencil, ruler or eraser 9
(each costs about RMB 2) were provided to the participants as compensation. After collecting original data from 1,700 participants, 1.5% of the participants were excluded because more than 10% of their items were missing. For the remaining 1,674 cases, about 24 participants had one missing value, and these missing values were substituted with their means when conducting data analysis. T tests were conducted to compare the original sample and final sample, and the means on different variables of these two samples were found to be the same. This final dataset used in all subsequent analyses had a mean age of 15.91 years old (SD = 0.9 years; range: 13-19 years). Of the participants, 851 were boys (mean age = 16.0 ± 0.9 years) and 823 were girls (mean age = 15.8 ± 0.9 years). 56.9% of the students’ fathers and 62.8% of the students’ mothers had at least some junior high school education. The majority of participants were from low income households: 55.9% were from low income households (￥0–3,000 per month), 31.3% were from middle income households (￥3,000–5,000 per month), and 12.4% were from high income households (above￥5,000 per month). No significant gender differences were found in the socio-demographic variables. Similar to age (F(6,1552) =1.216, p>0.05), mothers’ education (F(3,1459) =1.654, p>0.05) and fathers’ education (F(3,1462) =0.880, p>0.05) had no effect on loneliness in the current study, and thus, we did not include these sociodemographic variables (e.g., as covariates) in the following analyses. 2.2 Measures 2.2.1 Loneliness. The UCLA Loneliness Scale (University of California Los Angeles Loneliness Scale) is a Likert-type self-report scale used to measure the level of general subjective feelings of loneliness. It was developed by Russell, Peplau, and Ferguson (1978). The scale consists of 20 unidimensional items, including 11 positively and 9 negatively stated items (e.g., I have nobody to talk to). Participants were asked to rate the frequency of the experienced loneliness on a 4-point scale (1= I never feel this way, 2= I rarely feel this way, 3= 1
With rapid developments in China, more and more mothers leave their homes and go to work, which leads to the fact that no one can take care of the children. Many children go to school before the suggested age, and thus, the actual school starting age tends to be younger than the age suggested by the official guideline. 10
I sometimes feel this way, 4= I often feel this way). Item parceling strategies, which were introduced by Cattell (1956) and further developed by Cattell and Burdsal (1975), were used for the UCLA Loneliness Scale to improve reliability. Five items parcels were created, each of the item parcels consisted of four items. The composite score of each item parcel was calculated by adding the relevant item scores, resulting in composite scores ranging from 4 to 16, with higher scores indicating greater loneliness. Past research has found internal consistency as estimated by Cronbach’s alpha ranging from 0.89 to 0.94 (Russell, 1996). For the current study, Cronbach’s alpha was 0.86. 2.2.2 Perceived social support. The Multidimensional Scale of Perceived Social Support (MSPSS) is a 12-item self-report scale developed by Zimet et al. (1988) to assess perceived adequacy of social support. The MSPSS is composed of three subscales, including family (e.g., ‘I get the emotional help and support I need from my family’), friends (e.g., ‘I can talk about my problems with my friends’), and significant others (e.g., ‘There is a special person (i.e., teachers and other classmates) in my life who cares about my feelings’). Each subscale consists of 4 items. Participants were asked to rate the extent of their agreement with each statement on a 7-point scale, ranging from 1 (very strongly disagree) to 7 (very strongly agree). Since the objective of the current study was to examine the perceived social support in the context of school context, family support was excluded, and only support from friends and significant others were used. For each subscale, a total subscale score was generated by taking the sum of the relevant four item scores, with higher total subscale scores indicating greater levels of perceived support. Zimet et al. demonstrated that Cronbach’s alpha was 0.91 for significant other, 0.85 for friends, and 0.88 for the total scale (Zimet et al., 1988), suggesting excellent internal reliability. The present study obtained coefficient alphas of 0.80 for significant other, 0.77 for friends, and 0.86 for the total scale. 2.2.3 Interpersonal relationships. The Interpersonal Relationship Scale is developed by Wo et al. 11
