McArdle disease

Linville's self-complexity buffering hypothesis (Linville, 1985, 1987) posits that individual differences exist in the complexity of cognitive representations of the self. Individuals high in self-complexity organize knowledge about the self into numerous and distinct categories (e.g., self-aspects). On the other hand, individuals low in self-complexity have relatively undifferentiated and unidimensional self-views. According to Linville, high self-complexity serves as a buffer against the onset of depressive symptoms following negative events because it partly determines the impact of such events on thoughts and feelings about various self-aspects. Moreover, according to Linville, when an individual experiences a negative event, the self-aspect most closely related to the event is activated. Negative thoughts and feelings associated with the stressor thus are likely to become associated with the activated self-aspect. In addition, when a self-aspect is activated, activation spreads to associated self-aspects . Therefore, feelings and thoughts associated with the originally activated self-aspect are likely to become associated with related self-aspects as well. Because individuals high in self-complexity have self-aspects that are distinct from one another, they are less likely to experience such spillover effects. In other words, in high self-complexity individuals, thoughts and feelings evoked by negative events are likely to be confined to immediately salient self-aspects that make up a relatively small proportion of their self-knowledge. On the other hand, in individuals low in self-complexity, spillover effects are likely. Therefore, following negative events, a greater number and proportion of self-aspects are likely to be adversely affected in such individuals.

In a preliminary study examining the self-complexity buffering hypothesis, Linville (1985) found that individuals low in self-complexity exhibited greater levels of emotional lability than individuals high in self-complexity. In subsequent research, Linville (1987) found that individuals low in self-complexity were more likely than individuals high in self-complexity to experience increases in depressive symptoms following the occurrence of negative events.

Morgan and Janoff-Bulman (1994) extended the Linville (1985, 1987) model by proposing that positive and negative self-representations should be examined separately. According to their model, high positive self-complexity serves as a buffer against depressive symptoms following negative events. High negative self-complexity, on the other hand, serves as a vulnerability factor to depression following such events. In line with this hypothesis, in a cross-sectional study of college students, they found that high levels of positive self-complexity were associated with low levels of emotional and psychological distress following the occurrence of a traumatic event. At the same time, however, high levels of negative self-complexity were associated with high levels of distress. Subsequent research has supported Morgan and Janoff-Bulman's claim that self-complexity is not a unitary construct but rather is composed of at least two partly independent dimensions: positive and negative self-complexity (Morgan &Janoff-Bulm an, 1994). At the same time, however, results from such research have provided more consistent support for the hypothesis that high negative self-complexity serves as a vulnerability to depression factor than for the hypothesis that high positive self-complexity serves as a resiliency factor (e.g., Gara et al., 1993; Showers, Abramson, & Hogan, 1998; Woolfolk et al., 1999; Woolfolk, Novalany, Gara, Allen, & Polino, 1995; for exception see Atsushi, 1999).

Although Linville's self-complexity buffering hypothesis (Linville, 1984, 1987) and subsequent revisions of her model (e.g., Morgan & Janoff-Bulman, 1994) have received preliminary support in adult samples, no studies have examined the relationship between self-complexity and depressive symptoms in children. At the same time, however, based on results from a study examining the relationship between self-concept differentiation, one component of self-complexity, and depressive symptoms in fourth, sixth, and eight grade children, Jordan and Cole (1996) proposed that the relationship between self-complexity and depressive symptoms may be different in children. Five hundred and twelve children completed questionnaires assessing depressive symptoms and the occurrence of negative events. They then selected from a list of positive and negative adjectives those that best described them in five predesignated domains. High levels of both positive and negative self-concept differentiation were related to high levels of depressive symptoms and the occurrence of negative events. In addition, neither type of self-concept differentiation interacted with negative events to predict depressive symptoms. On the basis of these results, Jordan and Cole (1996) proposed that, in childhood, the occurrence of negative events is likely to lead to increases in both self-complexity and depressive symptoms. According to their theory, when children are faced with negative events, they are motivated to maintain a positive self-concept. Consequently, they develop new self-aspects so as to resist generalizing the negative self-relevant information to their entire self-concept. Thus, the development of high self-complexity in childhood is likely to reflect the development of coping styles and strategies that will moderate the impact of negative events later in life. In other words, in childhood, self-complexity may increase in response to the occurrence of negative events. During adulthood, those who developed high positive self-complexity in chi ldhood will be buffered against the impact of negative events. Thus, experiences of adversity during childhood may give rise to the organization of self-knowledge in a way that will buffer against depression later in life.

Although Jordan and Cole's theory offers intriguing explanations about both the development of self-complexity and the differing relationship between self-complexity and depressive symptoms in children and adults (Jordan & Cole, 1996), there are limitations to their study that prevent us from being able to interpret their results as providing support for their developmental model. The major limitation is that they assessed self-concept differentiation rather than self-complexity. In other words, they assessed the degree to which children's identities in five domains were differentiated from each other rather than integrated into a unitary self. Because self-complexity is defined by Linville (1985, 1987) to be a function of both the number of self-aspects and the degree to which self-aspects are differentiated from each other, Jordan and Cole's assessment procedure (Jordan & Cole, 1996) cannot be viewed as adequately measuring self-complexity. Therefore, in order to determine whether their results indicate the presence of developmental differences in the organization of self-knowledge and the relationship between such organization and depressive symptoms, we must compare their results to research examining the relationship between depressive symptoms and self-concept differentiation in adults rather than self-complexity.

