Abstract
The Relationship Profile Test is a widely used measure of dependency, detachment, and healthy dependency that has been examined in both clinical and nonclinical settings, though researchers have yet to validate this measure among conjugally bereaved adults. The present study examines the construct validity of a three-facet model of dependency–detachment by comparing relationships among self-report, semistructured interview–rated, and knowledgeable informant–rated functioning among conjugally bereaved adults. Participants (N = 112) included bereaved adults (M = 51.1 years; SD = 9.7) who had experienced the loss of a spouse 1.5 to 3 years prior to taking part in this study. Findings indicate adequate psychometric properties and theoretically expected associations with various measures of wellness and health including satisfaction with life, coping flexibility, somatic complaints, and ego resiliency. Results draw attention to adaptive correlates of dependency, suggesting potentially beneficial mental health interventions.
Decades of research on interpersonal dependency—reliance on others even in situations where autonomous functioning is warranted—has articulated its association with a wide net of important clinical correlates. For example, dependency is linked with increased risk for depression (Blatt & Zuroff, 1992), substance abuse (Nace, Davis, & Gaspari, 1991), and chronic grief (Johnson, Zhang, Greer, & Prigerson, 2007). Furthermore, dependency has been linked with health behaviors, such as increased treatment compliance (Fowler, Brunnschweiler, Swales, & Brock, 2005) and greater satisfaction with health care services (O’Neill & Bornstein, 2006). This multifaceted perspective on dependency has been informed by a cognitive/interactionist (C/I) model (Bornstein, 2011; Bornstein, Riggs, Hill, & Calabrese, 1996). According to the C/I model, the passive or active behaviors of dependent individuals are sensitive to context and stem from core beliefs and motives. As a result, dependency-related interpersonal functioning may have either positive or negative consequences.
Reflecting this multicomponent view of dependency, Bornstein and colleagues (Bornstein, Geiselman, Eisenhart, & Languirand, 2002; Bornstein et al., 2003) developed the Relationship Profile Test (RPT) to tap the full range of intra- and interpersonal qualities of dependency. 1 This measure represents dependency along three interrelated facets: Healthy Dependency (HD), Destructive Overdependence (DO), and Dysfunctional Detachment (DD). HD is characterized by flexible, mindful help, and support seeking that strengthens interpersonal ties rather than undermining them, DD is characterized by an inability to sustain social connections, rely on others for help or to use feedback effectively, and DO is characterized by a needy, clingy style of maladaptive dependency that tends to undermine relationships rather than strengthen them. Because this measure captures both overdependence and its theoretical counterpart (healthy dependency) in addition to detachment, the RPT affords the opportunity to evaluate both positive and negative consequences on mental and physical health outcomes. Factor analyses of the hypothesized RPT structure have lent some support to this three-factor structure (Morgan & Clark, 2010). This study aimed to examine how the adaptive and maladaptive facets of dependency articulated by the C/I model differ in their relations to coping behaviors and health outcomes among conjugally bereaved adults.
Extant research on RPT-assessed dependency qualities has suggested substantive links between health and coping behaviors. For example, Fiori, Consedine, and Magai (2008) found that HD was associated with lower levels of health impairments and greater positive affect among older adults. Conversely, both DO and DD have been associated with greater somatic symptoms and greater impairment because of chronic illness among a community sample of urban women (Porcerelli, Bornstein, Markova & Huprich, 2009). Though less research has focused on RPT-derived dependency and coping styles, some studies have suggested that HD is associated with lower levels of interpersonal distress compared with those individuals exhibiting higher levels of DO or DD (Haggerty, Blake, & Siefert, 2010). Related studies have confirmed that HD is associated with higher levels of life satisfaction and quality of life, whereas DO and DD are negatively associated with these indicators (Bornstein et al., 2003, Bornstein, Porcerelli, Huprich, & Markova, 2009). Dependency has also emerged as a correlate of chronic grief among bereaved spouses (Bonanno et al., 2002), and dependency on the deceased spouse has been found to mediate the relationship between perceived parental control in early childhood and the development of complicated grief (Johnson et al., 2007).
