Journal of Adolescence 43 (2015) 159e170
Contents lists available at ScienceDirect
Journal of Adolescence journal homepage: www.elsevier.com/locate/jado
Parental attachment, self-control, and depressive symptoms in Chinese and Italian adolescents: Test of a mediation model Jian-Bin Li a, *, Elisa Delvecchio a, Adriana Lis a, Yan-Gang Nie b, c, Daniela Di Riso a a
Department of Developmental Psychology and Socialization, University of Padua, 35131, Italy School of Education, Guangzhou University, 510006, PR China c Cantonese Psychological and Behavioral Research Center of Guangzhou, 510006, PR China b
a r t i c l e i n f o
a b s t r a c t
Article history: Available online xxx
The current study investigated the relationship between parental attachment and depressive symptoms as well as the mediating effect of self-control in two different cultures. Samples were 1305 Chinese and 1327 Italian adolescents. They completed the Inventory of Parental and Peer Attachment, the Self-Restraint Subscale of the Adolescents' Self-Consciousness Scale, and the Children's Depression Inventory that assessed parental attachment, self-control, and depressive symptoms, respectively. Results showed that: (1) Few cultural differences in depressive symptom were observed. (2) Parental attachment and self-control were negatively related to depressive symptoms in both cultures. (3) Selfcontrol mediated the relations between parental attachment and depressive symptoms in both cultures. (4) The direct and indirect effects were invariant across cultures. In conclusion, parental attachment and self-control are important for adolescents' depressive symptoms in Chinese and Italian adolescents. © 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
Keywords: Depressive symptoms Self-control Parental attachment Adolescents
Depression is an emotional problem frequently occurring during adolescence (Buist, Dekovi c, Meeus, & van Aken, 2004), with an estimated one year prevalence of 4e5% in adolescents (Costello, Egger, & Angold, 2005; Costello, Erkanli, & Angold, 2006). It is a worldwide mental health problem that disturbs adolescents and is a major risk factor for suicide for this age group (Lopez, Mathers, Ezzati, Jamison, & Murray, 2006). Therefore, investigation of this issue is highly necessary and important. Positive parenting process (e.g., parental warmth, autonomy granting, parental support) have been found to be negatively linked with adolescents' depressive symptoms (Donnelly, 1999; Herman-Stahl, 1999; Vazsonyi & Belliston, 2006). Among a number of parenting factors, parental attachment is an important one. It has been discovered that secure parental attachment (or high attachment quality with parents) is negatively related to adolescents' depressive symptoms in different cultures (Delvecchio, 2013; Laible, Carlo, & Raffaelli, 2000; Zhang, Zhang, Yang, & Zhang, 2010). In addition to this direct association, some studies have also revealed that parental attachment influences depressive symptoms through different self-related variables (Delvecchio, 2013; Kenny, Moilanen, Lomax, & Brabeck, 1993; Lu & Zhang, 2008).
* Corresponding author. Via Venezia, 15, Padua, Italy. Tel.: þ39 366 4645 602. E-mail address: [email protected] (J.-B. Li). http://dx.doi.org/10.1016/j.adolescence.2015.06.006 0140-1971/© 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
160
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
However, there is scarce study that has directly compared the relationships between parental attachment, the self, and depressive symptoms in adolescents in different cultures. To fill the gap in the literature, the current study aimed to investigate the relationships between parental attachment and depressive symptoms as well as the possible mediating effect of self-control in Chinese and Italian adolescents. We chose self-control as mediator with theoretical and practical concerns. First, Bowlby's (1969) attachment theory assumes that secure attachment with caregiver may help children establish a positive internal working model of the self, which therefore reduces subsequent internalizing problems such as depression. Some studies have supported this model by examining the role of self-esteem (Delvecchio, 2013), self-efficacy (Lu & Zhang, 2008), and self-view (Kenny et al., 1993). However, the self is a complicated construct which includes not only self-evaluation but it also comprises self-experience and self-control (Nie, Li, Dou, & Situ, 2014). Therefore, investigation of the role of self-control may provide further support for the attachment theory. Second, some studies have revealed that self-control can be improved (e.g., Muraven, 2010) and thus may decrease subsequent undesired problems (e.g., Denson, Capper, Oaten, Friese, & Schofield, 2011). In this sense, increasing one's self-control may be beneficial to the intervention for adolescents' depressive symptoms. Relationships between parental attachment, self-control, and depressive symptoms Parental attachment refers to an affectional bond with parents (Bowlby, 1969). Interactions with attachment figures are stored in one's memory as mental representations of attachment figures' response (working models of others) and representations of the self (working models of self), which may impact subsequent emotional and behavioral problems (Bowlby, 1969, 1980). There is strong empirical evidence that insecure attachment with parents (or low attachment quality) leads to increased internalizing problems, such as depression (Armsden, McCauley, Greenberg, Burke, & Mitchell, 1990; Delvecchio, 2013; Kobak, Sudler, & Gamble, 1991; for a review see Greenberg, 1999; Groh, Roisman, van IJzendoorn, BakermansKranenburg, & Fearon, 2012). These findings consistently suggest that secure parental attachment is a protective factor of adolescents' depression symptoms. Self-control refers to the ability to change or override one's inner responses (e.g., thought, feelings, and impulses) to interrupt the tendencies of acting on undesired behaviors (Tangney, Baumeister, & Boone, 2004). DSM-IV lists a set of disorders (including mood disorders) whose hallmark feature focuses on poor self-control (cf. DeWall, Gilman, Sharif, Carboni, & Rice, 2012). Prior studies have consistently found that people with high self-control report fewer depressive symptoms (DeWall et al., 2012; Nie et al., 2014; Tangney et al., 2004). Some scholars have also revealed that adolescents with depressed mood show lower self-control than the ones without such symptoms (Cataldo, Nobile, Lorusso, Battaglia, & Molteni, 2005). Collectively, these findings suggest that self-control is negatively related to depressive symptoms. The relationship between parental attachment and self-control could be understood in two theories. First, as noted above, attachment theory proposes that early attachment may influence later emotional problems through internal working model of the self (Bowlby, 1969). Several variables related to internal working model are thought to be potential mediators in the relations between attachment and adolescents' problem behaviors, and some of them (e.g., emotion regulation) particularly center on self-control (Fearon, Bakermans-Kranenburg, van IJzendoorn, Lapsley, & Roisman, 2010). Previous study has demonstrated that through repeated successful interaction with caregivers, individuals may develop a sense of self-worth about themselves and better emotional control ability, which in turn helps to regulate inappropriate emotions in later life (Cassidy, 1994). Recent study has supported the view that emotional regulation may mediate the relationships between attachment to parents and depression in adolescents (Kullik & Petermann, 2013). The second framework is Gottfredson and Hirschi's (1990) self-control theory. Although this criminological theory focuses on the effect of low self-control on criminal and deviant behavior, it appears to be very appropriate to understand the relationship between parenting and self-control. The self-control theory assumes that parents' monitoring and recognizing of the child's undesired behaviors and disciplining such behaviors help increase the child's self-control ability (Gottfredson & Hirschi, 1990). Furthermore, this theory also considers that a close parentechild relationship (i.e., parental attachment) may promote parents to spend more time with their children so that they have more opportunities to socialize them. This may make parents' monitoring, recognizing and disciplining of the child's unwanted behaviors more effectively, which suggests that parental attachment is a distal precursor of self-control. Several findings have demonstrated that secure parental attachment and positive parenting process promote self-control ability and self-regulation behaviors during adolescence (Eisenberg et al., 2005; Vazsonyi & Belliston, 2007). Cultural similarities in the relationships between parental attachment, self-control, and depressive symptoms in adolescents Although little has been done to directly compare the complete associations between adolescents' parental attachment, self-control, and depressive symptoms in different cultures, some comparative researches have addressed the links between parenting process, self-control, and psychological adjustment (e.g., Vazsonyi & Belliston, 2006, 2007), which shed light on the present study. First, previous parallel studies have examined the role of parental attachment in depressive symptoms in different cultures and nations, finding that secure attachment with parents is negative associated with reduced depression (Delvecchio, 2013; Lu & Zhang, 2008). Furthermore, Vazsonyi and Belliston (2006) directly compared the influence of parenting process on