(2001) to assess the quality of interpersonal relationships, which consists of five subscales, including opposite-sex relationships, same-sex relationships, parent-child relationships, teacher-student relationships and stranger relationships. For each item, participants were asked to rate the extent to which the statement was suitable to describe their perceived quality of the relationship on a 5-point scale, ranging from 1= “not suitable at all” to 5= “very suitable”. The aim of the current study was to explore interpersonal relationships in schools. Therefore, parent-child relationships and stranger relationships were excluded, and only three subscales (of 3 items each) were used, including opposite-sex relationships (e.g., ‘I cannot be friends with the opposite sex peers’), same-sex relationships (e.g., ‘I feel rejected and indifferent when I interact with the same sex peers’) and teacher-student relationships (e.g., ‘Sometimes, I feel my teachers get me in to trouble deliberately’). These three subscales assess the qualities of adolescents’ relationships with their opposite-sex peers, same-sex peers and teachers, respectively. Each of the three subscale scores was calculated by summing the relevant item scores, yielding scores ranging from 3 to 15, with higher scores indicating better relationship quality. The present study obtained Cronbach’s alphas of 0.62 for opposite-sex relationships, 0.66 for same-sex relationships, and 0.72 for teacher-student relationships. The significant correlations among the three sub-scales were 0.479 for same-sex and opposite-sex relationships, 0.390 for same-sex and teacher-student relationships, and 0.316 for opposite-sex and teacher-student relationships, indicating these three sub-scales are correlated each other. In summary, these three sub-scales not only assess the same constructs (interpersonal relationship), but also assess the different types of interpersonal relationships. 2.3 Data Analysis Firstly, in order to check whether the measurement structure of the scales was appropriate, confirmatory factor analysis was conducted to examine the measurement model and the measurement invariance. Secondly, a structural equation model was proposed to explore the associations between the quality of 12
interpersonal relationships, social support and loneliness, and further examine the mediating effect of social support in the associations between the quality of interpersonal relationships and loneliness. In this step, we obtained the optimal model that could best explain the data from several competing structural models by comparing model fit. Thirdly, we examined whether the optimal model was equal across different sex groups, namely whether the patterns of associations between the quality of interpersonal relationship, social support and loneliness were consistent in different adolescent gender groups. If the relationships were inconsistent across groups, we further explored which associations (pathways) between variables were different. Finally, several nested models were compared to assess whether various types of interpersonal relationships contributed differently to loneliness in adolescence. All SEM models mentioned above were estimated in AMOS 17.0, using maximum likelihood estimation. In order to estimate models successfully, all paths between error variables and observed indicators were set to be 1, and one of the unstandardized paths between the observed indicators and the latent variables was also set to be 1, randomly. In addition, the goodness of fit of relevant competing and nested models were also tested and compared in AMOS 17.0. The fit indices used to assess goodness of fit include GFI, CFI, RMSEA, AGFI and AIC. Those indices show good fit when GFI, CFI and AGFI are greater than 0.9, RMSEA is lower than 0.8 and AIC value are low. These fit indices and the differences in χ2 were used to determine the optimal model which fits the data best. 3 Results 3.1 Measurement Model and Measurement Invariance 3.1.1 Measurement model. Confirmatory factor analysis was used to evaluate factor loadings (Table 1). Factor loadings for the different types of interpersonal relationships items (i.e., opposite-sex relationship items, same-sex relationship items, and teacher-student relationship items), social support items and loneliness 13
parcel items were high, ranging from 0.45 to 0.84, indicating that the subscale scores and parcel scores were reasonable and high-quality indicators of the constructs we aimed to measure in the current study. Table 1 Factor loadings for confirmatory factor model Measures and variables
Factor loadings
SE
S1
0.55
0.057
S2
0.66
0.051
S3
0.68
0.046
Same-sex relationship
Opposite-sex relationship O1
0.45
0.037
O2
0.70
0.028
O3
0.56
0.032
T1
0.68
0.020
T2
0.73
0.021
T3
0.64
0.023
Parcel1
0.75
0.014
Parcel2
0.72
0.016
Parcel3
0.68
0.036
Parcel4
0.76
0.013
Parcel5
0.75
0.013
Teacher-student relationship
Loneliness
Social support Friends’ support
0.84
0.028
Significant other’s support
0.82
0.016
Note .Same-sex relationship is measured by S1 item, S2 item and S3 item. Opposite-sex relationship is measured by O1 item, O2 item and O3 item. Teacher-student relationship is measured by T1 item, T2 item and T3 item. Loneliness is measured by parcel1, parcel2, parcel3, parcel4 and parcel5. Social support is measured by friends’ support items and significant other’s support items. 3.1.2 Measurement invariance. When conducting cross-group comparisons, measurement invariance should be established as an important prerequisite to eliminate response bias and make sure the measurements have the same mental meanings in different groups (Vandenberg & Lance, 2000). In the current study, as model comparisons across different sex groups were conducted, we must be sure that the constructs assessed by the three scales (interpersonal relationships, perceived social support and loneliness) are measured in the same way across groups. Therefore, multi-group confirmatory factor analysis (CFA) was conducted to test the 14
