|Home | About | Journals | Submit | Contact Us | Français|
We examined concurrent and prospective associations of Behavioral Approach System (BAS)-relevant and non-BAS-relevant cognitive styles with bipolar spectrum disorders. Controlling for depressive and hypomanic/manic symptoms, 195 individuals with bipolar spectrum disorders scored higher than 194 demographically matched normal controls on BAS sensitivity and BAS-relevant cognitive dimensions of performance concerns, autonomy, and self-criticism, but not on BIS sensitivity and non-BAS-relevant dimensions of approval-seeking, sociotropy, and dependency. Moreover, group differences on autonomy fully mediated the association between higher BAS sensitivity and bipolar status. In addition, only BAS-related cognitive dimensions predicted the likelihood of onset of depressive and hypomanic/manic episodes among the bipolar individuals over a 3.2-year follow-up, controlling for initial symptoms and past history of mood episodes. Higher autonomy and self-criticism predicted a greater likelihood of hypomanic/manic episodes and higher autonomy predicted a lower likelihood of major depressive episodes. In addition, autonomy mediated the associations between BAS sensitivity and prospective hypomanic/manic episodes. These findings suggest that individuals with bipolar spectrum disorders may exhibit a unique profile of BAS-relevant cognitive styles that influence the course of their mood episodes.
Bipolar spectrum disorders are prevalent (4.4% of a nationally representative U.S. sample; Merikangas et al., 2007) and often produce significant impairment such as poorer academic achievement, erratic work history, divorce, suicide, and substance abuse (e.g., Angst, Stassen, Clayton, & Angst, 2002; Conway, Compton, Stinson, & Grant, 2006; Goodwin & Jamison, 1990; Grant et al., 2004; Nusslock, Alloy, Abramson, Harmon-Jones, & Hogan, 2008; Quackenbush, Kutcher, Robertson, Boulos, & Chaban, 1996; Strakowski, DelBello, Fleck, & Arndt, 2000). These disorders appear to form a spectrum of severity from the milder subsyndromal cyclothymia, to bipolar II disorder, to full-blown bipolar I disorder (e.g., Akiskal, Djenderedijian, Rosenthal, & Khani, 1977; Akiskal, Khani, & Scott-Strauss, 1979; Cassano et al., 1999; Depue et al., 1981; Goodwin & Jamison, 1990). Moreover, milder forms of bipolar disorder often progress to the more severe forms (e.g., Akiskal et al., 1977; 1979; Shen, Alloy, Abramson, & Grandin, 2008), providing support for the spectrum concept.
Recently, there has been increasing interest in psychosocial processes in the onset, course, and treatment of bipolar spectrum disorders (see Alloy et al., 2005; 2006a,c,d; in press for reviews). Psychological processes important in unipolar depression have been extended to research on bipolar depression specifically and bipolar spectrum disorders in general (see Alloy et al., 2005; 2006a,c,d; Cuellar, Johnson, & Winters, 2005; Johnson & Kizer, 2002 for reviews). In particular, cognitive theories of unipolar depression (e.g., Abramson, Metalsky, & Alloy, 1989; Beck, 1967; 1987) have been extended to bipolar disorder to address whether maladaptive cognitive styles similar to those seen among unipolar depressed individuals are also observed among bipolar individuals and predict the expression or course of bipolar disorder. Recent reviews of research on cognition in bipolar disorder (Alloy et al., 2005; 2006a,c,d; in press) concluded that individuals with bipolar spectrum disorders exhibit underlying cognitive patterns as negative as those of unipolar depressed persons overall, but with certain unique characteristics (see below). However, the degree of negativity of the observed cognitive styles of bipolar individuals depends in part on their current mood state and on whether the cognitive style assessment uses explicit or implicit measures.
In addition, there is some evidence that cognitive styles also predict the course of bipolar disorder, alone or in combination with relevant life events. However, these findings are also mixed. Whereas two studies (e.g., Johnson et al., 2000a; Johnson & Fingerhut, 2004) found that negative cognitions predicted subsequent depressive, but not manic, symptoms in bipolar I samples, Scott and Pope (2003) reported that negative self-esteem was the most robust predictor of any type of relapse (depressive or hypomanic) at 12-month follow-up in hypomanic bipolar patients.
Studies testing whether Beck’s (1987) sociotropic and autonomous cognitive styles combine with congruent stressful events (interpersonal events for sociotropic individuals and achievement events for autonomous individuals) to predict bipolar symptoms are also mixed. In a bipolar I sample, Hammen, Ellicott and Gitlin (1992) observed that the sociotropy x negative interpersonal events interaction, but not the autonomy x negative achievement events interaction, predicted subsequent symptom severity (although Hammen et al., 1989 only obtained a trend for this effect). In contrast, in a bipolar spectrum sample, Francis-Raniere, Alloy and Abramson (2006) found that autonomous/self-critical/perfectionistic cognitive styles interacted with congruent negative events to predict increases in depressive symptoms and with congruent positive events to predict increases in hypomanic symptoms over follow-up.
Finally, two additional longitudinal studies examined attributional style and dysfunctional attitudes as predictors of bipolar symptoms. Among both unipolar and bipolar spectrum participants, Alloy, Reilly-Harrington, Fresco, Whitehouse, and Zechmeister (1999) found that a negative attributional style for negative events interacted with later negative events to predict longitudinal increases in depressive symptoms and a positive attributional style for positive events interacted with later positive events to predict longitudinal increases in hypomanic symptoms. Dysfunctional attitudes did not predict symptom changes combined with life events. On the other hand, Reilly-Harrington, Alloy, Fresco, and Whitehouse (1999) reported that initial negative attributional styles, dysfunctional attitudes, and negative self-referent information processing each interacted with subsequent negative life events to predict increases in both depressive and hypomanic symptoms within a bipolar spectrum sample. Given that bipolar I and II individuals (in Reilly-Harrington et al.) have a course of disorder that includes major depressive episodes, they may be more responsive to negative life events than the bipolar individuals without major depression in the Alloy et al. study.