Interestingly, although Jordan and Cole's results (Jordan & Cole, 1996) contradict findings from research examining the relationship between self-complexity and depression in adult populations, they are congruent with findings from research examining the relationship between self-concept differentiation and depressive symptoms in adults. For example, in a sample of university students, Donahue, Robins, Roberts, and John (1993) found higher levels of self-concept differentiation were associated with higher levels of depressive symptoms and lower levels of self-esteem. Similarly, Jordan and Cole's findings (Jordan & Cole, 1996) are parallel to those obtained by past researchers examining self-concept differentiation in children. For example, Mullener and Laird (1971) reported that seventh grade students who gave less-differentiated self-evaluations of themselves rated themselves more positively in terms of achievement traits, intellectual skills, interpersonal skills, physical skills, and social responsibility than seventh grade students who had high levels of differentiation in self-ratings.

Because self-concept differentiation is a component of Linville's definition of self-complexity, should we believe that these two constructs are similar? Donahue et al. (1993) argue that we should not. They claim that measures of self-complexity reflect the number of distinct self-aspects individuals generate when providing descriptions of themselves and thus imply cognitive complexity and flexibility in the way people construe themselves. Measures of self-concept differentiation, in contrast, reflect the lack of interrelatedness of role identities and thus imply fragmentation and lack of coherence in the self-concept. To date, only one study has assessed both self-complexity and self-concept differentiation in the same sample (e.g., Campbell, Chew, & Scratchley, 1991). Although Campbell et al. (1991) did not report the correlation between these two measures, they found that higher levels of self-complexity were associated with higher levels of self-esteem whereas higher levels of self-concept differentiation were associated with lower levels of self-esteem. Their findings, therefore, suggest that these two constructs represent different aspects of self-concept structure and should not be equated.

Thus, an examination of research conducted on the relationship between self-concept differentiation and depressive symptoms indicates that (1) the results of Jordan and Cole's study (Jordan & Cole, 1996) are in line with past research using both adult and child samples and therefore are not indicative of a difference in the relationship between the organization of self-knowledge and depressive symptoms in children and adults and (2) self-concept differentiation and self-complexity represent different aspects of the structure of self-knowledge and therefore we cannot make inferences about the applicability of Linville's and Morgan and Janoff-Bulman's models of the relationship between self-complexity and depressive symptoms to children on the basis of the findings of Jordan and Cole's study (Jordan & Cole, 1996; Linville, 1985, 1987; Morgan & Janoff-Bulman, 1994).

At the same time, it is likely that an examination of Linville's and Morgan and Janoff-Bulman's models of the relationship between self-complexity and depressive symptoms (Linville, 1985, 1987; Morgan & JanoffBulman, 1994) will uncover age-related differences, for several reasons, when tested in a child sample. First, the complexity of children's self-concepts increases considerably during middle childhood (Harter, 1983; Montemayor & Eisen, 1977). For example, Harter (1990) reported that both the number and the nature of dimensions needed to account for children's self-perceptions changes dramatically between late childhood and early adolescence. An increase in the number of dimensions that make up a child's self-concept is likely to lead to higher levels of self-complexity. Second, young children tend to be globally positive about the self (Harter, 1988). As abstract reasoning abilities and formal operational thought emerge in late childhood and early adolescence, however, children begin to have difficulty m aintaining an overly simplified and globally positive view of themselves. Consequently, when exposed to feedback, they begin to acknowledge negative as well as positive self-relevant information and to realize that they are not entirely good in every area of their life (Harter & Monsour, 1992). As children begin to incorporate both positive and negative information into their self-concepts, self-concept differentiation is likely to increase, leading to higher self-complexity. In addition, as children begin incorporating negative information into their self-concept, negative self-complexity is likely to dramatically increase in some children, conferring them with a cognitive vulnerability to depression. Last, because different children are exposed to dramatically different self-relevant experiences during middle childhood, individual differences in the complexity of self-concepts are likely to emerge in this age group (e.g., Fiske & Linville, 1980; Pelham, 1993). As differences in self-complexity between child ren become more pronounced, self-complexity is likely to emerge as an individual difference predictive of resiliency and vulnerability to depressive symptoms following the occurrence of negative events.

The goal of this study was to test the diathesis-stress component of Linville (1985, 1987) and Morgan and Janoff-Bulman (1994) models of the relationship between self-complexity and depressive symptoms in children. Third and seventh grade children were chosen as participants because (1) they represent the outer limits of middle childhood and (2) previous research has shown that the complexity of children's self- concepts increases considerably during middle childhood (Harter, 1983, 1990; Harter & Monsour, 1992; Montemayor & Eisen, 1977). Therefore, age differences are likely to emerge between these two groups in the relationship between self-complexity and depressive symptoms following the occurrence of negative events. The procedure involved an initial assessment of self-complexity and depressive symptoms. The procedure also involved a follow-up assessment, 10 weeks later, in which depressive symptoms and the occurrence of negative events were assessed.

We tested the following hypotheses. First, in line with previous research, we hypothesize that seventh grade children will exhibit significantly higher levels of total, negative, and positive self-complexity than third grade children. Second, in line with Linville's self-complexity buffering hypothesis (Linville, 1985, 1987), seventh grade children who exhibit low total self-complexity will show increases in depressive symptoms following negative events. In contrast, seventh grade children with high total self-complexity will not show such increases. Third, in line with the revision by Morgan and Janoff-Bulman (1994) of Linville's theory, seventh grade children who exhibit high negative self-complexity will show increases in depressive symptoms following the occurrence of negative events. In contrast, seventh grade children with low negative self-complexity will not show such increases. Fourth, seventh grade children low in positive self-complexity will show increases in depressive symptoms following the occu rrence of negative events. In contrast, seventh grade children high in positive self-complexity will not show such increases. Last, because pronounced individual differences in the complexity of self-concept are hypothesized to not exist among third grade children, only negative events will predict increases in depressive symptoms in third graders.

