Abstract
According to Blatt and others (e.g., A. T. Beck), self-definition, or one’s sense of self, and one’s sense of relatedness to others represent core lifespan developmental tasks. This study examined the role of events pertaining to self-definition or relatedness in the development of personality traits from each domain (self-criticism and dependency), and their relationship to the development of depressive and anxiety symptoms. Two hundred seventy-six early adolescents completed a measure of self-criticism and dependency at baseline and again 24 months later, along with measures of depressive and anxiety symptoms. Every 3 months, participants completed a measure of life events, which were coded as self-definitional or relatedness oriented (80% rater agreement, kappa = .70). Structural equation models showed that self-definitional events predicted increases in self-criticism, which in turn predicted increases in depressive symptoms, whereas relatedness events predicted increases in dependency, although dependency was unrelated to change in symptoms.
Considerable research into personality or cognitive factors that confer vulnerability to depression, anxiety, and other psychopathologies has been prompted by evidence that most people do not develop clinical levels of psychopathology even when confronted by serious stressors, such as the end of a romantic relationship, failing an exam, or the death of a loved one (Coyne & Downey, 1991; Monroe, Bromet, Connell, & Steiner, 1986). Although several personality or cognitive variables have been studied as vulnerabilities to depression and anxiety (cf. Clark & Beck, 1999; Ingram, Miranda, & Segal, 1998), self-criticism and dependency (Blatt, D’Afflitti, & Quinlan, 1976; Blatt & Zuroff, 1992) are among the most widely studied. Blatt et al. (1976) proposed that highly self-critical individuals are characterized by “intense feelings of inferiority, guilt, and worthlessness and by a sense that one has failed to live up to expectations and standards” (pp. 383-384). Dependent individuals are characterized by “feelings of helplessness and weakness, by fears of being abandoned, and by wishes to be cared for, loved, and protected” (Blatt et al., 1976, p. 383). In adults, these personality factors have shown retest reliabilities of between .75 and .80 at 3 and 12 months, respectively (Zuroff, Igreja, & Mongrain, 1990; Zuroff, Moskowitz, Wielgus, Powers, & Franko, 1983), and display retest stability even following stressful life events (Zuroff, Mongrain, & Santor, 2004). In adolescence, however, and consistent with other research showing that personality is less stable in adolescence compared with adulthood (e.g., Costa & McCrae, 1994; Roberts, Caspi, & Moffitt, 2001), these traits show more modest levels of stability over a 2-year period (dependency r = .51, self-criticism r = .60; Thompson, Zuroff, & Hindi, 2012).
Evidence has confirmed that self-criticism and dependency not only increase risk for depression (see Blatt, 2004; Zuroff et al., 2004, for reviews), but are associated with a wide variety of psychopathological and negative psychosocial outcomes, including lower education and occupational status, dissatisfaction in relationships, and social maladjustment (Fichman, Koestner, & Zuroff, 1994; Lassri & Shahar, 2012; Zuroff, Koestner, & Powers, 1994). They are also associated with bipolar depression (Rosenfarb, Becker, Khan, & Mintz, 1998), social anxiety (Cox et al., 2000; Kopala-Sibley, Zuroff, Russell, & Moskowitz, 2014), self-harm (Babiker & Arnold, 1997), anger and aggression (Gilbert & Miles, 2000), post-traumatic stress disorder (Sharhabani-Arzy, Amir, & Swisa, 2005), borderline personality disorder (Kopala-Sibley, Zuroff, Russell, Moskowitz, & Paris, 2012; Levy, Edell, & McGlashan, 2007), and chronic pain (Lerman, Shahar, & Rudich, 2012). Furthermore, in children and adolescents, self-criticism and dependency are longitudinally associated with higher levels of depressive symptoms and increase the effect of negative events on depressive symptoms (Abela, Fishman, Cohen, & Young, 2012; Abela & Taylor, 2003; Abela, Webb, Wagner, Ho, & Adams, 2006).
In parallel, researchers have sought to understand the origins of self-criticism and dependency in negative developmental experiences (e.g., Blatt & Homann, 1992; Koestner, Zuroff, & Powers, 1991; Kopala-Sibley, Zuroff, Leybman, & Hope, 2012; see Kopala-Sibley & Zuroff, 2014, for a review). Although this literature has been informative, several methodological weaknesses have been common, including a reliance on cross-sectional and retrospective data to answer developmental questions (e.g., Campos, Besser, & Blatt, 2010; Soenens, Vansteenkiste, & Luyten, 2010) and a lack of repeated measurements over the course of longitudinal studies (e.g., Koestner et al., 1991; Thompson, Zuroff, & Hindi, 2012). In addition, Blatt’s (2008; Blatt & Luyten, 2009) models of the development of self-criticism and dependency are explicitly domain specific; experiences of self-definition and relatedness should, respectively, predict the development of self-criticism and dependency (e.g., Luyten & Blatt, 2011). No research that we are aware of has tested this possibility. Understanding the developmental antecedents of these personality factors could therefore contribute to the early identification of those at risk for psychopathology as well as the development of preventive interventions. Accordingly, the goals of this article are, first, to test a domain-specific model of the development of dependency and self-criticism and, second, to examine the role of change in each personality variable in the development of depression and anxiety symptoms in a community sample of early adolescents.
The Development of Self-Criticism/Self-Definition and Dependency/Relatedness
Blatt and colleagues (Blatt, 1974, 2007; Blatt & Homann, 1992; Blatt & Luyten, 2009; Blatt & Shichman, 1983; Blatt & Zuroff, 1992) articulated a two-polarities model of personality development in which self-definition, or one’s sense of self, and relatedness, or one’s sense of relationships with close others, are fundamental psychological factors integral to healthy functioning as well as the development of psychopathology. According to Blatt (2007), relatedness involves the development of increasingly mature, intimate, mutually satisfying, reciprocal, interpersonal relationships from infancy through adulthood (Blatt & Luyten, 2009). However, delays or deficits in the development of a healthy sense of relatedness may lead to high levels of a personality style labeled dependency. Blatt (1974; Blatt & Homann, 1992) argued that dependency stems from developmental experiences of over-protection or a lack of care, as well as experiences of care being contingent on the child’s expressing affection toward the caregiver. These experiences engender a delay in the development of relatedness and a dependent personality style characterized by high levels of insecurity regarding close others and a sense of self-worth that is contingent upon the care and support of others.