One factor that may contribute to the mixed findings with regard to the cognitive style - bipolar disorder concurrent and prospective associations is the particular types of cognitive styles examined. The Behavioral Approach System (BAS) is a psychobiological system that regulates approach behavior and appetitive motivation in response to goals and rewards (e.g., Davidson, 1999; Gray, 1981; 1982). When relevant external (e.g., an attractive goal object) or internal (e.g., expectancies of goal attainment) cues activate the BAS, the person increases movement toward attaining goals and cognitive activity (e.g., planning, self-efficacy, hope) aimed at promoting goal attainment. BAS activation has been associated with hope, elation, and happiness (Depue & Iacono, 1989; Gray, 1994). Based on work on BAS-mediated motivation, affect, and behavior, Depue and colleagues (Depue & Iacono, 1989; Depue, Krauss & Spoont, 1987) proposed a BAS dysregulation theory of bipolar disorder (see also Alloy et al., in press, Johnson, 2005, and Urosevic, Abramson, Harmon-Jones, & Alloy, 2008a for updates and reviews of evidence for this theory). According to this theory, bipolar individuals have a hypersensitive BAS that becomes dysregulated easily. Bipolar individuals’ hypersensitive BAS is hypothesized to respond with extremely elevated affect, high energy, excessive goal-seeking and self-confidence (i.e., hypomanic/manic symptoms) in response to BAS activation-relevant events involving themes of reward incentive, goal striving and attainment, and with depressed affect, low energy, anhedonia, and hopelessness (i.e., depressive symptoms) in response to BAS deactivation-relevant events such as definite failure or non-attainment of goals (Depue & Iacono, 1989; Depue et al., 1987; Fowles, 1988; 1993; Urosevic et al., 2008a). BAS sensitivity is a construct of hyper-responsiveness to incentive stimuli at all levels of analysis, including, but not limited to, cognition. From a BAS perspective, individuals with bipolar spectrum disorders should exhibit cognitive styles specific to the themes of high drive/incentive motivation associated with high BAS sensitivity. Indeed, high BAS sensitivity may be a temperament that contributes to the development of BAS-relevant cognitive styles (Alloy et al., in press) and BAS-relevant cognitive styles may at least partially mediate the association between high BAS sensitivity and mood episodes. Some of the mixed findings in the studies examining the bipolar - cognitive style association may be attributable to the failure to examine BAS-relevant styles specifically.
Several studies have obtained results consistent with distinctive BAS-relevant cognitive and personality styles in individuals with bipolar disorder. Rosenfarb, Becker, Khan, and Mintz (1988) reported that only remitted unipolar depressed women exhibited higher dependency than controls, whereas both remitted bipolar and unipolar depressed women exhibited higher self-criticism than controls on the Depressive Experiences Questionnaire (DEQ; Blatt, D’Afflitti & Quinlan, 1976). Both Scott, Stanton, Garland, and Ferrier (2000) and Goldberg, Gerstein, Wenze, Welker, and Beck (2008) also found that the perfectionism (performance evaluation), but not the approval by others, subscale of the Dysfunctional Attitudes Scale (DAS; Weissman & Beck, 1978) distinguished remitted or manic bipolar patients from normal controls. With current clinical state controlled, Lam, Wright and Smith (2004) found that bipolar disorder patients scored higher than unipolar depressed patients only on Goal Attainment dysfunctional attitudes (e.g., “A person should do well at everything”), but not dependent or achievement dysfunctional attitudes. In addition, higher Goal Attainment attitudes were associated with increased past hospitalizations for bipolar episodes in general and for mania in particular. In a later study, Wright, Lam, and Newsom-Davis (2005) found that whereas a positive mood induction led to a decrease of Goal Attainment dysfunctional attitudes among unipolar depressed patients, bipolar disorder patients did not exhibit this reduction in Goal Attainment attitudes following the same positive mood induction. Moreover, Lozano and Johnson (2001) found that an achievement-striving style predicted manic, but not depressive, symptoms in a 6-month follow-up of bipolar I patients. Finally, as described above, Francis-Raniere et al. (2006) found that a BAS-relevant cognitive style involving high autonomy, self-criticism, and performance-focus predicted increases in hypomanic symptoms in interaction with style-congruent positive events and increases in depressive symptoms in interaction with style-congruent negative events over follow-up in individuals with cyclothymic and bipolar II disorders. Thus, these studies suggest that the cognitive styles of bipolar individuals may be marked by the BAS-relevant traits of autonomy, perfectionism, self-criticism, and goal-striving, whereas there is less consistent evidence for sociotropic, dependent, or approval-seeking cognitive styles in bipolar individuals as one typically sees in unipolar depression (see Zuroff, Mongrain & Santor, 2004 for review).
This study examined the cross-sectional and prospective associations of BAS-relevant cognitive styles with bipolar spectrum disorders among participants in the Longitudinal Investigation of Bipolar Spectrum (LIBS) project (Alloy et al., 2008; Nusslock et al., 2007; Shen et al., 2008). We compared a large sample of individuals with bipolar spectrum disorders to demographically similar normal controls at baseline on self-report measures of cognitive/ personality styles commonly used in the unipolar depression literature (DAS [Weissman & Beck, 1978]; Sociotropy-Autonomy Scales [SAS; Beck, Epstein, Harrison, & Emery, 1983]; DEQ [Blatt et al., 1976]). We hypothesized that controlling for current depressive and hypomanic/manic symptoms, bipolar spectrum individuals would score higher than controls on BAS sensitivity and the BAS-relevant cognitive style dimensions of performance concerns, autonomy, and self-criticism, but not on the non-BAS-relevant dimensions of approval-seeking, sociotropy, and dependency. Moreover, we hypothesized that the group differences on BAS-relevant cognitive styles would partially mediate group differences on BAS sensitivity. We also examined whether the cognitive styles assessed at Time 1 predicted prospective onsets of hypomanic/manic and depressive episodes over follow-up among bipolar spectrum participants. We hypothesized that controlling for Time 1 levels of hypomanic/manic and depressive symptoms and past history of mood episodes, BAS-relevant cognitive styles would be more likely to predict prospective onsets of mood episodes than would non-BAS-relevant styles. Finally, we predicted that BAS-relevant cognitive styles would partially mediate the prospective prediction of mood episodes by BAS sensitivity. Predictions that BAS-relevant cognitive styles would mediate BAS sensitivity effects were novel and have never been examined before.