METHOD

Participants

Consent forms were sent to all parents of third and seventh grade children at participating schools. Consent rates varied from 75 to 85%, with a median rate of 80%. The final sample consisted of 91 seventh grade children (76 girls and 15 boys) and 67 third grade children (45 girls and 22 boys) from four Montreal area schools. Seventh grade participants came from one urban private girl school (18 children) and one suburban private school (73 children). Third grade participants came from one urban private girls school (19 children) and one suburban public school (48 children). The gender ratios in both grades reflect the gender ratios at participating schools. The final sample was 94.4% Caucasian, 2.3% African American, 1.1% Hispanic, 1.1% Asian, and 1.1% Native American of other descent. The mother tongue of 91.4% of the final sample was French. The mean age of the third grade students was 8 years and 3 months. The mean age of the seventh grade students was 12 years and 3 months. Although data on the social-ec onomic status of students' families is not available, all schools were located in predominantly upper middle class regions of the Montreal area.

Procedure

After consent forms were collected, researchers went to each school to meet with children. All questionnaires were read aloud by the researchers. At the start of each assessment, students were told that their participation was voluntary and they could choose to not participate if they desired. No children decided to not participate. During the first meeting, children completed the Children's Depression Inventory (Kovacs, 1981) and the Self-Complexity Scale for Children (Abela & Wronneau-McArdle, 2000). Ten weeks later, researchers met again with participating children. During this follow-up assessment, children completed the Children's Depression Inventory (Kovacs, 1981) and the Children's Life Events Scale (Coddington, 1972; Kanner, Feldman, Weinberger, & Ford, 1987; Kanner, Harrison, & Wertlieb, 1985). Following the completion of the questionnaires, the children were debriefed. They then participated in a modified version of the Penn Prevention Program (Jaycox, Reivich, Gillham, & Seligman, 1994).

Measures

All questionnaires were translated into French by a certified English-French translator. The French translations were then backtranslated into English by a certified French-English translator. The backtranslation was then compared to the original English version of the questionnaire. Any discrepancies between the original English version of the questionnaire and the backtranslation were examined to ensure that the French translation retained the intended meaning of the original item.

Self-Complexity Scale for Children (SCSC; Abela & Veronneau-McArdle, 2000)

The SCSC is a two-part questionnaire that was devised for the current study in order to assess total, negative, and positive self-complexity in children. In part one, children are asked to list all roles or activities (e.g., self-aspects) in their life that are important to them. Children are allowed to list up to 10. In part two, children are given a copy of a list of 20 adjectives (10 positive and 10 negative) for each role or activity they listed. Research assistants label each adjective list with the name of a separate role or activity the child provided. The 20 adjectives were selected by Hammen and Zupan (1984) from Carroll, Davies, and Richman's Word Frequency Book, which lists words taken from third to ninth grade reading materials (Carroll, Davies, & Richman, 1971). Examples of positive words are happy, proud, nice, and smart. Examples of negative words are bad, angry, unhappy, and lonely. These words were piloted for readability in a sample of third grade children by Jordan and Cole (1996). For each adjective list, children are asked to think about themselves in the role or activity written on the top of the list and to check off any words that accurately describe them in that role or activity. This procedure was completed for each role or activity the child listed.

As in previous work on self-complexity (e.g., Linville, 1985, 1987; Morgan & Janoff-Bulman, 1994), self-complexity was conceived of as a function of both the number and distinctness of self-aspects generated by the children. In line with previous researchers, in order to calculate self-complexity scores, we utilized the H statistic (e.g., Attneave, 1959; Scott, Osgood, & Peterson, 1979). This statistic is a measure of the amount of information contained in a participant's matrix of responses. For example, in a 10 x 20 matrix for total self-complexity (10 self-aspects x 20 adjectives), each cell contains either a 1 if the child chose that adjective as descriptive of a particular self-aspect and a 0 if the child did not. If the child chooses the same adjectives to describe every self-aspect, then the columns of the matrix are redundant. In such an instance, the H statistic would be small, indicating a low level of self-complexity. On the other hand, if the child chooses different adjectives to describe each sel f-aspect, the columns will be extremely different from one another. Therefore, the H statistic will be relatively large, indicating a higher-degree self-complexity.

The H statistic is calculated using the following formula:

H = [log.sub.2]x - ([x.sub.i] [log.sub.2][x.sub.1]/x

In this equation x is the total of number of adjectives (e.g., 20 if calculating total self-complexity and 10 if calculating negative or positive self-complexity) and [x.sub.i] is the number of adjectives that appear in a particular group combination. A group combination is defined by a unique pattern of "describes me" and "does not describe me" responses for a given adjective. For example, a particular adjective can be coded by a participant who generated four self-aspects as 1 (describes me in the first self-aspect), 1 (describes me in the second self-aspect), 0 (does not describe me in the third self-aspect) and 1 (describes me in the fourth self-aspect). The pattern for this adjective would therefore be 1-1-0-1. All adjectives with this pattern would make up one group combination. As recommended by Linville (1985, 1987), instances where [x.sub.i] was equal to 0 were excluded from the summation.

For each participant, we calculated three self-complexity scores: (1) a total self-complexity score based on all 20 adjectives; (2) a negative self-complexity score based on the 10 negative adjectives; and (3) a positive self-complexity score based on the 10 positive adjectives. Total self-complexity scores range from 0 to 4.32, with higher scores indicating higher levels of total self-complexity. Positive and negative self-complexity scores each range from 0 to 3.32, with higher scores indicating higher levels of positive or negative self-complexity.