In contrast, self-definition requires the development of the ability to hold a complex understanding of one’s self across different situations and domains of life (differentiated) in which one’s positive and negative qualities (e.g., failures or perceived inadequacies) as well as different selves across domains are reconciled into a cohesive sense of self (integrated), that is neither over-inflated nor overly self-deprecating (realistic) and that is, on the whole, positive (Blatt, 2007). That is, delays or a deficit in self-definition may lead both to pathologies characterized by low self-worth (e.g., depression), as well as to those characterized by an inflated sense of self (e.g., Narcissistic Personality Disorder; Blatt & Luyten, 2009; Ouimette, Klein, Anderson, Riso, & Lizardi, 1994). Delays or deficits in the development of self-definition may lead to high levels of a personality style labeled self-criticism. According to Blatt (1974), self-criticism stems from developmental experiences that undermine the emergence of autonomy via physical control, high performance expectations, and excessive criticism. The child experiences approval and acceptance as being contingent upon meeting those strict standards and becomes driven by the fear of loss of approval and acceptance. These developmental experiences may eventually lead to a deficit in self-definition and a highly self-critical personality style in which there is an excessive need to ascertain, confirm, and preserve status and value in the eyes of important others.
A large body of retrospective evidence suggests a role of negative developmental experiences as contributors to high levels of dependency and self-criticism (see Kopala-Sibley & Zuroff, 2014, for a review). For instance, dependent individuals are more likely to recall their parents as controlling (McCranie & Bass, 1984), over-protective (Campos et al., 2010), and as making love and approval contingent on the child’s expressing love for them (Soenens et al., 2010). Over and above parenting behaviors, better quality recalled peer attachments are related to lower adult levels of neediness and higher levels of connectedness, respectively, maladaptive and adaptive forms of dependency (Kopala-Sibley, Zuroff, Leybman, & Hope, 2012).
Highly self-critical individuals, too, are more likely to recall their parents as being strictly controlling or restrictive (Ahmad & Soenens, 2010; Klein et al., 2009; McCranie & Bass, 1984; Rosenfarb, Becker, & Mintz, 1994), critical (Whiffen & Sasseville, 1991), and as making approval contingent on meeting very high standards (Soenens et al., 2010). Over and above parenting, highly self-critical individuals recall more frequent overt and relational peer victimization during adolescence (Kopala-Sibley, Zuroff, Leybman, & Hope, 2013).
This body of research, although suggestive of an effect of developmental experiences on dependency and self-criticism, is limited by its retrospective nature. There are concerns about the accuracy of adult recall of developmental experiences, the inability to examine change over time, and the effects of other potentially confounding factors. Only three studies have examined this topic longitudinally. Koestner et al. (1991) showed that maternal reports of restrictive and rejecting parenting behavior at age 5 were related to self-criticism in the child at age 12, after controlling for infant temperament. More recently, Thompson et al. (2012) showed that, in a sample of adolescents followed from age 14 to 16, lower levels of instrumental and emotional support from parents at baseline predicted increases in dependency and self-criticism during that time period. Finally, Kopala-Sibley, Zuroff, Hermanto, and Joyal-Desmarais (2015) showed that the quality of one close friendship predicts changes in self-criticism over a 1-year period during emerging adulthood, while entering a new romantic relationship predicts decreases in neediness and increases in connectedness.
Beyond these limitations of past research, a core aspect of Blatt’s model of personality development is the role of domain-specific events or experiences. In the context of this article, “domain” refers to events falling under the category of either self-definition or relatedness, and does not refer to areas of life, such as school, peers, or family. From infancy to senescence, although particularly in adolescence, meaningful experiences in the domains of self-definition and relatedness should respectively affect the development of those aspects of personality. However, no research to date has attempted to examine the domain-specificity hypotheses of Blatt’s developmental model. Research has instead examined broad factors such as parental care and control or maltreatment and neglect. Results generally show broad effects of a variety of developmental experiences on both self-criticism and dependency (e.g., Campos et al., 2010; Kopala-Sibley, Zuroff, Leybman, & Hope, 2012; Kopala-Sibley, Zuroff, Leybman, & Hope, 2013; Thompson et al., 2012; see Kopala-Sibley & Zuroff, 2014, for a review). This is somewhat surprising, given that self-criticism and dependency are close to orthogonal, at least in adulthood (Zuroff et al., 2004), and are believed to stem from unique developmental experiences (Blatt & Homann, 1992). One possible way to resolve these findings is to consider the effects of domain-specific developmental experiences on the development of self-criticism and dependency.
Relatedness events are best understood as those that directly threaten close, supportive relationships with others. Self-definitional events, in contrast, directly threaten one’s sense of self by leading one to feel inferior or submissive to others, or by being controlled by others. Accordingly, the first goal of this article is to test a domain-specific model of personality development in which self-criticism and dependency change as a function of the respective occurrence of events pertaining to self-definition or relatedness. To be clear, however, we are not testing a diathesis-stress or congruency-vulnerability model of psychopathology (Zuroff et al., 2004); rather, we are examining how events in the domains of self-definition and relatedness affect the development of self-criticism and dependency, and how this, in turn, affects the development of depressive and anxiety symptoms. Our contention is that coding events based on their impacts on self-definition or relatedness will provide a theory-driven approach to understanding personality change and the development of symptoms of psychopathology in adolescence.