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
161
depressive symptom in adolescents in Hungary, Netherlands, Switzerland, and United States, finding that parenting process (i.e., closeness, support, monitoring, communication, conflict, and peer approval) were related to depression symptoms, and their magnitudes were highly similar across different nations. Second, drawing upon the self-control theory (Gottfredson & Hirschi, 1990), a number of studies have parallel examined the influences of parenting process on individuals' self-control in different cultures and nations, finding that positive parenting (e.g., parental attachment, parental monitoring) is positively related to self-control (Cretacci & Cretacci, 2012; Hay, 2001; Miller, Jennings, Alvarez-Rivera, & Lanza-Kaduce, 2009; Vazsonyi & Huang, 2010). Moreover, Vazsonyi and Belliston (2007) directly compared the effect of family process on self-control and deviance in adolescents from Hungary, Japan, Netherlands, Switzerland, and United States, revealing that associations between family process, low self-control, and deviance were highly similar across nations. Third, with respect to the relation between self-control and depressive symptoms, as far as we know, although there is no previous study that has directly compared this link across different cultural or national context, previous studies have found that adolescents' self-control is negatively related to depressive symptoms in different cultures (DeWall et al., 2012; Nie et al., 2014; Tangney et al., 2004). These findings suggest that adolescents' self-control is important for depressive symptoms in different cultures and different nations. The present study In the present study, we were interested in investigating two questions. The first question was the influence of adolescents' parental attachment on depressive symptoms and the mediating effect of self-control. The second question concerned about whether these direct and indirect associations showed similar pattern across cultures. We were aware that this study was explanatory because no prior study had examined this issue before. Regarding the first question, based on the relationships between parental attachment, self-control, and depressive symptoms reviewed above, it was reasonable for us to expect that secure parental attachment would be related to higher selfcontrol, which in turn would link with lower depressive symptoms. This expectation was aligned with previous studies which had demonstrated that self-regulation (or emotional regulation) partially mediated the relationships between parenting and depressive symptoms (Baker & Hoerger, 2012; Kullik & Petermann, 2013). Regarding the second question, we expected the mediating model would be equivalent across countries. Theoretically, attachment is considered to be a universal need to seek proximity, and failure to do so may lead to various internalizing and externalizing problems (Bowlby, 1969, 1973, 1980), as supported by recent studies (Fearon et al., 2010; Groh et al., 2012). Moreover, in self-control theory, it is also assumed that self-control is an important factor for people's adjustment independent of cultural context (Gottfredson & Hirschi, 1990). Empirically, prior cross-cultural studies have provided direct support that the relationships between parenting process and depressive symptoms (Vazsonyi & Belliston, 2006) and the link between parenting and self-control (Vazsonyi & Belliston, 2007) are highly similar across different cultures and nations. Furthermore, being able to control (or regulate) one's emotion (a crucial aspect of self-control) is also found to be linked with reduced depression in different cultures and countries (Bandura, Caprara, Barbaranelli, Gerbino, & Pastorelli, 2003; Cataldo et al., 2005; Nie et al., 2014). Taken together, these theoretical claims and empirical findings led us to expect that the observed direct and indirect links would show similar pattern across Chinese and Italian adolescents. In addition, previous studies have found that there are significant gender and age differences in parental attachment and depressive symptoms both in Chinese and Italian adolescents (Cheng, 1998; Delvecchio, 2013; Li, Delvecchio, Miconi, Salcuni, & Di Riso, 2014; Nie et al., 2014; Nolen-Hoeksema & Girgus, 1994; Tambelli, Laghi, Odorisio, & Notari, 2012; Tepper et al., 2008). Therefore, in order to rule out the possible demographic effects, we also tested whether there would be age and gender differences in parental attachment, self-control, and depressive symptoms. Collectively, we hypothesized that: (1) parental attachment and self-control would be negatively related to depressive symptoms; (2) self-control would mediate the association between parental attachment and depressive symptoms; and (3) the direct and indirect relationships would be invariant across Chinese and Italian adolescents. Method Participants A total of 2632 adolescents were recruited from middle and high schools with ages ranging from 12 to 16 years. Of these, 1305 Chinese adolescents (658 boys, 647 girls, M ¼ 14.07 years, SD ¼ 1.37 years) and 1327 Italian adolescents (623 boys, 704 girls, M ¼ 13.84 years, SD ¼ 1.46 years) were selected from Guangzhou (China) and Milan (Italy), respectively. Participants were collected in private and public cooperative schools which served mainly middle-class families (absolute SES, Hollingshead, 1975), with approximately similar basic quality of life, within urban and suburban school districts. All participants were residents of their respective countries, and the primary languages were Chinese and Italian for Chinese and Italian, respectively. About 93% of the families who received the leaflet agreed to participate in both countries. Those who refused to take part in the study indicated reasons such as lack of interest and concerns about sharing personal information. All participants indicated that they were from two-parent family and did not have psychological or psychiatric counselling over the past two years.
162
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
Measures Parental attachment The revised version of Inventory of Parent and Peer Attachment (IPPA-R) was used to measure participants' attachment to parents. This scale consists of 25 items rated on a 5-point scale (from “1 ¼ never” to “5 ¼ always”) assessing the magnitude of Trust, Communication, and Alienation separately towards mother and father with parallel wordings. Higher scores on Trust and Communication and lower score on Alienation indicate better parental attachment. Chinese and Italian participants answered the existing Chinese version and Italian version of IPPA-R, respectively (Li et al., 2014). Sample items include “I am angry at my mother” (reverse score) and “my mother respects my feelings”. This scale has been previously applied to Chinese (Song, Thompson, & Ferrer, 2009) and Italian adolescents (Pace, San Martini, & Zavattini, 2011), showing adequate psychometric properties. Prior studies have demonstrated the cross-cultural applicability of the IPPA-R in Chinese and Italian samples (Li et al., 2014). In this study, we carried out measurement invariance test to replicate this finding. Regarding maternal attachment, the three-factor model (i.e., Trust, Communication, and Alienation) showed adequate fit for Chinese (c2 (272) ¼ 1716.941, RMSEA ¼ .064, CFI ¼ .971) and Italian (c2 (272) ¼ 1824.328, RMSEA ¼ .066, CFI ¼ .980) samples. Models of configural invariance (c2 (544) ¼ 3553.660, RMSEA ¼ .065, CFI ¼ .970), weak invariance (c2 (569) ¼ 3752.980, RMSEA ¼ .065, CFI ¼ .968) and scalar invariance (c2 (594) ¼ 16.042, RMSEA ¼ .064, CFI ¼ .968) were demonstrated, respectively. As the values of the DCFI between two nested models (i.e., configural invariance and weak invariance, and weak invariance and scalar invariance) were both less than .01 and other fit indices showed good fit, measurement invariance could not be rejected (Cheung & Rensvold, 2002). The Cronbach's alpha for the overall scale of maternal attachment, maternal Trust, maternal Communication, and maternal Alienation were .91, .85, .81, and .69 respectively for Chinese and .92, .88, .83, and .77 respectively for Italian samples. With respect to paternal attachment, the same procedures of examination were applied. Similarly, the three-factor model indicated adequate fit for Chinese (c2 (272) ¼ 1921.322, RMSEA ¼ .068, CFI ¼ .970) and Italian (c2 (272) ¼ 1697.541, RMSEA ¼ .063, CFI ¼ .979) samples. Sequential examination of configural invariance (c2 (544) ¼ 3616.819, RMSEA ¼ .066, CFI ¼ .972), weak invariance (c2 (569) ¼ 3690.715, RMSEA ¼ .065, CFI ¼ .972) and scalar invariance (c2 (594) ¼ 3768.028, RMSEA ¼ .064, CFI ¼ .972) showed that paternal attachment was equivalent across Chinese and Italian samples, as the values of the DCFI between two nested models were less than .01 and other fit indexes were acceptable. The Cronbach's alpha for the overall scale of paternal attachment, paternal Trust, paternal