Children's Depression Inventory (CDI; Kovacs, 1981)

The CDI is a 27-item self-report questionnaire that measures the cognitive, affective, and behavioral symptoms of depression. It is the most widely used self-report measure of depression in children (Hammen & Gotlib, 1992). For each item, children are asked whether it describes how they were thinking and feeling in the past week. Items are scored from 0 to 2, with a higher score indicating greater symptom severity. At the request of school principals, we removed one item that inquires about suicidality. Total scores thus range from 0 to 52. A cutoff score of 13 has been suggested to indicate mild depression and 19 to indicate severe depression (Kovacs, 1983). During the initial and follow-up assessments respectively, we obtained alphas of .76-.83 in third graders and .57-.76 in seventh graders. These alphas indicate moderate to high internal consistency. Children's scores ranged from 0 to 33.

The Children's Life Events Scale (CLES; Coddington, 1972; Kanner et al., 1985, 1987)

The CLES is a list of 59 hassles (e.g., "Kids at school teased you"; "Your schoolwork was too hard") and negative life events (e.g., "You were suspended from school"; "A close friend or family member died") that children may experience. For each item, respondents are asked how often they have experienced that event in the past 10 weeks. If they have not experienced the event they receive a score of 0. If they experienced the event a few times, they receive a score of 1. If they have experienced the event often, they receive a score of 2. If they have experienced the event all the time, they receive a score of 3. Scores on the CLES range from 0 to 177, with higher numbers indicating greater frequency and duration of stressful events. Thirty-seven items on the CLES are from the Children's Hassles Scale (Kanner et al., 1985, 1987). The remaining 22 items are from the Coddington Life Stress Scale (Coddington, 1972).

RESULTS

Descriptive and Reliability Data

Means and standard deviations of all measures are presented in Table I. Several findings warrant attention. First, a Grade x Gender ANOVA on children's total self-complexity scores indicated that seventh graders exhibited higher levels of total self-complexity than third graders, F(l, 152) = 4.51, p < .05. In addition, girls exhibited higher levels of total self-complexity than boys, F(1, 152) = 7.45, p < .01. Second, a Grade x Gender ANOVA on children's positive self-complexity scores indicated that seventh graders exhibited higher levels of positive self-complexity than third graders, F(1, 152) = 15.32, p < .001. In addition, girls exhibited higher levels of positive self-complexity than boys, F(1, 152) = 7.61, p < .01. Third, a Grade x Gender ANOVA on children's Time 2 CDI scores indicated that boys exhibited higher levels of depressive symptoms at the follow-up assessment than girls, F(l, 148) = 5.15, p < .05. Fourth, a Grade x Gender ANOVA on children's CLES scores indicated that boys experienced a great er number of negative events during the follow-up period than girls, F(1, 147) = 10.51, p = .001.

Correlations between all Time 1 measures are presented in Table II. Several findings warrant attention. First, in third graders, higher levels of positive self-complexity were associated with higher levels of negative self-complexity. The strength of this association, however, was modest. In seventh graders, positive and negative self-complexity were not related to each other. Second, in both third and seventh graders, higher levels of negative self-complexity were associated with higher levels of depressive symptoms. In addition, in seventh graders, higher levels of positive self-complexity were associated with lower levels of depressive symptoms. Last, although not significant, in third graders, higher levels of positive self-complexity were associated with lower levels of depressive symptoms.

Overview of Statistical Analysis: Diathesis-Stress Component

To provide a test of the diathesis-stress component of the Linville's theory (Linville, 1985, 1987), we performed hierarchical multiple regression analyses (Cohen & Cohen, 1983). This approach allows for a stringent test of the contribution of each diathesis in explaining Time 2 depressive symptoms above and beyond both Time 1 depressive symptoms and the effect of stress itself. It also allows for a test of the diathesis--stress interaction. In all cases, the dependent variable was Time 2 CDI scores. First, Time 1 CDI scores were entered into the equation. This controlled for any initial differences between subjects in level of depressive symptoms at Time 1. Second, main effect variables (i.e., CLES and self-complexity) were entered into the equation. Finally, and most important, the diathesis--stress interaction was entered.

To examine the effect of grade, we first regressed Time 2 CDI scores on CLES scores, diathesis scores, grade, and all two- and three-way interactions. If the three-way interaction (CLES x Diathesis x Grade) was significant, we examined the diathesis--stress interaction separately in third and seventh graders. If the three-way interaction was not significant, we examined the diathesis--stress interaction in the entire sample simultaneously. In all analyses, the two-way interactions involving grade (e.g., Grade x CLES and Grade x Diathesis) were not significant. In a similar way, to examine the effect of gender, we regressed Time 2 CDI scores on CLES scores, diathesis scores, gender, and all two- and three-way interactions. If the three-way interaction (CLES x Diathesis x Gender) was significant, we examined the diathesis-stress interaction separately in boys and girls. If not, we collapsed data across gender. In all analyses, none of the interactions involving gender (e.g., Gender x CLES, Gender x Diathesis, a nd Gender x Diathesis x CLES) were significant. Therefore, date were collapsed across gender.

Test of the Diathesis--Stress Component: Total Self-Complexity

The regression of Time 2 CDI scores onto CLES scores, TOT-SCSC scores ,grade, and all two- and three-way interactions accounted for a significant proportion of the variance in CDI scores, [R.sup.2] = .57 (p < .001). The three-way interaction (TOT-SCSC x CLES x Grade), however, was not significant. This indicated that the TOT-SCSC x CLES interaction did not differ as a function of grade. Therefore, data were collapsed across grade.

Results pertaining to the TOT-SCSC x CLES interaction in the combined third and seventh grade sample are presented in Table III. CLES scores uniquely accounted for a significant 27.6% of the variance in residual change in CDI scores from Time 1 to follow-up. In contrast, TOT-SCSC scores were not a significant predictor of residual change scores. Similarly, the TOT-SCSC x CLES interaction was not a significant predictor of residual change scores.