In addition to the risk factors previously outlined, other investigations have identified potential resilience factors. One study found that healthy dependency is associated with asymptomatic functioning among bereaved midlife adults who experience the loss of a spouse (Denckla, Mancini, Bornstein, & Bonanno, 2011). In related findings, Blake-Mortimer, Koopman, Spiegel, Field, and Horowitz (2003) found that dependency among husbands anticipating the loss of their wife to metastatic breast cancer actually predicted more adaptive outcomes during bereavement. While the extant research reviewed suggests links between dependency, coping, and health behaviors among conjugally bereaved adults, how this maps on to dependency as conceptualized by the C/I model is unclear.
Aims and Hypotheses
The goal of this study is to assess the correlations between RPT subscales and multimethod indicators to test the hypothesis that RPT-derived dependency facets will be associated with different patterns of coping and health outcomes, as articulated more specifically in the hypotheses outlined below.
Method
Participants and Procedure
Participant data employed in the present report were taken from a larger study conducted to investigate reactions to loss. Recruitment and sampling procedures for bereaved participants have been detailed previously (Gupta & Bonanno, 2011). Briefly, participants who had experienced the loss of a spouse 1.5 to 3 years prior to participating in this study were recruited from a large, metropolitan area through newspaper ads and letters mailed to bereaved individuals based on public death record, flyers and medical and mental health professional referrals. Participants were reimbursed $100 for participation in the study. Ethnicity, years of education, gender, and age were broadly representative of the metropolitan area from which participants were recruited. Specifically, the average age of participants was 50.96 (SD = 9.73). The sample was split approximately equally between men (48%) and women (52%). Ethnicity of the full sample was 69% Caucasian, 29% African American, 4% Asian American, 8% Hispanic, and 1% other. Most of the sample had some college education (30%) and an additional 30% had a bachelor’s level degree. Fifteen percent of the sample had a high school education or less, and 27% of the sample had either some education at the master’s level or a doctoral degree. Finally, participants had been married an average of 18.15 years (SD = 12.19), and the average reported family income was $75,300 (SD = $111,090).
Once enrolled in the study, participants received a questionnaire packet in the mail, which they completed before arriving at the study location. They then participated in semistructured interviews at the study location; these included a structured clinical interview for the symptoms of major depressive disorder (MDD), posttraumatic stress disorder (PTSD), and grief-related symptoms. Finally, participants were provided three packets that contained questionnaires for friends and were instructed to distribute the packet to at least three people they felt knew them well. The packets contained a self-addressed and stamped envelope, and friends were instructed to return the envelope by mail to the study site. The final number of participants was N = 112.
Measures
Dependency–Detachment
The RPT (Bornstein et al., 2003) is a 30-item questionnaire containing items rated on a five-point scale ranging from 1 (not at all true of me) to 5 (very true of me). The RPT yields three, 10-item subscale scores: DO, DD, and HD. Information regarding RPT item development and subscale refinement is provided by Bornstein and coworkers (Bornstein et al., 2002; Bornstein et al., 2003). Sample items from the HD, DO, and DD scales, respectively, include I am comfortable asking for help; I am most comfortable when someone else takes charge; and I prefer making decisions on my own, rather than listening to others’ opinions. RPT scores have shown evidence of expected relationships with health-related behavior in college students, as well as adequate retest reliability (Bornstein & Huprich, 2006). RPT subscale scores also show acceptable levels of internal consistency (Bornstein et al., 2003); in the present sample, Cronbach’s alphas for DO, DD, and HD were good, representing values of .86, .76, and .75, respectively.
Ego Resiliency
The Ego Resiliency Scale (ER; Block & Kremen, 1996) is a 14-item measure assessing the capacity to contextually modify one’s level of control in response to external demands (Letzring, Block, & Funder, 2005). Ego resiliency is this study is captured using the ER89, with items scored on a four-step continuum, from 1 (does not apply to me at all) to 4 (applies very strongly), with increasing scores indicative of greater levels of ego resiliency. Studies have shown coefficient alpha reliability of .76 for adults aged 18 and 23 years, and 5-year retest reliability of .51 for females and .39 for males (Block & Kermen, 1996). In the present study, reliabilities for this scale were in the good range at α = .80.