Participants were from the LIBS Project and were selected based on a 2-phase screening procedure. In Phase I, approximately 20,500 18-24 year old students at Temple University and the University of Wisconsin were administered the revised General Behavior Inventory (GBI; Depue, Krauss, Spoont, & Arbisi, 1989) to identify potential bipolar spectrum and control participants. Students who met the initial GBI criteria (see Measures) for either the bipolar spectrum or control group were eligible for Phase II. In Phase II, 1,730 participants were administered a semi-structured diagnostic interview with an expanded Schedule for Affective Disorders and Schizophrenia-Lifetime (exp-SADS-L; Endicott & Spitzer, 1978) interview. Students who met Diagnostic and Statistical Manual for Mental Disorders, 4th Edition (DSM-IV; American Psychiatric Association, 1994) and/or Research Diagnostic Criteria (RDC; Spitzer, Endicott & Robins, 1978) for bipolar II or cyclothymia were eligible for the longitudinal LIBS Project.1 Based on the exp-SADS-L interview, control participants had no lifetime history of any Axis I psychopathology, with the exception that they could have a specific phobia. In recruiting control participants, we took steps to insure that the bipolar and control groups were similar on age, gender, and ethnicity.
Of the 285 eligible bipolar and 308 eligible control participants following Phase II, 227 (79.6%) bipolar spectrum (164 bipolar II and 63 cyclothymic) and 227 (73.7%) control participants completed the Time 1 assessment of the longitudinal study. Of these individuals, 195 bipolar spectrum (152 bipolar II, 43 cyclothymic) and 194 control participants also completed the initial symptom assessment and comprised the current study sample (see Table 1 for demographic characteristics). The bipolar and control groups did not differ on age, gender, or ethnicity. Among the bipolar participants, 30 (15.4%) were receiving formal treatment (medication and/or psychotherapy) at the outset of the study and 31 (15.9%) progressed to a bipolar I diagnosis (had a first onset of mania) over the follow-up. Thirty-two (74.4%) of the 43 cyclothymic participants progressed to a bipolar II diagnosis (had a first onset of major depression) during the follow-up. The current sample was representative of the Phase I screening sample on demographics and did not differ from Phase II eligible individuals who did not participate on demographics, diagnosis, treatment history, and GBI scores. Participants completed several measures of cognitive style at Time 1 and self-report measures of depressive and hypomanic/manic symptoms at an initial symptom assessment (see Measures below). In addition, a subset of participants (130 Bipolar, 160 control) completed the Behavioral Inhibition System/Behavioral Activation System Scales (BIS/BAS Scales; Carver & White, 1994) at Time 1.2 Bipolar spectrum individuals who completed Time 1 and the initial symptom assessment, entered the prospective phase of the LIBS Project, and were not in a mood episode at Time 1 (n = 167) were included in the longitudinal analyses. During the prospective follow-up, participants were administered an expanded SADS-Change diagnostic interview (exp-SADS-C; Spitzer & Endicott, 1978) approximately every 4 months to assess the occurrence of mood episodes.3 Prospective analyses were based on an average of 38.1 ± 19.1 (range = 73) months of follow-up.
The revised GBI (Depue et al., 1981; 1989) assesses chronic affective disorders in the general population. It contains 73 items that measure the frequency, intensity, and duration of core bipolar experiences on two subscales: Depression (D) and Hypomania and Biphasic (HB) items combined. As recommended by Depue et al. (1989), we used the case-scoring method to identify potential bipolar spectrum and control participants at the Phase I screening. Only items rated a “3” (“often”) or “4” (“very often or almost constantly”) on the GBI 4-point frequency scale counted toward the score on each subscale. Based on cutoffs recommended by Depue et al. (1989), participants who scored ≥ 11 on the D scale and ≥ 13 on the HB scale were identified as potential bipolar participants, whereas those who scored below these cutoffs formed a potential normal group. These criteria were based on Depue et al.’s findings (1989) and a pilot study for the LIBS Project in which these cutoffs were validated against diagnoses derived from exp-SADS-L interviews. The GBI has good internal consistency (αs = .90-.96), test-retest reliability (rs = .71-.74), high specificity (.99) and adequate sensitivity (.78) for bipolar spectrum conditions (Depue et al., 1981; 1989). Also, it has been validated extensively in college, psychiatric outpatient, and offspring of Bipolar I patient samples (Depue et al., 1981; 1989).
The exp-SADS-L (Endicott & Spitzer, 1978) is a semi-structured diagnostic interview that assesses current and lifetime history of Axis I disorders. The original SADS-L was expanded in several ways for the LIBS Project: 1) probes were added to allow for DSM-IV as well as RDC diagnoses; 2) questions were added to better capture the nuances of episodes and frequency and duration of symptoms for depression, hypomania, mania, and cyclothymia; 3) the order of interview questions was changed to increase the interview’s efficiency and comprehension; and 4) sections were added to assess eating disorders, ADHD, and acute stress disorder, additional probes were added in the anxiety disorders section, and organic rule-out and medical history sections were appended. An inter-rater reliability study based on 105 jointly rated exp-SADS-L interviews yielded κ’s ≥ .96 for bipolar spectrum diagnoses. Extensively trained research assistants, unaware of participants’ Phase I GBI scores, conducted the interviews. Training consisted of approximately 200 hours of reading and didactic instruction, watching videotaped interviews, role playing, discussing case vignettes, and extensive practice conducting live interviews with supervision and feedback. Consensus DSM-IV and RDC diagnoses were determined by a 3-tiered standardized diagnostic review procedure involving senior diagnosticians and an expert psychiatric diagnostic consultant.
