|Home | About | Journals | Submit | Contact Us | Français|
The authors conducted two studies to identify and validate potential personality subtypes in adolescent and adult children of alcoholics. As part of a broader NIMH-funded study, randomly selected psychologists and psychiatrists provided personality data on adolescent (n = 229) or adult (n = 359) children of alcoholics using a Q-sort procedure (SWAP-II-A for adolescents and SWAP-II for adults), which were subjected to a cluster-analytic procedure, Q-factor analysis. Q-factor analysis yielded five personality subtypes in both groups. Despite the different samples and age groups, four of the personality subtypes were highly similar, including externalizing, inhibited, emotionally dysregulated, and high-functioning. Providing initial data on their validity, the subtypes differed on Axis I and II pathology, adaptive functioning, and developmental and family history variables. These findings show heterogeneity among children of alcoholics and suggest the importance of addressing personality subtypes for research and practice in treating adolescent and adult children of alcoholics.
Estimates suggest that ten to fifteen percent of Americans suffer from alcoholism (Stinson et al., 2005) and over 28 million people in the United States alone are children of alcoholics (Obot, Wagner, & Anthony, 2001). Children of alcoholics (COAs) are at three to four times the risk for developing alcoholism than a child without an alcoholic parent, and daughters of alcoholics are more likely to marry alcoholic men, perpetuating the cycle to future generations (Obot et al., 2001). While not all COAs become alcoholic themselves, COAs are at an increased risk for many additional problems throughout their lives. Adult children of alcoholics (ACOAs) report increased difficulties at work, more interpersonal problems, and higher levels of emotional distress than comparison subjects (Baker & Stephenson, 1995; Coleman & Frick, 1994; Mathew, Wilson, Blazer, & George, 1993). As personality is a diathesis for many Axis I symptoms, a better understanding of the enduring cognitive, affective, and relational factors that place COAs at risk is warranted in order to develop more effective treatment and prevention.
Research indicates that personality may mediate the intergenerational link to alcoholism through multiple pathways (see Coleman & Frick, 1994). One such pathway towards alcoholism that is mediated by personality is genetics, as heredity accounts for up to 60% of the variance in the diffusion of alcoholism (Coleman & Frick, 1994; McGue, 1997; Prescott & Kendler, 1999). Another pathway is psychosocial factors, such as the development of ways to deal with growing up in an abusive or chaotic alcoholic household (Potter-Efron, 1990), modeling or identification with an alcoholic parent, or de-identification with an alcoholic parent (resolving never to be like him or her).
Research on personality characteristics of COAs is highly variable and often contradictory. The personality dimension most associated with COA is impulsivity/disinhibition, encompassing traits such as sensation seeking, aggressiveness, and impulsivity (Coleman & Frick, 1994; Sher, 1997; Sher, 1991; Windle, 1990). In addition, COAs show lower self-esteem, higher dependency, manipulative behaviors, neuroticism, and perfectionism (Baker & Stephenson, 1995; Carpenter, 1995; Fisher, Jenkins, Harrison, & Jesch, 1992; Roebuck, Mattson, & Riley, 1999; Stinson et al., 2005).
Contradictory findings in studies examining differences between COAs and comparison subjects has led researchers to question whether COAs are homogeneous with respect to personality (D’Andrea, Fisher, & Harrison 1994). Families of alcoholics vary on a continuum of dysfunction (Lease, 2002), as many other psychological disorders and disturbances in parenting (e.g., abuse) can coexist with parental alcoholism. Familial dysfunction can produce a variety of different consequences on the personality of COAs (Sher, 1997). As such, researchers have considered that differences between COAs and non-COAs may simply be masked by the heterogeneity of COAs (D’Andrea et al., 1994). In an effort to disentangle mixed and contradictory findings, several clinical and empirical studies have investigated the presence of personality subtypes among COAs.
Theory and research on subtyping COAs began decades ago with clinically-based taxonomies. Wegscheider (1981) proposed five personality styles in COAs based on clinical experience (i.e., The Enabler, Hero, Scapegoat, Lost Child, and Mascot). Black (1979) proposed similar subtypes of COAs, adding “The Placater.” However, empirical support for these clinical taxonomies is weak. Rhodes and Blackham (1987) developed four scales with internal consistency to quantify characteristics of the Hero, Lost Child, Scapegoat, and the Placater. However, significant differences between COAs and non-COAs only emerged on the Scapegoat scale. Devine and Braithwaite (1993), in a study attempting to validate the five roles described by Black and Wegscheider, found that the subtypes were able to discriminate COAs and non-COAs. However, these five roles were not exclusively linked to parental alcoholism. Context variables such as family disruption were a more potent predictor.
