Abstract
This study tested whether cumulative support (i.e., parental, interparental, peer, and teacher) was associated with decreases in youths’ loneliness and social anxiety during early adolescence. The unique associations of support sources and moderating effects of youth gender also were examined. Participants included 416 youth in sixth grade at the beginning of the study (mean age = 11.86, SD = .69; 91% European American). Cumulative support during sixth grade was associated with decreased loneliness and social anxiety across early adolescence, and it was found that the association with decreased social anxiety was stronger for boys than girls. Parental and peer support were uniquely associated with decreased loneliness and social anxiety, respectively. The protection offered by multiple sources of support for developmental tasks across early adolescence is discussed.
Socially supportive relationships offer protective benefits for individuals’ well-being across the life span (Chu, Saucier, & Hafner, 2010; Thoits, 2011), which may be particularly important during early adolescence when youth experience various transitions (e.g., school, peer networks, and cognitive and pubertal development; Steinberg et al., 2006) and heightened feelings of loneliness (Ladd & Ettekal, 2013) and social anxiety symptoms (Hale, Raaijmakers, Muris, Van Hoof, & Meeus, 2008). For some youth, feeling lonely or anxious around peers is normative and temporary, whereas for others these feelings are chronic and conflict with accomplishing important developmental tasks. Recent studies found that adolescents who experienced higher levels of loneliness were more likely to experience depressive symptoms and suicidal thoughts (Lasgarrd, Goosens, Bramsen, Trillingsgaard, & Elklit, 2011; Vanhalst et al., 2012). In addition to experiencing higher levels of depressive symptoms, socially anxious youth were more likely to report greater difficulty in peer relationships relative to their nonsocially anxious peers (e.g., deficits in social skills and conflict resolution; Epkins & Heckler, 2011; Kingery, Erdley, Marshall, Whitaker, & Reuter, 2010). Thus, identifying factors that protect against increases in loneliness and social anxiety is critical, and we propose that social support across family, peer, and school domains may be important sources of protection.
This study was guided by Call and Mortimer’s (2001) Arenas of Comfort Theory. A key proposition of arena of comfort theory that will be tested in this study is that accumulation of social support across multiple domains (e.g., family, peer, and school) offers youth protection through mechanisms such as self-enhancement, connectedness, and sense of mattering to others that, in turn, bolsters youths’ socioemotional well-being (Call & Mortimer, 2001; Rosenberg & McCullough, 1981). Despite the theoretical and developmental salience, there is a relative paucity of research examining the importance of cumulative support for adolescents’ socioemotional well-being. As such, the primary goal of the current study was to investigate the associations between multiple sources of social support (i.e., cumulative support) and changes in youths’ loneliness and social anxiety. Additional goals were addressed to examine the specificity of particular findings and their associated inferences, namely, (a) whether individual sources of social support have unique associations with changes in adolescent loneliness and social anxiety and (b) whether these various associations varied across boys and girls.
Conceptualization of loneliness and social anxiety
Although correlated, loneliness and social anxiety are distinct aspects of adolescent socioemotional distress. Adolescent loneliness is defined as the aversive experience that results when an individual’s social network is “quantitatively or qualitatively deficient” (Perlman & Peplau, 1981, p. 31), whereas social anxiety symptoms are defined as a continuum of feelings that range from worry to intense fear about negative evaluations and avoidance of social situations with age-mates (La Greca & Stone, 1993). Correlations between social anxiety and loneliness have ranged from .35 to .56, indicating shared variance of about 12% to 25% (Johnson, LaVoie, & Mahoney, 2001; Mahon, Yarcheski, Yarcheski, Cannella, & Hanks, 2006; Storch & Masia-Warner, 2004). Although some studies have examined the role of social support for loneliness or social anxiety (La Greca & Lopez, 1998; Tillfors, Persson, Willén, & Burk, 2012), relatively fewer studies have examined these outcomes within the same study (Johnson et al., 2001; Stednitz & Epkins, 2006), and no studies were identified that examined these two outcomes in relation to multiple sources of support across early adolescence.
Cumulative social support and adolescent adjustment
We adopted Chu, Saucier, and Hafner’s (2010) definition of social support that they extracted from Cohen (2004). In this meta-analysis of the association between social support and child and adolescent well-being, social support was defined as “the provision of both psychological and material resources with the intention of helping the recipients cope with stress” (p. 625). In the arena of comfort theory, Call and Mortimer (2001) defined sources of support as arenas of comfort that are soothing, accepting, and enduring qualities of youths’ relationships. Conceptually, arenas of comfort represent quality relationships, in which youth experience feelings of connection, openness, and trust. Moreover, youth do not need to worry about self-presentation within these relationships. A major tenet of the arena comfort theory is that youth who experience comfort across life domains (e.g., family, peer, and school) are provided with greater protection against the stress of transitions that occur during early adolescence. As such, multiple sources of social support across life domains act as coping resources to help youth efficaciously deal with stressors in multiple spheres of their lives (Call & Mortimer, 2001).