measurement invariance for these three scales across different sex groups. Traditionally, configural, metric, and scalar invariance should be examined in order to assess the measurement equivalence for structures, factor loadings, and mean values, respectively. In the configural invariance model, no equality constraints are imposed on the intercepts, factor loadings, or residual variances across different groups. In the metric invariance model, the equality constraints are imposed on factor loadings across samples. And in the scalar invariance model, factor loadings and intercepts are set to be equal across groups. The results suggested that measurement invariance of all scales showed reasonable model fits (Table 2). For the interpersonal relationship, the configural invariance model showed acceptable model fits: χ2
(48)
=144.057, p
S0049-089X(15)00093-9 http://dx.doi.org/10.1016/j.ssresearch.2015.05.003 YSSRE 1790
To appear in:
Social Science Research
Received Date: Revised Date: Accepted Date:
21 July 2014 3 April 2015 10 May 2015
Please cite this article as: Zhang, B., Gao, Q., Fokkema, M., Alterman, V., Liu, Q., Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences, Social Science Research (2015), doi: http://dx.doi.org/10.1016/j.ssresearch.2015.05.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
title page
Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences Baoshan Zhanga · Qianyun Gaoa · Marjolein Fokkemab · Valeria Altermanc · Qian Liua a
School of Psychology, Shaanxi Normal University, China
b
Department of Clinical Psychology and EMGO Institute, VU University Amsterdam, Netherlands
c
Department of Management, University of Florida, America
Q. Gao Corresponding author School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
phone numbers: 8615529055201 B. Zhang School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
M. Fokkema Department of Clinical Psychology and EMGO Institute, VU University Amsterdam, Amsterdam, Netherlands, 1081HV e-mail:
[email protected]
V. Alterman Department of Management, University of Florida, PO BOX 117165, Gainesville, FL 32611 e-mail:
[email protected]
Q. Liu School of Psychology, Shaanxi Normal University, 199 South Chang’an Road, Xi'an, China, 710062 e-mail:
[email protected]
Adolescent Interpersonal Relationships, Social Support and Loneliness in High Schools: Mediation Effect and Gender Differences Loneliness is a common unpleasant experience resulting from deficiencies in a person’s social relations in either quality or quantity (Perlman & Peplau, 1984), which is prevalent during adolescence (Goossens, 2006; Hawthorne, 2008). It has been well established that adolescent individuals have to cope with a number of poignant developmental challenges (Laursen & Hartl, 2013), of which the two most important are establishing intimate social relationships and developing concepts of the self (Sippola & Bukowski, 1999). But these two developmental tasks are conflicting, and the contradiction between separating the self as unique from others and increasing need for wanting intimacy with others gives rise to a high risk for loneliness (Sippola & Bukowski, 1999). As a consequence, almost 80 percent of adolescents constantly feel lonely (Hawkley & Cacioppo, 2010), and numerous studies have suggested that the prevalence of loneliness reaches a peak in early adolescence (Heinrich & Gullone, 2006; Roekel, Scholte, Verhagen, Goossens, & Engels, 2010). Loneliness has negative influences on adolescents’ mental health and their behaviors (Heinrich & Gullone, 2006). The sharp increase in loneliness during adolescence is associated with lower global life satisfaction (Kapıkıran, 2013), poorer academic performance (Benner, 2011), poorer adjustment (Larson, 1999), higher school dropout rates (Asher & Paquette, 2003), poorer health practices (Stickley, Koyanagi, Koposov, Schwab-Stone, & Ruchkin, 2014), lower self-esteem (Vanhalst, Goossens, Luyckx, Scholte, & Engels, 2013), fewer health-promoting behaviors (Lasgaard, Goossens, Bramsen, Trillingsgaard, & Elklit, 2011) and more severe internet addiction (Huan, Ang, & Chye, 2014; Özdemir, Kuzucu, & Ak, 2014). Additionally, serious loneliness could precipitate a greater crisis for adolescents (Harris, Qualter, & Robinson, 2013; Hawkley & Cacioppo, 2010; Qualter et al., 2013), such as depression (Ladd & Ettekal, 2013), anxiety (Lasgaard et al., 2011), social phobia (Schinka, Dulmen, Mata, Bossarte, & Swahn, 2013) as well as suicide (Dulmen & Goossens, 1
2013), resulting in a greater burden to family and society (Heinrich & Gullone, 2006), as well as to social resources (e.g., greater use of the health care system (Geller, Janson, McGovern, & Valdini, 1999)). As a consequence, chronic loneliness has a vital influence on adolescents’ emotional (Qualter, Brown, Munn, & Rotenberg, 2010; Schinka, VanDulmen, Bossarte, & Swahn, 2012; Vanhalst et al., 2012), mental and physical well-being (Harris et al., 2013; Vanhalst et al., 2013), and on their behavior (Schinka et al., 2013). Therefore, it is very important to further study adolescents’ loneliness and its related protective factors. Previous studies have demonstrated that social support and interpersonal relationships are two important predictors of adolescents’ loneliness (Bowker & Spencer, 2010; Eldeleklioğlu, 2008; Merz & Jak, 2013). For example, there is clear evidence that social support is negatively correlated with loneliness (Kong & You, 2013), and decreased social support is strongly linked to increased feelings of loneliness in adolescence (Hudson, Elek, & Campbell-Grossman, 2000; Kapıkıran, 2013). In