The limitations of the clinically derived composite portraits of COAs have led researchers to search for other personality subtypes. D’Andrea and colleagues (1994) found three personality subtypes among COAs. The first subtype showed elevations on sociability, self-acceptance, well-being, responsibility, self-control, achievement, and intellectual efficiency. Subtype two scored slightly below subtype one on all the previously mentioned scales but still maintained scores characteristic of a normal population. Subtype three produced low scores on all of these scales, and was significantly associated with higher rates of parental death, sexual abuse, and physical abuse. Although these findings suggest the possibility of personality subtypes, they largely discriminate relatively healthy from relatively disturbed individuals, with more troubled histories compounding parental alcoholism.
In an attempt to bridge the previously inconsistent findings on COA personality subtypes, we undertook two studies designed to identify personality subtypes of adolescent and adult COAs. Rather than start from a theoretically-derived taxonomy of COAs, we report on two studies that attempted to identify personality subtypes using an instrument with adolescent and adult versions (Shedler-Westen Assessment Procedure; SWAP) designed to capture subtle aspects of a range of personality, thereby elucidating or corroborating existing personality subtypes. Statistical procedures that produce dimensional (i.e., the extent to which the patient matches a personality prototype, which can vary from not at all to substantially) or categorical subtypes were implemented on SWAP results of COAs. We used Q-factor analysis on two separate samples, one of adolescents (n = 229) and the other, adults (n = 359), to determine if personality subtypes are similar across two different age groups. No previous research has examined ACOA personality subtypes or compared personality subtypes of COAs across two age groups. Therefore, this section of the study was exploratory. Differences among the subtypes within each sample on criterion variables indicative of a valid taxonomy, specifically (a) Axis I diagnoses, (b) Axis II pathology, (c) adaptive functioning, and (d) etiological variables such as developmental and family history were examined.
The adolescent sample was collected as part of a NIMH-funded project on the nature and classification of adolescent personality pathology broadly defined, including not only the severe disturbances captured on Axis II of DSM-IV (APA, 1994), but the wider range of personality problems seen in patients currently described as having “subthreshold” pathology. Patients were not preselected for parental history of alcoholism; rather, parental history was one variable of many recorded as part of the broader study. The studies involving both adolescents and adults received approval by the university ethics board (IRB) and participation was informed and voluntary.
The methods for both Study 1 and Study 2 have been described elsewhere in detail (Westen & Shedler, 2007). We ascertained the sample as follows: 950 doctoral-level clinicians who treat adolescents completed a battery of psychometric instruments on a randomly selected patient aged 13–18 in their care. Each clinician provided data on only one patient; thus, there was no overlap among informants or patients across the study. All clinicians completed measures of personality pathology, adaptive functioning, and developmental history that we used as criterion variables to assess validity and association with personality pathology.
A practice network approach was used, in which randomly selected, experienced clinicians provide data on patients that can be aggregated for large participant pool research on the nature, classification, and treatment of psychopathology (Dutra, Campbell, & Westen, 2004; Westen & Shedler, 1999a, 1999b; Westen, Shedler, Durrett, Glass, & Martens, 2003; Westen & Weinberger, 2004). Multiple studies using this method suggest that (a) clinician descriptions of patients strongly correlate with interview-based assessments, typically ranging from r = .50 to .80 (Hilsenroth et al., 2000; Westen, Muderrisoglu, Fowler, Shedler, & Koren, 1999); (b) clinician descriptions predict criterion variables such as developmental and family psychiatric history in ways that are theoretically predictable and unrelated to clinicians’ theoretical orientation or training (Dutra et al., 2004; Westen et al., 2003); (c) the factor structure of clinician-report versions of established instruments such as the Child Behavior Checklist (CBCL; Achenbach, 1991) tends to be similar to the factor structure that emerges with other informants (Dutra et al., 2004) and (d) clinicians tend to use appropriately conservative rules of inference in assessing etiologically relevant variables such as family and developmental history (e.g., requiring intact pre-treatment memories or other corroborating data before ascribing confidence to a history of childhood sexual abuse; DeFife et al., 2010; Wilkinson-Ryan & Westen, 2000).