Previous research has provided tangential evidence for the theoretical proposition that adolescents who experience social support in multiple domains of their lives are better adjusted than other youth (Call & Mortimer, 2001; Levitt et al., 2005; Rueger, Malecki, & Demaray, 2008). Call and Mortimer (2001) found that multiple sources of social support were associated with improved adolescent adjustment, including higher personal well-being and positive self-concept, and lower depressive symptoms across 1 year of high school. In a sample of pre- and early adolescents, Levitt et al. (2005) found that youth receiving support from both close family and extended family or friends reported more positive self-concepts and less loneliness and internalizing symptoms 1 year later than those with fewer sources of support. In a cross-sectional study, Rueger, Malecki, and Demaray (2008) also found that cumulative support was associated with fewer internalizing symptoms and greater adaptive social skills.
Unique sources of support and adolescent loneliness and social anxiety
An important context in which adolescents may experience support is within their family (Chu et al., 2010). The most proximal source of support, parental support, has consistently predicted better adolescent adjustment (De Goede, Branje, Delsing, & Meeus, 2009; Swanson, Valiente, Lemery-Chalfant, & O’Brien, 2011) and, in particular, less loneliness (van Roekel, Scholte, Verhagen, Goossens, & Engels, 2010) and social anxiety (Hutcherson & Epkins, 2009). Rueger, Malecki, and Demaray (2010) found that parental support was uniquely associated with adolescents’ psychological adjustment, above and beyond other support domains. Chu et al. (2010) found that the average, significant effect size between familial social support and child and adolescent well-being was .19.
Interparental support, that is, the support that spouses provide to and receive from each other, is another familial source of support that may impact adolescents’ socioemotional functioning. Davies, Cummings, and Winter (2004) proposed that positive interactions in the interparental system can serve as a source of support and contribute to a sense of emotional security for the adolescent within the family context and thus improved adolescent functioning. Colarossi and Eccles (2000) found that interparental support was associated negatively with daughters’ depressive symptoms and positively with self-esteem, however, no studies were identified that examined the association between interparental support and adolescent loneliness or social anxiety.
Extrafamilial sources of support represent more distal, albeit important, support domains as adolescents are spending increasing time with peers and teachers (Reddy, Rhodes, & Mulhall, 2003). Previous studies have found evidence for the protective effect of peer support for adolescent well-being (Call & Mortimer, 2001), depressive symptoms and self-esteem (Colarossi & Eccles, 2003), and anxiety (social anxiety, Epkins & Heckler, 2011; general anxiety, Rueger, Malecki, & Demaray, 2010; and social phobias, van Oort, Greaves-Lord, Ormel, Verhulst, & Huizink, 2011). La Greca and Harrison (2005) also found that adolescents who reported having more positive interactions (e.g., approval) with best friends had fewer social anxiety symptoms; less is known, however, about whether peer support protects adolescents from loneliness. Chu et al. (2010) found that the average, significant effect size between peer support and child and adolescent well-being was .10.
With regard to teachers, Roeser, Eccles, and Sameroff (1998) found that adolescents who viewed their teachers as providing both academic and emotional support were less likely to experience emotional distress and alienation from school. Teacher support also predicted decreased depressive symptoms and improved self-esteem during middle school (Reddy et al., 2003). Other studies found that perceived support from teachers was associated with higher levels of mastery (Call & Mortimer, 2001) and academic achievement (Malecki & Demaray, 2003). Chu et al. (2010) found that the average, significant effect size between teacher and school support and child and adolescent well-being was .20.
The potential role of gender
Previous research has reported mean-level differences in social support with girls tending to report more received social support than boys (Call & Mortimer, 2001), such as greater support from classmates and close friends (Rueger et al., 2008). Yet girls also report higher levels of social anxiety symptoms than do boys (Epkins & Heckler, 2011; Tillfors et al., 2012). In contrast, studies typically have not found evidence of gender differences in adolescent loneliness (Mahon et al., 2006; van Roekel et al., 2010). Despite the potential mean-level differences in social support and social anxiety, however, little research is available that addresses whether associations between support domains and adolescent loneliness and social anxiety vary across boys and girls. Chu et al. (2010) found some evidence of a stronger association between social support and child and adolescent well-being for girls than for boys.
