Rorschach Variables and Dysfunctional Attitudes as

This article was downloaded by: [Universitetsbiblioteket i Tromsø] On: 20 August 2012, At: 23:43 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Personality Assessment Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/hjpa20

Rorschach Variables and Dysfunctional Attitudes as Measures of Depressive Vulnerability: A 9-Year FollowUp Study of Individuals With Different Histories of Major Depressive Episodes a

Ellen Hartmann , Marianne Halvorsen a

b c

& Catharina E. A. Wang

d

Department of Psychology, University of Oslo, Norway

b

Department of Pediatric Rehabilitation, University Hospital of North Norway, Tromsø, Norway c

North-Norway Rehabilitation Center, Tromsø, Norway

d

Department of Psychology, University of Tromsø, Norway

Version of record first published: 20 Aug 2012

To cite this article: Ellen Hartmann, Marianne Halvorsen & Catharina E. A. Wang (2012): Rorschach Variables and Dysfunctional Attitudes as Measures of Depressive Vulnerability: A 9-Year Follow-Up Study of Individuals With Different Histories of Major Depressive Episodes, Journal of Personality Assessment, DOI:10.1080/00223891.2012.713881 To link to this article: http://dx.doi.org/10.1080/00223891.2012.713881

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Journal of Personality Assessment, 1–12, 2012 C Taylor & Francis Group, LLC Copyright ISSN: 0022-3891 print / 1532-7752 online DOI: 10.1080/00223891.2012.713881

Rorschach Variables and Dysfunctional Attitudes as Measures of Depressive Vulnerability: A 9-Year Follow-Up Study of Individuals With Different Histories of Major Depressive Episodes ELLEN HARTMANN,1 MARIANNE HALVORSEN,2,3 AND CATHARINA E. A. WANG4 1

Department of Psychology, University of Oslo, Norway Department of Pediatric Rehabilitation, University Hospital of North Norway, Tromsø, Norway 3 North-Norway Rehabilitation Center, Tromsø, Norway 4 Department of Psychology, University of Tromsø, Norway

Downloaded by [Universitetsbiblioteket i Tromsø] at 23:43 20 August 2012

2

Forty-six individuals with different histories of major depressive episodes (MDEs) completed the Rorschach (Exner, 2003) and the Dysfunctional Attitude Scale (DAS; Weissman & Beck, 1978) at 2 assessment points (T1, T2) over a 9-year follow-up. At T1, history of MDE and the Rorschach variable MOR (associated with negative self-image) emerged as significant predictors of number of MDEs over the follow-up. At T2, Rorschach markers of depressive vulnerability and scars were identified (i.e., WSum6, related to illogical thinking; X+%, related to conventional perception and social adjustment; X–%, linked to erroneous judgments; MQ–, associated with impaired social relations; and MOR). Test–retest analyses displayed significant temporal stability in Rorschach variables, with r ranging from .34 to .67 and in the DAS, r = .42. Our findings highlight MDE as a recurrent and serious disorder, number of MDEs as a risk factor for future depressions, and Rorschach variables as markers of depressive vulnerability and scars.

Depression is a highly recurrent disorder that frequently takes a chronic course (see, e.g., Andrade et al., 2003; Halvorsen, Wang, Eisemann, & Waterloo, 2010; Kessler et al., 2003; Monroe & Harkness, 2011; Solomon et al., 2000) and past depression is one of the strongest predictors of future depression (Beevers, Rohde, Stice, & Nolen-Hoeksema, 2007; Solomon et al., 2000). Despite their origins in different theoretical traditions, cognitive and psychodynamic theories have many similar basic properties. Both traditions have proposed variants of the vulnerability-stress model, which accentuates certain preexisting, dysfunctional cognitive schemas and personality constellations rooted in negative relational experiences in childhood (see, e.g., Beck, 2008; Weiner & Bornstein, 2009). With time, such dysfunctional characteristics are thought to develop into relatively stable, trait-like, maladaptive organizations of mental elements that are assumed to antedate and predict the first onset of a major depressive episode (MDE) and to give rise to vulnerability to recurrence or maintenance of MDE in conjunction with stressful life events (Abramson, Metalsky, & Alloy, 1989; Bieling & Grant, 2007; Halvorsen et al., 2010; Hankin, 2008; Ormel, Oldehinkel, & Volleberg, 2004; Shapiro, 1991; Zuroff, Blatt, Sanislow, Bondi, & Pilkonis, 1999). Scar theories of depression argue that whether or not maladaptive cognition and personality traits antedate depression, there are also changes, presumably encoded at the biological level, that happen during an MDE, which are long-lasting and make future episodes more likely (Burcusa & Iacono, 2007; Ingram, Miranda, & Segal, 1998; Kendler, Thornton, & Gardner,