A random national sample of psychiatrists and psychologists with at least 5 years experience post-residency (M.D.s) or post-licensure (Ph.D.s) from the membership registers of the American Psychiatric and American Psychological Associations was contacted. We selected all clinicians whose membership records indicated an interest in children or adolescents, and supplemented this where necessary with a general sample, given that many clinicians treat adolescents as well as adults. Over one-third of clinicians agreed to participate in the study. Participating clinicians received a consulting fee of $200 for a procedure that required approximately 2 hours of time.
Clinicians were asked to describe “an adolescent patient you are currently treating or evaluating who has enduring pattern of thoughts, feeling, motivation or behavior—that is, personality problems—that cause distress or dysfunction.” To obtain a broad range of personality pathology, from relatively minimal to substantial, we emphasized that patients must have problematic personality traits but need not have a PD diagnosis. To avoid biasing the sample one way or another, we instructed clinicians to disregard the caveats in the DSM-IV regarding the application of Axis II diagnoses to adolescents and simply to select a patient with any degree or form of personality pathology as defined above. To further minimize selection bias, we directed clinicians to consult their calendars to select the last patient they saw during the previous week who met study criteria. We obtained a stratified random sample, stratifying on age (13–18) and sex. The only exclusion criteria were chronic psychosis and mental retardation. In addition, we asked clinicians to select a patient whose personality they felt they knew, using as a guideline ≥ 6 clinical contact hours but ≤ 2 years (to minimize confounds imposed by personality change with treatment).
The CDF (which has adult and adolescent versions, the CDF and CDF-A) is a clinician-report form developed over several years that assesses a range of variables relevant to demographics, diagnosis, and etiology (Westen et al., 2003; Westen & Shedler, 1999a, 1999b). Clinicians first provide demographic data on themselves and the patient. They then rate the patient’s adaptive functioning using a number of indices, such as ratings of school performance and peer relations, as well as relatively objective indicators such as history of arrests, suicide attempts, and psychiatric hospitalizations. Research has demonstrated that clinician ratings of adaptive functioning variables show high inter-rater reliability and validity (e.g., correlations with the same data obtained by independent interview r > .60) (Hilsenroth et al., 2000; Westen et al., 1997).
The next section of the CDF assesses aspects of the patient’s developmental and family history with which clinicians who have met with adolescents and/or their parents over several sessions are likely to be familiar. The CDF assesses a wide range of variables of potential etiological relevance, such as history of foster care, family stability, and physical or sexual abuse. Clinicians working with adolescent patients generally have relatively direct access to such information, from having met with parents and/or other collateral sources. The adolescents who were studied in this investigation were selected based on the clinician’s response to a question in the CDF-A: “Has the patient had an alcoholic parent or step-parent living in the home for a substantial period of time?” If the clinician answered yes, the participant was included in our sample of adolescent children of alcoholics. This includes adolescent children of an alcoholic mother, father, or stepparent. Clinicians were instructed on all such questions to answer “no” unless confident in their response.
To diagnose DSM-IV Axis I disorders common to children of alcoholics without relying on clinicians to apply DSM-IV diagnostic algorithms, we asked clinicians to circle “no” or “yes” in response to a randomly ordered diagnostic criteria taken directly from the DSM-IV diagnoses of major depressive disorder (MDD), general anxiety disorder (GAD), substance abuse disorder, and social phobia.
To generate both categorical and dimensional DSM-IV PD diagnoses (without depending on clinicians to make free-form diagnoses, which tend to be unreliable), we presented clinicians with a randomly ordered checklist of the criteria for all Axis II disorders. In prior studies, this method has produced results that mirror findings based on structured interviews such as the SCID-II (Blais & Norman, 1997; Morey, 1988; Westen et al., 2003). To create categorical diagnoses, we applied DSM-IV decision rules to the present/absent data. To generate DSM-IV dimensional diagnoses that mirror those widely used in the PD literature, we summed the number of criteria judged present for each disorder. Prior research with both adolescents and adults finds that Axis II ratings collected in this manner have similar external correlates and produce similar patterns of comorbidity to those obtained in structured interviews Westen et al., 2003; Zittel & Westen, 2005).