addition, the negative correlation between interpersonal relationships and loneliness is also quite robust (Cheng & Furnham, 2002), and research has suggested that poor-quality interpersonal relationships are closely linked to loneliness (Vanhalst, Luyckx, & Goossens, 2014). Furthermore, the predictors mentioned above may be manifested differently depending on gender. Therefore, another important issue is the patterns of the correlations between interpersonal relationships, social support and loneliness may be different in different sex groups. Previous studies have shown that the focus of interpersonal relationships is different for girls and boys: girls tend to establish dyadic relationships and value intimacy, self-disclosure, and emotional support, while boys prefer to develop relationships with larger groups, involving companionship and joint activity (Gabriel & Gardner, 1999; Maccoby, 1990). As for social support and loneliness, there are also significant differences between boys and girls. For example, Salomon and Strobel (1997) found that girls looked for social support more readily and got more emotional support than boys. In 2
addition, Stokes and Levin (1986) found that different standards were used to evaluate the feeling of loneliness in different sex groups: boys tend to use group-oriented criteria, whereas girls tend to value the quality of dyadic relationships. Although prior studies have established associations between the quality of interpersonal relationships, social support and loneliness (e.g., Bowker & Spencer, 2010; Eldeleklioğlu, 2008; Merz & Jak, 2013), few studies have focused on the individual effects of different types of interpersonal relationships on social support and on loneliness, the role of social support in the association between the quality of different types of interpersonal relationships and loneliness, as well as the topic of whether the associations between the quality of various types of interpersonal relationships, social support and loneliness are consistent across gender. It is important to explore the above patterns of relationships for the following two reasons: firstly, if we know the possible factors that mediate the association between interpersonal relationships and loneliness for boys and girls separately, we can target these factors in future interventions when reducing loneliness in different sex groups; secondly, it is not clear how different types of interpersonal relationships influence loneliness in adolescence and whether there exists any gender differences in the relationship between loneliness and its related protective factors. Therefore, examining this possible mediation mechanism in different groups could greatly contribute to the body of knowledge in this area. Based on the above arguments, the main emphasis of the current study is placed on examining the associations between the quality of various types of interpersonal relationships in school settings (i.e., opposite-sex relationships, same-sex relationships and teacher-student relationships), social support and loneliness. The purposes of the present study are as follows: Firstly, we will explore the mediating effect of social support on the association between the quality of different types of interpersonal relationships and loneliness. Secondly, we will examine the gender difference in 3
the patterns of the associations between the quality of various types of interpersonal relationships, social support and loneliness. Finally, we will examine whether the individual effects of different types of interpersonal relationships on social support and loneliness are consistent across gender. In the following paragraphs, we first introduce our theoretical framework, from which we derived hypotheses regarding the associations between the quality of interpersonal relationships, social support, loneliness, and gender. 1 Theoretical Framework and Hypothesis Development 1.1 The Quality of Interpersonal Relationships and Loneliness Interpersonal relationships can be defined as the patterns of emotion, thought and behavior when individuals interact with others (Plutchik, 1997). Prior studies have demonstrated the bi-directional association between the quality of interpersonal relationships and loneliness (Stensland, Thoresen, Wentzel-Larsen, Zwart, & Dyb, 2014; Woodhouse, Dykas, & Cassidy, 2012). On one hand, the quality of interpersonal relationships is strongly associated with loneliness during adolescence (Cheng & Furnham, 2002; Woodhouse et al., 2012). For example, there is evidence that self-disclosure (an important indicator of interpersonal relationships) is negatively correlated with loneliness (Stokes & Levin, 1986), and within the context of negative interpersonal relationships, researchers have found that rejected students show higher levels of loneliness (Crick & Ladd, 1993; Woodhouse et al., 2012). On the other hand, prior research has also demonstrated that loneliness predicts poorer social interactions (Hawkley, Burleson, Berntson, & Cacioppo, 2003). In summary, the quality of interpersonal relationships is a significant predictor of loneliness, and a higher level of loneliness will result in deficiencies in social relationships during adolescence. Although both the quality of interpersonal relationships and loneliness have influences on adolescent students (Giordano, 2003;Heinrich & Gullone, 2006), they do so from quite different aspects: the quality of 4