Prospective onsets of mood episodes were assessed with an expanded SADS-C diagnostic interview administered approximately every 4 months during the follow-up. The exp-SADS-C was used to assess onsets, remissions, and relapses and recurrences of diagnosable episodes of Axis I disorders, including DSM-IV and RDC major depression (MD) and hypomanic and manic (HYP/MA) episodes, during each prospective interval. Interviewers were blind to participants’ cognitive style scores, BIS/BAS scores, Phase I GBI scores, and Phase II diagnostic status. The SADS-C was expanded like the SADS-L. In addition, features of the Longitudinal Interval Follow-up Evaluation (LIFE II; Shapiro & Keller, 1979) were added to the exp-SADS-C to systematically track the course of symptoms and episodes during the follow-up. The exp-SADS-C inquired about the presence of each symptom on a daily basis during the prospective interval. Inter-rater reliability (Alloy et al., 2008) for the exp-SADS-C in joint ratings of 60 interviews was good (κ > .80). In a validity study, participants dated their symptoms on the exp-SADS-C with at least 70% accuracy compared with daily symptom ratings made over a 4-month interval (Alloy et al., 2008). Nusslock et al. (2007) provide further details about exp-SADS-L and SADS-C mood episode diagnoses.
The original Beck Depression Inventory (BDI; Beck, Rush, Shaw, & Emery, 1979) was used to assess initial levels of depressive symptoms. The BDI has been validated in student samples and the internal and test-retest reliabilities are good in both clinical and non-clinical samples (Beck, Steer & Garbin, 1988). Initial levels of hypomanic/manic symptoms were assessed with the Halberstadt Mania Inventory (HMI; Alloy et al., 1999). This 28-item self-report measure is modelled after the BDI and assesses the affective, motivational, cognitive, and somatic symptoms of (hypo)mania. Like the BDI, the HMI asks participants to choose one of 4 statements graded in severity that best describes their experience, e.g., “I do not feel particularly happy,” “I feel happy, “ “I feel so happy and cheerful it’s like a high,” or “I am bursting with happiness and I’m on top of the world.” The HMI has good internal consistency (α = .82), and has demonstrated convergent validity with the MMPI-Mania scale (r = .32, p < .001), as well as discriminant validty with the MMPI-Depression scale (r = -.26, p < .001) and the BDI (r = -.12, p < .001) (Alloy et al., 1999). The HMI also correlated (r = .46) with hypomanic symptoms rated from the exp-SADS-C interview in the LIBS Project (Alloy et al., 2008) and had an internal consistency of α = .78. Finally, the HMI also shows expected changes as cyclothymic individuals cycle through hypomanic, euthymic, and depressed mood states (Alloy et al., 1999).
The Dysfunctional Attitudes Scale (DAS), Form A (Weissman & Beck, 1978) is a 40-item questionnaire that assesses dysfunctional beliefs regarding concerns about others’ approval and performance expectations on 7-point scales ranging from “totally agree” to “totally disagree.” Two DAS factors have been extracted that are hypothesized to be related to the sociotropic and autonomous subscales of the SAS and the dependent and self-critical subscales of the DEQ (Cane, Olinger, Gotlib, & Kuiper, 1986; Segal, Shaw & Vella, 1989): Approval by Others (AO - 10 items: “My value as a person depends greatly on what others think of me”) and Performance Evaluation/Perfectionism (PE - 15 items; “If I fail partly, it is as bad as being a complete failure”), respectively. We viewed PE as a BAS-relevant dimension of dysfunctional attitudes, but AO as not relevant to BAS sensitivity. In the current sample, α’s for the PE and AO subscales were .89 and .78, respectively. Both factors have shown good construct validity (Francis-Raniere et al., 2006; Segal et al., 1989).
The Sociotropy-Autonomy Scale (SAS; Beck et al., 1983) is a 60-item questionnaire designed to assess Beck’s (1987) depressive personality modes, with 30 items each on the Sociotropy (“I am afraid of hurting other people’s feelings”) and Autonomy (“It is more important to get a job done than to worry about other people’s reactions”) subscales. Autonomy assesses valuing of achievement, mobility, and freedom from control and is BAS-relevant, whereas Sociotropy measures valuing of attachment and fears of abandonment and rejection by others and is not BAS-relevant. Each item is rated on 5-point scales (0%, 25%, 50%, 75%, 100%). The Sociotropy and Autonomy scales have shown good internal consistency (α = 0.90 and 0.93, respectively) and high retest reliability (Beck et al., 1983; Zuroff, Mongrain & Santor, 2004). In the present sample, the α’s for Sociotropy and Autonomy were .93 and .92, respectively. The Sociotropy scale also has high concurrent validity with other measures of dependency and affiliation (Clark, Beck & Brown, 1992), whereas the Autonomy scale is moderately correlated with an autonomy subscale of the Personality Research Form (Clark et al., 1992).
The Depressive Experiences Questionnaire (DEQ; Blatt et al., 1976) is 66 items, rated on 7-point scales (from “strongly disagree” to “strongly agree”), and has 3 factors measuring the depressive personality styles hypothesized by Blatt et al. (1976): Dependency, Self-Criticism, and Efficacy. We only used the Dependency and Self-Criticism subscales in this study. Self-Criticism (“I have a difficult time accepting weaknesses in myself”) is BAS-relevant, whereas Dependency (“Without support from others who are close to me, I would be helpless”) is not. We used the DEQ factor scores as recommended by Blatt et al. (1976). The DEQ has shown high internal and retest reliability (Blatt et al., 1976; Zuroff et al., 1983). In this sample, α’s for Dependency and Self-Criticism were .64 and .87, respectively. The factors have shown good construct validity as well (Zuroff et al., 2004).