The SWAP-II-A, the most recent version of the SWAP-200-A for adolescents (Westen et al., 2003; Westen, Dutra, & Shedler, 2005), is a 200-item personality pathology Q-sort designed for use by clinically experienced observers based on longitudinal information over the course of treatment or a systematic clinical diagnostic interview of the patient and parents (Westen & Muderrisoglu, 2003). The clinician sorts (rank-orders) the 200 personality statements into eight categories based on their applicability to the patient, from those that are not descriptive (assigned a value of 0) to those that are highly descriptive (assigned a 7). Statements that apply to a greater or lesser degree are placed in intermediate categories.
Both the adult and adolescent versions of the SWAP show considerable evidence of reliability and validity (Westen & Shedler, 2007), predicting a range of measures of adaptive functioning (e.g., history of hospitalizations, school performance, violence), psychopathology (e.g., the CBCL), etiological variables (e.g., childhood history of physical and sexual abuse, family history of internalizing and externalizing disorders), and personality as assessed by independent interviewers blind to clinician data (Nakash-Eisikovits et al., 2002; Westen & Shedler, 1999a, 1999b; Westen et al., 2003). Empirically, clinicians’ theoretical orientation and professional degree (psychology or psychiatry) has little impact on the way they use the instrument (Shedler & Westen, 2004a, 2004b).
The sample of adolescent COAs (N=229) was 50.7% male with an average age of 15.5 years old (SD= 1.6). The sample was 75.1% Caucasian, 10% Hispanic, and 8.7% African American, with SES rated as 14.4% poor, 32.9% working class, 34.9% middle class, and 14.4% upper class. Patients with complete data on each of the overlapping measurement tools were selected for the control group (N=266). They were 48% male with an average age of 15.6 years old (SD=1.61). The control was 80.5% Caucasian, 5.6% Hispanic, and 6.8% African American with 2.3% considered poor, 15.4% working class, 43.2% middle class, and 30.5% upper class. Respondents (26.6% psychiatrists) were highly experienced, with a mean of 17.4 years (SD 8.8) of post-training experience.
To identify potential subgroups of patients based on personality profiles, we used Q-factor analysis (also called Q-analysis), a technique that has been used effectively in studies of normal personality as well as PDs in adults and adolescents (Westen & Shedler, 1999a, 1999b; Westen et al., 2003; Westen & Harnden-Fischer, 2001). Whereas conventional factor analysis identifies items that share a common underlying dimension, Q-analysis as applied to personality data identifies patients who share a core personality style or organization (i.e., patients who have similar profiles across items). Two advantages of Q-factor analysis relative to other clustering procedures are, first, that it does not require the assumption of mutually exclusive types (i.e., patients can load on multiple factors, to differing degrees); and second, that it does not assume that subgroups exist (i.e., like conventional factor analysis, it can identify unidimensional as well as multidimensional constructs).
We followed standard factor-analytic procedures, first entering the data into a principal components analysis, specifying eigenvalues ≥ 1 (Kaiser’s criteria), and using the scree plot, percent of variance accounted for, and parallel analysis (Horn, 1965; O’Connor, 2000) to determine the number of Q-factors to rotate. These procedures suggested a 5–7-factor solution. We applied multiple estimation procedures with 4–7 factors to increase the likelihood of identifying robust and coherent factors. We created dimensional and categorical scores for each patient as follows. For dimensional scores, we used Q-correlations, which reflect the correlation or “match” between each patient’s 200-item profile and the empirically generated 200-item Q-factor profiles. We created categorical subgroup diagnoses by assigning patients to the group for which they had the highest Q-correlation (similar to a factor loading) ≥ .40. Although continuous data produced essentially the same pattern of results, we treated Q-factors categorically in several analyses (e.g., percent of patients meeting criteria for comorbid Axis I and Axis II diagnoses) for ease of interpretation.