Satisfaction With Life
The Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985) is a five-item scale that assesses global domains of satisfaction, allowing each individual to respond according to his or her personal value system. Each item is scored from 1 (strongly disagree) to 7 (strongly agree). The possible range of scores on the test is from 5 (low satisfaction) to 35 (high satisfaction). Diener et al. (1985) reported a coefficient alpha of .87 for the scale and a 2-month retest stability coefficient of .82. In the present study, reliability for the SWLS was good, at α = .86.
Somatic Complaints
Physical symptoms were assessed using an 18-item checklist of physical symptoms (Marmot et al., 1991). Participants were instructed to indicate whether or not they had experienced specific physical difficulties included in the checklist within the past 14 days; symptoms included aches and pains, sore throats, gastrointestinal pain, headaches, fatigue, skin problems, and difficulty sleeping. The last item was a free response item that allowed participants the opportunity to identify any other symptom not included in the list, and if completed was added to the total count of symptoms for a maximum score of 18. In the present study, reliability was good, at α = .83.
Coping Flexibility
The Perceived Ability to Cope Scale (PACT; Bonanno, Pat-Horenczyk, & Noll, 2011) contains 20 items, each rated using a 1 (not at all able) to 7 (extremely able) scale. Factor analysis has confirmed the presence of two domains: The perceived ability to focus on processing the trauma (trauma focus) and the ability to focus on moving beyond the trauma (forward focus). The measure also yields a single flexibility score that represents the ability to use both types of coping. A flexibility score is captured by employing a three-step algorithm. First, a sum coping ability score is created by standardizing scores for the forward focus and trauma focus scales and then adding the scales; next, a coping polarity score is calculated as the absolute value of the discrepancy between the standardized scores for each scale; finally, a flexibility score is calculated as total coping ability minus coping polarity. The resulting score of high flexibility represents the individual’s perceived ability to engage in both forward-focused and trauma-focused coping. In the present study, reliabilities for trauma-focused coping were in the acceptable range at α = .65 and in the good range for forward-focused coping at α = .88.
Grief Symptoms
Trained interviewers administered selected items from a semistructured questionnaire (SCID; First, Spitzer, Gibbon, & Williams, 2002) corresponding to suggested symptom criteria set for Prolonged Grief (Prigerson et al., 1999). Specific items included the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) criteria for symptoms of MDD (9 items, α = .84), symptoms of PTSD that did not overlap with the MDD items (14 items, α = .85), and grief symptoms adapted from Horowitz, Siegel, Holen, and Bonanno (1997): Strong yearning for the deceased; preoccupation with thoughts about the loss; recurrent regrets or self-blame about own behavior toward the deceased; recurrent regrets or blame regarding the behavior of others toward the deceased; difficulty accepting the finality of the loss; marked loneliness or sense of emptiness; pervasive sense that life is meaningless; unusual difficulty developing new relations (eight items, α = .81). Nine doctoral candidates in clinical psychology performed the interviews. Each interviewer coded a randomly selected set of five additional videotaped interviews. Interrater reliability among coders with respect to all prolonged grief items was very high (κ = .92).
Informant Ratings
Anonymous ratings of participants’ functioning across a number of health domains were provided by close friends according to methods established previously among conjugally bereaved adults (Bonanno et al., 2007). The materials in the packets distributed to friends asked informants for basic information about the friends’ relationship with the participants: (a) how many years they had known the participant in years, (b) how close they felt to the participants on a 5-point scale (1 = very close to 5 = not very close), (c) and how often they interacted with the participant on a 5-point scale (1 = less than once a month to 5 = almost daily). Then, informants answered five questions about the participant’s functioning in five domains of health (mental health, physical health, quality of social interactions, ability to accomplish goals, and, coping ability) compared with “most other people” prior to the bereavement (usual level of health, that is, before the loss) using a 7-point scale (0 = much worse than most; 4 = about the same as most people; 7 = much better than most; α = .84). Similar ratings were made on five items regarding a change in current level of health (change in participant’s current level of adjustment, “in comparison with his or her usual level of health”) using a 7-point scale (0 = much worse than usual; 4 = about the same as usual; 7 = much better than usual α = .89). All 10 items were averaged to create a composite functioning score, consistent with previously established methods (α = .91). In the final step, if ratings from more than one friend were obtained, the values across all participants were averaged to create a mean composite functioning score; at least one friend rating was obtained from 111 bereaved participants. Correlations between ratings from different friends for the same participant were all significant and in the moderate range (rs = .46-.58).