Of the few studies that examined the moderating effect of adolescent gender on multiple sources of support and adolescent internalizing symptoms, findings have been mixed. For example, Call and Mortimer (2001) found that the association between cumulative sources of support and internalizing symptoms was stronger for girls than boys (i.e., depressive symptoms and self-derogation), whereas Rueger and colleagues (2008) found that cumulative support was significantly associated with boys’ anxiety and unrelated to girls’ anxiety. Inconsistent patterns across studies also were found regarding the moderating role of adolescent gender on the association between unique sources of support and internalizing symptoms: (a) with stronger effects found for girls than boys across parent, peer, and teacher support (Call & Mortimer, 2001) and (b) none of the sources of support (parent, peer, and teacher) were uniquely associated with boys’ or girls’ anxiety (Rueger et al., 2008, 2010). Given little theory and research to draw upon specific to adolescent loneliness and social anxiety, we examined the moderating effect of gender as an exploratory step across analyses and did not propose explicit hypotheses.
Present study
Using a multi-method and multi-informant design, we extend previous research by examining associations between cumulative and unique sources of social support and changes in adolescent loneliness and social anxiety across early adolescence. First, guided by the arena of comfort theory, we examined whether having cumulative support was associated with decreases in loneliness and social anxiety. The advantage of using a cumulative support index, which is an unweighted score, is that we were able to test the idea that accumulation of support across multiple domains of adolescents’ lives (e.g., family, peer, and school) matters for adolescent adjustment (Call & Mortimer, 2001). Second, to test the relative importance of each source of support, we examined whether parental, interparental, peer, and teacher support were uniquely associated with changes in loneliness and social anxiety, above and beyond support from other sources. Finally, we examined whether these associations vary across girls and boys.
Method
Sampling procedure and characteristics
The sample was drawn from a relatively large cross-sectional study of family life during the transition into adolescence. Participants included sixth-grade adolescents from 13 middle schools in a large, geographically diverse county in the southeastern U.S. Ninety-six percent of teachers participated. Letters inviting adolescents’ participation were distributed during homeroom and two additional invitations were mailed directly to parents. Of the 71% who returned consent forms, 80% of these adolescents had parental permission to complete a questionnaire on family life during school, resulting in a sample of 2,346 sixth graders. Youth resembled the demographic profile of the county (contact author for statistical details).
Families were recruited from this larger study for the longitudinal study if (a) parents were married or long-term cohabitants and (b) no stepchildren were in or out of the home. Of the 1,131 eligible families from the larger study, 416 (37%) agreed to participate. Time constraints or unwillingness to be videotaped were primary reasons for nonparticipation. Studies using similar intensive procedures with three family members had comparable response rates (Sweet & Bumpass, 2005, National Survey of Families and Households (NSFH): 34%; Updegraff et al., 2004, NSFH: 37%). Preliminary analyses indicated that eligible participating families were similar to eligible nonparticipating families on all study variables, suggesting minimal selection bias (contact author for statistical details).
At Wave 1 (W1), when adolescents were in the sixth grade, they ranged in age from 11 to 14 years (M = 11.86 years, SD = 0.69) and 51% (n = 211) were girls. The majority of families (91%) were European American and 3% were African American. This percentage (3%) is lower than the percentage of married African American couples with their own children younger than 18 years in the county (5%) and in the U.S. (7.8%; U.S. Census, 2000a, table PCT27 of summary file 4). In this sample, the average level of parents’ education was an associate’s degree. Educational attainment was similar to that of European American adults in the county who were older than 24 years (county mean category was some college, no degree; U.S. Census, 2000c, table P148A of summary file 4). The median level of 2001 household income for study families was about US$70,000, which was higher than the median 1999 income for married couple families in the county (US$59,548; U.S. Census, 2000b, table PCT40 of summary file 3; US$64,689 inflation-adjusted dollars through 2001).
Data collection procedures
Adolescents completed a questionnaire about family life at school during sixth grade, which, in this study, included the Achenbach Youth Self-Report Survey (YSR; Achenbach, 1991) and the Children’s Report of Parental Behavior Inventory (CRPBI; Schaefer, 1965). This questionnaire completion was done during the mornings in the cafeteria in groups of 40–50 students. Typically, seven research assistants were available to answer questions. Youth were treated to a pizza party following data collection. Questionnaires also were mailed home and family members were asked to complete their booklets independently. Questionnaires were collected during a home visit. Observational tasks included a semi-structured problem-solving discussion activity and a marital discussion task. The problem-solving task involved the mother, father, and adolescent and focused on trying to solve issues of contention selected by family members (28-item Issues Checklist; Melby, Ge, Conger, & Warner, 1995). The second, 20-min task included only the wife and husband and focused on the marital relationship, daily interaction patterns, and co-parenting. Interactions were videotaped and rated by trained coders using the Iowa Family Interaction Rating Scales (Melby & Conger, 2001). Within each family, different coders rated the interaction from the two tasks to minimize coder carryover effects.