Q-factor analysis identified five subgroups (Q-factors). We report here the unweighted least squares (ULS) solution with Promax (oblique) rotation, which accounted for 39.9 % of the variance (18%, 10.3%, 4.6%, 3.8%, and 3.2% for each Q-factor, respectively). The median correlation among factors was r = .08. Table 1 shows the items most characteristic of each subgroup. The items are arranged in descending order based on factor scores, expressed in standard deviation units, which reflect the item’s centrality to the construct relative to the other items in the item set (In Q-factor analysis, as opposed to conventional factor analysis, patients load on factors and items receive factor scores). Q-factors were labeled Angry/Externalizing, Awkward/Inhibited, Hyperconscientious/High-functioning, Emotionally Dysregulated, and Sexualized/Self-defeating.
Patients who matched the Angry/Externalizing subtype were characterized by their tendency to be rebellious, angry, manipulative, impulsive, and deceitful. They also demonstrated psychopathic features such as remorselessness, lack of empathy, and unresponsive to consequences. Patients who matched the Awkward/Inhibited prototype were characterized by lack of social skills and relationships and feelings of inadequacy, depression, neglect, and powerlessness. Hyperconcientious/High-functioning adolescents were liked by others, articulate, and assertive. However, adolescents who matched this subtype also tended to exhibit feelings of depression, guilt, and anxiety about living up to their own moral and ethical standards. Emotionally Dysregulated patients had many borderline personality disorder (BPD) features, notably a tendency for emotions to spiral out of control, to have difficulty soothing themselves in times of emotional distress, and to have problems with identity formation and impulsivity. Finally, Sexualized/Self-defeating adolescents tended to abuse alcohol, acted sexually provocative and promiscuous, became attached quickly and often in abusive relationships, and engaged in self-mutilating behavior.
The first step in taxonomic work is to identify potential diagnostic groupings. The second is to assess their validity or external correlates. In this study, we examined the relation between the personality subtypes and three sets of criterion variables: diagnostic variables, adaptive functioning variables, and variables of potential etiologic significance. For diagnostic criterion variables, we treated the subgroups categorically for ease of presentation; for all other analyses we report correlational (dimensional) analyses.
We tested the validity of the subtypes by examining their external correlates using contrast analysis and correlational analysis. Where appropriate, we used contrast analysis, which maximizes power, minimizes the likelihood of spurious findings resulting from running multiple analyses, and tests focal, one-tailed hypotheses (about the relative ordering of group means) rather than global questions that are usually of less theoretical interest (e.g., whether group means differed in some unspecified way; Rosenthal, Rosnow, & Rubin, 1999). We generated contrast weights (i.e., hypotheses about the relative ordering of means) based on prior research and examination of the item content of the subgroups, prior to examining their association with external criterion variables. For nondiagnostic criterion variables, to maximize power, we analyzed the data dimensionally, correlating patients’ scores on each personality dimension with variables of interest. To maximize reliability of external correlates, we aggregated ratings of relevant constructs to create composite variables related to adaptive functioning and developmental history (e.g., mean number of adverse childhood events), first standardizing where appropriate.
Table 2 presents correlations between personality subtypes and Axis I and II comorbidity. We report in the table both our a priori hypotheses (in the form of contrast weights) and the findings. Treated categorically, the rates of Axis I disorders among personality subtypes did not differ for social phobia disorder, major depressive disorder, or generalized anxiety disorder. However, the Angry/Externalizing and Sexualized/Self-defeating subtypes differed significantly from the other subtypes in the presence of substance abuse disorder (41% and 79% respectively). For Axis II pathology, as predicted, the Awkward/Inhibited subtype had the highest rates of avoidant personality disorder, the Emotionally Dysregulated and Sexualized/Self-defeating groups had the highest rates of BPD, the Angry/Externalizing subtype had the highest rates of antisocial personality disorder, and the Hyperconcientious/High-functioning group had the least Axis II pathology.
Table 3 presents correlations between the personality subtypes (treated dimensionally) and composite measures of adaptive functioning. Both the Angry/Externalizing and Emotionally Dysregulated dimensions were associated with poor global functioning. The Awkward/Inhibited and Angry/Externalizing dimensions were associated with poor school functioning, whereas the Emotionally Dysregulated dimension was uniquely associated with suicide attempts and hospitalizations. Table 3 also includes data on variables of potential etiologic significance (developmental and family history). Predictions were made based on previous personality pathology research. For example, the Emotionally Dysregulated dimension was predicted to be positively correlated with sexual abuse, disrupted childhood attachment, and a family history of suicide. Predicted positive correlations are in bold in Table 3. The Angry/Externalizing dimension was highly associated with an adverse childhood environment, childhood attachment disruption, physical abuse, childhood psychopathy, and a family history of criminality. The Emotionally Dysregulated dimension was also associated with an adverse childhood environment and physical abuse as well as sexual abuse. As expected, the Sexualized/Self-defeating dimension was highly correlated with sexual abuse. Overall, the Hyperconscientious/High-functioning dimension was negatively associated with most developmental and familial risk factors.