2001; Lewinsohn, Steinmetz, Larson, & Franklin, 1981). Potential scars have been examined within cognitive, emotional, personality, and other areas. In two follow-up studies, Lewinsohn et al. (1981) and Rohde, Lewinsohn, and Seeley (1990) found no evidence that psychological scars were present in previously depressed patients as a consequence of having been depressed. However, in a later 1-year follow-up study of adolescents, Rohde, Lewinsohn, and Seeley (1994) found psychological scars including depressive symptoms, anxiety, and intense emotional reliance on others. Furthermore, in a 5-year followup, Nolen-Hoeksema, Girgus, and Seligman (1992) found that a pessimistic attribution style might be a scar effect. More recently, however, Beevers et al. (2007) tested the scar hypothesis in a 7-year longitudinal study of female adolescents and provided limited support for the hypothesis, and in a review of the research literature, Burcusa and Iacono (2007) found virtually no support for the scar hypothesis. Accordingly, Wichers, Geschwind, van Os, and Peeters (2010) proposed that scars could arise gradually after suffering from multiple episodes. They defined scars as all possible maladaptive changes in cognition, emotion, personality characteristics, behavior, or biology that develop in the aftermath of having suffered from depression and that persist after remission and recovery and render the individual vulnerable to future MDE. The strong relationship between depression and cognition (see, e.g., Beck, 2008) and between depression and personality traits like neuroticism (see, e.g., Butcher, Mineka, & Hooley, 2004; Duggan, Sham, Lee, Minne, & Murray, 1995) has led us and many researchers to focus on possible changes in cognition and personality in the aftermath of MDE. A solid and frequently used measure of cognitive vulnerability is the Dysfunctional Attitude Scale (DAS; Weissman & Beck, 1978), which assesses a variety of rigid and unrealistic attitudes regarding personal adequacy, acceptability, and

Received April 5, 2011; Revised June 22, 2012. Address correspondence to Ellen Hartmann, Department of Psychology, University of Oslo, PB 1094 Blindern, 0317 Oslo, Norway; Email: ellen.hartmann @psykologi.uio.no

1


2 worthiness. Studies examining stability and change of DAS scores have typically found that DAS scores are elevated during MDE and decrease to a normal level after remission, signifying dysfunctional attitudes as a state rather than a trait characteristic (see, e.g., Beevers & Miller, 2004; Clark & Beck, 1999; Haaga, Dyck, & Ernst, 1991; Ingram et al., 1998; Zuroff et al., 1999). However, some follow-up studies have demonstrated a moderate to high stability of elevated DAS scores even after remission of MDE (Beevers & Miller; Wang, Halvorsen, Eisemann, & Waterloo, 2010; Zuroff et al.). Furthermore, in a community-based sample of 750 women, Otto et al. (2007) examined elevation of dysfunctional attitudes as measured by the DAS both as a vulnerability factor for depression and as a scarring effect of depression. The authors found that DAS levels for previously depressed were in between those of clinically depressed and never depressed, which are consistent with both the vulnerability and the scarring perspective. On the other hand, in the prospective part of this study no evidence indicated DAS scores to serve as either depressive scar effects or as unique vulnerability factors. In line with these findings, Zuroff et al. (1999) put forward a state–trait vulnerability model including both stable, trait-linked differences in availability of dysfunctional attitudes and fluctuating, state-linked differences in accessibility of those attitudes dependent on present levels of depressive symptoms. Prospective behavioral high-risk studies have, however, demonstrated that initially never depressed individuals classified as high risk compared to low risk on cognitive depressive vulnerability showed a greater probability of experiencing a first onset of MDE (Alloy et al., 2006). High-risk relative to low-risk initially never depressed individuals also showed a greater chance of recurrence, a larger number of MDEs, and more chronic course (Iacoviello, Alloy, Abramson, Whitehouse, & Hogan, 2006). Thus, negative cognitive style seems to be a risk factor both for first onset and recurrence of MDE. However, these researchers did not test the scar hypothesis. A review by Christensen and Kessing (2006), including longitudinal high-risk and population-based studies, examined whether personality traits predicted first-onset MDE (i.e., the vulnerability perspective) and whether personality might change by experiencing an MDE (i.e., the scar perspective) and found that most studies supported the vulnerability and fewer the scar perspective. The majority of the studies examined MDE according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders (4th ed. [DSM–IV]; American Psychiatric Association, 1994) but used different measures of personality characteristics. The strongest evidence was found for a relation between MDE and neuroticism as assessed by the Eysenck Personality Questionnaire (EPQ; Eysenck & Eysenck, 1975). Among the reviewed studies, however, three longitudinal studies (Clayton, Ernst, & Angst, 1994; Hirschfeld et al., 1989; Wilhelm, Parker, Dewhurst-Savellis, & Asghari, 1999; all cited in Christensen & Kessing, 2006) showed evidence in support of both perspectives. The mixed evidence for the scar hypothesis might be explained by the Wichers et al. (2010) hypothesis that scars develop “gradually along the life cycle, proportionally to the severity and duration of the depressive symptoms experienced” (p. 361). In line with this dimensional view of depression, they also argued that cross-sectional designs could obscure the effects of depressive vulnerability and scarring.