Self-Criticism, Dependency, Negative Events, and the Development of Psychopathology in Adolescence
No research to date has examined the role of personality change, as predicted by domain-specific events, in the development of psychopathology by classifying developmental experiences along the lines of self-definition and relatedness. In adults, self-criticism and dependency as well as negative events have shown well-replicated longitudinal relationships with symptoms of both depression and anxiety (see Blatt, 2004; Blatt & Zuroff, 1992; Zuroff et al., 2004, for reviews). Relatively few studies have examined the relationships between self-criticism, dependency, and internalizing symptoms in adolescents. In the few studies to examine these issues, self-criticism was found to be both cross-sectionally (Blatt, Schaffer, Bers, & Quinlan, 1992; Fichman et al., 1994) and longitudinally associated with higher levels of depressive symptoms in adolescent samples (Abela, McIntyre-Smith, & Dechef, 2003; Abela & Taylor, 2003; Cohen et al., 2013). Whereas dependency generally predicts depressive symptoms in older adolescents and adults (Blatt et al., 1992; Blatt & Zuroff, 1992; Fichman et al., 1994; Luthar & Blatt, 1995), it does not generally predict depressive symptoms in younger samples (Fichman, Koestner, & Zuroff, 1996), although some contradictory evidence exists (e.g., Abela et al., 2012).
Thus, evidence suggests an important role for self-criticism in understanding early adolescent depressive symptoms, whereas the role of dependency is less clear. One possibility is that in early adolescence or late childhood, dependency could be related to other internalizing problems, such as anxiety, rather than being related to depressive symptoms. For instance, dependency, which taps concerns about attachment and the support and availability of caring others, could be a particularly important issue in early adolescence as youth transition from having their parents as their main attachment figures to developing attachments to peers and romantic partners (Furman & Buhrmester, 1992; Markiewicz, Lawford, Doyle, & Haggart, 2006). Moreover, depression and anxiety are differentiated in that only depression predicts low levels of positive affect. Similarly, dependency and self-criticism are both associated with negative affect, whereas only self-criticism is inversely associated with positive affect (e.g., Kopala-Sibley & Santor, 2009; Mongrain & Zuroff, 1994). A recent study found that, in a sample of Chinese adolescents, whereas self-criticism predicted depressive symptoms, dependency predicted social anxiety symptoms (Cohen et al., 2013). It is therefore possible that dependency may predict anxiety symptoms in early adolescents rather than depressive symptoms. Finally, although evidence has shown that self-criticism longitudinally predicts depressive symptoms, no research has examined whether the development of these personality factors, or change in them over time, predicts symptoms of psychopathology. The second goal of this article, therefore, is to test the relationships between negative self-definitional and relatedness-oriented events, changes in self-criticism and dependency in early adolescence, and the relationship of these changes to the development of both depressive and anxiety symptoms.
Distinguishing Self-Definition and Relatedness From Neuroticism
Concerns have been raised over whether self-definition and relatedness are distinct from broader personality traits, such as neuroticism (e.g., Coyne & Whiffen, 1995). Although this issue has been addressed elsewhere (Zuroff et al., 2004), it bears repeating here. Self-criticism is indeed moderately related to neuroticism (Pearson rs of approximately .4-.6; Mongrain, 1993; Zuroff, 1994), although dependency is typically less strongly related to neuroticism (Bagby & Rector, 1998; Clara, Cox, & Enns, 2003). Despite the overlap of self-criticism with neuroticism, across several studies self-criticism has shown incremental utility in predicting outcomes. For instance, controlling for neuroticism, self-criticism predicts depressive symptoms, the occurrence of major depression, and global psychosocial functioning in depressed patients (Dunkley, Blankstein, Zuroff, Lecce, & Hui, 2006) and in college students (Clara et al., 2003; Mongrain & Leather, 2006), although one recent study showed that self-criticism moderates the effects of stress on depression, whereas neuroticism does not (Békés et al., 2015). Over and above neuroticism, self-criticism also predicts social phobia (Cox et al., 2000), interpersonal behavior (Kopala-Sibley et al., 2013), and negative affect in Borderline Personality Disorder patients (Kopala-Sibley et al., 2012). Finally, both self-criticism and dependency show age-related decreases from age 18 to 59, even after adjusting for age-related differences in neuroticism (Kopala-Sibley, Mongrain, & Zuroff, 2013). In sum, although self-criticism, dependency, and neuroticism show substantial shared variance, self-criticism and dependency are best considered unique risk factors for various psychosocial outcomes, over and above neuroticism.
Overview and Hypotheses
Based on Blatt’s (1974, 2007, 2008; Blatt & Luyten, 2009; Blatt & Zuroff, 1992) developmental model of self-definition and relatedness and their role in psychopathology in adolescence, this article had two goals. First, we examined a domain-specific model of the role of self-definition and relatedness-oriented experiences in change in self-criticism and self-definition over a 2-year period in early adolescence. The second goal was to examine the independent contributions of change in dependency and self-criticism and the occurrence of negative events in the domains of self-definition and relatedness over this 2-year period to the development of symptoms of anxiety and depression. Our model allowed us to test the possibility that over a relatively long time period (i.e., 2 years), negative events may result in personality changes, and it is these changes in personality that account for the effects of chronic negative life events on the development of internalizing symptoms.
We expected that higher levels of negative self-definitional events over a 2-year period would be specifically related to greater increases in self-criticism over that time period, whereas higher levels of negative relatedness events would be related to greater increases in dependency. We further expected change in self-criticism to predict symptoms of both anxiety and depression. We also expected change in dependency to predict change in anxiety symptoms, although we had no a priori prediction regarding the effect of dependency on depressive symptoms, given the mixed past evidence on this topic in adolescent samples. Finally, we expected higher levels of negative self-definitional and relatedness-oriented events to predict both depressive and anxiety symptoms. We expected the effects of self-definitional events to be at least partially mediated by change in self-criticism, whereas we expected the effects of relatedness events to be at least partially mediated by change in dependency.
Method
Participants
Participants were recruited from Montreal, Quebec, Canada, and Chicago, Illinois. Participants were recruited through ads placed in local newspapers and around the local community. Ads asked if readers “would like to help us understand the factors that impact an adolescent’s development.” No other details of the purposes of the study were stated in the advertisements. Participants in the current study were drawn from a larger sample of 382 youth (see Abela & Hankin, 2011), of whom a subset completed the Depressive Experiences Questionnaire for Adolescents (DEQ-A) at baseline. Due to the use of newspaper advertisements for recruitment, we cannot know the response rate to our ads. There were no differences on any demographic variable or on any measure of internalizing symptoms or levels of hassles experienced over our follow-up between those who completed and those who did not complete the DEQ-A (all ps > .05). Our effective sample therefore comprised 241 adolescents (138 boys) between the ages of 11 and 14 (M = 12.57 years, SD = 1.11 years). Full details of the demographic characteristics of both samples are reported in Abela and Hankin (2011). The Montreal participants were 76.8% Caucasian, whereas the Chicago participants were 54.7% Caucasian. The Montreal and Chicago samples were comparable in terms of adolescent gender, age, grade, highest level of education completed by mother, highest level of education completed by father, and family income. At the same time, the Chicago sample consisted of a greater proportion of ethnic minority youth, and youth from single-parent households (Abela & Hankin, 2011).