The BIS/BAS Scales (Carver & White, 1994) quantify individual differences in sensitivity of the BIS and BAS and are the most frequently used self-report measures for this purpose. The scales include 20, 4-point items, ranging from “strongly disagree” to “strongly agree,” and consist of one BIS subscale, and three BAS subscales: Reward Responsiveness (RR), Drive (D), and Fun-Seeking (FS). The BIS scale has 7 items and assesses sensitivity to potential punishment cues, e.g., “If I think something unpleasant is going to happen, I usually get pretty ‘worked up’.” The BAS-RR scale has 5 items that assess positive responses to rewards, such as, “When I get something I want, I feel excited and energized.” The D scale has 4 items that index vigor and persistence in pursuit of rewards, such as, “When I want something, I usually go all-out to get it.” The FS scale has 4 items that index willingness to impulsively approach rewards, such as, “I will often do things for no other reason than that they might be fun.” Carver and White (1994) reported internal consistencies (α’s) from .59 to .74 for the BIS/BAS scales, and Urosevic et al. (2008b) found good test-retest reliabilities and stabilities in both bipolar spectrum and normal samples. In this sample, the α’s were higher and were .75 for BIS, .81 for BAS Total, .81 for BAS D, .72 for BAS FS, and .66 for BAS RR. Confirmatory factor analyses of the BIS/BAS Scales have confirmed the latent structure of one BIS scale and three correlated BAS subscales (Carver & White, 1994; Campbell-Sills, Liverant, & Brown, 2004). Numerous studies support the construct validity of the BIS/BAS Scales, including their relation to asymmetrical prefrontal cortical activity, affect, personality traits, and performance on reaction-time and learning tasks involving incentives (e.g., Alloy et al., 2006b; Harmon-Jones & Allen, 1997; Sutton & Davidson, 1997; ZInbarg & Mohlman, 1998).
Table 2 displays the bivariate correlation matrix for all variables. Some of these correlations are directly relevant to the study hypotheses and are discussed below. However, several other patterns of correlations are worth noting. Overall, the BAS-relevant (PE, AUT, SC) cognitive styles correlated with each other more strongly than they did with the non-BAS-relevant (AO, SOC, DEP) cognitive styles, which also inter-correlated with each other. Depressive (BDI) and hypomanic/manic (HMI) symptoms showed more positive correlations with the BAS-relevant dimensions than the non-BAS-relevant dimensions of each cognitive style measure, with the exception of the correlation between the BDI and the AUT vs. SOC scales of the SAS. The 3 subscales of the BAS scale inter-correlated moderately with each other. Thus, we examined both BAS Total and each of the BAS subscales in our main analyses.
In describing the cognitive style measures, we proposed that certain cognitive style dimensions (PE, AUT, SC) are BAS-relevant and other dimensions (AO, SOC, DEP) are not BAS-relevant based on their item content. Before testing the study hypotheses, we first sought to verify empirically this categorization of the cognitive style dimensions. The simple correlations in Table 2 show that BAS sensitivity tended to correlate more strongly and positively with the BAS-relevant than the non-BAS-relevant cognitive dimensions, whereas BIS sensitivity correlated more strongly with the non-BAS-relevant dimensions. However, these simple correlations do not control for participants’ current mood state. Thus, we conducted a series of hierarchical regression analyses in which we examined whether BIS and BAS sensitivity were associated with each of the cognitive style dimensions, controlling for levels of depressive (BDI) and hypomanic/manic symptoms (HMI). Hierarchical regression has the advantage that it allows for the examination of the unique contribution of predictors. Table 3 displays the results of these analyses. As expected, controlling for symptom levels, higher BAS scores were significantly associated with higher DAS PE, SAS AUT, and DEQ SC scores, but not with DAS AO, SAS SOC, and DEQ DEP scores. On the other hand, controlling for symptoms, higher BIS scores were significantly associated with higher DAS AO, SAS SOC, and DEQ DEP scores, but were not related to DAS PE and DEQ SC scores. Higher BIS was associated significantly with lower SAS AUT scores. Thus, our initial categorization of the cognitive style dimensions with respect to BAS-relevance was supported.
Next, we proceeded to test the first hypothesis. Diagnostic group was more strongly correlated (see Table 2 for the correlations) with DAS PE than AO (t(1) = 5.75, p < .001), with SAS AUT than SOC (t(1) = 3.35, p < .001), and with DEQ SC than DEP (t(1) = 8.08, p < .001). Although consistent with our first hypothesis, these differences in correlations do not control for current symptom levels. To examine group differences in cognitive styles not attributable to associations between levels of depressive and hypomanic/manic symptoms and cognitive styles, we conducted hierarchical regression analyses in which the various cognitive style scores were regressed onto BDI and HMI scores in Step 1 and Diagnostic Group in Step 2. Table 4 presents the results of these analyses, as well as the means and standard deviations (SDs) of the cognitive style scores for each group. As hypothesized, controlling for depressive and hypomanic/manic symptoms, the bipolar spectrum group scored significantly higher than the control group on DAS PE, SAS AUT, and DEQ SC. The groups did not differ on DAS AO, SAS SOC, or DEQ DEP. We also conducted hierarchical regression analyses to examine diagnostic group differences in BAS and BIS sensitivities controlling for concurrent symptom levels (see Table 4).4 Controlling for depressive and hypomanic/manic symptoms, the bipolar spectrum group scored significantly higher than the control group on BAS Total, BAS D, and BAS FS, but not on BAS RR or BIS.
Given that the bipolar and control groups differed significantly on BAS sensitivity (BAS Total, D, and FS) and on the BAS-relevant cognitive styles (PE, AUT, and SC), we next examined whether the group differences in BAS sensitivity were mediated, at least in part, by the group differences on BAS-relevant cognitive styles. According to Baron and Kenny (1986), to demonstrate mediation we need to show that: 1) BAS sensitivity (Total, D, and FS) predicts the potential mediators (PE, AUT, and SC); 2) the potential mediators predict diagnostic group; 3) the potential mediators still predict diagnostic group with BAS sensitivities controlled; and 4) the associations between BAS sensitivities and diagnostic group are significantly smaller when the mediators are controlled. Steps 1 and 2 of Baron and Kenny’s approach were already demonstrated above.
We conducted a series of hierarchical logistic regressions in which diagnostic group was regressed on BDI and HMI scores in Step 1, a BAS score in Step 2, and a BAS-relevant cognitive style (the mediator) in Step 3. Given that only a subset of participants completed the BIS/BAS Scales, we first reconducted the hierarchical regression analyses examining the association between Diagnostic Group and the BAS-relevant cognitive styles controlling for BDI and HMI scores (shown in Table 4) within this subsample to be sure the significant relationships still held in the smaller sample. Diagnostic Group was significantly related to the BAS-relevant cognitive styles within the smaller sample (p’s < .01), so we proceeded with the mediation analyses.