The second study focuses on adult children of alcoholics (over the age of 18), using comparable methods.
We collected an adult sample as part of a NIMH-funded project on the nature and classification of adolescent personality pathology broadly defined, as in Study 1. A total of 1201 doctoral-level clinicians completed a battery of psychometric instruments on a randomly selected patient over the age of 18 in their care. To obtain a broad range of personality pathology, we emphasized that patients need not have a PD diagnosis. Patients had to meet the following additional inclusion criteria: ≥ age 18, not currently psychotic, and known well by the clinician (once again using the guideline of ≥ 6 clinical contact hours but ≤ 2 years). To minimize selection biases, we once again directed clinicians to consult their calendars to select the last patient they saw during the previous week who met study criteria.
Measures were identical to Study 1, except that we used the adult forms of all measures (the CDF and SWAP-II).
The sample of ACOAs (N=356) was 47.6% male with an average age of 41.14 years old (SD= 11.1). The sample was 79.4% Caucasian, 8.4% Hispanic, and 7.8% African American with 11.4% considered poor, 39.8% working class, 30.6% middle class, and 15.6% upper class. The comparison sample (N=266) was 42.8% male with an average age of 41.8 years old (SD=11.6). The comparison sample was 80.8% Caucasian, 4.5% Hispanic, and 9.4% African American with 3.8% considered poor, 20.7% working class, 44% middle class, and 26.3% upper class. Respondents (27.9% psychiatrists) were highly experienced, with a mean of 18.1 years (SD 7.8) of post-training experience.
To identify potential subgroups of patients based on personality profiles, we once again used Q-factor analysis and followed standard factor-analytic procedures, which identified five subgroups (Q-factors). We report here the unweighted least squares (ULS) solution with Promax (oblique) rotation, which accounted for 41 % of the variance (18%, 10.7%, 5.5%, 3.7%, and 2.8% for each Q-factor, respectively). The median correlation among factors was r = .04. Table 4 shows the items most characteristic of each subgroup. We labeled the Q-factors Inhibited, High-functioning, Externalizing, Emotionally Dysregulated, and Reactive/Somatizing.
Patients who matched the Inhibited subtype were characterized by their tendency to be passive, self-conscious, depressed, ashamed, and feel inadequate, guilty, anxious, and uncomfortable in social situations. Patients who matched the High-functioning subgroup were articulate, conscientious, energetic, creative, and empathic. Externalizing adults tended to abuse drugs/alcohol, act impulsively, take advantage of others, and be manipulative, critical, deceitful, and angry. Emotionally Dysregulated adults had many BPD features, notably a tendency for emotions to spiral out of control, to have difficulty soothing themselves, and to have problems with identity and impulsivity. Finally, Reactive/Somatizing adults tended to exhibit physical and emotional reactivity to interpersonal slights and conflicts. They develop somatic symptoms in response to stress, ruminate on relational problems, react severely to slights or insults, hold grudges, and tend to be critical, controlling, anxious, and angry.
We tested the validity of the subtypes by examining their external correlates using contrast analysis and correlational analysis. All analyses were identical to those used in Study 1. Table 5 presents correlations between personality subtypes and Axis I and II comorbidity. Treated categorically, the rates of Axis I disorders between personality subtypes differed significantly on three of the four disorders we tested. The disorders were chosen due to their high frequency in children of alcoholics in previous research. The Emotionally Dysregulated and Reactive/Somatizing adults had the highest rates of major depression. The Externalizing subtype had the lowest rates of major depression and the highest rates of substance abuse disorder. Adults matching the Inhibited prototype had the highest rates of generalized anxiety disorder. For Axis II pathology, the Inhibited subtype had the highest rates of avoidant personality disorder, the Emotionally Dysregulated adults had the highest rates of BPD, the Externalizing adults had the highest rates of antisocial personality disorder, and the High-functioning group had the lowest rates of all personality disorders tested.