Test of the Diathesis--Stress Component: Negative Self-Complexity

The regression of Time 2 CDI scores onto CLES scores, NEG-SCSC scores ,grade, and all two- and three-way interactions accounted for a significant proportion of the variance in CDI scores, [R.sup.2] = .57 (p < .001). The three-way interaction (NEG-SCSC x CLES x Grade) was significant, F(l, 132) = 6.6l, p <.05. This effect indicated that the NEG-SC x CLES interaction differed as a function of grade. Therefore, the NEG-SCSC x CLES relationship was examined separately in each grade.

Results pertaining to the NEG-SCSC x CLES interaction in third graders are presented in Table IV. In the third graders, CLES scores uniquely accounted for a significant 32.0% of the variance in residual change in CDI scores from Time 1 to Time 2. In contrast, NEG-SCSC scores were not a significant predictor of residual change scores. In a similar manner, the NEG-SCSC x CLES interaction was not a significant predictor of residual change scores.

Results pertaining to the NEG-SCSC x CLES interaction in seventh graders are presented in Table V. CLES scores uniquely accounted for a significant 21.3% of the variance in residual change in CDI scores from Time 1 to follow-up. In contrast, NEG-SCSC scores were not a significant predictor of residual change scores. Last, and most central to Linville's theory (Linville, 1985, 1987), the NEG-SCSC x CLES interaction accounted for a significant 8.1% of the variance in residual change in CDI scores from Time 1 to Time 2. To examine the form of the NEG-SCSC x CLES interaction in the seventh grade sample, as recommended by Cohen and Cohen (1983), residual CDI change scores were calculated by inserting specific values for SCSC and CLES scores (i.e., 1 standard deviation above and below the mean) into the regression equation summarized in Table V. The results of this analysis are presented in Fig. 1. As can be seen, among seventh graders who experienced high levels of negative events, those with high levels of negati ve self-complexity showed increases in CDI scores whereas those with low levels of negative self-complexity did not. Seventh graders who did not experience high levels of negative events did not show increases in CDI scores regardless of their level of negative self-complexity.

Last, we examined whether the NEG-SCSC x CLES interaction continued to predict increases in CDI scores in seventh graders after controlling for the degree of negative content in the children's self-concepts. This was done by computing both (1) the number of negative adjectives each child chose as self-descriptive (NEGADJ) and (2) the ratio of negative to positive adjectives each child chose as self-descriptive (RATIO). After controlling for NEGADJ, the NEG-SCSC x CLES interaction continued to account for a significant proportion of the variance in residual change in CDI scores from Time 1 to Time 2, F(1, 71) = 4.041, p < .05. In a similar manner, after controlling for RATIO, the NEG-SCSC x CLES interaction continued to account for a significant proportion of the variance in residual change scores, F(l, 71) = 6.572, p < .05.

Test of the Diathesis--Stress Component: Positive Self-Complexity

The regression of Time 2 CDI scores onto CLES scores, POS-SCSC scores, grade, and all two- and three-way interactions accounted for a significant proportion of the variance in CDI scores [R.sup.2] = .55 (p < .001). The three-way interaction (POS-SCSC x CLES x Grade), however, was not significant. This effect indicated that the POS-SCSC x CLES interaction did not differ as a function of grade. Therefore, data were collapsed across grade.

Results pertaining to the POS-SCSC x CLES interaction in the combined third and seventh grade sample are presented in Table VI. CLES scores uniquely accounted for a significant 27.6% of the variance in residual change in CDI scores from Time 1 to follow-up. In contrast, POS-SCSC scores were not a significant predictor of residual change scores. Similarly, the POS-SCSC x CLES interaction was not a significant predictor of residual change scores.

DISCUSSION

Several findings emerge from the current study. First, seventh graders exhibited higher levels of total and positive self-complexity than third graders. At the same time, however, seventh and third graders did not differ in their level of negative self-complexity. Second, correlational analyses indicated that, in both third and seventh graders, positive and negative self-complexity were two partly independent dimensions of self-complexity. Last, neither total nor positive self-complexity served as a buffer against the onset of depressive symptoms following the occurrence of negative events in either seventh or third graders. Negative self-complexity, however, interacted with the occurrence of negative events to predict increases in depressive symptoms in seventh but not third grade children.

Regarding the development of self-complexity, as hypothesized, seventh grade children exhibited significantly higher levels of total and positive self-complexity than third grade children. This finding is in line with previous research showing that the complexity of children's self-concepts increases considerably during middle childhood (Harter, 1983, 1990; Harter & Monsour, 1992; Montemayor & Eisen, 1977). At the same time, however, contrary to hypotheses, seventh graders did not exhibit significantly higher levels of negative self-complexity than third graders. One possible explanation is that age-related differences in levels of negative self-complexity do not emerge until early adolescence. Perhaps even late in middle childhood, when exposed to feedback, children are still more likely to incorporate positive than negative information into their self-concept, leading to increases in positive but not negative self-complexity. As abstract reasoning abilities and formal operational thought emerge in early ado lescence, however, children may begin to acknowledge negative as well as positive self-relevant information and realize that they are not entirely good in every area of their life (Harter & Monsour, 1992). As early adolescents begin incorporating negative information into their self-concept, negative self-complexity is likely to increase. Future research is needed to examine this hypothesis.