HARTMANN, HALVORSEN, WANG Furthermore, scars as well as vulnerability factors might not be accessible by self-report instruments or interviews (see, e.g., Riso et al., 2006; Wichers et al., 2010) as such methods assess the respondent’s conscious, controlled processes that the individuals are capable and willing to accurately report (Meyer, 1996; Viglione, 1999). Such instruments are susceptible to denial, selfdeception, or self-presentation (Westen & Shedler, 2007) and, moreover, responses are found to fluctuate with clinical state and are therefore often characterized as state dependent (e.g., Zuroff et al., 1999). Depressive vulnerability factors and scar effects such as cognitive schemas and various personality traits are implicit structures. Thus, indirect and less controllable assessment instruments might be more appropriate. The Rorschach method (Exner, 2003; Rorschach, 1921/1942) is a more indirect approach to assessing healthy and disturbed personality characteristics, thus less susceptible to the participant’s control. It is one of the most commonly used performancebased instruments in clinical settings (Camera, Nathan, & Puente, 2000; Hogan, 2005). The mental processes that are activated in the production of Rorschach responses tap into and trigger the test person’s underlying cognitive schemas and personality structures, many of which are normally out of conscious representation, whereas others are not (Meyer, 2002; Meyer & Kurtz, 2006; Weiner, 2003). The behaviors coded during the Rorschach method provide data about the person’s way of functioning in the immediate present, and the assumption is that this way of responding has implications for the person’s approach to events outside the testing situation (McGrath, 2008) as well as how he or she is likely to cope with more critical and stressful challenges in the future (Hartmann & Grønnerød, 2009). The Rorschach response process (Exner, 2003) could therefore disclose built-in mechanisms that might activate and identify dormant and often inaccessible dysfunctional schemas and personality constellations, which might in turn constitute underlying vulnerability factors that dispose toward depression, represent latent scar or trait effects of having experienced depression, or both. Exner, Armbruster, and Viglione (1978) proposed that most Rorschach data are associated with durable trait-like personality characteristics and therefore should remain consistent over time. Zuroff et al. (1999) argued that in prospective studies it might be misleading to exclusively examine changes in outcome mean scores; that is, absolute stability by using analysis of variance (ANOVA) statistics. A sample can demonstrate significant changes in mean scores and still show moderate to high test–retest coefficients; that is, relative or rank order stability as exposed in correlational analyses. Thus, both change and stability need to be addressed. Accordingly, a meta-analysis by Grønnerød (2003) on temporal stability for the Rorschach method, including 36 samples, indicated stability levels ranging from .77 to .97 for immediate retesting and from .65 to .90 after 5 years, which is considerable by any standard. These findings are in accordance with Hankin’s (2008) argument that the pattern of test–retest correlations over time for measures of personality characteristics should be relatively constant regardless of the length of the follow-up period.