Procedure
Phase 1 of the study involved an initial laboratory assessment. Two research assistants met with one adolescent–parent pair at a time. Parents completed a consent form for themselves and their child; the adolescent completed an assent form. Parents also completed a demographics form. A research assistant verbally administered the Children’s Depression Inventory (CDI; Kovacs, 2004), the 10-item Multidimensional Anxiety Scale for Children (MASC-10; March, Parker, Sullivan, Stallings, & Conners, 1997), and the Depressive Experiences Questionnaire for Adolescents–Short Form (DEQ-A-SF; Fichman et al., 1994) to the adolescent while the adolescent completed their own copy.
Phase 2 of the study involved a series of eight telephone follow-up assessments. Assessments occurred every 3 months during the 2 years following the initial assessment. At each assessment, a research assistant verbally administered the Adolescent Life Events Questionnaire (ALEQ; Hankin & Abramson, 2002) to adolescents. At baseline, 241 youth participated. Of these, 208 (85.2%) completed measures at the 2-year follow-up. The average number of follow-up assessments completed by participants was 6.74 (SD = 1.61). The percentage of follow-up assessments completed by participants was distributed as follows: 0.8% completed none, 1.6% completed one, 1.8% completed two, 1.0% completed three, 3.1% completed four, 6.0% completed five, 15.7% completed six, 30.4% completed seven, and 39.5% completed eight. We used Little’s Missing Completely at Random (MCAR; Little, 1988; Little & Rubin, 1987) test in SPSS 20.0 and included baseline and follow-up self-criticism and dependency, depression and anxiety, and total self-definitional and relatedness hassles. Results confirmed that data were MCAR, χ2(83) = 96.51, p = .147. Data were thus viewed as missing at random for analyses. Full Information Maximum Likelihood (FIML) procedures in AMOS 5.0 were used to estimate the means and intercepts in the presence of missing data. This approach to missing data is generally acknowledged to be preferable to other methods for dealing with missing data, such as listwise deletion or data imputation, as these latter approaches are more likely to yield severely biased estimates, whereas estimates based on FIML are less biased relative to these other procedures (Little & Rubin, 1989; Muthen, Kaplan, & Hollis, 1987; Schafer, 1997). At the final assessment (i.e., Time 2) 24 months after baseline, the DEQ-A-SF, the CDI, and the MASC were again verbally administered to each participant who completed his or her own form manually. Parents and adolescents were compensated US$200 for participation.
Materials
Dependency and self-criticism
Dependency and self-criticism were assessed with the DEQ-A-SF (Fichman et al., 1994). In the full version of the DEQ-A, the 66 items of the adult Depressive Experiences Questionnaire (Blatt et al., 1976) were rephrased to make the items more appropriate for adolescents (Blatt et al., 1992). Fichman et al. (1994) derived a 20-item version from the 66-item DEQ-A using the 16 items on the DEQ-A with the highest loadings on self-criticism and dependency that also showed low cross-loadings. Four items from a third factor, efficacy, were also included as buffers. An example of an item loading on the dependency factor is “I worry a lot about upsetting or hurting someone who is close to me.” An example of a self-criticism item is “I often find that I fall short of what I expect of myself.” Participants are asked to judge the extent to which they agree or disagree with each statement on a 7-point scale. In a sample of 12- and 13-year-olds, the dependency and self-criticism subscales of the DEQ-A-SF showed acceptable internal consistency, with alphas of .65 and .70, respectively, while correlating highly with the original full scales from the adult DEQ (r = .77 for self-criticism and r = .65 for dependency; Fichman et al., 1994). The scales showed only a modest correlation (r = .30; Fichman et al., 1994), suggesting they are acceptably independent from one another. In the current study, they correlated at .51 and .40 at Times 1 and 2, respectively. The dependency and self-criticism scales have also shown good convergent validity in several studies (e.g., Blatt et al., 1992; Fichman et al., 1994) and, despite their moderate correlation with each other, are uniquely associated with different psychosocial outcomes in adolescents (Fichman, Koestner, & Zuroff, 1996, 1997). In the current study, alphas for dependency were .75 and .74 at Times 1 and 2, respectively. For self-criticism, alphas for Times 1 and 2, respectively, were .82 and .72.
Depressive symptoms
Symptoms of depression were assessed with the CDI (Kovacs, 2004). The CDI is a widely used 27-item self-report measure of the cognitive, affective, and behavioral symptoms of depression in children and youth. Each item asks children to indicate the extent to which they were thinking, feeling, or acting in a particular way in the past week. Item response options range from 0 to 2, with higher scores indicating greater symptom severity. The CDI reliably distinguishes children with major depression from those without (Saylor et al., 1984). In the current study, coefficient alpha ranged from .87 to .91 across administrations.
Anxiety symptoms
The MASC-10 is a 10-item version of the 39-item MASC. The MASC-10 items, which tap physiological symptoms, social anxiety, harm avoidance, and separation/panic, are summed to form a global anxiety symptom score. Children are asked to rate the extent to which each of the 10 statements is true about them on a scale from 0 (never true about me) to 3 (often true about me). Total scores range from 0 to 30, with higher scores indicating higher levels of anxiety. The MASC-10 has adequate internal consistency (α = .60-.85), good test–retest reliability (r = .79-.93), and good convergent and discriminant validity (March et al., 1997). In the current study, alphas for total MASC-10 scores were .64 at Time 1 and .77 at Time 2.