SAS AUT was significantly associated with Diagnostic Group controlling for BAS Total (Wald = 10.93, p < .001, OR = 1.06, CI = 1.02 - 1.10), BAS D (Wald = 11.74, p < .001, OR = 1.06, CI = 1.03 - 1.10), and BAS FS (Wald = 11.84, p < .001, OR = 1.06, CI = 1.03 - 1.10) and these BAS sensitivities were no longer associated with Group with SAS AUT controlled (p’s changed from .028 to .298 for BAS Total, from .064 to .494 for BAS D, and from .014 to .131 for BAS FS). The Sobel tests for mediation were significant (p’s < .01). Thus, SAS Autonomy fully mediated the BAS sensitivity - Diagnostic Group associations. Although DAS PE and DEQ SC were still significantly associated with Diagnostic Group controlling for symptom levels and BAS sensitivities (DAS PE: Wald = 7.50, p < .006, OR = 2.04, CI = 1.22 - 3.40 with BAS Total controlled; Wald = 6.31, p < .012, OR = 1.90, CI = 1.15 - 3.13 with BAS D controlled; Wald = 7.80, p < .005, OR = 2.06, CI = 1.24 - 3.41 with BAS FS controlled; DEQ SC: Wald = 18.18, p < .001, OR = 3.17, CI = 1.87 - 5.39 with BAS Total controlled; Wald = 17.37, p < .001, OR = 3.03, CI = 1.80 - 5.10 with BAS D controlled; Wald = 19.96, p < .001, OR = 3.41, CI = 1.99 - 5.85 with BAS FS controlled), PE and SC did not mediate the BAS sensitivity - Diagnostic Group associations (BAS Total, D, and FS were still significantly associated with Group controlling for PE and SC). In sum, Autonomy was a mediator of the BAS - Diagnostic Group associations, but Performance Evaluation and Self-Criticism were not.
To examine whether any of the cognitive style dimensions predicted the likelihood of onset of mood episodes among bipolar spectrum participants, we conducted a series of hierarchical logistic regression analyses with the occurrence (yes/no) of MD and HYP/MA episodes during the follow-up as the dependent variables. Bipolar participants currently in a mood episode at Time 1 were excluded from these analyses to insure that episodes were truly prospective. In each logistic regression (n = 167 for these analyses), the length of follow-up (in days) was entered in Step 1, past history of MD or HYP/MA episodes, respectively, was entered in Step 2, initial depressive (BDI scores) and hypomanic/manic (HMI scores) symptoms were entered together in Step 3, and a cognitive style score was entered in Step 4. We included a past history of MD or HYP/MA episodes as a control variable to account for any effects of past mood episodes on the prospective occurrence of new mood episodes. Table 5 displays the results of these analyses.
As shown in Table 5, SAS AUT was the only cognitive style that significantly predicted the likelihood of onset of MD, controlling for length of follow-up, past MD, and initial depressive and hypomanic/manic symptoms. Bipolar participants with higher autonomy scores were less likely to develop a MD episode than those with lower autonomy. Both SAS AUT and DEQ SC significantly predicted onset of HYP/MA episodes, controlling for length of follow-up, past HYP/MA, and initial symptoms. Higher autonomy and self-criticism both predicted a greater likelihood of HYP/MA episode occurrence.
Alloy et al. (2008) previously found that controlling for follow-up time and initial depressive and hypomanic/manic symptoms, higher BAS Total and RR scores predicted a greater likelihood of and shorter time to onset of HYP/MA episodes in the LIBS Project bipolar sample. Likewise, higher BAS RR and BIS scores marginally predicted a greater likelihood and shorter time to onset of MD episodes, controlling for time in study and initial symptoms. Given that Alloy et al. (2008) did not also control for past history of mood episodes, we reconducted these analyses also controlling for past history of MD or HYP/MA episodes, respectively. Consistent with our prior findings, both BAS Total (Wald = 4.24, p < .04, OR = 1.14, CI = 1.01 - 1.29) and BAS RR (Wald = 4.27, p < .04, OR = 1.35, CI = 1.02 - 1.78) predicted a greater likelihood of onset of prospective HYP/MA episodes controlling for past HYP/MA episodes and initial symptoms. BIS (Wald = 3.21, p < .075, OR = 1.13, CI = .99 - 1.29) continued to marginally predict a greater likelihood of prospective MD episodes, controlling for past MD episodes and initial symptoms, but BAS RR was no longer a marginal predictor of MD episodes.
Thus, we tested whether the Time 1 cognitive styles that predicted prospective onsets of MD or HYP/MA (namely, SAS AUT and DEQ SC) mediated the predictive associations between BAS Total and RR and the prospective onset of HYP/MA episodes, and between BIS and the prospective onset of MD episodes. Following Baron and Kenny (1986), to examine mediation, we added the SAS AUT or DEQ SC score on the last step of the hierarchical logistic regression analyses predicting the likelihood of onset of MD or HYP/MA. Time in days (Step 1), past history of MD or HYP/MA, respectively (Step 2), Time 1 BDI and HMI scores (Step 3), and either BAS Total, BAS RR, or BIS (Step 4) were entered on the previous steps before adding in the potential mediator (SAS AUT or DEQ SC) on Step 5. With BAS Total or RR controlled, SAS AUT significantly predicted the likelihood of onset of HYP/MA episodes (Wald = 5.37, p < .02, OR = 1.14, CI = 1.02 - 1.26 with BAS Total; Wald = 6.20, p < .02, OR = 1.15, CI = 1.03 - 1.28 with BAS RR), and these BAS sensitivities no longer predicted HYP/MA episodes with SAS AUT controlled. The Sobel tests for mediation indicated that Autonomy was a mediator of BAS sensitivities (p’s < .05, one-tailed). DEQ SC no longer predicted prospective onsets of HYP/MA significantly with BAS Total or RR controlled; thus, it did not mediate these BAS sensitivities. SAS AUT also significantly predicted a lower likelihood of onset of MD even with BIS (Wald = 3.82, p = .05, OR = .95, CI = .91 - .1.0) controlled, but SAS AUT did not mediate the associations between BIS sensitivity and prospective onset of MD (because BIS sensitivity continued to significantly predict onset of MD even with SAS AUT controlled).