THE INDEX STUDY Hartmann, Wang, Berg, and Sæther (2003, called the index study from now on) examined 16 clinically depressed, 19


THE RORSCHACH AND THE DAS: A FOLLOW-UP STUDY previously depressed, and 18 never depressed individuals and found that Rorschach variables identified cognitive and aggressive disturbances in clinically depressed, but not in previously depressed and never depressed individuals. Other Rorschach variables revealed affective and coping disturbances in clinically depressed and to some degrees also in the previously depressed, but not in never depressed individuals. In sum the results gave evidence of vulnerability related to self-esteem, stress, and coping as assessed by SumY (associated with stress-related feelings of helplessness and anxiety), MOR (associated with negative self-image or self-esteem), and EBPer (related to rigid and maladaptive coping strategies), but not to dysfunctions in thinking, judgments, and social cognition as assessed by WSum6 (related to illogical and incoherent thinking), X+% (related to conventional perception and social adjustment), and X–% (associated with erroneous judgments that lead to inadequate adjustment). The index study took place in the years 1998 and 1999 and the participants were recruited from a study by Wang, Brennen, and Holte (2005), which consisted of 61 clinically depressed, 42 previously depressed, and 46 never depressed individuals, all individually diagnosed according to the DSM–IV (American Psychiatric Association, 1994) using the SCID–I interview (First, Spitzer, Gibbon, & Williams, 1997). The participants were a mixture of undergraduate students and patients consulting their general practitioner, all located in Tromsø, Norway. All the SCID interviews were audiotaped and subsequently 30 of these interviews, 10 from each group, were randomly sampled for reliability testing. The interrater agreement (kappa) between two raters was 0.9, p < .0001, indicating a highly satisfactory reliability of the group classification (Wang et al., 2005). None of the SCID interviewers or test administrators took part in the baseline and follow-up assessments of this study. When Hartmann et al. (2003) began their data collection in 1998, Wang et al. (2005) had already tested about 70 individuals who were therefore never invited to the index study. None of the Rorschach administrators of the index study took part in the data collection of Wang et al., but they used the same grouping of individuals, which was released when the data collection of the index study was finished.

THE FOLLOW-UP STUDY Nine years after the end of the data collections of the index study and the study of Wang et al. (2005), Halvorsen et al. (2010) managed to enlist 115 of the individuals who had taken part in the Wang et al. study. The main findings from this follow-up study were that a majority of the clinically and previously depressed participants had experienced one or several MDEs over the 9-year follow-up interval and that the Young Scale Questionnaire (YSQ; Young & Brown, 1990) scales, contrary to any DAS scales, seemed to be promising as depressive vulnerability markers (Halvorsen et al., 2010). In addition, Wang et al. (2010) found significant moderate test–retest correlations for both the DAS and the YSQ over the 9-year-follow-up period. At the end of the test administration of Halvorsen et al. (2010), all their subjects who had participated in the index study were invited to participate in our follow-up study and everyone accepted the invitation. This study reports on this 9-year follow-up study of clinically depressed, previously depressed, and never depressed individuals. The following aims were addressed.