As part of the longitudinal design, assessments were conducted again 1 year later (W2) and 2 years later (W3). Most adolescents were in the seventh grade at W2 (M = 12.84 years, SD = 0.68) and in the eighth grade at W3 (M = 13.83 years, SD = 0.67). There were 366 participating families at W2 and 340 families at W3 (82% retention of W1 families). Attrition analyses using multivariate analysis of variance were conducted using the W1 data, and there were no differences between the retained and attrited families on any of the study variables. Families were paid US$100 for their participation in W1, US$120 for W2, and US$135 for W3; funds were allocated to parents to divide among family members who participated. Parental consent and adolescent assent were obtained at enrollment of the study, and any time data were collected.
Measurement
Adolescent loneliness and social anxiety were measured at W1 (sixth grade) and 2 years later, at W3 (eighth grade). Baseline controls were included in the analytic models to examine change in adolescent loneliness and social anxiety across early adolescence. Social support domains were assessed at W1 of the study with the exception of peer support, which was not measured during this wave; thus peer support was measured at W2 of the study (seventh grade).
Adolescent loneliness
Loneliness was measured by a latent construct using three manifest indicators. Youth completed these measures. Two parcels were created by randomly assigning 4 items to each parcel from the 8-item, UCLA Loneliness Scale (Russell, Peplau, & Ferguson, 1978). One sample item from each parcel was “I am no longer close to anyone” and “I lack companionship” and the 4-point response scale ranged from 1 (never) to 4 (a lot). Internal consistencies for Parcel 1 (α = .71, W1; α = .82, W3) and Parcel 2 (α = .67, W1; α = .77, W3) were adequate. Loneliness also was measured using a single item, “I feel lonely,” which was taken from the Achenbach YSR survey completed at W1 and W3 (Achenbach, 1991). This YSR item had a 3-point response format, which ranged from 0 (not true) to 2 (very true or often true).
Parcels are an aggregate-level indicator comprising the average of two or more items (Little, Cunningham, Shahar, & Widaman, 2002). With regard to procedures used, 4 items from the loneliness measure were randomly assigned to each parcel and then the items were averaged for Parcel 1 and Parcel 2, respectively. The same 4-item parcels were created as indicators for adolescent loneliness at W1 and W3. The benefits of using parcels in latent variable modeling include reduced bias of the effects of measurement error, more valid parameter estimates, and simplification of models by reducing the number of parameters to be estimated (Coffman & MacCallum, 2005). Additionally, parcels allowed for a measurement design with three indicators of loneliness when no other substantive indicators were available.
Adolescent social anxiety symptoms
Social anxiety symptoms were represented by a latent construct with two manifest variables. Adolescents’ social anxiety symptoms were measured at W1 and W3 via adolescent reports of their fears of negative evaluation (3-item subscale) and social avoidance of new peers (3-item subscale) using the Social Anxiety Scale for Children–Revised Questionnaire (La Greca & Stone, 1993). For both subscales, youth indicated the frequency of social experiences on a 5-point response format ranging from 1 (not at all) to 5 (all the time). A sample item from the fear of negative evaluation subscale was “I worry about what other kids think of me.” A sample item from the social avoidance subscale was “I feel shy around kids I don’t know.” Internal consistency of the fear of negative evaluation (α = .90, W1; α = .91, W3) and social avoidance measures (α = .81, W1; α = .84, W3) were high.
Cumulative support index
Similar to measurement strategies used in cumulative risk research (Gerard & Buehler, 2004), we created a cumulative support index by summing dichotomized social support domains that have been theorized to act as protective factors. Each domain-specific measure of social support was standardized (multiple measures were used in each domain to increase content validity) and then indicators were averaged to create the respective social support domain (e.g., parental support). Then, frequencies were examined for each social support domain and the top 75th quartile cutoff represented the presence of social support whereas the bottom 25th quartile represented a lack of support. The top 75% of the distribution received the value of “1” indicating support from a given domain. Dichotomized social support domains were then summed to create the cumulative social support index that could range from 0 (i.e., bottom quartile in each of the four support domains) to 4 (i.e., top 75th percentile in each of the four support domains).
Parental support
Parental support was measured at W1 with (a) youth report of each parent and (b) maternal and paternal reports using the 10-item acceptance subscale of the CRPBI (Schaefer, 1965). A sample item for the youth measure was “My mother (father) is a person who gives me a lot of care and attention.” The response format ranged from 1 (not like her/him) to 3 (a lot like her/him). Because the construct represented parental support rather than support from just one parent, youth reports of maternal and paternal support were averaged to create a summary score (α = .88). The same 10 items were used for parental reports of their own behavior toward their child. A sample item was “I am a person who believes in showing love for this child.” The response format ranged from 1 (not like me) to 3 (a lot like me). Internal consistency estimates of maternal (α = .76) and paternal report (α = .83) of support were adequate. Parental support at W1 also was measured using six observational ratings from the Iowa Family Interaction Rating Scale (Melby & Conger, 2001). During the family problem-solving task, mothers and fathers were coded based on ratings of listener responsiveness, communication, and prosocial behavior shown toward the youth. The 9-point coding format ranged from 1 (not at all characteristic) to 9 (mainly characteristic). Observational ratings of mother’s and father’s behavior toward youth were averaged together to create a composite parental support score (α = .80). Twenty percent of the interactions were coded by two coders, and the percentage agreement was greater than .80 for the ratings and the intraclass correlations (ICCs) ranged from .50 to .55, which is acceptable for these rating scales and consistent with previous research (Franck & Buehler, 2007; Melby & Conger, 2001).