Life events/hassles
The ALEQ (Hankin & Abramson, 2002) was used to assess the occurrence of hassles and major events that are likely to occur in the lives of adolescents, including school problems (e.g., “You got into trouble with the teacher or principal”), relationship difficulties (e.g., “You found out your boyfriend/girlfriend was cheating on you”), and family problems (e.g., “You had an argument with a close family member [parent, sibling]”). Each of the 57 events is rated for frequency in the past 3 months on a Likert-type scale ranging from A (never) to E (always). The ALEQ shows good internal consistency (α = .94; Hankin & Abramson, 2002) and predicts symptoms of both depression and anxiety in youth (e.g., Hankin, 2008; Hankin, Stone, & Wright, 2010).
Events and hassles were coded based on definitions derived from theoretical models of self-definition and relatedness as articulated by Blatt and colleagues (Blatt, 2004, 2007, 2008; Blatt & Luyten, 2009; Luyten & Blatt, 2011). Definitions (see the appendix) were written by the first two authors of this study. Seven independent raters at the doctoral or post-doctoral level in clinical psychology were asked to code the 57 items of the ALEQ as self-definition related, relatedness related, or neither. No authors of this article were involved in the coding of events. Of the 57 items, 36 were agreed upon by five, six, or all seven raters, and those items were retained. One of those items (“Had a bad teacher”) was coded as “neither” by all seven raters and so was dropped. The remaining 35 items as well as the category to which they were coded are shown in Table 1. Inter-rater reliability was established by computing free-marginal multi-rater kappa (Randolph, 2005). This kappa is appropriate for assessing inter-rater reliability as it adjusts for the amount of agreement that could be expected to occur through chance. Coding in the current study resulted in a free-marginal kappa of .70 based on the 35 items that had adequate agreement across raters, suggesting good inter-rater reliability based on standard cutoffs (e.g., Brennan & Prediger, 1981; Landis & Koch, 1977). Overall rater agreement was 80%. Scores were summed across time points to create a total measure of life events in each domain across the first seven measurement time points. This was done because we were interested in the total, cumulative effects of self-definitional and relatedness hassles over the 24-month interval. The eighth time point was dropped from analyses to avoid any potentially confounding effects that may arise from concurrently measuring dependency, self-criticism, depression, anxiety, and negative events.
Final Items Used as a Measure of Self-Definitional and Relatedness Events, and Into Which Domain They Were Coded.
Data Analyses
Primary data analyses consisted of cross-lagged structural equation panel models (Figure 1). Cross-lagged panel models are appropriate for examining predictors of change over time. By controlling the effect of Time 1 scores (e.g., self-criticism) on Time 2, predictors of that Time 2 score are then predicting the residual, or change, in that score from Time 1 to Time 2. Thus, models simultaneously examined both the effects of life events assessed over the 24-month study period on change in dependency and self-criticism, as well as the effects of change in dependency and self-criticism on change in depressive and anxiety symptoms from baseline to the 24-month follow-up. By “change” in self-criticism and dependency, we refer to the effects of Time 2 self-criticism and dependency on Time 2 depression and anxiety while adjusting for the effects of Time 1 self-criticism and dependency on depression and anxiety. We also adjust for the effects of Time 1 depressive and anxiety symptoms on change in personality. This path model also allowed us to examine the effects of life events in each domain on change in depression and anxiety symptoms, and whether this effect was mediated by change in dependency or self-criticism. Our initial model also included the effects of gender as well as age at baseline on change in self-criticism, dependency, and depression and anxiety symptoms. However, neither was related to change in any dependent variable, and so age and gender were dropped from the final model. Multi-group models would be preferable for examining gender differences in our models; we analyzed our model within each gender, but these models did not converge, and yielded a non-positive definite covariance matrix. This may be due to an inadequate sample size within each gender for our model.

Structural Modeling Equation (SEM) results.
Analyses were conducted using AMOS 5.0 (Arbuckle, 2003). To measure the latent variables in the model, we created item parcels following the recommendations of Little, Cunningham, Shahar, and Widaman (2002). For our life events as well as depressive and anxiety symptoms measures, items were randomly assigned to parcels. This approach is appropriate for unidimensional measures that are unit weighted (Little et al., 2002). For both measures, and at each time point, items were randomly assigned to one of three parcels. The same randomization was used at each time point. For the DEQ-A-SF, at both time points, three item parcels were created based on the factor loadings found in Fichman et al. (1994) so that the average loadings were nearly equal across parcels. The approaches of random assignment of items to parcels for unidimensional item-weighted measures and of the distribution of items to parcels based on their factor loadings both yield parcels that are equally balanced in terms of the factor loadings of the items that make up the parcels, and are appropriate when mean levels of indicators are of little concern and when analyses are specified a priori (Little et al., 2002).
Power analyses found that a sample of 200 would be adequate to detect moderate effects and to adequately assess our measurement model given 12 latent variables and 36 observed variables (Cohen, 1988; Soper, 2014; Westland, 2010).
As measures of goodness of fit, we present chi-square, ratio of chi-square to degrees of freedom, comparative fit index (CFI), and root mean square error of approximation (RMSEA). Generally, CFI values greater than .90 (Hoyle & Panter, 1995), a χ2/df less than 2 (Carmines & McIver, 1981), and a RMSEA less than .08 (Kline, 1998) indicate acceptable fit. Figure 1 shows our model. Our model was fully saturated, allowing all independent variables to predict all dependent variables, to confirm that paths that were expected to be non-significant were in fact not significant (e.g., from relatedness hassles to change in self-criticism). Although bootstrapping would be preferable for examining the significance of indirect effects of self-definitional and relatedness hassles on depression and anxiety via self-criticism and dependency, missing data in the current sample preclude this option. Instead, we used the Monte Carlo Method for Assessing Mediation (MCMAM; MacKinnon, Lockwood, & Williams, 2004). In this approach, random draws from distributions of the two indirect effects are simulated and the product of these values is computed. This procedure is repeated a very large number of times and the resulting distribution of the indirect effect is used to estimate a confidence interval (CI) around the observed value of the indirect effect. MCMAM performs better than the Sobel test, and comparably with bootstrap approaches (MacKinnon, Lockwood, & Williams, 2004). In the current study, we computed a 95% CI with 20,000 repetitions. We also present CIs for each significant regression effect.