3

Aim 1: Prediction of Number of New MDEs During the Follow-Up Interval Due to the finding of Halvorsen et al. (2010) that the majority of the clinically and previously depressed individuals had suffered one or several MDEs during the follow-up interval, we wanted to explore the incremental validity achieved by combining the Rorschach and the DAS as administered at Time 1 (T1) by Hartmann et al. (2003) and Wang et al. (2005), respectively for predicting number of new MDEs over the 9-year interval. Aim 2: Prediction of Rorschach and DAS Outcome Scores at Time 2 In line with the dimensional view on depression and the development of scars (Wichers et al., 2010), we assumed that alterations in cognition and personality traits will be a function of number of MDEs and therefore probably cannot be identified until the individual has experienced several MDEs. Thus, we anticipated that history of MDE at Time 2 (T2) would significantly predict Rorschach scores at T2. In line with research indicating that the DAS does not discriminate well between previously and never depressed (see, e.g., Haeffel et al., 2005; Halvorsen et al., 2010; Zuroff et al., 1999), we anticipated that history of MDE at T2 would not predict DAS scores at T2. Aim 3: Change and Stability in Mean Scores From T1 to T2 on the Rorschach and the DAS As most of the clinically and previously depressed had suffered new MDEs over the follow-up period, in agreement with the scar perspective we expected that at T2 contrary to at T1, the previously depressed (T1 group status) would reveal similar Rorschach scores as the clinically depressed (T1 group status), which would mostly be located within the psychopathological range, whereas the Rorschach scores of the never depressed (T1 group status) would still be within the normal range. In accordance with research indicating that DAS does not discriminate well between previously and never depressed, we anticipated that only the individuals who were currently depressed at T2 would have disturbed DAS scores. Aim 4: Test–Retest Stability Test–retest or relative stability of depressive vulnerability is a phenomenon that has received relatively little attention (Zuroff et al., 1999), and to our knowledge has never been investigated by the Rorschach. Consistent with the view of Exner et al. (1978), that most Rorschach data are related to trait-like personality characteristics, and the test–retest findings reported by Grønnerød (2003), we anticipated that the Rorschach variables would reveal relatively moderate to high test–retest stability over the 9-year interval, although many of the participants might have experienced various numbers of MDEs and others none during the follow-up interval. Based on the studies of Zuroff et al. (1999), Hankin (2008), Riso et al. (2006), and the recent findings of Wang et al. (2010), we expected moderate test–retest correlations for the DAS. METHOD Participants We were able to sign up 46 (86%) of the 53 Norwegian individuals who had participated in the index study (T1) 9 years

HARTMANN, HALVORSEN, WANG

4 TABLE 1.—Demographic and clinical characteristics at Time 1 and Time 2 for the CD, the PD, and the ND.

Variable Gender (Female/Male) T1 Age T2 Age T2 Years of education T1 Single/recurrent episode T2 Single/recurrent episode T1 Antidepressant T2 Antidepressant

T1 CDa M SD 11/3 28.64 37.41 14.51 3/11 0/14 1 4

10.65 10.54 3.17

T1 PDb M SD 16/1 26.53 35.42 15.23 6/11 0/17

10.52 10.37 2.83

T1 NDc M SD 13/2 21.80 32.04 16.09

2.77 2.64 3.36

1/0

Note. N = 46. T1 = Time 1; T2 = Time 2; CD = clinically depressed; PD = previously depressed; ND = never depressed. a n = 14. bn = 17. cn = 15.


of 22 individuals with recurrent, 9 with single, and 15 with no past MDE. At T2, Halvorsen et al. (2010) through the SCID interview assessed the number of new MDEs during the followup interval as a dimensional variable. In addition, we calculated the T2 history of MDE as the weighted sum of past MDE at T1 (two or more recurrent episodes = 2, a single past episode = 1, and no past episode = 0) plus the number of new MDEs during the follow-up interval.

earlier. At the follow-up assessment (T2), the participants first took part in the study of Halvorsen et al. (2010) where they were rediagnosed and interviewed about the number of MDEs between T1 and T2 according to the criteria of the DSM–IV–TR (American Psychiatric Association, 2000) using the SCID–I interview (First et al., 1997) by Halvorsen and her researchers. For a more detailed description of the design, sample, and group classification at T1 and T2 see Hartmann et al. (2003), Wang et al. (2005), and Halvorsen et al. (2010). Our T2 sample consisted of 39 women and 7 men, aged 28 to 58 (M = 34.61, SD = 8.84); 14 clinically depressed (CD), 17 previously depressed (PD), and 15 never depressed (ND); all T1 diagnostic status. Among the 7 dropouts (all women, 3 CD, 1 PD, and 3 ND), 1 person had died, 1 was not traceable, and 5 were unwilling to take part. The T1 assessment was conducted in the years 1998 and 1999 and the T2 assessment in 2007 and 2008. The mean period of time from T1 to T2 was approximately 9 years (M = 9.45, SD = 1.24). The T1 sample included a mixture of undergraduate students (n = 42) and patients (n = 11) consulting their general practitioner. At both T1 and T2 none of the participants were inpatients and few of the CDs and PDs were on antidepressant medication, indicating that their MDE was mild to moderate. Demographical and clinical characteristics of the participants are presented in Table 1. As shown in Table 1, we used the participants’ original T1 group status at T2. One-way betweengroups ANOVA showed no significant differences between the subgroups according to gender, age, and years of education (p varied between .06 and .38 and the effect sizes calculated using eta-squared varied between .04 and .12). The timing of the T1 and T2 assessments was, however, arbitrary points of time. The group classifications were based on snapshots of symptoms at T1 and T2. As shown in Table 1, some of the CDs and PDs had, for example, suffered one or more previous MDEs before T1 and all CDs and PDs had experienced one or more new MDEs during the follow-up interval. Therefore, we decided that it would be more optimal to use history of MDE at T1 and T2 as the criterion in addition to the diagnostic status CD, PD, and ND at T1 and not use any diagnostic group classification at T2. At T1, Wang et al. (2005) assessed past MDE based on the SCID interview as an ordinal variable, T1 history of MDE, with three classifications: two or more recurrent episodes, a single past episode, and no past episode. Based on this classification, our sample at T1 consisted