Given the success of cognitive models of unipolar depression in elucidating important cognitive processes in the onset, course, and treatment of depression, there has been much interest in applying these models to bipolar spectrum disorders. However, the evidence for the occurrence of negative cognitive styles independent of current mood state in individuals with bipolar disorders and the role of such cognitive patterns in predicting mood symptoms and episodes in the course of bipolar disorder is mixed (Alloy et al., 2005; 2006a,c,d; Cuellar et al., 2005). The present study was designed to examine whether the concurrent and prospective associations between cognitive styles and bipolar spectrum disorders would be more consistent for a subset of cognitive styles that are BAS-relevant.
Based on a BAS dysregulation model of bipolar disorders (e.g., Alloy et al., 2008; in press; Depue & Iacono, 1989; Depue et al., 1987; Johnson, 2005; Urosevic et al., 2008a), we hypothesized that individuals with bipolar spectrum disorders would differ from normal control individuals on BAS-relevant, but not non-BAS-relevant cognitive styles, and that the group differences on BAS-relevant styles would at least partially mediate group differences in BAS sensitivity. The findings were supportive of the first, and partially supportive of the second, hypothesis. As expected, the correlations between diagnostic group and the BAS-relevant cognitive styles were significantly greater than the correlations between diagnostic group and the non-BAS-relevant styles from the same measures. Moreover, controlling for concurrent levels of depressive and hypomanic/manic symptoms, bipolar participants exhibited significantly higher scores than controls on the BAS-relevant cognitive dimensions of performance evaluation (DAS), autonomy (SAS), and self-criticism (DEQ). However, they did not differ from controls on the non-BAS-relevant dimensions of approval by others (DAS), sociotropy (SAS), and dependency (DEQ). That the associations between bipolar status and elevated scores on BAS-relevant cognitive dimensions held despite controlling for concurrent depressive and hypomanic/manic symptoms suggests that symptomatic state is unlikely to provide a plausible explanation for the group differences. Indeed, higher state levels of hypomanic symptoms (HMI) were associated with higher autonomy and lower need for approval by others, sociotropy, and dependency. In contrast, with state hypomanic symptoms controlled, a tendency to experience recurrent hypomania/mania as reflected in a bipolar diagnosis was not associated with lower scores on the BAS-irrelevant dimensions of approval by others, sociotropy, and dependency.
In addition, our findings are consistent with several other studies that also have reported that individuals with disorders in the bipolar spectrum only exhibit dysfunctional cognitive patterns with BAS-relevant features (Goldberg et al., 2008; Lam et al., 2004; Rosenfarb et al., 1988; Scott et al., 2000; Wright et al., 2005). Thus, individuals with bipolar disorders may exhibit a unique profile of cognitive styles consistent with the goal-striving, drive, and incentive motivation associated with high BAS sensitivity (Alloy et al., 2005; 2006a,c,d; Johnson, 2005), but not dependency, approval-seeking, and attachment attitudes typically observed among individuals with unipolar depression (Zuroff et al., 2004). Interestingly, and consistent with the past unipolar depression findings, we did observe that higher state levels of depressive symptoms (BDI) were associated significantly with higher scores on all of the cognitive style dimensions except SAS autonomy.
Moreover, our findings also go beyond prior research on BAS-relevant cognitive styles in bipolar disorder by examining whether individuals’ BAS-related cognitive styles mediated their elevated scores on BAS sensitivity. Consistent with the mediation hypothesis, bipolar spectrum participants scored higher on self-reported BAS sensitivity (Total, Drive and Fun-Seeking) than the controls and higher BAS sensitivity was significantly associated with elevated scores on the three BAS-relevant cognitive styles that differentiated the bipolar and control groups (DAS performance evaluation, SAS autonomy, and DEQ self-criticism). In addition, we found that autonomy fully mediated the association between bipolar status and BAS sensitivity. In contrast, although performance evaluation and self-criticism continued to be significantly associated with bipolar status controlling for BAS sensitivity, neither mediated the bipolar status - BAS sensitivity association. Although these findings are cross-sectional and not able to demonstrate causal relationships, they are consistent with the idea that high BAS sensitivity contributes to the development of an autonomous cognitive style, which, in turn, may contribute risk for bipolar disorder. Whereas high BAS sensitivity may also contribute to self-critical and perfectionistic cognitive styles and these styles are related to bipolar status, they did not seem to provide a mechanism by which BAS sensitivity is associated with bipolarity (at least as measured here).
We also evaluated whether BAS-related cognitive styles were more likely than non-BAS-relevant styles to predict the onset of mood episodes among bipolar spectrum participants during a 3.2-year prospective follow-up and whether BAS-relevant cognitive styles mediated predictive associations between BAS sensitivity and prospective mood episodes. Consistent with hypothesis, some of the BAS-related cognitive styles predicted significantly the likelihood of onset of major depressive and hypomanic/manic episodes, controlling for initial levels of depressive and hypomanic/manic symptoms and past history of mood episodes. None of the non-BAS-related cognitive dimensions predicted mood episode onsets in our bipolar sample. Whereas higher levels of self-criticism and autonomy predicted a greater likelihood of hypomanic/manic episode onset, higher autonomy was associated with a smaller likelihood of major depressive episode onset. Again, the fact that some BAS-related cognitive dimensions predicted the likelihood of mood episode onset prospectively controlling for initial symptom levels and past history of mood episodes suggests that the prospective associations between BAS-relevant cognitive styles and mood episodes are not readily attributable to residual symptoms or scarring by past episodes associated with elevated cognitive styles. However, it should be noted that the associations between autonomy and self-criticism and prospective mood episode onsets were small to moderate in magnitude. It may be that these BAS-relevant cognitive styles would have greater predictive power for bipolar mood episode onsets in combination with BAS-relevant life events (see Alloy et al., in press; Urosevic et al., 2008).