Results
Descriptive Statistics and Bivariate Correlations
Descriptive statistics and bivariate correlations are presented in Table 2. Older age was associated with more self-definitional and relatedness events, higher symptoms of depression at Time 2, and higher levels of self-criticism at Time 2. Times 1 and 2 self-criticism correlated positively, as did Times 1 and 2 dependency. Although higher levels of self-definitional hassles were associated with higher levels of both self-criticism and dependency, self-definitional hassles tended to correlate more strongly with self-criticism at both Times 1 and 2 than they did with dependency at either Time 1 (z = 1.84, p = .06, 95% CI [.03, .27]) or Time 2 (z = 1.87, p = .05, 95% CI [.02, .28]). Relatedness hassles, however, were equally strongly associated with both dependency and self-criticism at Times 1 (z = −.248, p = .80, 95% CI [−.10, .14]) and 2 (z = .125, p = .90, 95% CI [−.12, .14]). Higher levels of both Time 2 dependency and self-criticism were associated with higher levels of both depressive and anxiety symptoms at Time 2. Finally, higher levels of both self-definitional and relatedness hassles were associated with higher levels of depressive and anxiety symptoms at Time 2.
Bivariate Correlations and Descriptive Statistics for all Variables.
Note. Correlations and means presented for total, raw, observed scores. DEP = dependency; SC = self-criticism; CDI = Children’s Depression Inventory; MASC = Multidimensional Anxiety Scale for Children.
p <.05. **p < .001.
Effects of Self-Definition and Relatedness Events
Results are presented in Figure 1. This model yielded the following acceptable fit indices: χ2(366, N = 241) = 606.346, p < .05, χ2/df = 1.66, CFI = .940, RMSEA = .052. All factor loadings exceeded .57, and all were significant (p < .001). Standardized regression coefficients are presented in Figure 1, although they are also provided here (β), along with unstandardized regression estimates (b) and 95% CIs. Results showed that higher levels of self-definitional events specifically predicted increases in self-criticism, b = .05, β = .31, 95% CI [.001, .096], whereas higher levels of relatedness events specifically predicted increases in dependency, b = .05, β = .26, 95% CI [.001, .10]. Higher levels of self-criticism at Time 2 predicted increases in depressive symptoms, b = 22, β = .42, 95% CI [.07, .37], although neither personality trait predicted increases in anxiety symptoms. Finally, lower levels of self-definitional, b = −.07, β = −.69, 95% CI [−.13, −.01], and higher levels of relatedness hassles, b = .096, β = .61, 95% CI [.01, .18], independently predicted increases in anxiety symptoms. The effect of self-definitional hassles on anxiety symptoms was negative, but this is likely due to suppression effects, given the positive zero-order correlation between these two variables (Table 2). Finally, the indirect effect of self-definitional events on depressive symptoms at Time 2 via changes in self-criticism was significant, MCMAM 95% CI [.001, .02].
Figure 1 presents the standardized parameters for all significant effects. The model is best understood by starting with Time 1 self-criticism and dependency. These personality factors were both related to higher levels of self-definitional and relatedness events over the 2-year study interval. However, these events then uniquely contributed to change in self-criticism versus dependency. Specifically, and in support of our primary hypotheses, self-definitional events assessed over 24 months uniquely predicted increases in self-criticism, which in turn predicted increases in depressive symptoms, whereas relatedness events uniquely predicted increases in dependency and in anxiety symptoms, although dependency was unrelated to the development of anxiety symptoms.
Ancillary Analyses Examining the Specificity of Effects
To further examine whether results showed specific effects of events on personality and of personality on symptomology, we examined whether or not self-definitional and relatedness events affect their congruent personality domains more strongly than their non-congruent personality domains. We computed pairwise comparisons of regression paths. Pairwise comparisons compute the critical ratio for comparisons between two parameters. These critical ratios follow a standard normal distribution (i.e., values more than 1.96 are significant at p = .05). We tested a model in which self-definitional and relatedness events both predicted change in both personality factors. Results showed that the effect of self-definitional events on self-criticism was significantly stronger than its non-significant effect on dependency, z = −2.23, p < .05. The effect of relatedness events on dependency trended toward being significantly greater than its non-significant effect on self-criticism, z = 1.86, p = .06.
Discussion
This is the first study to examine the role of domain-specific events in changes in self-criticism and dependency over a 2-year period in a group of early adolescents, and whether changes in self-criticism and dependency predicted change in depressive and anxiety symptoms over the same period. This is also the first study to examine the impact of developmental experiences from a variety of areas of life (e.g., parents, siblings, romantic partners, peers, academic success) on these personality factors, as opposed to solely parents (e.g., Thompson & Zuroff, 2012) or parents and peers (e.g., Kopala-Sibley et al., 2012, 2013).
Some initial findings merit mention prior to the primary findings. First, we were able to reliably code events into self-definition and relatedness domains. This approach to coding life events is in accordance with a sizeable body of theory pertaining to self-definition and relatedness (e.g., Beck et al., 1983; Blatt, 1974; Blatt & Homann, 1992; Blatt & Luyten, 2009). Second, we found only moderate test–retest correlations for both dependency (r = .42) and self-criticism (r = .48) over a 2-year period during adolescence. These are slightly lower, although not statistically different (both ps > .15), than 2-year stabilities from age 14 to 16 (dependency r = .51, self-criticism r = .60; Thompson & Zuroff, 2012) and significantly lower than 1-year stabilities found in college students (.79 for both dependency and self-criticism, Zuroff et al., 1990, z delta for dependency = 3.90, p < .001, z delta for self-criticism = 3.43, p < .001). Although we would need to follow a sample of adolescents from age 12 to early adulthood to truly compare age-related changes in self-criticism in dependency, retest reliabilities in the current study taken together with others (Thompson & Zuroff, 2012; Zuroff et al., 1990) suggest that both self-criticism and dependency become increasingly stable from early adolescence to early adulthood. The finding of moderate stability in personality traits during adolescence is also consistent with the broader personality literature (e.g., Roberts & DelVecchio, 2000; Roberts et al., 2001). Although not the purpose of this article, results were also consistent with prior research examining the relationship between self-criticism, dependency, and negative events (e.g., Cohen et al., 2013; Mongrain & Zuroff, 1994; Starrs et al., 2010) in that baseline self-criticism and dependency were associated with elevated levels of both relatedness and self-definitional events, although dependency was more strongly associated with relatedness events than with self-definitional events.