Interparental support
Interparental support was measured at W1, with wife and husband report using the 8-item warmth subscale from the Iowa Youth and Family Project (Melby & Conger, 2001). Each spouse reported support given to and received from their partner during the past month. A sample item asked partners to indicate whether they “act supportive and understanding toward their spouse.” The response format ranged from 1 (never) to 7 (always). Wives’ self-reports and spouse reports (α = .82) were averaged to create a composite score, as were husbands’ self-reports and spouse reports (α = .85). Higher scores indicated greater support. Interparental support also was measured using six observational ratings from the Iowa Family Interaction Rating Scale (Melby & Conger, 2001). Ratings of listener responsiveness, communication, and prosocial behavior shown toward their spouse during the marital discussion task were rated (α = .90). The 9-point coding format ranged from 1 (not at all characteristic) to 9 (mainly characteristic). Interrater agreement was greater than .80 and the ICCs averaged .50, which is acceptable for these rating scales and consistent with the previous research (Melby & Conger, 2001).
Peer support
Peer support was assessed using the 5-item friend support scale measured at W2 (Richman & Bowen, 1997). A sample item was “I can trust my friends.” The response format ranged from 0 (not like me) to 2 (a lot like me). Items were averaged and higher scores indicated greater support from close peers (α = .79). Peer support at W2 also was measured with youth self-report of best friendships using the 7-item children’s friendship scale (Berndt & Perry, 1986; Vernberg, Abwender, Ewell, & Beery, 1992). A sample item was “when you do a good job on something how often does this friend praise and congratulate you?” The response format ranged from 1 (never) to 5 (every day). Items were averaged (α = .84), and higher scores indicated greater support.
Teacher support
Teacher support was measured by youth report using the 4-item teacher support and satisfaction scale (Richman & Bowen, 1997). A sample item was “How many of your teachers believe you can do well in school?” The response format ranged from 1 (none) to 5 (all). Items were averaged (α = .77), and higher scores indicated greater perceived support from teachers. Teacher support also was assessed by comfort in adolescents’ interactions with teachers using two, single-item questions (Barber, personal communication, “How comfortable do you feel talking to your teachers in class” and “How comfortable do you feel talking to your teachers outside of class?”). The response format ranged from 1 (not comfortable) to 3 (very comfortable). These single-item measures allowed for an adequate number of manifest indicators of the latent construct of teacher support.
Analytic procedures
Hypotheses were tested using structural equation modeling (AMOS 20). Model fit was assessed by examining the model χ2 statistic, the comparative fit indices (CFIs), and the root mean square error of approximation (RMSEA). A good fit is typically indicated by a nonsignificant χ2 statistic, CFI values greater than .95, and RMSEA values less than .05; however, with relatively large sample sizes, significant χ2 statistics often point to only small model misspecifications (Hu & Bentler, 1999). An adequate fitting model includes CFI values of .90 and RMSEA values between .06 and .08 (Browne & Cudeck, 1993; Hu & Bentler, 1999).
We also examined the moderating effects of gender across the models. Using multigroup structural equation modeling, a model was estimated across boys and girls with all of the statistical parameters constrained to equality. This analysis provided a χ2 estimate of the overall fit of the constrained model. Then, a new model was estimated in which the regression paths (i.e., structural paths) were allowed to vary across boys and girls. The second analysis provided a χ2 estimate of the overall fit for the “partially freed” model across boys and girls. These two χ2 estimates were compared using the χ2 difference test (i.e., constrained versus the partially freed model). A significant difference in the χ2 estimates indicated that one of the structural paths differed across boys and girls. Critical ratio estimates were used to identify the specific path coefficients that were different; critical ratios greater than 1.96 were statistically significant (p < .05).
There were little missing data within each wave (less than 3%). Missing data within and across waves (i.e., attrition) were addressed using full information maximum likelihood (FIML) estimation methods. FIML method was used because it produces less biased estimates than other methods such as imputing the sample mean or dropping cases for data missing within and across waves (Newman, 2003).
Results
Of the 416 adolescents and using this set of measurement procedures, 39% (n = 164) had support from all four domains, 35% (n = 147) had support from three domains, 17% (n = 71) had support from two domains, 7% (n = 28) had support from one domain, and 1% (n = 6) had no support from any of these domains. The mean cumulative support score was 3.09 (SD = .92).