Table 6 presents correlations between the personality subtypes (treated dimensionally) and composite measures of adaptive functioning. Both the Externalizing and Emotionally Dysregulated dimensions were associated with poor adaptive functioning. The High-functioning dimension was associated with high adaptive functioning and employment. Table 6 also includes data on variables of potential etiologic significance (developmental and family history). The Emotionally Dysregulated dimension was associated with a poor childhood environment, sexual and physical abuse, and suicide. The Externalizing dimension was uniquely associated with poor childhood attachment, childhood psychopathy, and a family history of criminality. Overall, the High-functioning dimension was negatively associated with most developmental and familial risk factors.
The results corroborate previous research suggesting that COAs are not a homogeneous group. Rather, in both studies we found five distinct personality patterns in adolescent and adult COAs, four of which showed strong similarities across the two age-groups. The subtypes identified were similar to, but non-overlapping with, personality subtypes previously identified in prior work with the SWAP (Westen, Dutra, & Shedler, 2005; Westen & Shedler, 2007). The emergence of non-overlapping subtypes with significantly different external correlates than a comparison group supports the implication that these personality subtypes are unique to adolescent and adult children of alcoholics.
Among adolescents, the first subtype is Angry/Externalizing. These adolescents show psychopathic features such as taking advantage of others, expressing little empathy, and appearing impervious to consequences. These characteristics have been associated with conduct problems at a younger age of onset (Wooten, Frick, Shelton, & Silverthorn, 1997). Patients who matched this profile also tended to be deceitful, manipulative, angry, impulsive, and critical. As expected, these adolescents showed significantly higher rates of substance abuse disorder and antisocial personality disorder. In addition, adolescents who matched this prototype showed poor global adaptive functioning, poor school functioning, and an adverse childhood environment, childhood attachment disruption, physical abuse, childhood psychopathy, and a family history of criminality. The antisocial and delinquent nature of adolescents in this subtype is quite similar to those described in Wegscheider’s (1981) clinical descriptions of ‘The Scapegoat’.
Adolescents in the Awkward/Inhibited subtype showed social deficits and internalizing problems. These adolescents have a tendency to be passive, avoidant, depressed, and may fear rejection or abandonment. Our hypothesis was supported in that they also showed the highest rates of avoidant personality disorder and poor school functioning. Adolescents in this subtype resemble Wegscheider’s (1981) ‘Lost Child.’
The third adolescent personality subtype was Hyperconscientious/High-functioning. This subtype includes a range of healthy attributes such as a tendency to be conscientious, responsible, articulate, empathic, and able to respond to humor; however, this group also includes variables associated with negative affectivity, such as a tendency to feel guilty, to be self-critical, and to expect themselves to be perfect. As expected, the Hyperconscientious/High-functioning adolescents were associated with high global adaptive functioning, school functioning, and a negative association with most developmental and familial risk factors.
The fourth subgroup, Emotionally Dysregulated, is strongly reminiscent of BPD, and has been identified in other studies of adults and adolescents using the SWAP (Westen et al., 2003). Patients who match this prototype are characterized by intense emotional experiences and relational instability. As expected, these adolescents had high rates of BPD. High loadings on this factor were also associated with poor global adaptive functioning, suicide attempts and hospitalizations, an adverse childhood environment, and physical and sexual abuse.
The last adolescent subtype was Sexualized/Self-defeating. This subtype tends to abuse alcohol, act promiscuously and sexually seductive while becoming quickly attached in abusive relationships, act impulsive and engage in thrill-seeking behaviors, and engage in self-mutilating behavior. As expected, these adolescents showed high rates of substance abuse disorder and BPD, and were more likely to have been sexually abused.
Study 2 examined personality subtypes in a separate adult sample. Four of the five subtypes seem to be adult analogues of the adolescent subtypes. The first subtype, Inhibited, is similar to the Awkward/Inhibited adolescents, characterized by passive and constricted behaviors with feelings of guilt, depression, and anxiety. As expected, these adults had the highest rates of generalized anxiety disorder.