Instruments The Rorschach Method (Exner, 2003) is a personality assessment method that induces the participants to use available cognitive, perceptual, affective, problem-solving, and coping resources when solving tasks in an unfamiliar situation. We refined our variable set by replacing Appropriate Human Movement (MA) with Human Movement with minus quality (MQ–) as MA did not discriminate among CDs, PDs, and NDs in the index study. Probably, number of MQ– is a better indication of dysfunctional processes than number of M or MA (associated with empathy and adequate social perception and interpersonal relations), as the presence of MQ– indicates psychopathology (Exner, 2003), and more specifically impairment of social perception, deficient empathy, and poor interpersonal relationship (Weiner, 2003). To reduce the number of Rorschach variables, we dropped the variables that did not discriminate between CDs, PDs, and NDs in the index study (i.e., (CF+C)-FC, C,’ SumV, and DEPI). All the Rorschach variables were originally selected on the basis of standard clinical interpretations that made it reasonable to anticipate the identification of current depression as well as underlying cognitive, perceptual, interpersonal, affective, and coping disturbances that might be risk factors for recurrent depression or scar effects of depression (see, e.g., Exner, 2003; Weiner, 2003). The following nine Rorschach variables from the Comprehensive System (CS) were used as dependent variables: 1. Three cognitive variables: The weighted sum of cognitive Special Scores (WSum6; related to illogical and incoherent thinking), Conventional Perception (X+%; related to conventional perception and social adjustment), and Distorted Form (X–%; associated with erroneous judgments that lead to inadequate adjustment). 2. One interpersonal variable: Human Movement with minus quality (MQ–; associated impairment of social perception, deficient empathy, and poor interpersonal relationships). 3. Four affective variables: Pure color response (C; indicating lack of emotional modulation, which could include intense and uncontrolled affective/aggressive discharge), All diffuse shading responses (SumY; associated with uncontrollable stress, stress-related feelings of helplessness, anxiety, and hopelessness), aggressive movement (AG; related to assertive and competitive behavior and verbal and physical aggression and hostility), and morbid content (MOR; associated with negative self-image or self-esteem). 4. One coping variable: The Pervasive Experience Balance (EBPer; indicating rigid and maladaptive coping strategies; only calculated when there is a marked style in the Experience Balance).