Cumulative support change model
To test whether cumulative support was associated with decreases in adolescent loneliness and social anxiety, we estimated a model with W1 baseline controls for loneliness and social anxiety (Figure 1). The associations between (a) W1 loneliness and W3 social anxiety and (b) W1 social anxiety and W3 loneliness also were controlled. The model fit to these data was good, χ2(35) = 68.40, p = .001; CFI = .98; RMSEA = .048. As hypothesized, cumulative support was associated with decreases in loneliness (β = −.16, p = .004) and social anxiety (β = −.14, p = .017). These results indicate that adolescents who reported experiencing greater support across the four support domains experienced decreased loneliness and social anxiety across early adolescence.
Direct effects model of cumulative support predicting loneliness and social anxiety. Model fit, χ2(35) = 68.40, p = .001; CFI = .98; RMSEA = .05. The model differed across boys and girls, Δχ2(2) = 6.17, p < .05, with boys’ unstandardized estimate displayed first, followed by girls for the pathway from cumulative support to social anxiety. Standardized estimates are displayed in parentheses. Dotted lines indicate nonsignificant pathways and solid lines indicate significant pathways. Boldface estimates are significant at p < .05. P1: Parcel 1; P2: Parcel 2; Obs: observed; YSR: Achenbach Youth Self-Report; SAS: Generalized Social Avoidance and Distress subscale; FNE: Fear of Negative Evaluation; CFI: comparative fit index; RMSEA: root mean square error of approximation.
Moderating effects of youth gender
The model differed across boys and girls, Δχ2(2) = 6.17, p < .05; the pathway between cumulative support and decreased social anxiety was stronger for boys (b = −.23, p < .001) than for girls (b = −.16, p < .001), although statistically significant for both. The pathway between cumulative support and changes in loneliness was not moderated by youth gender (boys: b = −.14, p < .001; girls: b = −.12, p < .001).
Unique sources of support
As a follow-up test to the cumulative support model, we examined a model testing the specific unique pathways from parental, interparental, peer, and teacher support to adolescent loneliness and social anxiety across time. The model fits these data well, χ2(182) = 320.66, p = .00; CFI = .95; RMSEA = .04 (Figure 2). Parental support was uniquely associated with decreased loneliness across early adolescence (β = −.49, p = .008). Peer support was uniquely associated with decreased social anxiety across early adolescence (β = −.20, p = .03).
Direct effects model with latent constructs predicting loneliness and social anxiety from multiple social support domains. Model fit, χ2(182) = 320.66, p = .00; CFI = .95; RMSEA = .04. Standardized coefficients are displayed in parentheses. Boldface estimates are significant at p < .05. P1: Parcel 1; P2: Parcel 2; Obs: observed; Peer: close peer; BF: best friend. YSR: Achenbach Youth Self-Report. SAS: Generalized Social Avoidance and Distress subscale; FNE: Fear of Negative Evaluation.
Descriptive Statistics and Correlations.
Note. P1 = Parcel 1; P2 = Parcel 2; YR = Youth report; YSR = I feel lonely; SAS = Generalized Social Avoidance and Distress Subscale; FNE = Fear of Negative Evaluation; Par Sup = Parental support; MR = Maternal report; PR = Paternal report; Obs = Observed; Int Sup = Interparental support; WR = Wife report; HR = Husband report; Peer = Close Peer; BF = Best Friend; Tea sup = Teacher support; Comfort1 = How comfortable do you feel talking to your teachers in class? Comfort2 = How comfortable do you feel talking to your teachers outside of class? W1 = 6th grade, W2 = 7th grade, W3 = 8th grade.
*p < .05; **p < .01; ***p < .001.
Moderating effects of youth gender
The model examining associations between the individual domains of parental, interparental, peer, and teacher support and changes in adolescent loneliness and social anxiety did not differ across gender, Δχ2(8) = 12.20, p = .14.
Discussion
This study contributed to the literature by examining whether cumulative social support is associated with decreases in early adolescents’ loneliness and social anxiety. Drawing upon the arena of comfort theory and previous research on social support and adolescent adjustment (Call & Mortimer, 2001; Chu et al., 2010; Rueger et al., 2008), we hypothesized that having multiple sources of support is associated with decreased adolescent loneliness and social anxiety. Consistent with this expectation, we found that cumulative support was associated with decreased adolescent loneliness and social anxiety. With regard to the moderating role of youth gender, we found that the association between cumulative support and adolescents’ social anxiety was stronger for boys than for girls, whereas the association with loneliness characterized boys and girls similarly. The significant associations between unique sources of support and adolescent loneliness and social anxiety did not vary for boys and girls.