interpersonal relationships influence adolescents from an external aspect derived from their social interpersonal environment, whereas loneliness impacts them from an internal aspect derived from their inner experiences (Heinrich & Gullone, 2006). The dissatisfaction with the quality and quantity of their external social relationships results in the inner feeling of loneliness (Perlman & Peplau, 1984). Therefore, the current study focuses on exploring the effect of external interpersonal relationship on internal loneliness, concerning the protective factors to the experience of loneliness. The interactions with family members, teachers, opposite-sex peers, same-sex peers and romantic partners form the main interpersonal relationships during adolescence (Galanaki & Vassilopoulou, 2007; Giordano, 2003). Although romantic relationships and peer relationships are closely correlated for adolescents in high school (Connolly, Craig, Goldberg, & Pepler, 1999), previous research has suggested that these two types of interpersonal relationships are significantly different: romantic relationships are characterized with passion, commitment and exclusivity, while peer friendships are characterized with affiliation and based on similarity and shared perspectives (Connolly et al., 1999; Giordano, 2003). In addition, while most adolescents experience peer-relationships during high school (Mortimer & Call, 2001), not all adolescents experience romantic relationships (Carver, Joyner, & Udry, 2003). As a consequence, peer relationships influence adolescent individuals from a more universal aspect (Mortimer & Call, 2001), and thus, although the quality of romantic relationships is also a predictor of loneliness (Moore & Leung, 2002), most studies did not take romantic relationships into consideration when exploring the effect of interpersonal relationships on adolescents (e.g.,Dong, Liu, & Ding, 2012). For most adolescents, school life is the most important part of their lives, and it significantly influences their psychological well-being, in ways such as general social-emotional health (Eccles et al., 1993) and self-esteem (Shiu, 2001). Based on the argument stated above, interpersonal relationships in school consist 5
mainly of opposite-sex peers relationships, same-sex peers relationships and teacher-student relationships (Dong et al., 2012; Wo, Lin, Ma, & Li, 2001). These three relationships are vital in adolescents’ mental development (Giordano, 2003), and play an important role in adolescent academic performance (Dong et al., 2012), emotion distress (Kenny, Dooley, & Fitzgerald, 2013) and risk behaviors (Rudasill, Reio, Stipanovic, & Taylor, 2010). Based on the strong association between general interpersonal relationships and loneliness, as well as the predicting effect of general interpersonal relationships on loneliness (e.g. Heinrich & Gullone, 2006), we expect that the qualities of different types of interpersonal relationships in school are significant predictors of loneliness in adolescence. Hypothesis 1: The qualities of opposite-sex relationships, same-sex relationships and teacher-student relationships will be three significant predictors of loneliness during adolescence. That is, increased qualities of these three relationships will be linked with decreased feelings of loneliness. 1.2 Mediating Effects of Perceived Social Support Social support can be characterized as “the perceived or actual instrumental and/or expressive provisions supplied by the community, social networks, and confiding partners” (Lin, 1986), and may come from family, friends, or a significant other (Zimet, Dahlem, Zimet, & Farley, 1988). It has been suggested that the quality of interpersonal relationships could influence social support. For example, a large body of research has revealed that adolescents with good quality of interpersonal relationships obtained more support from their teachers and peers, whereas adolescents with poor quality of relationships were less likely to get that support from their relationships (Gumpel & Ish-Shalom, 2003; Malecki & Demaray, 2003). Additionally, prior research has shown that social support is also strongly correlated with loneliness. A meta-analytic study focusing on loneliness has demonstrated that social support is an important predictor for loneliness in adolescence (Mahon, Yarcheski, Yarcheski, Cannella, & Hanks, 2006). More specifically, social support from parents, classmates, 6
and close friends is negatively correlated with the risk of loneliness (Lasgaard, Nielsen, Eriksen, & Goossens, 2010). Similarly, it has been found that enhanced social support could reduce the level of loneliness (Masi, Chen, Hawkley, & Cacioppo, 2011). Based on the above arguments, it is likely that social support would be a mediator in the association between the quality of interpersonal relationships and loneliness. Hypothesis 2: Perceived social support will be a mediator in the association between the qualities of different types of interpersonal relationships and loneliness in the school context. That is, better qualities of opposite-sex relationships, same-sex relationships and teacher-student relationships in school will be associated with more social support, and the increasing support will result in reduced feelings of loneliness. 