This latter finding suggests that self-critical and dependent adolescents may generate negative events in a non-domain-specific manner. That is, although Blatt’s developmental model is domain specific, the associations between both personality factors and both types of events suggests that these individuals may generate negative events in both domains. Furthermore, self-definitional and relatedness events correlated substantially. This is consistent with a variety of prior research, which has generally found that negative events in one area of life are strongly and positively associated with negative events in other areas of life (e.g., Cohen et al., 2013; Gini, 2008; Kopala-Sibley et al., 2013; Storch, Crisp, Roberti, Bagner, & Masia-Warner, 2005). However, the developmental effects of self-definition and relatedness events on personality, despite their high correlation, then appear to be domain congruent. Thus, although the generation and occurrence of negative events in each domain may not be specific, their developmental effects on personality and internalizing symptoms are unique and domain specific.
The Development of Self-Criticism and Dependency
Findings from this study suggest that experiences in self-definitional and relatedness domains, from a variety of areas of life, influence the development of self-criticism and dependency over 2 years in early adolescence. Results support Blatt’s recent revisions (Blatt, 2007; Blatt & Luyten, 2009) to his earlier models (Blatt, 1974) in that current theory emphasizes that self-definition and relatedness are influenced not only by early experiences of parenting (Blatt, 1974), but by developmental experiences from a variety of domains throughout early adolescence and, indeed, the entire lifespan (Kopala-Sibley, Mongrain, & Zuroff, 2013). Self-definitional events may exert their influence by restricting the adolescent’s still emerging autonomy; that is, individuals may feel controlled or coerced into behaving a certain way or achieving to a certain degree to garner the respect and admiration of others, be they parents, peers, or romantic partners. This may result in the failure to separate one’s sense of self from the views of others, resulting in a sense of self-worth that is contingent upon others’ respect or admiration as well one’s perceived status relative to others. Self-definitional events, such as doing poorly in school, may challenge one’s view of oneself as competent and, over time, ultimately result in the aspects of self-criticism that motivate individuals to avoid failure so that others will respect them. Adolescents may further learn that others will only respect or admire them if they meet the standards set by others, whether those pertain to academic achievement or behaving in specific ways. This may ultimately lead to increases in self-criticism, with concomitant needs for social status and beliefs that achievement and acting in specific prescribed ways may garner this status.
According to Blatt’s model, relatedness requires the development of the capacity for stable, mature, and reciprocal close relationships with others. Consistent with this hypothesis, our results suggest that this process may be derailed when close relationships are unstable or others are perceived as unavailable or uncaring. It is important to note here that our results suggest that threats to relationships that may be the fault of neither the youth nor the other person, such as serious illness or the death of an emotionally significant person, may also engender dependency over time. For instance, the death or illness of a parent may leave the adolescent with a lasting fear that others will leave or abandon him or her. That is, both intentional and unintentional withdrawal of care or support appear to ultimately result in the development of a more dependent personality style characterized by self-worth based on receiving love, care, or support from others, as well as feelings of helplessness without that support.
Finally, we cannot know from the current study whether or not influences of different domains of events would predict personality change independent of or in interaction with youths’ genetics (Holden, 1987; Krueger & Markon, 2006; Plomin, Caspi, Pervin, & John, 1999). It is possible that our personality variables, as well as the occurrence of events, may be influenced by children’s genetics (e.g., Kandler, Bleidorn, Riemann, Angleitner, & Spinath, 2012), and future research would do well to delineate the relative contribution of both to the development of self-criticism and dependency.
Self-Criticism, Dependency, and the Development of Depressive and Anxiety Symptoms
The results of this study also contribute to our understanding of the role of hassles and personality development in the development of depressive symptoms in youth. Self-definitional events over the course of 2 years uniquely predicted increases in depressive symptoms at follow-up via change in self-criticism. That increased self-criticism mediates the effects of developmental experiences on depressive symptoms is consistent with several studies that have found the effects of negative developmental experiences on depressive symptoms to be accounted for by levels of self-criticism (e.g., Campos et al., 2010; Enns, Cox, & Larsen, 2000; Soenens et al., 2010). Given that self-definitional events and self-criticism uniquely predicted increases in depressive symptoms but not anxiety, depressive symptoms in early adolescence may stem primarily from concerns about self-worth, achievement, and social status.
However, adjusting for the effects of self-criticism and relatedness and self-definitional hassles, dependency was unrelated to the development of depressive or anxiety symptoms, although dependency showed zero-order correlations with both. These results may be consistent with prior evidence that dependency is relatively less maladaptive compared with self-criticism (Zuroff et al., 2004), and with at least one study in which dependency did not mediate the effects of developmental experiences on depression (e.g., Campos et al., 2010). However, relatedness hassles did show a significant effect on change in anxiety symptoms, whereas there was no effect of self-definitional hassles or self-criticism or dependency. This may suggest that generalized anxiety symptoms in early adolescence may stem primarily from hassles or negative events that threaten the quality and existence of close attachments, be they to parents, peers, or romantic partners, although this relationship does not appear to be mediated by personality development.
These results also provide an alternative lens through which to understand internalizing symptoms compared with diathesis-stress or personality-vulnerability models (e.g., Blatt & Zuroff, 1992), according to which personality factors are “triggered” by and increase the effects of negative events on symptoms of psychopathology. Our results suggest that negative events, rather than only serving as triggers, may also influence the development of personality (i.e., self-criticism) over time, which in turn influences internalizing symptoms. However, this interpretation must be considered tentative as personality and symptoms were assessed concurrently at Time 2, so we cannot establish the temporal order of the relationship between personality and symptoms. That being said, increases in self-criticism over time and concomitant increases in concerns about failure, social status, and low self-worth appear to have main effects on depression symptoms. This is not to suggest that self-criticisms should not still be conceptualized in part as a diathesis, which moderates the effects of events. Rather, it may be understood as having both moderating and main effects on the development of internalizing symptoms (Zuroff et al., 2004). This conceptualization is consistent with both a sizable diathesis-stress literature (see Blatt, 2004) as well as more recent stress-generation research (e.g., Cohen et al., 2012; Starrs et al., 2010) and studies of “scar” models, in which events or symptoms contribute to increases in the psychological vulnerability itself (see Klein, Kotov, & Bufferd, 2011, for a review).