Another way in which negative self-complexity may increase in early adolescents is through experiences with depression (Woolfolk et al., 1995). Depression rates have been shown to increase significantly from childhood to adolescence (Angold, 1988). Depressed individuals have been shown to exhibit higher levels of self-focused attention than nondepressed individuals (Ingram, 1990; Pyszczynski, Greenberg, Hamilton, & Nix, 1991). In addition, Salovey (1992) reported that self-complexity was higher in individuals in whom self-focus had been induced. Consequently, experiences of depression may lead early adolescents to repeated episodes of self-focus on their shortcomings (e.g., Abramson, Alloy, & Metalsky, 1989; Beck, 1967; Nolen-Hoeksema, 1991). Extensive reflection on negative self-aspects may lead early adolescents to become more aware of and familiar with their flaws and failures. This may lead them to develop a more differentiated, fine-grained system of negative self-aspects and consequently higher levels o f negative self-complexity. Such increases in negative self-complexity are consequently likely to even further increase their risk of experiencing future depressive episodes. Further research is needed for examining this possibility.

Results from the current study fail to provide support for Jordan and Cole's hypothesis that positive and negative self-complexity do not begin to differentiate until adolescence or adulthood (Jordan & Cole, 1996). Although positive and negative self-complexity were positively correlated in third grade children, the strength of this association was weak. In addition, positive and negative selfcomplexity were not correlated in seventh grade children. Last, in both third and seventh grade children, higher levels of negative self-complexity were associated with higher levels of depressive symptoms. Higher levels of positive self-complexity, on the other hand, were associated with lower levels of depressive symptoms. These findings indicate that in both third and seventh graders, self-complexity is not a unitary construct but rather comprises at least two partly independent dimensions: positive self-complexity and negative self-complexity.

In contrast to both Linville's self-complexity buffering hypothesis (Linville, 1985, 1987) and Morgan and Janoff-Bulman's positive self-complexity buffering hypothesis (Morgan & Janoff-Bulman, 1994), neither high levels of total nor positive self-complexity served as a buffer against the onset of depressive symptoms following the occurrence of negative events. At the same time, however, in support of Morgan and Janoff-Bulman's negative self-complexity vulnerability model (Morgan & Janoff-Bulman, 1994), high levels of negative self-complexity interacted with the occurrence of negative events to predict increases in depressive symptoms in seventh grade children. One possible interpretation for failure to find support for the buffering effects of high total and positive self-complexity is that positive self-complexity does not begin to emerge as a resiliency to depression factor until adolescence and adulthood. Such an interpretation is indeed in line with Jordan and Cole's developmental model (Jordan & Cole, 19 96). At the same time, however, the current findings appear to be in line with the majority of studies conducted with adult populations illustrating greater support for the negative self-complexity vulnerability hypothesis than the positive self-complexity buffering hypothesis. Therefore, it is possible that in both adults and children there is a stronger connection between depression and the complexity with which negative information about the self is organized than with the complexity with which positive information about the self is organized.

Given that third and seventh grade children's levels of negative self-complexity did not significantly differ but yet negative self-complexity served as a vulnerability to depression factor only in the seventh graders, it is important for future research to examine why. One possible explanation is that although overall level of negative self-complexity does not differ between the two age groups, other features of the self-structure do. One possible feature is the importance ascribed to different self-aspects, a feature which Pelham and Swann (1989) have labelled as differential importance. Perhaps the relative importance of certain negative self-aspects is greater in seventh graders and therefore such self-aspects serve as more potent vulnerability to depression factors. For example, the relative importance of parents, peers, and teachers changes during middle childhood (Cole, Martin, & Powers, 1997). Younger children (i.e., 9-11 years old) depend more on parents for social support and appraisals of competenc e, whereas older children rely more on their peers (Furman & Buhrmester, 1985, 1992). Such age-related shifts in the relative importance of parents and peers as sources of support and competency appraisals are likely to lead to shifts in the relative importance of self-aspects pertaining to the domains of family and peer relationships. If children possess negative self-aspects in domains that become increasingly important to them as they enter adolescence, their level of vulnerability to depression may be heightened during this time period. Future research should assess the relative importance of the self-aspects children generate so as to examine this hypothesis.

In addition, other changes in cognitive development during early adolescence might increase the risk for depression in individuals who already possess high levels of negative self-complexity. For example, Elkind (1967, 1978) has speculated that increases in egocentrism and self-consciousness during early adolescence may contribute to increases in depression rates in this age group. According to Elkind, early adolescents maintain the false belief that others are as concerned about their thoughts and behaviors as they are. One consequence of the belief in an "imaginary audience" is heightened concern about what others think about them. Elkind has also suggested that adolescents who are critical of themselves will anticipate a critical audience. On the other hand, adolescents who admire themselves will expect others to admire them as well. On the basis of Elkind's theory, Garber, Weiss, and Shanley (1993) have hypothesized that self-esteem might moderate the relationship between self-consciousness and depression in early adolescents. In other words, although most adolescents show increases in self-focused attention, such increases will only increase risk for depression in those who possess a negative self-view. It is likely that negative self-complexity moderates the relationship between increases in self-focused attention and depression in early adolescents as well. Future research should assess egocentrism and self-consciousness so as to examine this hypothesis.