The High-functioning adults were similar to the Hyperconscientious/High-functioning adolescents, showing healthy attributes such as a tendency to be conscientious, responsible, and empathic. The pathological features of guilt and perfectionism found in the adolescent subtype appear less frequently in the adult subtype. A possible hypothesis for this change might be that adults were able to work out some of their conflicts from adolescents by no longer living with the alcoholic parent. As expected, high loadings on the High-functioning Q-factor was associated with high global adaptive functioning and employment, and negatively associated with most developmental and familial risk factors.
The Externalizing adult subtype is very similar to the adolescent Angry/Externalizing subtype, with psychopathic features and a tendency to abuse alcohol. These adults fit the model of children of alcoholics who themselves become alcoholics. As expected, the extent to which patients matched this dimension predicted poor adaptive functioning, poor childhood attachment, childhood psychopathy, and a family history of criminality.
The Emotionally Dysregulated subgroup is almost identical to the adolescent Emotionally Dysregulated subtype. Emotionally Dysregulated adults had the highest rates of major depression and BPD. High loadings on this Q-factor were associated with poor adaptive functioning, a poor childhood environment, sexual and physical abuse, and suicide, all of which have been found to be associated with BPD (Zanarini, 1997).
Finally, the Reactive/Somatizing subtype tended to be sensitive and reactive to interpersonal conflicts, develop somatic symptoms in response to stress, ruminate on problems, hold grudges, and be critical, controlling, anxious, and angry. These adults also showed a significantly high rate of major depression. These adults diverged from the adolescent Sexualized/Self-defeating subtype. One hypothesis would be that some of the histrionic aspects of the adolescent subtype were expressed in adults through somatic symptoms. However, this hypothesis is speculative and deserves further research.
The study has several limitations. First, patients were included as part of a study of adolescent and adult personality pathology, and those with an alcoholic parent were identified by clinician report. A more accurate method to identify COAs would involve administering an assessment measure to the parent (i.e., SCID-II) to determine clinical levels of alcohol use. However, clinicians knew their patients on average 20 weeks, those who worked with adolescents are likely to have known the parents, and clinicians followed a highly conservative decision rule in answering developmental history questions, namely to check “absent” unless they were “certain.” Thus, our control groups were more likely to contain false negatives than our COA groups were to include false positives. Further, most prior research has been based on self-reports of parental alcohol use, which are unlikely to be more reliable. With respect to the representativeness of the sample, the identification of a high-functioning subgroup in both populations suggests that clinicians did indeed follow our directions to select patients with any degree of personality pathology, from relatively mild to relatively severe, although the percent of high-functioning patients in non-patients samples would likely be higher. Future research should attempt to replicate these findings using broader samples and better validated measures of parental substance use and abuse.
Second, we relied on a single informant (the treating clinician), which could potentially have led to biases in patient descriptions, although the fact that we found subtypes unknown to clinician-informants and that they predicted external criterion variables in predictable ways mitigates the extent to which such biases could have influenced the results. Nevertheless, future research should rely on data from multiple informants.
Third, relying on clinician report prohibited the collection of other important types of data, including laboratory measures of impulsivity or effortful control, neuroimaging data, or molecular genetic variables that might distinguish the groups (e.g., DAT1, SERT, MAOA). Thus, using a direct interview method in future research would allow for easier access to external correlates beyond those studied here.
Children of alcoholics are a heterogeneous, not homogeneous, group. While we cannot assume that parental alcoholism causes the identified subtypes, we can assume that research investigating COAs as a homogeneous group is inappropriate given that the heterogeneity is not random. It seems advisable to include personality subtyping in all research on children of alcoholics, given the clear heterogeneity within and across disorders that can be accounted for by them. A sample that mixes emotionally dysregulated, psychopathic, inhibited, and high-functioning adolescents and adults is likely to yield findings that are difficult to generalize and may offset each other to show null findings (e.g., inhibited and psychopathic). The data presented here also have clinical implications, in pointing clinicians to different personality patterns they are likely to encounter in working with adolescent and adult COAs.
This research was supported by NIMH grant MH62377 and MH62378 to the last author.
Disclosure of financial interest: The primary instrument used in this study for personality assessment and diagnosis, the Shedler Westen Assessment Procedure, is likely to have commercial applications, available at www.swapassessment.org, although no funding was provided by any commercial entity for this research.
Jonathan Hinrichs, Saint Louis University.
Jared DeFife, Emory University.
Drew Westen, Emory University.