Results
Relationship Profile Test Scale Properties
First, we calculated a series of zero-order correlations to examine the convergent and discriminant validity of the RPT subscale scores. Consistent with previous reports (Bornstein et al., 2003), HD was moderately negatively correlated with both DO r(112) = −.23, p < .05 and DD r(112) = −.27, p < .01, and there was no significant correlation between DO and DD r(112) = .14, p = ns.
Next, we examined the pattern of correlations between the three RPT scale scores and validating variables separately for men and women, given previous studies that have identified gender differences in self-reported dependency and detachment (Bornstein, 1992; Bornstein et al., 2004). Results are reported in Table 1. We then compared mean effect sizes with weighted contrasts to test a priori predictions about differences in correlations among study variables according to methods outlined in Rosenthal (1991).
RPT Correlations With Validating Variables Among Bereaved Adults (n = 112).
Note. RPT = Relationship Profile Test; DO = Destructive Overdependence; DD = Dysfunctional Detachment, HD = Healthy Dependency. Bivariate correlations between dependency and validating variables that differ across gender are indicated in boldface, with the first value corresponding to that for women and second for men. Shaded cells highlight hypothesized associations tested in this study.
As expected, HD was positively correlated with SWL among women r(59) = .53, p < .01 and men r(53) = .43, p < .01. SWL was also significantly negatively correlated with DD among women r(59) = −.45, p < .01, though the association among men did not reach significance r(53) = −.16, p = ns. Finally, the correlation between DO and SWL did not achieve significance for either gender. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DO and DD, respectively, differed from that for HD confirmed that these correlation coefficients did differ significantly among both women (Z = −3.17, p < .001; Z = −5.19, p < .001) and men (Z = −3.35, p < .001; Z = −3.10, p < .001).
As expected, HD was significantly and positively associated with ego resiliency for both women r(59) = .47, p < .01 and men r(53) = .52, p < .01, and DO was significantly and negatively associated with ego resiliency for both women r(59) = −.35, p < .01 and men r(53) = −.47, p < .01. Correlations between DD and ego resiliency did not reach significance for either gender. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DO and DD, respectively, differed from that for HD confirmed that these correlation coefficients did differ significantly among both women (Z = −4.63, p < .001; Z = −3.12, p < .001) and men (Z = −5.43, p < .001; Z = −3.93, p < .001).
Next, we examined associations among the measure of somatic complaints and subscales of the RTP. As hypothesized, HD was inversely associated with somatic complaints among men r(53) = −.38, p < .01, but the association among women did not reach significance. Correlations between DD and somatic complaints showed the opposite gender pattern: The relationship was significant and positive among women, r(59) = .32, p < .05, but not among men. Finally, there were no significant correlations between DO and somatic complaints for either gender. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DO and DD, respectively, differed from that for HD suggested that these correlation coefficients did not differ significantly among women (Z = −0.37, p < ns; Z = −1.59, p = ns). However, correlations between DO and HD did differ significantly among men (Z = 1.65, p < .05), but those between DD and HD did not (Z = 1.45, p = ns).
Next, we examined the pattern of relationships between the RPT subscales and PACT scales of trauma-focused coping, forward-focused coping, and coping flexibility. First, DO, DD, or HD did not demonstrate significant associations with trauma-focused coping among women or men. However, forward-focused coping did have a significant positive correlation with HD among both women r(59) = .52, p < .01 and men r(53) = .29, p < .05. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DO and DD, respectively, differed from that for HD confirmed that these correlation coefficients did differ significantly among both women (Z = −4.29, p < .001; Z = −3.05, p < .001) and men (Z = −2.11, p < .05; Z = −1.95, p < .05). Finally, HD and coping flexibility did correlate significantly among women, r(59) = .45, p < .01, but not among men r(53) = .20, p = ns. Neither DO nor DD demonstrated significant relationships between coping flexibility for men or women. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DO and DD, respectively, differed from that for HD with respect to coping flexibility confirmed that these correlation coefficients did differ significantly among women (Z = −3.58, p < .001; Z = −2.46, p < .01). However, only DO differed from that for HD among men (Z = −1.82, p < .05), with DO and HD not reaching significance (Z = −0.42, p = ns).