THE RORSCHACH AND THE DAS: A FOLLOW-UP STUDY The DAS–Form A (Weissman & Beck, 1978) is a 40-item self-report inventory that was designed to assess dysfunctional attitudes and beliefs underlying cognitive depressive symptoms that are held to be common among depressed individuals (e.g., “I cannot be happy unless most people I know admire me”). For each item, there is a 7-point scale ranging from totally agree, through neutral, to totally disagree. Possible scores range from 40 to 280. Scores above 125 are considered high. The scale is also intended to measure cognitive vulnerability to depression (Zuroff et al., 1999). In several mood-priming studies, the DAS has been able to distinguish depression-prone individuals from individuals who are not vulnerable to depression (Scher, Ingram, & Segal, 2005). The Beck Depression Inventory–First Edition (BDI–I; Beck, Rush, Shaw, & Emery, 1979) and the Beck Depression Inventory–Second Edition (BDI–II; Beck, Steer, & Brown, 1996) are versions of a 21-item self-report inventory designed to assess the presence and severity of depressive symptoms. Both are rated on a 4-point Likert-type scale ranging from 0 to 3, reflecting the severity of each item. These scales were included in this study to assess depression severity when responding to the Rorschach and the DAS at T1 and T2, respectively. Beck and Steer (1987) classified BDI–I scores as follows: 0–9, minimal; 10–18, mild; 19–29, moderate; and 30–63, severe. BDI–II scores are classified somewhat differently: 0–13, minimal; 14–19, mild; 20–28, moderate; and 29–63, severe (Beck, Steer, Ball, & Ranierei, 1996). Although similar to the BDI–I, in the BDI–II earlier items linked to changes in body image, somatic preoccupation, and work difficulties were replaced by items related to depressive symptoms such as worthlessness, concentration difficulties, and loss of energy in the BDI–II (Beck et al., 1996). Due to these differences in the two versions of the BDI, we converted BDI–I to BDI–II, using the scoring adjustment recommended in the BDI–II manual (Beck et al., 1996) and called it AdjBDI.

Procedure The data collection in this study was time consuming and entailed traveling all over Norway and visiting Copenhagen (Denmark), Stockholm (Sweden), and Alicante (Spain). When Halvorsen et al. (2010) had to terminate their data collection; we continued our collection and were able to recruit 3 additional individuals (1 CD, 1 PD, and 1 ND) from the index study, who only took part in this study and not in the follow-up of Halvorsen et al. The test administrator collected the data about new MDEs during the follow-up interval from these 3 participants after having administered the tests. For the other participants the data about new MDE during the follow-up interval were given to us from Halvorsen et al. after all our data had been collected and coded. In the index study the range of R varied between 11 and 45 and differed significantly between CDs and PDs (effect size in the medium range), thus requiring control for R. There were also eight brief records (R < 14), which for various reasons were not excluded from the index study. Thus, we tried to evade brief records, reduce the range of R, and decrease the variation in average R between the subgroups by using an alternative administration recommended by Dean, Viglione, Perry, and Meyer (2007), and which, according to their findings, seem to have improved CS utility and validity. Following this procedure, the

5

examiner prompted the participants for another response whenever only one response was provided to a card, and allowed only four responses to any card. The prompts were identical to the one outlined by Exner (2003) and the restriction in number of responses was achieved by removing a given card when the participant had provided four responses to it. Otherwise, the participants underwent the Rorschach procedure as outlined in the CS. We administered the DAS and the BDI–II after the Rorschach on the same occasion and according to standard instructions. Due to financial margins, just one well-trained Rorschach researcher (Hartmann) who was blind to the diagnostic status of the participants and to their results on the outcome measures at T1 administered the tests. The local medical research ethics committee endorsed the study. The participants gave written informed consent and were treated in accordance with the Ethical Principles of Psychologists and Code of Conduct (American Psychological Association, 1992).

Control Analyses We calculated effect sizes as Cohen’s (1988) d, and evaluated their levels according to his guidelines as effects around d = .20 are small; effects around d = .50 are medium; and effects around d = .80 are large. Effect sizes were also calculated as eta-squared and partial eta-squared and evaluated according to Cohen’s guidelines as small (.01), moderate (.06), and large (.14). We also followed Cohen’s suggestions for determining the size of the value of the correlation coefficient as small (r = .10), moderate (r = .30), and large (r = .50). An alpha level of .05 was used for all statistical tests when nothing else is stated. The statistics for the outcome variables are presented in Table 2 together with the means and standard deviations for the composite international normative Comprehensive System Rorschach reference data, based on 21 samples of adult data from 17 countries (Meyer, Erdberg, & Shaffer, 2007). A one-sample Kolmogorov–Smirnov test indicated that various Rorschach variables did not meet the assumption of normal distribution. Meyer, Viglione, and Exner (2001) argued that distributions that have a skew smaller than 2.0 or kurtosis smaller than 7.0 can be viewed as a moderately nonnormal distribution. Using these criteria, all Rorschach variables, except C and SumY were suitable for parametric tests. As nonnormal distributions are not so problematic for predictor variables, we decided to use C and SumY as predictor variables and to use both parametric and nonparametric tests for these variables in the test–retest analyses. The range of R varied from 14 to 40, thus there were no Brief protocols (R < 14), and t tests showed that there were no significant differences between the mean number of R between CD, PD, and ND, and the effect sizes were all small. The altered administration had an equalizing effect by evading brief records, decreasing the range of R, and reducing the variability in mean R between the groups. Furthermore, there were no outliers or marked disparities of variance. Scoring and Interrater Reliability Hartmann scored the Rorschach protocols and another assistant coscored 21 of the protocols (seven from each group, randomly selected). Scoring followed the guidelines presented by the CS. The coscorer was blind to the design and hypotheses