It is interesting that SAS autonomy predicted a greater likelihood of hypomanic/manic episodes, but a smaller likelihood of major depressive episodes. This particular finding raises the intriguing possibility that some BAS-relevant cognitive styles may not always be maladaptive. Research suggests that bipolar disorder is characterized by both high levels of impairment and achievement (see Nusslock et al., 2008). Thus, an important question is what psychological traits or mechanisms are associated with achievement among individuals with bipolar spectrum disorders. It is possible that some BAS-related cognitive styles, such as high autonomy, also contribute to BAS-mediated adaptive outcomes such as positive goal striving and achievement. Future research needs to test this speculation.
Although autonomy and self-criticism predicted the likelihood of mood episode onsets prospectively, only autonomy mediated the predictive association between BAS sensitivity and prospective hypomanic/manic episodes. Indeed, self-criticism no longer predicted hypomania/mania with BAS sensitivity controlled. Thus, autonomy mediated the associations between BAS sensitivity and both bipolar diagnosis (cross-sectional analyses) and hypomania/mania onset (prospective analyses). The prospective mediation findings for autonomy are particularly noteworthy because prospective data allow for a stronger test of mediation than do cross-sectional analyses. These prospective findings for autonomy provide further support for the idea that high BAS sensitivity may contribute to the development of an autonomous cognitive style, which, in turn, increases risk for bipolar disorder and hypomanic/manic episodes. Given that an autonomous cognitive style as measured by the SAS involves an emphasis on individualistic achievement, this finding is consistent with Lozano and Johnson’s (2001) report that an achievement-striving style predicted manic symptoms in a 6-month follow-up of bipolar I patients. Moreover, our results suggest that an autonomous cognitive style mediates the effects of a temperament characterized by high drive and incentive motivation on bipolarity.
This investigation has several strengths. These include the inclusion of a large sample of individuals with bipolar spectrum disorders and demographically similar normal controls, the use of standardized diagnostic interviews and criteria, interviewers blinded to cognitive style and BIS/BAS scores, a prospective longitudinal design, conservative statistical tests of the study hypotheses, and an examination of whether elevated BAS sensitivity among bipolar individuals is mediated by BAS-relevant cognitive styles.
However, it is important to recognize this investigation’s limitations as well. First, the study sample consisted of undergraduates, which although ethnically and socioeconomically diverse, may not be representative of community or clinical samples. Replication of our findings in a community sample with bipolar spectrum disorders and in samples with more severe bipolar I disorder is important. However, bipolar II and cyclothymia tend to be understudied relative to bipolar I disorder, and are often risk factors for the progression to bipolar I disorder (e.g., Shen et al., 2008), suggesting the value of the present study as well. Second, cognitive styles were assessed with self-report instruments only. Although the self-report measures chosen for this study are reliable and valid assessments of cognitive style, future tests of associations between BAS-related cognition and bipolar disorder may benefit from use of task-based measures of cognition as well. Similarly, although the BIS/BAS Scales have been validated against behavioral (Zinbarg & Mohlman, 1998) and neurobiological (Harmon-Jones & Allen, 1997; Sutton & Davidson, 1997) indices of BAS sensitivity, future studies of the relationship between cognitive styles and BAS sensitivity would also benefit from use of multiple indicators of BAS (e.g., EEG).
Taken together, the present findings suggest that individuals with disorders in the bipolar spectrum may be characterized by a unique profile of cognitive styles that are relevant to and may mediate BAS sensitivity. Such BAS-related cognitive styles may also influence the course of bipolar disorder and contribute some degree of vulnerability to onsets of affective episodes among bipolar individuals. In conclusion, this investigation suggests that a BAS dysregulation model of bipolar disorder may be promising for understanding the nature of cognitive functioning in bipolar disorder and warrants further study.
This research was supported by National Institute of Mental Health grants MH52617 to Lauren B. Alloy and MH52662 to Lyn Y. Abramson.
1Participants who met criteria for bipolar I disorder were excluded because an aim of the LIBS Project was to examine the understudied “softer” bipolar conditions and to identify risk factors that predicted progression to bipolar I status over time.
2The BIS/BAS Scales were administered along with other measures not used in the present analyses at Time 1. Since participants had a right to refuse any portion of the longitudinal study, only a subset of the longitudinal sample completed this extensive set of measures. There were no differences between participants who did and did not complete the BIS/BAS Scales on diagnosis, age, gender, ethnicity, GBI, or cognitive styles scores.
3Although prospective assessments were planned every 4 months, some participants missed planned assessments for a variety of reasons (e.g., out of the country). In such cases, the next assessment was completed as soon as possible and it covered the time since the previously completed assessment. Most participants who missed a planned assessment did not actually attrit from the study. Over the first 4 years of follow-up, there was 10.7% attrition across the bipolar and control groups. The majority (71%) of attrition occurred within the first 8 months of follow-up because the major reason for attrition was the time commitment required by the study, which participants discovered by the first couple of assessments. The likelihood of attrition among participants who made it to the 5th follow-up assessment was very low.
4Alloy et al. (2008) presented similar diagnostic group differences in BAS and BIS scores based on a smaller subset of the LIBS Project sample.
Lauren B. Alloy, Department of Psychology, Temple University.
Lyn Y. Abramson, Department of Psychology, University of Wisconsin-Madison.
Patricia D. Walshaw, Department of Psychiatry, University of California - Los Angeles.
Rachel K. Gerstein, Department of Psychology, Temple University.
Jessica D. Keyser, Department of Psychology, Temple University.
Wayne G. Whitehouse, Department of Psychology, Temple University.
Snezana Urosevic, Department of Psychology, University of Wisconsin-Madison.
Robin Nusslock, Department of Psychology, University of Wisconsin-Madison.
Michael E. Hogan, Department of Psychology, University of Wisconsin-Madison.
Eddie Harmon-Jones, Department of Psychology, Texas A & M University.