Finally, we sought to examine the relationship between cross-method assessment modalities and RPT-derived dependency. First, we examined correlations among RPT-assessed dependency, knowledgeable informant–rated functioning, and semistructured interview–rated symptom levels. Results are presented in Table 1. HD was negatively associated with semistructured interview–rated symptom levels, r(111) = −.36, p < .01, but there was no significant correlation between knowledgeable informant–rated adjustment and HD. We also found a positive relationship between DD and interview-rated symptom levels, r(111) = .39, p < .01, but no significant association between knowledgeable informant–rated adjustment. Finally, there were no significant associations between DO and either symptom level or knowledgeable informant–rated adjustment.
Discussion
Results support the construct validity of the RPT (Bornstein et al., 2003) in assessing self-reported dependency and detachment among bereaved community adults. First, scrutiny of Table 1 confirms that the pattern of RPT correlations in this sample was similar to that of previous studies (e.g., Bornstein et al., 2002; Bornstein et al., 2003; Haggerty et al., 2010). Second, the majority of our a priori hypotheses with respect to associations between RPT subscales and criterion variables were in expected directions and suggest a heretofore unexplored association between dependency, coping, and health behaviors among conjugally bereaved adults. Results contribute to extant findings of the construct validity of the RPT among various populations and samples, including low-income urban women (Bornstein et al., 2009; Porcerelli et al., 2009), psychotherapy patients (Huprich et al., 2010), substance-abuse patients (Bornstein, Gottdiener, & Winarick, 2010), and college students (Bornstein et al., 2002; Haggerty et al., 2010).
The positive and expected associations between HD and health and coping behaviors extends the construct validity of dependency by documenting convergent and discriminant validity with measures of SWL, ego resiliency, somatic complaints, forward-focused and trauma-focused coping, as well as coping flexibility. Given the many adaptive qualities implied in the conceptualization of HD (Bornstein, 1998), it is not surprising that these associations were confirmed in the current study. The pattern of findings suggests that these adaptive individual difference qualities, alone or in combination, may be contributing factors to asymptomatic functioning following the loss of a loved one.
Study findings also yielded some unexpected patterns. While we did find a significant negative association between somatic complaints and HD among men as predicted, we did not find such an association among women. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for HD and Somatic complaints differed between genders confirmed that these correlation coefficients did differ significantly (Z = 1.90, p < .05). Among women, the absence of an expected negative correlation between HD and somatic complains suggests that HD may not buffer against health complaints in women as it appears to do for men. However, given that women tend to report somatic complaints at a higher rate than men (Kroenke & Spitzer, 1998), it may be that women were more likely to report physical complaints regardless of their dependency style. Also, although women—but not men—demonstrated a statistically significant positive association between DD and somatic complaints, a comparison of effect sizes with weighted contrasts between men and women confirmed that these correlation coefficients did not differ (Z = 1.19, p = ns).
We also found a significant negative correlation between DD and satisfaction with life among women but not men. A comparison of effect sizes with weighted contrasts to ascertain whether correlations for DD and SWL differed between genders confirmed that these correlation coefficients did differ significantly (Z = −1.67, p < .05). This finding could potentially reflect the fact that some degree of detachment in men is normative, whereas detachment in women is more unusual (so perhaps brings more negative social consequences; see Birtchnell, 1995). Though speculative, it may also be that unhappiness (in both genders) leads to greater detachment, but that this pattern is masked somewhat in men because of their higher overall/baseline level of detachment.