Several limitations of this study should be noted. First, self-report questionnaires were used to assess depressive symptoms. Although the CDI possesses high degrees of reliability and validity, it is difficult to make conclusions about clinically diagnosed depression on the basis of self-report questionnaires. Future studies should utilize clinical interviews to see if findings extend to the development of clinical depression. Second, self-report questionnaires were also used to assess stress. Although measures of life events and daily hassles that require subjects solely to indicate how frequently an event occurred are less likely to be influenced by informant bias than those that ask subjects to rate the subjective impact of each event, more sophisticated methods of assessing stress are likely to give us more precise measurements of stress levels. Third, future studies should utilize larger sample sizes. Such samples would allow for more powerful tests of the diathesis-stress component of Linville (1985, 1 987) and Morgan and Janoff-Bulman (1994) models. Fourth, although the Self-Complexity Scale for Children represents an improvement over previous measures used to assess self-complexity in children in that it allows children to generate their own self-aspects, more sophisticated measures for assessing self-complexity in children should be developed. Last, it is important to note that the effect size for the interaction of negative self-complexity and negative events in the present study was in the small to medium range. Although effect sizes of this magnitude are common and even expected in the social science literature (Rosenthal, 1984), they also suggest that factors other than negative self-complexity play a role in the etiology of depressive symptoms in children as well. Future research should focus on identifying such factors and examining how they interact with negative self-complexity so as to develop a more comprehensive understanding of the etiology of depression in children.

In conclusion, past research on cognitive vulnerability to depression in children and adolescents has focused more heavily on the content rather than the structure of self-knowledge as a vulnerability to depression factor. Results from the current study suggest that the structure of negative information about the self may be equally important and can serve as a vulnerability to depression factor following the occurrence of negative events in early adolescents. Future research using more sophisticated assessments of stress, depressive symptoms, and self-complexity as well as larger sample sizes is likely to help us gain a deeper understanding of the role of the structure of self-knowledge in the etiology of depression in children and adolescents. In addition, future research should assess additional factors such as the relative importance of self-aspects, self-consciousness, and egocentrism to help us gain a deeper understanding of the emergence of negative self-complexity as a cognitive vulnerability to depre ssion during early adolescence.

[Graph omitted]

ACKNOWLEDGMENTS

The research reported in this paper was supported by a Research Development Grant awarded to John R. Z. Abela by the Faculty of Graduate Studies and Research at McGill University.

We would like to thank the students, principals, and teachers at the following schools for participation in this project: College Notre-Dame-de-Lourdes, College Regina-Assumpta, Ecole Gentilly, and Villa Sainte-Marcelline. In addition, we would like to thank Caroline Brunelle, Raquel Dorsaim-Pierre, Jacinthe Fortier, Toni Fried, Alyson Gentes-Hawn, Marie-Amelie Guilbault, Daniel Lavoie, Anne Lussier, Alexandra McIntyre-Smith, Tania Mazzarello, Adele Rochon, Caroline Sullivan, and Claudia Zambrana for conducting assessments and running depression prevention workshops. We would also like to thank Valerie Azoulay, Laurie Clark, Dean Elterman, Emily Haigh, Alison Leary, Dessa Miller, Zohreen Murad, Sharmeen Shah, Jane Simpson, Elizabeth Taxel, and Ella Vanderbilt for scoring questionnaires and entering data. Last, we would like to thank Chantale Bousquet for overseeing translation of the questionnaires and Susan Campbell, Barbara Sherwin, and two anonymous reviewers for comments on earlier versions of this manusc ript.

Received March 12, 2001; revision received July 10, 2001; accepted July 10, 2001

REFERENCES

Abela, J. R. Z., & Vdronneau-McArdle, M. (2000). Self-Complexity Scale for Children. Unpublished questionnaire.

Abramson, L. Y., Metalsky, G. L., & Alloy, L. B. (1989). Hopelessness depression: A theory based subtype of depression. Psychological Review, 96, 358-372.

Angold, A. (1988). Childhood and adolescent depression: I. Epidemiological and etiological aspects. British Journal of Psychiatry, 152, 601-617.

Atsushi, S. (1999). Complexity of self-representations as a cognitive buffer against depression and affective responses following life events. Japanese Journal of Educational Psychology, 47, 131-140.

Attneave, F. (1959). Applications or information theory to psychology New York: Holt-Dryden.

Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Harper & Row.

Campbell, J. D., Chew, B., & Stratchley, L. S. (1991). Cognitive and emotional reactions to daily events: The effects of self-esteem and complexity. Journal of Personality, 59, 473-505.

Carroll, J. B., Davies, P., & Richman, B. (1971). Word frequency book. New York: American Heritage.

Coddington, R. D. (1972). The significance of life events as the etiologic factors in the diseases of children, II: A study of a normal population. Journal of Psychosomatic Research, 16, 205-213.

Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences (2nd ed.), Hillsdale.

Cole, D. A., Martin, J. M., & Powers, B. (1997). A competency-based model of child depression: A longitudinal study of peer, parent, teacher, and self-evaluations. Journal of Child Psychology and Psychiatry, 38, 505-514.

Donahue, E. M., Robins, R. W., Roberts, B. W., & John, 0. P. (1993). The divided self: Concurrent and longitudinal effects of psychological adjustment and social roles on self-concept differentiation. Journal of Personality and Social Psychology, 64, 834-846.

Elkind, D. (1967). Egocentrism in adolescence. Child Development, 38, 1025-1034.

Elkind, D. (1978). Understanding the young adolescent. Adolescence, 49, 127-134.

Fiske, S., & Linville, P. (1980). What does the schema-concept buy us? Personality and Social Psychology Bulletin, 6, 543-557.

Furman, W., & Buhrmester, D. (1985). Children's perceptions of the personal relationships in their social network. Developmental Psychology, 21, 1016-1024.

Furman, W., & Buhrmester, D. (1992). Age and sex differences in perceptions of networks of personal relationships. Child Development, 63, 103-115.

Gara, M. A., Woolfolk, R. L., Cohen, B. D., Goldstein, R. B., Allen, L.A., & Novalany, J. (1993). Perception of self and other in major depression. Journal of Abnormal Psychology, 102, 93-100.

Garber, J., Weiss, B., & Shanley, N. (1993). Cognitions, depressive symptoms, and development in adolescents. Journal of Abnormal Psychology, 102, 47-57.