1.3 Gender Differences in the Associations between Qualities of Interpersonal Relationships, Social Support and Loneliness Gender differences have been found in the levels of various types of interpersonal relationships. For example, girls experience better relationships with their peers in general (Cheng & Chan, 2004; Kenny et al., 2013), and report more positive qualities in same-sex relationships (Blyth & Foster-Clark, 1987) and teacher-student relationships (Salomon & Strobel, 1997). Boys, on the other hand, report more positive qualities in their opposite-sex relationships (Blyth & Foster-Clark, 1987). Furthermore, prior studies also revealed that different types of interpersonal relationships play different roles for boys and girls. For example, adolescent girls obtain more mental support from their peers, such as greater levels of intimacy and emotional support, whereas boys obtain more instrumental support (Buhrmester & Furman, 1987; Frey & Röthlisberger, 1996; Furman & Buhrmester, 1992; Salomon & Strobel, 1997). Based on these arguments, boys and girls will not only show different levels of the different types of interpersonal relationships, but these relationships will also provide different support for them. Thus, it is possible that the patterns of associations between different types of interpersonal relationships and loneliness should be different across gender groups. 7
Hypothesis 3: Gender differences will exist in the patterns of associations between the quality of interpersonal relationships, social support and loneliness among adolescent students. 1.4 Individual Effects of Different Types of Interpersonal Relationships on Loneliness Different types of interpersonal relationships offer different resources to adolescents (Weiss, 1974). Positive teacher-student relationships provide students with the scaffold and instruction to develop key social, behavioral and learning competencies (Brophy, 1985; Pianta, 1999), while peer-relationships (same-sex and opposite-sex relationships) serve as significant sources of intimacy (Ladd, 2005; Way & Greene, 2006), providing an emotional comfort to buffer developmental challenges in adolescence (Mortimer & Call, 2001). In addition, adolescents tend to establish relationships with their same-sex peers more easily and faster than their opposite-sex peers, and thus same-sex relationships are prominent in peer-relationships (Bukowski, Sippola, & Hoza, 1999; Hartup & Stevens, 1999). Furthermore, teachers often change yearly for adolescents, and consequently, it is common for teachers and students to distant each other before they can establish a stable relationship (Galanaki & Vassilopoulou, 2007). Due to the fact that teacher-students relationships tend to be more unstable than same-sex or opposite-sex relationships, and same-sex relationships are more prominent than opposite-sex relationships, it is likely that these three relationships would contribute differently to loneliness. Hypothesis 4: The qualities of different types of interpersonal relationships will show different associations with loneliness in adolescence. That is, same-sex relationships will contribute more to loneliness than opposite-sex relationships or teacher-student relationships, and opposite-sex relationships will contribute more to adolescents’ loneliness than teacher-student relationships.
8
opposite-sex
a
relationship
social support
b same-sex
g
relationship e c teacher-student
f
d
loneliness
relationship
Figure 1. Interrelationships between different types of interpersonal relationships, social support and loneliness. 1.5 Model Specifications To test our hypotheses, we formulated the structural equation model presented in Figure 1. This model specifies that the quality of opposite-sex relationships, same-sex relationships, and teacher-student relationships have direct effects on social support and loneliness. Social support served as an endogenous variable with a direct effect on loneliness. The three interpersonal relationships in school (same-sex relationships, opposite-sex relationships and teacher-student relationships), which served as exogenous variables in the model, were allowed to correlate with each other and also have an indirect effect on loneliness through social support. 2 Method 2.1 Participants Participants were 10th and 11th grade students recruited from four mixed-sex high schools in three cities from mainland China. The expected average age of high school students in China is 17 years (range from 16 to 18 years) according to the official guideline. For all participants, written participant informed assent and parental informed consent for participation were obtained. The current study was also approved by the university's Institutional Review Board and the Committee on Ethics of Research in Humans at each authors’ institution. The self-reports were administered by researchers, and small gifts such as a pencil, ruler or eraser 9