Results from this study may also have practical implications. Given that increases in self-criticism predicted increases in depression, results support early interventions designed to identify and treat those youth who show elevated levels of self-criticism prior to the development of clinically significant levels of internalizing symptoms. On one hand, youth may benefit from interventions designed to directly reduce levels of dependency or self-criticism, such as self-compassion-focused exercises (e.g., Kelly, Zuroff, Foa, & Gilbert, 2010). On the other hand, rather than targeting personality directly, interventions may seek to bolster interpersonal effectiveness as well as academic success to diminish the chances of developing a self-critical personality. More generally, fostering experiences of mastery and competence as well as relationships that are non-critical and autonomy supportive may prevent the development of self-criticism.
Limitations and Future Directions
Although this study had some notable strengths, including its 2-year, longitudinal design, as well as the frequent assessment of developmental experiences during the interim, some limitations should be noted. One limitation is the exclusive use of self-report measures to assess depression and anxiety symptoms as well as negative events, which may be over-reported by those with higher levels of self-criticism, dependency, depression, or anxiety. Similarly, the study of developmental experiences would be strengthened with more objective, interview-based contextual threat data. For instance, it is plausible that an event such as failing an exam would affect relatedness if one feels that others will be less caring because of it. Furthermore, interview-based measurement of life events would likely yield stronger inter-rater agreement in terms of coding events, as well as a better distinction between the two categories of events.
We should also acknowledge that we cannot make strong causal statements about the effects in this study as there may be other, unmeasured factors that account for our findings. However, we were able to separate and control for the covariation of baseline personality and negative events over the follow-up, and results still confirmed an effect of events on change in personality. By using SEM models, we were able to adjust for the covariance between the two domains of life events, as well as for the overlap between symptoms of depression and anxiety. We would also emphasize that the follow-up measurements of symptoms were taken after the measurement of cumulative negative events as we did not include those events that were measured at the same time points as symptoms in our measure of events.
Furthermore, although our assessment of negative events every 3 months is a substantial improvement over studies that ask participants to recall events over the course of years (e.g., Thompson et al., 2012), we cannot completely rule out problems with retrospective recall in terms of accuracy, as well as the possibility that the recall of more highly dependent or self-critical participants may be colored by their personality or symptomatology. However, it may be impractical to assess youth much more frequently than was the case in this study while still following them for a period of years. Furthermore, analyses of repeated measures such as our life events measure would ideally be analyzed using multi-level models, which could examine within-subject effects. However, because self-criticism and dependency were measured twice, we could not assess within-subject changes in them, and so panel models were the most appropriate analyses. Related to this, personality and symptoms were measured concurrently at Time 2. Ideally, to establish mediation of the effects of events on symptoms via change in personality, depression and anxiety would be measured at some time point subsequent to the second wave of personality assessment. Finally, although we controlled for the effects of gender, differential effects across genders would be ideally tested using multi-group SEMs. We tested these, but they did not converge. This issue may be resolvable using larger sample sizes of each gender. Future studies would also benefit from measuring self-criticism and dependency multiple times over the study interval to examine within-subject change.
Another issue is the moderate correlations between self-criticism and dependency found in this sample. Although this is likely due to the unit-weighted scoring approach of the short form of the DEQ-A, it raises concerns about the overlap of self-criticism and dependency in adolescents, which Blatt’s theory portrays as essentially independent (Blatt, 2004). Ideally, the full-length, factor-scored measure of dependency and self-criticism in adolescents (Blatt et al., 1992) would have been used, which shows minimal association between the two traits, similar to what is typically found using the adult DEQ (Zuroff et al., 2004).
One issue not addressed in this study is the role of genetics in the development of self-definition and relatedness. That personality is heritable, and influenced by specific genetic polymorphisms has been well-established in the literature (e.g., Holden, 1987; Krueger & Johnson, 2008; Plomin et al., 1999). No research to date has examined the heritability of self-definition and relatedness, their relation to specific genetic polymorphisms, or whether they may represent the phenotypic expression of an interaction between specific genetic polymorphisms and negative developmental experiences, although theorists have suggested these possibilities (Blatt & Luyten, 2009; Luyten & Blatt, 2013). Similarly, specific genetics may evoke more negative environments via gene-environment correlations (e.g., Jaffee & Price, 2007). Thus, genetic factors may influence the occurrence of negative developmental experience via their effects on maladaptive personality development (Kopala-Sibley & Zuroff, 2014). These are possibilities that will likely prove to be fruitful avenues for research in the future.
Last, because of our use of newspaper ads for recruitment, we cannot know the response rate to our ads, and therefore cannot examine whether there were differences between those who volunteered to participate and those who did not, raising concerns about possible biases in our sample. Future research may mitigate this concern by recruiting via letters mailed to families, which would allow researchers to take into account the percentage of potential participants who actually agreed to participate.
Conclusion
In a sample of early adolescents, greater levels of negative self-definitional events over 2 years uniquely predicted increases in the personality trait of self-criticism, which in turn predicted increases in depressive symptoms over the same interval. Greater levels of negative relatedness-oriented events over 2 years uniquely predicted the development of dependency as well as increases in anxiety symptoms, although dependency was unrelated to change in anxiety symptoms. These results further our understanding of both personality development in early adolescence as well as the role of personality development in psychopathology, and suggest avenues for early interventions with youth.
Footnotes
Appendix
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported in this article was supported by research grants from the Social Sciences and Humanities Research Council of Canada and the National Alliance for Research on Schizophrenia and Depression awarded to John R. Z. Abela. The specific grant numbers are unknown at this point.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