Hammen, C. L., & Gotlib, I. H. (1992). Psychological aspects of depression. New York: Wiley.

Hammen, C. L., & Zupan, B. A. (1984). SeIf-schemas, depression, and the processing of personal information in children. Journal of Experimental Child Psychology, 37, 598-608.

Harter, S. (1983). Developmental perspectives on the self-system. In E. M. Hetherington (Ed.), Handbook of child psychopathology: Vol. 4. Socialization, personality, and social development (pp. 275-385). New York: Wiley.

Harter, S. (1988), Developmental processes in the construction of the self. In T. Yawkey & J. Johnson (Eds.), Integrative processes and socialization: Early to middle childhood (pp.45-78). Hillsdale, NJ: Erlbaum.

Harter, S. (1990). Causes, correlates, and the functional role of global self-worth: A life-span perspective. In R. J. Steinberg & J. Kolligan, Jr. (Eds.), Competence considered (pp. 67-97). New Haven: Yale University Press.

Harter, S., & Monsour, A. (1992). Development analysis of conflict caused by opposing attributes in the adolescent self-portrait. Developmental Psychology, 28, 251-260.

Ingram, R. E. (1990). Self-focused attention in clinical disorders: Review and a conceptual model. Psychological Bulletin, 107, 156-176.

Jaycox, L. H., Reivich, K. J., Gillham, J., & Seligman, M. E. P. (1994). Prevention of depressive symptoms in school children. Behaviour Research and Therapy, 32, 801-816.

Jordan, A., & Cole, D. A. (1996). Relation of depressive symptoms to the structure of self-knowledge in childhood. Journal of Abnormal Psychology, 105, 530-540.

Kanner, A. D., Feldman, S. S., Weinberger, D. A., & Ford, M. E. (1987). Uplifts, hassles, and adaptational outcomes in early adolescents. Journal of Early Adolescence, 7, 371-394.

Kanner, A. D., Harrison, A., & Wertlieb, D. (1985, August). Development of the children's hassles and uplifts scales. Poster session presentation at the meeting of the American Psychological Association, Los Angeles, CA.

Kovacs, M. (1981). Rating scales to assess depression in school children. Acta Paedopsychiatrica, 46, 305-315.

Kovacs, M. (1983). The Children's Depression Inventory: A self-rated depression scale for school-aged youngsters. Unpublished manuscript, University of Pittsburgh School of Medicine, Pittsburgh.

Linville, P. W. (1985). Self-complexity and affective extremity: Don't put all of your eggs in one cognitive basket. Social Cognition, 3, 94-120.

Linville, P. W. (1987). Self-complexity as a cognitive buffer against stress-related illness and depression. Journal of Personality and Social Psychology, 52, 663-676.

Montemayor, R., & Eisen, M. (1977). The development of self-conceptions from childhood to adolescence. Developmental Psychology, 13, 314-319.

Morgan, H. J., & Janoff-Bulman, R. (1994). Positive and negative self-complexity: Patterns of adjustment following traumatic versus nontraumatic life experiences. Journal of Social and Clinical Psychology, 13, 63-85.

Mullener, N., & Laird, J. D. (1971). Some developmental changes in the organization of self-evaluations. Developmental Psychology, 5, 233-236.

Nolen-Hoeksema, S. (1991). Responses to depression and their effects on depressive episodes. Journal of Abnormal Psychology, 100, 569-582.

Pelham, B. W. (1993). The idiographic nature of human personality: Examples of the idiographic self-concept. Journal of Personality and Social Psychology, 64, 665-677.

Pelham, B. W., & Swann (1989). From self-conceptions to self-worth: On the sources and structure of global self-esteem. Journal of Personality and Social Psychology, 57, 672-680.

Pyszczynski, T., Greenberg, J., Hamilton, J., & Nix, G. (1991). On the relation between self-focused attention and psychological disorder: A critical appraisal. Psychological Bulletin, 110, 53 8-543.

Rosenthal, R. (1984). Mera-analytic procedures for social research. London: Sage.

Salovey, P. (1992). Mood-induced self-focus attention. Journal of Personality and Social Psychology, 62, 699-707.

Scott, W. A., Osgood, D. W., & Peterson, C. (1979). Cognitive structure: Theory and measurement of individual differences. Washington DC: Winston.

Showers, C. J., Abramson, L. Y., & Hogan, M. E. (1998). The dynamic self: How the content and structure of the self-concept change with mood. Journal of Personality and Social Psychology, 75, 478-493.

Woolfolk, R. L., Gain, M. A., Ambrose, T. K., Williams, J. E., Allen, L. A., Irvin, S.I., et al. (1999). Self-complexity and the persistence of depression. Journal of Nervous and Mental Disease, 187, 393-399.

Woolfolk, R. L., Novalany, J., Gara, M. A., Allen, L. A., & Polino, M. (1995). Self-complexity, self-evaluation, and depression: An examination of form and content within the self-schema. Journal of Personality and Social Psychology, 68, 1108-1120.

John R. Z. Abela (1,2) and Marie-Helene Veronneau-McArdle (1)

(1.) Department of Psychology, McGill University, Stewart Biological Sciences Building, Montreal, Quebec, Canada.

(2.) Address all correspondence to John R. Z. Abela, Department of Psychology, McGill University, Stewart Biological Sciences Building, 1205 Dr. Penfield Avenue, Montreal, Quebec, Canada H3A 1B1.

0091-0627/02/0400-0155/0 [c] 2002 Plenum Publishing Corporation

COPYRIGHT 2002 Plenum Publishing Corporation
COPYRIGHT 2002 Gale Group