The second part of the study examined the construct validity of RPT-assessed dependency among conjugally bereaved adults using cross-method assessment. That is, in addition to comparing the construct validity of dependency scales with other self-report scales, we also sought to compare the construct validity of the self-report scales with both semistructured expert-rated symptom levels and knowledgeable informant–rated functioning. We did find evidence for expected patterns of convergence between HD, DD, and total symptom levels. That is, HD was associated with lower symptoms levels, and DD associated with higher symptom levels in this population. However, DO showed no significant association with symptom levels. Though inconsistent with previous studies linking DO with increased psychopathology (Bornstein, 1992), findings are consistent with other studies that suggest that overly dependent individuals may mobilize social support in times of duress to cope with adversity (Besser & Priel, 2003; Mongrain, 1988). On the other hand, we did not find significant correlations between the dependency variables and knowledgeable informant rating of participant functioning. Previous studies have documented strong associations between self- and other ratings of daily behavior (Funder, 2012) and judgments about personality traits (Vazire & Mehl, 2008), so there is some evidence suggesting we may have expected to find a stronger association between informant-rated functioning and adaptive and maladaptive dependency traits. However, the relatively small magnitude of the correlation between facets of dependency (both HD and DO) and expert-rated assessment of symptom levels suggests that the strength of the association between personality traits and overall functioning may have been too small to detect in this sample.
Implications for Assessment and Clinical Intervention
The experience of losing a loved one is likely to touch all individuals at least once in their lifetimes, and understanding the individual difference elements that serve as resilience and risk factors has important health applications (Stroebe, Schut, & Stroebe, 2007). Our findings may shed some light on the complexity of dependency’s role in adjusting to loss, building on prior studies that have identified risk factors (Bonanno et al., 2002; Johnson et al., 2007). A growing body of evidence suggests that the particular benefit of any one regulatory strategy is sensitive to the context, the person employing a particular strategy, and that person’s ability to integrate feedback (Bonanno & Burton, 2013; Morf, 2006). This implies that rather than any one particular regulatory strategy (e.g., dependency) being universally beneficial, it is the ability to flexibly deploy various regulatory strategies that is best associated with adjustment after loss. As a better understanding of the interplay between self-regulatory processes and context-specific experiences develops, interventions that foster flexibility in self-regulatory strategies, rather than those that target one putatively maladaptive trait, may better assist individuals in adjusting to the loss of a loved one. This perspective on personality and motivation is also consistent in many ways with emerging consensus on the utility of a dimensional model of personality pathology (Widiger & Samuel, 2005), and with circumplex models of personality and personality pathology that capture both the adaptive and maladaptive features of personality traits and disorders (Hopwood, Wright, Ansell, & Pincus, 2013). The complex pattern of associations among dependency, coping, and health behaviors suggests that dependency is a multifaceted individual difference variable that encompasses both risk and resilience factors for bereaved adults, and results therefore warrant further study.
Certain limitations in this study should be noted, some of which affect the generalizability of the primary findings. Most important, the cross-sectional nature of this investigation limits our ability to make causal inferences. Prospective designs will be necessary to ascertain the degree to which individual differences in affiliative styles predict adjustment to loss. Importantly, prospective methodological approaches will serve to better delineate those individual difference factors that serve as risk and/or resilience factors in coping with life’s challenges, and the manner in which the positive and negative consequences of these risk and resilience factors may vary over time and across situation. Second, we could not confirm the factor structure of the RPT because of our modest sample size. Given the importance of a confirmatory analysis in providing more substantive evidence for the construct validity of RPT-derived dependency facets, future studies should employ larger sample sizes to investigate the hypothesized structure of the RPT among clinical samples. This is particularly important given some studies have found somewhat mixed support for a three-factor structure for the RPT (Morgan & Clark, 2010), whereas other studies have found evidence for a four-factor structure, with the fourth factor representing an adaptive variant of detachment (Fiori et al., 2008). The later study suggests the interesting possibility that in addition to HD representing an adpative pole of overdepdenence, maladaptive detachment may also have an adpative variant.
In summary, this report highlights the importance of individual differences in dependency for adaptive functioning among bereaved adults. The study offers broad support for the use of the RPT among conjugally bereaved adults as a measure that taps the correlates of HD, DO, and DD, and offers some insight into the complexity with which individual difference factors influence the process of coping with loss. In so doing, this study complements and extends extant findings that point toward the importance of individual difference factors in adaptive and maladaptive bereavement trajectories.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The primary study was supported by a grant from NIMH awarded to G.A. Bonanno, and preparation of the manuscript was supported by a grant from the International Psychoanalytic Association awarded to C.A. Denckla.