(each costs about RMB 2) were provided to the participants as compensation. After collecting original data from 1,700 participants, 1.5% of the participants were excluded because more than 10% of their items were missing. For the remaining 1,674 cases, about 24 participants had one missing value, and these missing values were substituted with their means when conducting data analysis. T tests were conducted to compare the original sample and final sample, and the means on different variables of these two samples were found to be the same. This final dataset used in all subsequent analyses had a mean age of 15.91 years old (SD = 0.9 years; range: 13-19 years). Of the participants, 851 were boys (mean age = 16.0 ± 0.9 years) and 823 were girls (mean age = 15.8 ± 0.9 years). 56.9% of the students’ fathers and 62.8% of the students’ mothers had at least some junior high school education. The majority of participants were from low income households: 55.9% were from low income households (￥0–3,000 per month), 31.3% were from middle income households (￥3,000–5,000 per month), and 12.4% were from high income households (above￥5,000 per month). No significant gender differences were found in the socio-demographic variables. Similar to age (F(6,1552) =1.216, p>0.05), mothers’ education (F(3,1459) =1.654, p>0.05) and fathers’ education (F(3,1462) =0.880, p>0.05) had no effect on loneliness in the current study, and thus, we did not include these sociodemographic variables (e.g., as covariates) in the following analyses. 2.2 Measures 2.2.1 Loneliness. The UCLA Loneliness Scale (University of California Los Angeles Loneliness Scale) is a Likert-type self-report scale used to measure the level of general subjective feelings of loneliness. It was developed by Russell, Peplau, and Ferguson (1978). The scale consists of 20 unidimensional items, including 11 positively and 9 negatively stated items (e.g., I have nobody to talk to). Participants were asked to rate the frequency of the experienced loneliness on a 4-point scale (1= I never feel this way, 2= I rarely feel this way, 3= 1
With rapid developments in China, more and more mothers leave their homes and go to work, which leads to the fact that no one can take care of the children. Many children go to school before the suggested age, and thus, the actual school starting age tends to be younger than the age suggested by the official guideline. 10
I sometimes feel this way, 4= I often feel this way). Item parceling strategies, which were introduced by Cattell (1956) and further developed by Cattell and Burdsal (1975), were used for the UCLA Loneliness Scale to improve reliability. Five items parcels were created, each of the item parcels consisted of four items. The composite score of each item parcel was calculated by adding the relevant item scores, resulting in composite scores ranging from 4 to 16, with higher scores indicating greater loneliness. Past research has found internal consistency as estimated by Cronbach’s alpha ranging from 0.89 to 0.94 (Russell, 1996). For the current study, Cronbach’s alpha was 0.86. 2.2.2 Perceived social support. The Multidimensional Scale of Perceived Social Support (MSPSS) is a 12-item self-report scale developed by Zimet et al. (1988) to assess perceived adequacy of social support. The MSPSS is composed of three subscales, including family (e.g., ‘I get the emotional help and support I need from my family’), friends (e.g., ‘I can talk about my problems with my friends’), and significant others (e.g., ‘There is a special person (i.e., teachers and other classmates) in my life who cares about my feelings’). Each subscale consists of 4 items. Participants were asked to rate the extent of their agreement with each statement on a 7-point scale, ranging from 1 (very strongly disagree) to 7 (very strongly agree). Since the objective of the current study was to examine the perceived social support in the context of school context, family support was excluded, and only support from friends and significant others were used. For each subscale, a total subscale score was generated by taking the sum of the relevant four item scores, with higher total subscale scores indicating greater levels of perceived support. Zimet et al. demonstrated that Cronbach’s alpha was 0.91 for significant other, 0.85 for friends, and 0.88 for the total scale (Zimet et al., 1988), suggesting excellent internal reliability. The present study obtained coefficient alphas of 0.80 for significant other, 0.77 for friends, and 0.86 for the total scale. 2.2.3 Interpersonal relationships. The Interpersonal Relationship Scale is developed by Wo et al. 11
(2001) to assess the quality of interpersonal relationships, which consists of five subscales, including opposite-sex relationships, same-sex relationships, parent-child relationships, teacher-student relationships and stranger relationships. For each item, participants were asked to rate the extent to which the statement was suitable to describe their perceived quality of the relationship on a 5-point scale, ranging from 1= “not suitable at all” to 5= “very suitable”. The aim of the current study was to explore interpersonal relationships in schools. Therefore, parent-child relationships and stranger relationships were excluded, and only three subscales (of 3 items each) were used, including opposite-sex relationships (e.g., ‘I cannot be friends with the opposite sex peers’), same-sex relationships (e.g., ‘I feel rejected and indifferent when I interact with the same sex peers’) and teacher-student relationships (e.g., ‘Sometimes, I feel my teachers get me in to trouble deliberately’). These three subscales assess the qualities of adolescents’ relationships with their opposite-sex peers, same-sex peers and teachers, respectively. Each of the three subscale scores was calculated by summing the relevant item scores, yielding scores ranging from 3 to 15, with higher scores indicating better relationship quality. The present study obtained Cronbach’s alphas of 0.62 for opposite-sex relationships, 0.66 for same-sex relationships, and 0.72 for teacher-student relationships. The significant correlations among the three sub-scales were 0.479 for same-sex and opposite-sex relationships, 0.390 for same-sex and teacher-student relationships, and 0.316 for opposite-sex and teacher-student relationships, indicating these three sub-scales are correlated each other. In summary, these three sub-scales not only assess the same constructs (interpersonal relationship), but also assess the different types of interpersonal relationships. 