Communication, and paternal Alienation were .91, .86, .83, and .71 respectively for Chinese and .92, .88, .84, and .73 respectively for Italian samples. Collectively, the present findings suggested that the IPPA-R was applicable in Chinese and Italian adolescents. Self-control The self-restraint subscale of Adolescent Self-Consciousness questionnaire (ASC) was utilized to assess respondents' self-control ability. This scale includes 11 items rated on a 5-point scale (from “1 ¼ not like me at all” to “5 ¼ like me very much”). A total score can be obtained by summing up all the items, and a higher score indicate greater self-control ability. This scale was developed in Chinese context (Nie et al., 2014) and has been validated in Italian adolescents, both showing good psychometric properties (Delvecchio, Mabilia, Lis, Mazzeschi, Nie, & Li, 2014; Delvecchio, Mabilia, Miconi, Chirico, & Li, 2015). Chinese and Italian participants answered the existing Chinese (Nie et al., 2014) and Italian (Delvecchio, Mabilia, Lis, et al., 2014; Delvecchio, Mabilia, Miconi, et al., 2015) version of this scale, respectively. Sample items include “I fail in overcoming my bad habits even though I have tried many times” and “I can control my emotion”. In the current study, the one-factor model showed good fit both in Chinese (c2 (44) ¼ 242.689, RMSEA ¼ .059, CFI ¼ .969) and Italian (c2 (44) ¼ 246.784, RMSEA ¼ .059, CFI ¼ .969) samples. A measurement invariance test demonstrated that configural invariance (c2 (88) ¼ 489.473, RMSEA ¼ .059, CFI ¼ .969), weak invariance (c2 (99) ¼ 511.136, RMSEA ¼ .056, CFI ¼ .968), and scalar invariance (c2 (110) ¼ 534.737, RMSEA ¼ .054, CFI ¼ .967) were obtained, as the values of the DCFI between two nested models were less than .01 and other fit indexes showed good fit, suggesting that the self-restraint subscale was invariant across cultures. The Cronbach's alpha of this scale was .79 and .65 for Chinese and Italian samples, respectively. In sum, this indicated that the self-restraint subscale was applicable in both countries. Depression The Children's Depression Inventory (CDI) was administered to subjects to assess their depressive symptoms. This instrument contains 27 items. Kovacs (1992) found that this scale included five dimensions e anhedonia (8 items), negative emotion (6 items), negative self-esteem (4 items), ineffectiveness (4 items), and interpersonal problem (4 items). Each item consists of three statements graded in severity (e.g., “0 ¼ I am sad once in a while”, “1 ¼ I am sad many times”, and “2 ¼ I am sad all the time”) and higher scores reflect more severe depressive symptoms. In this study, item 9 e an item about suicide ideation e was excluded in data analyses because it belongs to none of the above five subscales. CDI has been used both in Chinese and Italian adolescents and it shows good psychometric properties (e.g., Delvecchio, 2013; Nie et al., 2014). Sample items include “I hate/do not like/like myself” (reverse score) and “I feel like
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
163
crying everyday/many days/once in a while” (reverse score). Previous studies showed that the factor structure of the CDI was inconsistent in adolescents (Delvecchio, 2013; Poli, Sbrana, Marcheschi, & Masi, 2003). Therefore, we first investigated the five-factor structure as described in Kovacs's (1992) report. The five-factor structure showed adequate fit in Chinese (c2 (289) ¼ 1037.418, RMSEA ¼ .045, CFI ¼ .981) and good fit (c2 (287) ¼ 2377,134, RMSEA ¼ .074, CFI ¼ .961) in Italian sample.1 To test the measurement invariance, a serial of nested models were performed. Configural invariance2 (c2 (574) ¼ 3601.358, RMSEA ¼ .063, CFI ¼ .961), weak invariance (c2 (600) ¼ 4282.954, RMSEA ¼ .068, CFI ¼ .953), and strong invariance (c2 (626) ¼ 4427.575, RMSEA ¼ .068, CFI ¼ .952) were obtained sequentially. The Cronbach's alpha for the total scale, anhedonia, negative emotion, negative self-esteem, ineffectiveness, and interpersonal problem were .86, .69, .64, .60, .56, and .28 respectively for Chinese sample, and .88, .60, .73, .70, .59, and .26 respectively for Italian sample.3 Overall, these findings suggested that the CDI was generally applicable in Chinese and Italian adolescents. Procedures This research was part of collaboration project between the University of Padua (Italy) and Guangzhou University (China). Administration was carried out in compliance with the ethical standards for research outlined in the Ethical Principles of Psychologists and Code of Conduct (American Psychological Association, 2010). This study was approved by the Ethical Committee for psychological research from both universities. Participation in the study was solicited through leaflets. Written consents were first obtained from the heads of school and parents, and then participants provided their oral consent and were willing to participate. Voluntary participation and anonymity were highlighted and no monetary reward was given. Participants completed the instruments during regular class and they had the right to withdraw if they did not want to continue. Participants were asked to be open and honest and to refrain from sharing answers with others. Confidentiality was assured by replacing respondents' information with a numeric code. Data analysis The .05 significance level was used throughout the paper. Before analyzing the data, imputation of missing values for all variables were carried out for both samples using PASW Statistics (version 18.0), which was based on values observed in other cases which showed a similar response pattern. Subsequently, we used MANOVA to examine the cultural and gender differences in parental attachment and depressive symptoms, whereas ANOVA was employed to investigate the cultural and gender differences in self-control. The p value and partial eta squared (hp 2 ) were used to judge the meaningfulness of the significance. Specifically, meaningful significance referred to F ratio was significant (i.e., p < .05) and the value of hp 2 .01 and .02 respectively for univariate and multivariate analyses of variance (Cohen, 1992). Only results that reached meaningful significance were interpreted. Afterward, we used Mplus 7.0 to carry out structural equation model to examine the relations between parental attachment, self-control, and depressive symptoms, with possible confounding demographic variables controlled. A robust method of maximum likelihood (MLR) was used to obtain the fit index because the data were ordinal. The values of RMSEA and CFI were used to judge the model fit, with the values of RMSEA no larger than .08 and CFI no less than .90 indicating good fit (Bentler, 1990; Hu & Bentler, 1999). In addition, we drew 5000 bootstrap sample from the full data and used 95% confidence interval to determine the significance of the mediating effect.4 Significant mediating effect would be identified if the confidence interval excluded 0. Finally, we examined whether the hypothesized model was invariant across cultures by comparing the free model (where all main paths were freely estimated) and the constrained model (where all main paths were constrained to be equal). Because c2 has been criticized for its sensitivity to sample size, we used DCFI to determine whether the coefficients of main paths were invariant as previous study had recommended (Cheung & Rensvold, 2002). Invariance would be identified if DCFI was below the threshold of .01 and other index showed good fit (i.e., RMSEA .08 and CFI .90, Cheung & Rensvold, 2002).
1 For the Chinese sample, no errors were correlated, but for Italian sample, errors of item 20 and item 22, and errors of item 1 and items 20 were correlated. If without these correlations, then the fit index was c2 (289) ¼ 3180.862, RMSEA ¼ .087, CFI ¼ .946. We were aware that CDI might consist of other factor structures for normal adolescent sample, such as the six-factor model (Craighead, Smucker, Craighead, & Ilardi, 1998) and the three-factor model (Cole, Hoffman, Tram, & Maxwell, 2000). However, we tried these models and found that the fit index for these factor structures was even worse (RMSEA > .10). Considering that the present study mainly focused on the relationships between parental attachment, self-control, and depressive symptoms, we adopted the five-factor model as proposed by Kovacs (1992) but allowed these errors to be correlated. 2 We used the Italian model pattern (i.e., five-factor with errors correlated) as the baseline model. We tested this model pattern in Chinese sample, finding that this pattern also showed adequate fit in Chinese sample, c2 (289) ¼ 1005.094, RMSEA ¼ .044, CFI ¼ .982. This suggested that this model could be used as the baseline model for subsequent tests of measurement invariance. 3 We noticed that the reliability of the interpersonal problem subscale was very low. This is consistent with previous studies (e.g., Delvecchio, 2013) which have also found this subscale problematic. Because there were few items in this subscale and other scales, we considered that the total CDI in general was reliable for both samples based on the Cronbach's alpha of the total scale and intercorrelations between subscales as shown in Table 3. 4 When bootstrapping technique was applied, maximum likelihood method was used because MLR does not support bootstrapping technique in the current version of Mplus.