Overall, these results indicated that adolescents who perceived higher support across multiple contexts (family, peers, and school) experienced improved socioemotional functioning. Thus, quality social support, characterized primarily in the current study as perceived emotional support from their family, peer, and teacher relationships may provide youth with a greater sense of belonging, companionship, and comfort (Thoits, 2011), thereby decreasing youths’ feelings of loneliness and social anxiety. Similarly, Call and Mortimer (2001) found that adolescents with at least 2 or more supportive relationships reported higher self-esteem and well-being and lower depressive symptoms. Other studies also found that cumulative support is important for healthy adjustment during the transition from early to middle adolescence (Levitt et al., 2005) and we extend this research to loneliness and social anxiety.
In addition, we found that the association between cumulative support and social anxiety was stronger for boys than for girls. It appears that having multiple sources of support serves an important protective function, in general, but our results also indicate that boys experience greater protection against social anxiety than do girls in the context of cumulative support. Rueger and colleagues (2008) found that the association with cumulative support was significant for boys’ anxiety and unrelated to girls’ anxiety. We extend this correlational finding to suggest that cumulative support may be particularly important for boys’ social anxiety over time. This finding from the current study diverges from a previous study that found that the association between cumulative support and internalizing symptoms was stronger for girls than boys (Call & Mortimer, 2001). Possible reasons for the discrepancy are that Call and Mortimer’s study focused on a high-school-age sample and that social anxiety was not included in their analyses. Social anxiety differs from outcomes such as loneliness and depressive symptoms in that youths’ social relationships with peers and nonfamilial adults are more directly impacted by their social fears. Future studies should examine whether associations between social support and social anxiety vary across gender to see whether this finding holds across samples and additional social outcomes such as social skills and leadership capacities.
As a follow up to the test of cumulative support, we also examined the unique associations among parental, interparental, peer, and teacher support and adolescent loneliness and social anxiety. We found that parental support offered unique protection of decreases in youths’ loneliness across early adolescence, above and beyond support from additional domains. One explanation for this finding may be that parents are better able to detect their child’s behaviors as symptoms of loneliness. For example, adolescents may feel lonely for various reasons including separation from family or friends, feelings of sadness and emptiness, or longing for emotional connections with others (Russell et al., 1978). Given the proximal nature of parent–adolescent relationships, parents also may be more likely than other supportive relationships to provide additional opportunities for their child to be included and engage with others when they may have otherwise felt alienated or excluded. Previous scholars have argued that proximal relationships provide a more meaningful type of “emotional sustenance” that includes bolstering (a) an individual’s self-worth, (b) a sense of mattering to others, and (c) a belief that he or she is embedded within a network of supportive relationships (Cobb, 1976; Thoits, 2011; Uchino, 2009). As such, the protective effect of parental support for youths’ decreased loneliness (Hutcherson & Epkins, 2009; van Roekel et al., 2010) may be explained by their constant and directed support that includes both emotional (e.g., validation and acceptance) and instrumental support (e.g., scheduling time with friends or relatives).
Our effect size of parental support was stronger (.49 vsersus .19) than that reported by Chu et al. (2010), perhaps because of the strong measurement design used in the current study (e.g., self-report, parent report, and lab observation), whereas Chu and colleagues assessed studies using self-report only. Other potential reasons are differences in samples and the measurement of social support. This study included primarily European American adolescents (age range: 11–14 years) and primarily measures of perceived emotional support whereas Chu et al. included adolescents (age range: 3–20 years) from diverse racial and ethnic backgrounds and several types of support (i.e., perceived, size of network, enacted, and support seeking). Given these differences, comparisons across studies may be more limited; however, both studies suggest the importance of parental support for adolescents’ socioemotional functioning.
For social anxiety, peer support emerged as a unique predictor. One explanation for this association may be that there is an optimal match between source of support and youths’ developmental needs. Adolescents, for example, may be more likely to draw upon support from their peers to decrease their fears in social situations with age-mates as their peers tend to be available in these situations and offer an important source of validation. Similar to our results, Festa and Ginsburg (2011) found that social support, acceptance, and validation from a friend were associated with lower social anxiety among younger children. Other studies provided evidence for the important role of peers for youths’ social anxiety (Epkins & Heckler, 2011; van Oort et al., 2011). Existing literature also has shown the detrimental effects of a lack of peer support for youths’ increased social anxiety (Miers, Blöte, de Rooij, Bokhorst, & Westenberg, 2013; Tillfors et al., 2012). Moreover, our findings add to existing research by providing evidence of the unique, protective effect of youths’ peer support and decreased social anxiety, above and beyond other sources of support.
Although we argue that source of support is important in addition to the quality of the relationship, an alternative explanation of the unique associations of parental and peer support for loneliness and social anxiety, respectively, is that different functions of social support are provided by parents and peers. Given that we did not operationalize each source of support with the same measure, we cannot entirely rule out this possibility, however, examining the item content of support measures provided some evidence that the functions of emotional and instrumental support assessed in this study are distributed evenly across sources. An additional strength of this study is the multi-informant and multi-method design, which minimizes the shared method variance that could otherwise inflate associations. Nonetheless, future studies should examine whether the functionality of social support varies by social support providers (Munsch & Blyth, 1993).