2.3 Data Analysis Firstly, in order to check whether the measurement structure of the scales was appropriate, confirmatory factor analysis was conducted to examine the measurement model and the measurement invariance. Secondly, a structural equation model was proposed to explore the associations between the quality of 12
interpersonal relationships, social support and loneliness, and further examine the mediating effect of social support in the associations between the quality of interpersonal relationships and loneliness. In this step, we obtained the optimal model that could best explain the data from several competing structural models by comparing model fit. Thirdly, we examined whether the optimal model was equal across different sex groups, namely whether the patterns of associations between the quality of interpersonal relationship, social support and loneliness were consistent in different adolescent gender groups. If the relationships were inconsistent across groups, we further explored which associations (pathways) between variables were different. Finally, several nested models were compared to assess whether various types of interpersonal relationships contributed differently to loneliness in adolescence. All SEM models mentioned above were estimated in AMOS 17.0, using maximum likelihood estimation. In order to estimate models successfully, all paths between error variables and observed indicators were set to be 1, and one of the unstandardized paths between the observed indicators and the latent variables was also set to be 1, randomly. In addition, the goodness of fit of relevant competing and nested models were also tested and compared in AMOS 17.0. The fit indices used to assess goodness of fit include GFI, CFI, RMSEA, AGFI and AIC. Those indices show good fit when GFI, CFI and AGFI are greater than 0.9, RMSEA is lower than 0.8 and AIC value are low. These fit indices and the differences in χ2 were used to determine the optimal model which fits the data best. 3 Results 3.1 Measurement Model and Measurement Invariance 3.1.1 Measurement model. Confirmatory factor analysis was used to evaluate factor loadings (Table 1). Factor loadings for the different types of interpersonal relationships items (i.e., opposite-sex relationship items, same-sex relationship items, and teacher-student relationship items), social support items and loneliness 13
parcel items were high, ranging from 0.45 to 0.84, indicating that the subscale scores and parcel scores were reasonable and high-quality indicators of the constructs we aimed to measure in the current study. Table 1 Factor loadings for confirmatory factor model Measures and variables
Factor loadings
SE
S1
0.55
0.057
S2
0.66
0.051
S3
0.68
0.046
Same-sex relationship
Opposite-sex relationship O1
0.45
0.037
O2
0.70
0.028
O3
0.56
0.032
T1
0.68
0.020
T2
0.73
0.021
T3
0.64
0.023
Parcel1
0.75
0.014
Parcel2
0.72
0.016
Parcel3
0.68
0.036
Parcel4
0.76
0.013
Parcel5
0.75
0.013
Teacher-student relationship
Loneliness
Social support Friends’ support
0.84
0.028
Significant other’s support
0.82
0.016
Note .Same-sex relationship is measured by S1 item, S2 item and S3 item. Opposite-sex relationship is measured by O1 item, O2 item and O3 item. Teacher-student relationship is measured by T1 item, T2 item and T3 item. Loneliness is measured by parcel1, parcel2, parcel3, parcel4 and parcel5. Social support is measured by friends’ support items and significant other’s support items. 3.1.2 Measurement invariance. When conducting cross-group comparisons, measurement invariance should be established as an important prerequisite to eliminate response bias and make sure the measurements have the same mental meanings in different groups (Vandenberg & Lance, 2000). In the current study, as model comparisons across different sex groups were conducted, we must be sure that the constructs assessed by the three scales (interpersonal relationships, perceived social support and loneliness) are measured in the same way across groups. Therefore, multi-group confirmatory factor analysis (CFA) was conducted to test the 14
measurement invariance for these three scales across different sex groups. Traditionally, configural, metric, and scalar invariance should be examined in order to assess the measurement equivalence for structures, factor loadings, and mean values, respectively. In the configural invariance model, no equality constraints are imposed on the intercepts, factor loadings, or residual variances across different groups. In the metric invariance model, the equality constraints are imposed on factor loadings across samples. And in the scalar invariance model, factor loadings and intercepts are set to be equal across groups. The results suggested that measurement invariance of all scales showed reasonable model fits (Table 2). For the interpersonal relationship, the configural invariance model showed acceptable model fits: χ2
(48)
=144.057, p