164
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
Results Demographic differences in parental attachment, self-control, and depressive symptoms MANOVA were performed to investigate cultural and gender differences in parental attachment and depressive symptoms and ANOVA was conducted to examine cultural and gender differences in self-control (Tables 1 and 2). Regarding maternal attachment, MANOVA revealed a significant multivariate main effect for culture (Wilk's l ¼ .914, F (3, 2626) ¼ 82.16, p < .001, hp 2 ¼ .086) and gender (Wilk's l ¼ .979, F (3, 2626) ¼ 18.57, p < .001, hp 2 ¼ .021), but not for their interaction (Wilk's l ¼ .998, F (3, 2626) ¼ 1.41, p ¼ .238, hp 2 ¼ .002). Given the significance of the overall test, the univariate main effects were examined. Results showed that Italian adolescents had substantially higher maternal trust (F (1, 2628) ¼ 229.90, p < .001, hp 2 ¼ .080), maternal communication (F (1, 2628) ¼ 195.13, p < .001, hp 2 ¼ .069) and lower maternal alienation (F (1, 2628) ¼ 90.53, p < .001, hp 2 ¼ .033) than Chinese adolescents. Girls showed substantially higher maternal communication (F (1, 2628) ¼ 29.53, p < .001, hp 2 ¼ .011) than boys. Although there was also significant gender difference in maternal alienation (F (1, 2628) ¼ 7.19, p ¼ .007, hp 2 ¼ .003), the effect size was too trivial to interpret. No significant interaction was found for maternal attachment. With respect to paternal attachment, results of MANOVA demonstrated a significant multivariate a meaningful main effect for culture (Wilk's l ¼ .967, F (3, 2626) ¼ 30.28, p < .001, hp 2 ¼ .033), but not for gender (Wilk's l ¼ .987, F (3, 2626) ¼ 11.70, p < .001, hp 2 ¼ .013) or for their interaction (Wilk's l ¼ .995, F (3, 2626) ¼ 4.83, p ¼ .002, hp 2 ¼ .005). Based on this, a subsequent univariate main effect was examined. Results showed that Italian adolescents had substantially higher paternal trust (F (1, 2628) ¼ 76.66, p < .001, hp 2 ¼ .028) and lower alienation (F (1, 2628) ¼ 54.25, p < .001, hp 2 ¼ .020) than their Chinese counterparts. Italian adolescents also showed higher paternal communication than Chinese adolescents (F (1, 2628) ¼ 22.78, p < .001, hp 2 ¼ .009), but the effect size did not reach the threshold (i.e., .01). As for gender difference, boys reported more paternal communication (F (1, 2628) ¼ 22.61, p < .001, hp 2 ¼ .009) and less paternal alienation (F (1, 2628) ¼ 6.72, p ¼ .010, hp 2 ¼ .003) than girls, but their effect sizes were too small to be noticed. There was no significant gender difference in paternal trust (F (1, 2628) ¼ 1.31, p ¼ .253, hp 2 < .001). Although interaction was significant for paternal trust (F (1, 2628) ¼ 4.96, p ¼ .026, hp 2 ¼ .002) and paternal alienation (F (1, 2628) ¼ 12.84, p < .001, hp 2 ¼ .005), the effect sizes were also too trivial to worth interpreting. No significant interaction was found for paternal communication (F (1, 2628) ¼ 1.09, p ¼ .297, hp 2 < .001). Regarding self-control, results of ANOVA demonstrated that the main effect of culture was not significant, F (1, 2628) ¼ .90, p ¼ .343, hp 2 < .001. The main effect of gender was significant, with girls reporting higher self-control than boys (F (1, 2628) ¼ 16.01, p < .001, hp 2 ¼ .006), but the effect size was too small to deserve interpretation. No significant interaction was found (F (1, 2628) ¼ 1.72, p ¼ .190, hp 2 < .001). As for depressive symptoms, results of MANOVA indicated a significant multivariate main effect for culture (Wilk's l ¼ .948, F (5, 2624) ¼ 29.00, p < .001, hp 2 ¼ .052). The multivariate main effect for gender was significant but the effect size was not substantial (Wilk's l ¼ .991, F (5, 2626) ¼ 5.02, p < .001, hp 2 ¼ .009). No significant multivariate interaction was found (Wilk's l ¼ .998, F (3, 2626) ¼ 1.30, p ¼ .261, hp 2 ¼ .002). Subsequently, univariate main effects were examined. Results showed that Chinese adolescents reported higher anhedonia (F (1, 2628) ¼ 8.51, p ¼ .004, hp 2 ¼ .003) and negative emotion (F (1, 2628) ¼ 5.86, p ¼ .016, hp 2 ¼ .002) than Italian adolescents, but these effect sizes were too small to explain. Chinese adolescents showed substantially lower negative self-esteem than their Italian counterpart (F (1, 2628) ¼ 39.33, p < .001, hp 2 ¼ .015). No significant cultural difference in ineffectiveness (F (1, 2628) ¼ .35, p ¼ .852, hp 2 < .001) or interpersonal problem (F (1, 2628) ¼ 2.83, p ¼ .093, hp 2 ¼ .001) was found. Regarding gender difference, among the five subscales, boys showed higher interpersonal problem than did girls (F (1, 2628) ¼ 15.22, p < .001, hp 2 ¼ .006), but this effect size was too small to be noticed. No significant gender difference in other subscales was found, and no significant interaction was found for any of the five subscales.