Finally, we did not find that either interparental support or teacher support was uniquely associated with adolescent loneliness or social anxiety. One explanation may be that the amount of interparental support (support provided and received from each spouse) may not directly impact adolescents’ socioemotional functioning but may be indirectly linked to youths’ outcomes. Colarossi and Eccles (2000) found that interparental support was associated with youths’ outcomes indirectly through the parent–adolescent relationship. An alternative explanation may be that youths’ appraisals or beliefs about their parents’ marital relationship has a more direct impact on youths’ social anxiety or loneliness (Davies, Cummings, & Winter, 2004) or that youths’ perceptions of interparental support are more salient than observed or spouse-reported support. In terms of teacher support, one possible explanation for the null findings is that youth may spend less time with their teachers and experience less support as a result of transitioning into a new school system and changing classrooms throughout the school day. In accordance with this explanation, Martínez and colleagues (2011) found that teacher support decreased during the transition from elementary to middle school.
Similar to this study, previous research has reported fewer unique effects relative to the independent effects of each source of support (Rueger et al., 2010). That is, each source of support may bolster adolescents’ socioemotional functioning independent of other sources of support, whereas fewer sources of support may be uniquely associated with changes in adolescents’ socioemotional functioning above and beyond other sources of support. Our results also indicated that interparental and teacher support were important when combined with parental and peer support because cumulative support across these four sources of support bolstered adolescents’ socioemotional functioning during a developmental period of transition.
Limitations and future directions
Several limitations of this study also should be considered. This study included primarily European American adolescents, so it is not known whether these results would be similar across different ethnic groups. Future studies should examine variation within ethnic groups across multiple sources of support and adolescents’ socioemotional functioning, particularly as youth of color tend to experience additional sources of adversity (e.g., racism and discrimination) that are not experienced by their European American counterparts (Greene, Way, & Pahl, 2006). Additional research also should include adolescents from diverse families (e.g., unmarried parents and stepparents) to examine potential differences and further inform prevention work. Third, we were not able to examine other potentially important support domains such as siblings, extended family members, fictive kin, coaches, and religious leaders. Future research also may profit by examining the effects of cumulative support from a different theoretical perspective using a latent variable with indicators from multiple sources. Finally, this study emphasized the temporal precedence of social support for later loneliness and social anxiety, although reciprocal associations between social support and socioemotional functioning may be present (Storch & Masia-Warner, 2004).
Conclusion
Despite these limitations, this study extends current research by examining cumulative and unique associations of four important sources of support for adolescent loneliness and social anxiety. Consistent with the arena of comfort theory, protection provided by support from these various sources helped reduce the chances that youth would experience some of the debilitating effects that both loneliness and social anxiety have on their social and emotional development during this important developmental period. The findings also contributed to understanding of the importance of source of support for adolescents’ loneliness and social anxiety. That is, the unique importance of parental and peer support for adolescent loneliness and social anxiety, respectively.
Several implications of this study can inform prevention and intervention work. First, universal prevention programs aimed at improving socioemotional functioning should help youth establish supportive relationships across multiple domains of their lives. One approach used by community organizations is strengthening the partnership between families and schools while connecting youth and their families to additional community resources (Bandy & Moore, 2011). A second implication is that parents and adults working with youth should be mindful that boys also need intimate relationships and benefit from having a supportive network. Third, parents and peers are proximal socializing agents and key points for intervention to decrease youths’ loneliness and social anxiety. For example, Bandy and Moore (2011) found that intervention programs that focused on increasing socioemotional support and skill building (e.g., problem solving, social skills—conflict resolution and helping others) within these proximal relationships helped youth develop healthy relationships in other domains of their lives and improved youths’ socioemotional functioning. Another effective intervention technique is teaching youth and their parents cognitive coping strategies to prevent increases in socioemotional distress (e.g., fear, negative thoughts, and withdrawal behavior) and encourage youths’ participation in daily activities and relationships across important contexts of their lives (Stormont, Reinke, Herman, & Lembke, 2012). Moreover, findings from this study suggest that intervening on youths’ perceptions of support through techniques like cognitive reframing may be an important point of intervention to reduce feelings of social anxiety and loneliness, and increase feelings of connectedness and belonging within youths’ relationship network. All youth require supportive social relationships and youth, families, and schools each have a role in fostering healthy relationships for adolescents’ positive adjustment.
Footnotes
Funding
This research was supported by a grant from The National Institute of Mental Health, R01-MH59248 to the second author. We thank the staff of the Family Life Project for their unending contributions to this work and the youth, parents, teachers, and school administrators who made this research possible.