Table 1 Means and standard deviation of parental attachment, self-control, and depressive symptoms. Overall (N ¼ 2632)
Maternal trust Maternal communication Maternal alienation Paternal trust Paternal communication Paternal alienation Self-control CDI-anhedonia CDI-negative emotion CDI-negative self-esteem CDI-ineffectiveness CDI-interpersonal problem
Chinese boys (N ¼ 658)
Chinese girls (N ¼ 647)
Italian boys (N ¼ 623)
Italian girls (N ¼ 704)
M
SD
M
SD
M
SD
M
SD
M
SD
39.36 31.25 13.82 38.64 28.18 14.38 36.40 3.60 2.70 2.02 2.39 1.21
7.30 7.12 4.68 7.55 7.42 4.69 6.62 2.67 2.34 1.74 1.71 1.18
36.93 28.49 15.08 37.22 28.05 15.12 35.83 3.75 2.72 1.84 2.44 1.32
6.85 6.46 4.42 7.27 6.95 4.75 7.01 2.82 2.09 1.53 1.71 1.20
37.62 30.23 14.28 37.53 26.98 14.95 37.20 3.77 2.89 1.79 2.33 1.19
6.99 7.06 4.15 7.84 7.65 4.52 6.61 2.68 2.15 1.44 1.65 1.05
41.55 32.50 13.05 40.41 29.72 13.15 35.93 3.53 2.63 2.23 2.45 1.29
6.65 6.41 4.79 6.80 6.89 4.27 6.35 2.81 2.76 2.12 1.76 1.32
41.28 33.66 12.90 39.43 28.05 14.26 36.62 3.37 2.55 2.24 2.35 1.06
7.43 7.30 4.96 7.76 7.84 4.90 6.40 2.35 2.30 1.80 1.70 1.13
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
165
Table 2 Significance of demographic differences in parental attachment, self-control, and depressive symptoms. F Maternal trust Maternal communication Maternal alienation Paternal trust Paternal communication Paternal alienation Self-control CDI-anhedonia CDI-negative emotion CDI-negative self-esteem CDI-ineffectiveness CDI-interpersonal problem
country
229.90 195.13 90.53 76.66 22.78 54.25 .90 8.51 5.86 39.33 .04 2.83
(1,2628)
p
hp 2
F
Contents lists available at ScienceDirect
Journal of Adolescence journal homepage: www.elsevier.com/locate/jado
Parental attachment, self-control, and depressive symptoms in Chinese and Italian adolescents: Test of a mediation model Jian-Bin Li a, *, Elisa Delvecchio a, Adriana Lis a, Yan-Gang Nie b, c, Daniela Di Riso a a
Department of Developmental Psychology and Socialization, University of Padua, 35131, Italy School of Education, Guangzhou University, 510006, PR China c Cantonese Psychological and Behavioral Research Center of Guangzhou, 510006, PR China b
a r t i c l e i n f o
a b s t r a c t
Article history: Available online xxx
The current study investigated the relationship between parental attachment and depressive symptoms as well as the mediating effect of self-control in two different cultures. Samples were 1305 Chinese and 1327 Italian adolescents. They completed the Inventory of Parental and Peer Attachment, the Self-Restraint Subscale of the Adolescents' Self-Consciousness Scale, and the Children's Depression Inventory that assessed parental attachment, self-control, and depressive symptoms, respectively. Results showed that: (1) Few cultural differences in depressive symptom were observed. (2) Parental attachment and self-control were negatively related to depressive symptoms in both cultures. (3) Selfcontrol mediated the relations between parental attachment and depressive symptoms in both cultures. (4) The direct and indirect effects were invariant across cultures. In conclusion, parental attachment and self-control are important for adolescents' depressive symptoms in Chinese and Italian adolescents. © 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
Keywords: Depressive symptoms Self-control Parental attachment Adolescents
Depression is an emotional problem frequently occurring during adolescence (Buist, Dekovi c, Meeus, & van Aken, 2004), with an estimated one year prevalence of 4e5% in adolescents (Costello, Egger, & Angold, 2005; Costello, Erkanli, & Angold, 2006). It is a worldwide mental health problem that disturbs adolescents and is a major risk factor for suicide for this age group (Lopez, Mathers, Ezzati, Jamison, & Murray, 2006). Therefore, investigation of this issue is highly necessary and important. Positive parenting process (e.g., parental warmth, autonomy granting, parental support) have been found to be negatively linked with adolescents' depressive symptoms (Donnelly, 1999; Herman-Stahl, 1999; Vazsonyi & Belliston, 2006). Among a number of parenting factors, parental attachment is an important one. It has been discovered that secure parental attachment (or high attachment quality with parents) is negatively related to adolescents' depressive symptoms in different cultures (Delvecchio, 2013; Laible, Carlo, & Raffaelli, 2000; Zhang, Zhang, Yang, & Zhang, 2010). In addition to this direct association, some studies have also revealed that parental attachment influences depressive symptoms through different self-related variables (Delvecchio, 2013; Kenny, Moilanen, Lomax, & Brabeck, 1993; Lu & Zhang, 2008).
* Corresponding author. Via Venezia, 15, Padua, Italy. Tel.: þ39 366 4645 602. E-mail address: [email protected] (J.-B. Li). http://dx.doi.org/10.1016/j.adolescence.2015.06.006 0140-1971/© 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
160
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
However, there is scarce study that has directly compared the relationships between parental attachment, the self, and depressive symptoms in adolescents in different cultures. To fill the gap in the literature, the current study aimed to investigate the relationships between parental attachment and depressive symptoms as well as the possible mediating effect of self-control in Chinese and Italian adolescents. We chose self-control as mediator with theoretical and practical concerns. First, Bowlby's (1969) attachment theory assumes that secure attachment with caregiver may help children establish a positive internal working model of the self, which therefore reduces subsequent internalizing problems such as depression. Some studies have supported this model by examining the role of self-esteem (Delvecchio, 2013), self-efficacy (Lu & Zhang, 2008), and self-view (Kenny et al., 1993). However, the self is a complicated construct which includes not only self-evaluation but it also comprises self-experience and self-control (Nie, Li, Dou, & Situ, 2014). Therefore, investigation of the role of self-control may provide further support for the attachment theory. Second, some studies have revealed that self-control can be improved (e.g., Muraven, 2010) and thus may decrease subsequent undesired problems (e.g., Denson, Capper, Oaten, Friese, & Schofield, 2011). In this sense, increasing one's self-control may be beneficial to the intervention for adolescents' depressive symptoms. Relationships between parental attachment, self-control, and depressive symptoms Parental attachment refers to an affectional bond with parents (Bowlby, 1969). Interactions with attachment figures are stored in one's memory as mental representations of attachment figures' response (working models of others) and representations of the self (working models of self), which may impact subsequent emotional and behavioral problems (Bowlby, 1969, 1980). There is strong empirical evidence that insecure attachment with parents (or low attachment quality) leads to increased internalizing problems, such as depression (Armsden, McCauley, Greenberg, Burke, & Mitchell, 1990; Delvecchio, 2013; Kobak, Sudler, & Gamble, 1991; for a review see Greenberg, 1999; Groh, Roisman, van IJzendoorn, BakermansKranenburg, & Fearon, 2012). These findings consistently suggest that secure parental attachment is a protective factor of adolescents' depression symptoms. Self-control refers to the ability to change or override one's inner responses (e.g., thought, feelings, and impulses) to interrupt the tendencies of acting on undesired behaviors (Tangney, Baumeister, & Boone, 2004). DSM-IV lists a set of disorders (including mood disorders) whose hallmark feature focuses on poor self-control (cf. DeWall, Gilman, Sharif, Carboni, & Rice, 2012). Prior studies have consistently found that people with high self-control report fewer depressive symptoms (DeWall et al., 2012; Nie et al., 2014; Tangney et al., 2004). Some scholars have also revealed that adolescents with depressed mood show lower self-control than the ones without such symptoms (Cataldo, Nobile, Lorusso, Battaglia, & Molteni, 2005). Collectively, these findings suggest that self-control is negatively related to depressive symptoms. The relationship between parental attachment and self-control could be understood in two theories. First, as noted above, attachment theory proposes that early attachment may influence later emotional problems through internal working model of the self (Bowlby, 1969). Several variables related to internal working model are thought to be potential mediators in the relations between attachment and adolescents' problem behaviors, and some of them (e.g., emotion regulation) particularly center on self-control (Fearon, Bakermans-Kranenburg, van IJzendoorn, Lapsley, & Roisman, 2010). Previous study has demonstrated that through repeated successful interaction with caregivers, individuals may develop a sense of self-worth about themselves and better emotional control ability, which in turn helps to regulate inappropriate emotions in later life (Cassidy, 1994). Recent study has supported the view that emotional regulation may mediate the relationships between attachment to parents and depression in adolescents (Kullik & Petermann, 2013). The second framework is Gottfredson and Hirschi's (1990) self-control theory. Although this criminological theory focuses on the effect of low self-control on criminal and deviant behavior, it appears to be very appropriate to understand the relationship between parenting and self-control. The self-control theory assumes that parents' monitoring and recognizing of the child's undesired behaviors and disciplining such behaviors help increase the child's self-control ability (Gottfredson & Hirschi, 1990). Furthermore, this theory also considers that a close parentechild relationship (i.e., parental attachment) may promote parents to spend more time with their children so that they have more opportunities to socialize them. This may make parents' monitoring, recognizing and disciplining of the child's unwanted behaviors more effectively, which suggests that parental attachment is a distal precursor of self-control. Several findings have demonstrated that secure parental attachment and positive parenting process promote self-control ability and self-regulation behaviors during adolescence (Eisenberg et al., 2005; Vazsonyi & Belliston, 2007). Cultural similarities in the relationships between parental attachment, self-control, and depressive symptoms in adolescents Although little has been done to directly compare the complete associations between adolescents' parental attachment, self-control, and depressive symptoms in different cultures, some comparative researches have addressed the links between parenting process, self-control, and psychological adjustment (e.g., Vazsonyi & Belliston, 2006, 2007), which shed light on the present study. First, previous parallel studies have examined the role of parental attachment in depressive symptoms in different cultures and nations, finding that secure attachment with parents is negative associated with reduced depression (Delvecchio, 2013; Lu & Zhang, 2008). Furthermore, Vazsonyi and Belliston (2006) directly compared the influence of parenting process on
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
161
depressive symptom in adolescents in Hungary, Netherlands, Switzerland, and United States, finding that parenting process (i.e., closeness, support, monitoring, communication, conflict, and peer approval) were related to depression symptoms, and their magnitudes were highly similar across different nations. Second, drawing upon the self-control theory (Gottfredson & Hirschi, 1990), a number of studies have parallel examined the influences of parenting process on individuals' self-control in different cultures and nations, finding that positive parenting (e.g., parental attachment, parental monitoring) is positively related to self-control (Cretacci & Cretacci, 2012; Hay, 2001; Miller, Jennings, Alvarez-Rivera, & Lanza-Kaduce, 2009; Vazsonyi & Huang, 2010). Moreover, Vazsonyi and Belliston (2007) directly compared the effect of family process on self-control and deviance in adolescents from Hungary, Japan, Netherlands, Switzerland, and United States, revealing that associations between family process, low self-control, and deviance were highly similar across nations. Third, with respect to the relation between self-control and depressive symptoms, as far as we know, although there is no previous study that has directly compared this link across different cultural or national context, previous studies have found that adolescents' self-control is negatively related to depressive symptoms in different cultures (DeWall et al., 2012; Nie et al., 2014; Tangney et al., 2004). These findings suggest that adolescents' self-control is important for depressive symptoms in different cultures and different nations. The present study In the present study, we were interested in investigating two questions. The first question was the influence of adolescents' parental attachment on depressive symptoms and the mediating effect of self-control. The second question concerned about whether these direct and indirect associations showed similar pattern across cultures. We were aware that this study was explanatory because no prior study had examined this issue before. Regarding the first question, based on the relationships between parental attachment, self-control, and depressive symptoms reviewed above, it was reasonable for us to expect that secure parental attachment would be related to higher selfcontrol, which in turn would link with lower depressive symptoms. This expectation was aligned with previous studies which had demonstrated that self-regulation (or emotional regulation) partially mediated the relationships between parenting and depressive symptoms (Baker & Hoerger, 2012; Kullik & Petermann, 2013). Regarding the second question, we expected the mediating model would be equivalent across countries. Theoretically, attachment is considered to be a universal need to seek proximity, and failure to do so may lead to various internalizing and externalizing problems (Bowlby, 1969, 1973, 1980), as supported by recent studies (Fearon et al., 2010; Groh et al., 2012). Moreover, in self-control theory, it is also assumed that self-control is an important factor for people's adjustment independent of cultural context (Gottfredson & Hirschi, 1990). Empirically, prior cross-cultural studies have provided direct support that the relationships between parenting process and depressive symptoms (Vazsonyi & Belliston, 2006) and the link between parenting and self-control (Vazsonyi & Belliston, 2007) are highly similar across different cultures and nations. Furthermore, being able to control (or regulate) one's emotion (a crucial aspect of self-control) is also found to be linked with reduced depression in different cultures and countries (Bandura, Caprara, Barbaranelli, Gerbino, & Pastorelli, 2003; Cataldo et al., 2005; Nie et al., 2014). Taken together, these theoretical claims and empirical findings led us to expect that the observed direct and indirect links would show similar pattern across Chinese and Italian adolescents. In addition, previous studies have found that there are significant gender and age differences in parental attachment and depressive symptoms both in Chinese and Italian adolescents (Cheng, 1998; Delvecchio, 2013; Li, Delvecchio, Miconi, Salcuni, & Di Riso, 2014; Nie et al., 2014; Nolen-Hoeksema & Girgus, 1994; Tambelli, Laghi, Odorisio, & Notari, 2012; Tepper et al., 2008). Therefore, in order to rule out the possible demographic effects, we also tested whether there would be age and gender differences in parental attachment, self-control, and depressive symptoms. Collectively, we hypothesized that: (1) parental attachment and self-control would be negatively related to depressive symptoms; (2) self-control would mediate the association between parental attachment and depressive symptoms; and (3) the direct and indirect relationships would be invariant across Chinese and Italian adolescents. Method Participants A total of 2632 adolescents were recruited from middle and high schools with ages ranging from 12 to 16 years. Of these, 1305 Chinese adolescents (658 boys, 647 girls, M ¼ 14.07 years, SD ¼ 1.37 years) and 1327 Italian adolescents (623 boys, 704 girls, M ¼ 13.84 years, SD ¼ 1.46 years) were selected from Guangzhou (China) and Milan (Italy), respectively. Participants were collected in private and public cooperative schools which served mainly middle-class families (absolute SES, Hollingshead, 1975), with approximately similar basic quality of life, within urban and suburban school districts. All participants were residents of their respective countries, and the primary languages were Chinese and Italian for Chinese and Italian, respectively. About 93% of the families who received the leaflet agreed to participate in both countries. Those who refused to take part in the study indicated reasons such as lack of interest and concerns about sharing personal information. All participants indicated that they were from two-parent family and did not have psychological or psychiatric counselling over the past two years.
162
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
Measures Parental attachment The revised version of Inventory of Parent and Peer Attachment (IPPA-R) was used to measure participants' attachment to parents. This scale consists of 25 items rated on a 5-point scale (from “1 ¼ never” to “5 ¼ always”) assessing the magnitude of Trust, Communication, and Alienation separately towards mother and father with parallel wordings. Higher scores on Trust and Communication and lower score on Alienation indicate better parental attachment. Chinese and Italian participants answered the existing Chinese version and Italian version of IPPA-R, respectively (Li et al., 2014). Sample items include “I am angry at my mother” (reverse score) and “my mother respects my feelings”. This scale has been previously applied to Chinese (Song, Thompson, & Ferrer, 2009) and Italian adolescents (Pace, San Martini, & Zavattini, 2011), showing adequate psychometric properties. Prior studies have demonstrated the cross-cultural applicability of the IPPA-R in Chinese and Italian samples (Li et al., 2014). In this study, we carried out measurement invariance test to replicate this finding. Regarding maternal attachment, the three-factor model (i.e., Trust, Communication, and Alienation) showed adequate fit for Chinese (c2 (272) ¼ 1716.941, RMSEA ¼ .064, CFI ¼ .971) and Italian (c2 (272) ¼ 1824.328, RMSEA ¼ .066, CFI ¼ .980) samples. Models of configural invariance (c2 (544) ¼ 3553.660, RMSEA ¼ .065, CFI ¼ .970), weak invariance (c2 (569) ¼ 3752.980, RMSEA ¼ .065, CFI ¼ .968) and scalar invariance (c2 (594) ¼ 16.042, RMSEA ¼ .064, CFI ¼ .968) were demonstrated, respectively. As the values of the DCFI between two nested models (i.e., configural invariance and weak invariance, and weak invariance and scalar invariance) were both less than .01 and other fit indices showed good fit, measurement invariance could not be rejected (Cheung & Rensvold, 2002). The Cronbach's alpha for the overall scale of maternal attachment, maternal Trust, maternal Communication, and maternal Alienation were .91, .85, .81, and .69 respectively for Chinese and .92, .88, .83, and .77 respectively for Italian samples. With respect to paternal attachment, the same procedures of examination were applied. Similarly, the three-factor model indicated adequate fit for Chinese (c2 (272) ¼ 1921.322, RMSEA ¼ .068, CFI ¼ .970) and Italian (c2 (272) ¼ 1697.541, RMSEA ¼ .063, CFI ¼ .979) samples. Sequential examination of configural invariance (c2 (544) ¼ 3616.819, RMSEA ¼ .066, CFI ¼ .972), weak invariance (c2 (569) ¼ 3690.715, RMSEA ¼ .065, CFI ¼ .972) and scalar invariance (c2 (594) ¼ 3768.028, RMSEA ¼ .064, CFI ¼ .972) showed that paternal attachment was equivalent across Chinese and Italian samples, as the values of the DCFI between two nested models were less than .01 and other fit indexes were acceptable. The Cronbach's alpha for the overall scale of paternal attachment, paternal Trust, paternal Communication, and paternal Alienation were .91, .86, .83, and .71 respectively for Chinese and .92, .88, .84, and .73 respectively for Italian samples. Collectively, the present findings suggested that the IPPA-R was applicable in Chinese and Italian adolescents. Self-control The self-restraint subscale of Adolescent Self-Consciousness questionnaire (ASC) was utilized to assess respondents' self-control ability. This scale includes 11 items rated on a 5-point scale (from “1 ¼ not like me at all” to “5 ¼ like me very much”). A total score can be obtained by summing up all the items, and a higher score indicate greater self-control ability. This scale was developed in Chinese context (Nie et al., 2014) and has been validated in Italian adolescents, both showing good psychometric properties (Delvecchio, Mabilia, Lis, Mazzeschi, Nie, & Li, 2014; Delvecchio, Mabilia, Miconi, Chirico, & Li, 2015). Chinese and Italian participants answered the existing Chinese (Nie et al., 2014) and Italian (Delvecchio, Mabilia, Lis, et al., 2014; Delvecchio, Mabilia, Miconi, et al., 2015) version of this scale, respectively. Sample items include “I fail in overcoming my bad habits even though I have tried many times” and “I can control my emotion”. In the current study, the one-factor model showed good fit both in Chinese (c2 (44) ¼ 242.689, RMSEA ¼ .059, CFI ¼ .969) and Italian (c2 (44) ¼ 246.784, RMSEA ¼ .059, CFI ¼ .969) samples. A measurement invariance test demonstrated that configural invariance (c2 (88) ¼ 489.473, RMSEA ¼ .059, CFI ¼ .969), weak invariance (c2 (99) ¼ 511.136, RMSEA ¼ .056, CFI ¼ .968), and scalar invariance (c2 (110) ¼ 534.737, RMSEA ¼ .054, CFI ¼ .967) were obtained, as the values of the DCFI between two nested models were less than .01 and other fit indexes showed good fit, suggesting that the self-restraint subscale was invariant across cultures. The Cronbach's alpha of this scale was .79 and .65 for Chinese and Italian samples, respectively. In sum, this indicated that the self-restraint subscale was applicable in both countries. Depression The Children's Depression Inventory (CDI) was administered to subjects to assess their depressive symptoms. This instrument contains 27 items. Kovacs (1992) found that this scale included five dimensions e anhedonia (8 items), negative emotion (6 items), negative self-esteem (4 items), ineffectiveness (4 items), and interpersonal problem (4 items). Each item consists of three statements graded in severity (e.g., “0 ¼ I am sad once in a while”, “1 ¼ I am sad many times”, and “2 ¼ I am sad all the time”) and higher scores reflect more severe depressive symptoms. In this study, item 9 e an item about suicide ideation e was excluded in data analyses because it belongs to none of the above five subscales. CDI has been used both in Chinese and Italian adolescents and it shows good psychometric properties (e.g., Delvecchio, 2013; Nie et al., 2014). Sample items include “I hate/do not like/like myself” (reverse score) and “I feel like
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
163
crying everyday/many days/once in a while” (reverse score). Previous studies showed that the factor structure of the CDI was inconsistent in adolescents (Delvecchio, 2013; Poli, Sbrana, Marcheschi, & Masi, 2003). Therefore, we first investigated the five-factor structure as described in Kovacs's (1992) report. The five-factor structure showed adequate fit in Chinese (c2 (289) ¼ 1037.418, RMSEA ¼ .045, CFI ¼ .981) and good fit (c2 (287) ¼ 2377,134, RMSEA ¼ .074, CFI ¼ .961) in Italian sample.1 To test the measurement invariance, a serial of nested models were performed. Configural invariance2 (c2 (574) ¼ 3601.358, RMSEA ¼ .063, CFI ¼ .961), weak invariance (c2 (600) ¼ 4282.954, RMSEA ¼ .068, CFI ¼ .953), and strong invariance (c2 (626) ¼ 4427.575, RMSEA ¼ .068, CFI ¼ .952) were obtained sequentially. The Cronbach's alpha for the total scale, anhedonia, negative emotion, negative self-esteem, ineffectiveness, and interpersonal problem were .86, .69, .64, .60, .56, and .28 respectively for Chinese sample, and .88, .60, .73, .70, .59, and .26 respectively for Italian sample.3 Overall, these findings suggested that the CDI was generally applicable in Chinese and Italian adolescents. Procedures This research was part of collaboration project between the University of Padua (Italy) and Guangzhou University (China). Administration was carried out in compliance with the ethical standards for research outlined in the Ethical Principles of Psychologists and Code of Conduct (American Psychological Association, 2010). This study was approved by the Ethical Committee for psychological research from both universities. Participation in the study was solicited through leaflets. Written consents were first obtained from the heads of school and parents, and then participants provided their oral consent and were willing to participate. Voluntary participation and anonymity were highlighted and no monetary reward was given. Participants completed the instruments during regular class and they had the right to withdraw if they did not want to continue. Participants were asked to be open and honest and to refrain from sharing answers with others. Confidentiality was assured by replacing respondents' information with a numeric code. Data analysis The .05 significance level was used throughout the paper. Before analyzing the data, imputation of missing values for all variables were carried out for both samples using PASW Statistics (version 18.0), which was based on values observed in other cases which showed a similar response pattern. Subsequently, we used MANOVA to examine the cultural and gender differences in parental attachment and depressive symptoms, whereas ANOVA was employed to investigate the cultural and gender differences in self-control. The p value and partial eta squared (hp 2 ) were used to judge the meaningfulness of the significance. Specifically, meaningful significance referred to F ratio was significant (i.e., p < .05) and the value of hp 2 .01 and .02 respectively for univariate and multivariate analyses of variance (Cohen, 1992). Only results that reached meaningful significance were interpreted. Afterward, we used Mplus 7.0 to carry out structural equation model to examine the relations between parental attachment, self-control, and depressive symptoms, with possible confounding demographic variables controlled. A robust method of maximum likelihood (MLR) was used to obtain the fit index because the data were ordinal. The values of RMSEA and CFI were used to judge the model fit, with the values of RMSEA no larger than .08 and CFI no less than .90 indicating good fit (Bentler, 1990; Hu & Bentler, 1999). In addition, we drew 5000 bootstrap sample from the full data and used 95% confidence interval to determine the significance of the mediating effect.4 Significant mediating effect would be identified if the confidence interval excluded 0. Finally, we examined whether the hypothesized model was invariant across cultures by comparing the free model (where all main paths were freely estimated) and the constrained model (where all main paths were constrained to be equal). Because c2 has been criticized for its sensitivity to sample size, we used DCFI to determine whether the coefficients of main paths were invariant as previous study had recommended (Cheung & Rensvold, 2002). Invariance would be identified if DCFI was below the threshold of .01 and other index showed good fit (i.e., RMSEA .08 and CFI .90, Cheung & Rensvold, 2002).
1 For the Chinese sample, no errors were correlated, but for Italian sample, errors of item 20 and item 22, and errors of item 1 and items 20 were correlated. If without these correlations, then the fit index was c2 (289) ¼ 3180.862, RMSEA ¼ .087, CFI ¼ .946. We were aware that CDI might consist of other factor structures for normal adolescent sample, such as the six-factor model (Craighead, Smucker, Craighead, & Ilardi, 1998) and the three-factor model (Cole, Hoffman, Tram, & Maxwell, 2000). However, we tried these models and found that the fit index for these factor structures was even worse (RMSEA > .10). Considering that the present study mainly focused on the relationships between parental attachment, self-control, and depressive symptoms, we adopted the five-factor model as proposed by Kovacs (1992) but allowed these errors to be correlated. 2 We used the Italian model pattern (i.e., five-factor with errors correlated) as the baseline model. We tested this model pattern in Chinese sample, finding that this pattern also showed adequate fit in Chinese sample, c2 (289) ¼ 1005.094, RMSEA ¼ .044, CFI ¼ .982. This suggested that this model could be used as the baseline model for subsequent tests of measurement invariance. 3 We noticed that the reliability of the interpersonal problem subscale was very low. This is consistent with previous studies (e.g., Delvecchio, 2013) which have also found this subscale problematic. Because there were few items in this subscale and other scales, we considered that the total CDI in general was reliable for both samples based on the Cronbach's alpha of the total scale and intercorrelations between subscales as shown in Table 3. 4 When bootstrapping technique was applied, maximum likelihood method was used because MLR does not support bootstrapping technique in the current version of Mplus.
164
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
Results Demographic differences in parental attachment, self-control, and depressive symptoms MANOVA were performed to investigate cultural and gender differences in parental attachment and depressive symptoms and ANOVA was conducted to examine cultural and gender differences in self-control (Tables 1 and 2). Regarding maternal attachment, MANOVA revealed a significant multivariate main effect for culture (Wilk's l ¼ .914, F (3, 2626) ¼ 82.16, p < .001, hp 2 ¼ .086) and gender (Wilk's l ¼ .979, F (3, 2626) ¼ 18.57, p < .001, hp 2 ¼ .021), but not for their interaction (Wilk's l ¼ .998, F (3, 2626) ¼ 1.41, p ¼ .238, hp 2 ¼ .002). Given the significance of the overall test, the univariate main effects were examined. Results showed that Italian adolescents had substantially higher maternal trust (F (1, 2628) ¼ 229.90, p < .001, hp 2 ¼ .080), maternal communication (F (1, 2628) ¼ 195.13, p < .001, hp 2 ¼ .069) and lower maternal alienation (F (1, 2628) ¼ 90.53, p < .001, hp 2 ¼ .033) than Chinese adolescents. Girls showed substantially higher maternal communication (F (1, 2628) ¼ 29.53, p < .001, hp 2 ¼ .011) than boys. Although there was also significant gender difference in maternal alienation (F (1, 2628) ¼ 7.19, p ¼ .007, hp 2 ¼ .003), the effect size was too trivial to interpret. No significant interaction was found for maternal attachment. With respect to paternal attachment, results of MANOVA demonstrated a significant multivariate a meaningful main effect for culture (Wilk's l ¼ .967, F (3, 2626) ¼ 30.28, p < .001, hp 2 ¼ .033), but not for gender (Wilk's l ¼ .987, F (3, 2626) ¼ 11.70, p < .001, hp 2 ¼ .013) or for their interaction (Wilk's l ¼ .995, F (3, 2626) ¼ 4.83, p ¼ .002, hp 2 ¼ .005). Based on this, a subsequent univariate main effect was examined. Results showed that Italian adolescents had substantially higher paternal trust (F (1, 2628) ¼ 76.66, p < .001, hp 2 ¼ .028) and lower alienation (F (1, 2628) ¼ 54.25, p < .001, hp 2 ¼ .020) than their Chinese counterparts. Italian adolescents also showed higher paternal communication than Chinese adolescents (F (1, 2628) ¼ 22.78, p < .001, hp 2 ¼ .009), but the effect size did not reach the threshold (i.e., .01). As for gender difference, boys reported more paternal communication (F (1, 2628) ¼ 22.61, p < .001, hp 2 ¼ .009) and less paternal alienation (F (1, 2628) ¼ 6.72, p ¼ .010, hp 2 ¼ .003) than girls, but their effect sizes were too small to be noticed. There was no significant gender difference in paternal trust (F (1, 2628) ¼ 1.31, p ¼ .253, hp 2 < .001). Although interaction was significant for paternal trust (F (1, 2628) ¼ 4.96, p ¼ .026, hp 2 ¼ .002) and paternal alienation (F (1, 2628) ¼ 12.84, p < .001, hp 2 ¼ .005), the effect sizes were also too trivial to worth interpreting. No significant interaction was found for paternal communication (F (1, 2628) ¼ 1.09, p ¼ .297, hp 2 < .001). Regarding self-control, results of ANOVA demonstrated that the main effect of culture was not significant, F (1, 2628) ¼ .90, p ¼ .343, hp 2 < .001. The main effect of gender was significant, with girls reporting higher self-control than boys (F (1, 2628) ¼ 16.01, p < .001, hp 2 ¼ .006), but the effect size was too small to deserve interpretation. No significant interaction was found (F (1, 2628) ¼ 1.72, p ¼ .190, hp 2 < .001). As for depressive symptoms, results of MANOVA indicated a significant multivariate main effect for culture (Wilk's l ¼ .948, F (5, 2624) ¼ 29.00, p < .001, hp 2 ¼ .052). The multivariate main effect for gender was significant but the effect size was not substantial (Wilk's l ¼ .991, F (5, 2626) ¼ 5.02, p < .001, hp 2 ¼ .009). No significant multivariate interaction was found (Wilk's l ¼ .998, F (3, 2626) ¼ 1.30, p ¼ .261, hp 2 ¼ .002). Subsequently, univariate main effects were examined. Results showed that Chinese adolescents reported higher anhedonia (F (1, 2628) ¼ 8.51, p ¼ .004, hp 2 ¼ .003) and negative emotion (F (1, 2628) ¼ 5.86, p ¼ .016, hp 2 ¼ .002) than Italian adolescents, but these effect sizes were too small to explain. Chinese adolescents showed substantially lower negative self-esteem than their Italian counterpart (F (1, 2628) ¼ 39.33, p < .001, hp 2 ¼ .015). No significant cultural difference in ineffectiveness (F (1, 2628) ¼ .35, p ¼ .852, hp 2 < .001) or interpersonal problem (F (1, 2628) ¼ 2.83, p ¼ .093, hp 2 ¼ .001) was found. Regarding gender difference, among the five subscales, boys showed higher interpersonal problem than did girls (F (1, 2628) ¼ 15.22, p < .001, hp 2 ¼ .006), but this effect size was too small to be noticed. No significant gender difference in other subscales was found, and no significant interaction was found for any of the five subscales.
Table 1 Means and standard deviation of parental attachment, self-control, and depressive symptoms. Overall (N ¼ 2632)
Maternal trust Maternal communication Maternal alienation Paternal trust Paternal communication Paternal alienation Self-control CDI-anhedonia CDI-negative emotion CDI-negative self-esteem CDI-ineffectiveness CDI-interpersonal problem
Chinese boys (N ¼ 658)
Chinese girls (N ¼ 647)
Italian boys (N ¼ 623)
Italian girls (N ¼ 704)
M
SD
M
SD
M
SD
M
SD
M
SD
39.36 31.25 13.82 38.64 28.18 14.38 36.40 3.60 2.70 2.02 2.39 1.21
7.30 7.12 4.68 7.55 7.42 4.69 6.62 2.67 2.34 1.74 1.71 1.18
36.93 28.49 15.08 37.22 28.05 15.12 35.83 3.75 2.72 1.84 2.44 1.32
6.85 6.46 4.42 7.27 6.95 4.75 7.01 2.82 2.09 1.53 1.71 1.20
37.62 30.23 14.28 37.53 26.98 14.95 37.20 3.77 2.89 1.79 2.33 1.19
6.99 7.06 4.15 7.84 7.65 4.52 6.61 2.68 2.15 1.44 1.65 1.05
41.55 32.50 13.05 40.41 29.72 13.15 35.93 3.53 2.63 2.23 2.45 1.29
6.65 6.41 4.79 6.80 6.89 4.27 6.35 2.81 2.76 2.12 1.76 1.32
41.28 33.66 12.90 39.43 28.05 14.26 36.62 3.37 2.55 2.24 2.35 1.06
7.43 7.30 4.96 7.76 7.84 4.90 6.40 2.35 2.30 1.80 1.70 1.13
J.-B. Li et al. / Journal of Adolescence 43 (2015) 159e170
165
Table 2 Significance of demographic differences in parental attachment, self-control, and depressive symptoms. F Maternal trust Maternal communication Maternal alienation Paternal trust Paternal communication Paternal alienation Self-control CDI-anhedonia CDI-negative emotion CDI-negative self-esteem CDI-ineffectiveness CDI-interpersonal problem
country
229.90 195.13 90.53 76.66 22.78 54.25 .90 8.51 5.86 39.33 .04 2.83
(1,2628)
p
hp 2
F