6

TABLE 2.—Descriptive statistics for the outcome variables at T1 and T2 and mean and standard deviation for the composite international Comprehensive System reference data for the variables used in this study.


T1

T2

M

SD

Skew

R WSum6 X+% X–% MQ– C SumY AG MOR EBPer

23.39 8.76 .60 .22 .57 .52 .50 .85 2.80 1.18

9.43 10.15 .15 .09 .69 .86 .69 1.41 2.82 1.86

1.03 1.82 –.37 –.19 .83 1.56 1.48 1.96 1.75 1.89

DAS BDI–I AdjBDIa BDI–II

112.11 8.20 9.17

23.36 7.49 8.72

.25 .77 1.24

Kurtosis

M

SD

The Rorschach variables .47 25.76 7.82 3.13 7.81 10.23 –.51 .65 .12 .67 .17 .07 –.46 .83 .81 1.47 .46 .81 2.56 .46 .81 3.55 .80 1.05 2.91 2.50 3.01 3.30 1.57 1.72 The self-report instruments –.40 111.13 25.49 –.31 .84 9.59 8.21

Skew

Kurtosis

.46 1.40 –.52 .47 1.57 2.56 .39 .93 1.34 .52

–.79 1.18 .91 –.28 2.10 2.91 7.87 .21 1.03 –.93

.38

–.35

.48

–.73

CS Reference Data M SD

22.31 7.63 .79 .19 .63 .34 1.34 .54 1.26

7.90 7.75 .11 .11 1.05 .66 1.63 .86 1.43

Note. N = 46. T1 = Time 1; T2 = Time 2; CS = Comprehensive System; R = number of responses; WSum6 = weighted sum of cognitive special scores; X+% = conventional form; X–% = distorted form; MQ– = human movement with minus quality; C = pure color response; SumY = all diffuse shading responses; AG = aggressive movement; MOR = morbid content; EBPer = EB Pervasive; DAS = Dysfunctional Attitude Scale; BDI = Beck Depression Inventory. a Adjusted BDI–I to BDI–II.

of the study. The scoring took place 1 year after the administration of the tests. Accordingly, primary scorer did not remember which protocols belonged to participants who were clinically depressed during the testing. Traditionally, intraclass correlation (ICC) values between .60 and .74 have been interpreted as good reliability, and values between .75 and 1.00 as excellent reliability (Cicchetti, 1994; Meyer et al., 2002). We calculated two-way ICC (2,1) for single judge reliability on protocol-level scoring. ICC values ranged from .74 (MQ–) to .97 (MOR), indicating good to excellent interrater agreement in all variables (see Table 3). As the ICC values were mostly excellent, we used the original scoring for all protocols.

Gender Differences We used a t test to analyze for gender effects (39 female vs. 7 male participants). No t test was significant, which could be due to the very low number of male participants. TABLE 3.—Protocol-level interrater reliability on summary scores (21 Rorschach protocols). Variable

Protocol-Level (ICC)a

WSum6 X+% X–% MQ– C SumY AG MOR EBPer

.90 .79 .78 .74 .81 .83 .86 .97 .93

Note. ICC = intraclass correlation; WSum6 = weighted sum of cognitive special scores; X+% = conventional form; X–% = distorted form; MQ– = human movement with minus quality; C = pure color response; SumY = all diffuse shading responses; AG = aggressive movement; MOR = morbid content; EBPer = EB Pervasive. a The ICC model used is ICC(2,1) and it is only calculated for the outcome variables included in the statistical analyses.

RESULTS Prediction of Number of New Major Depressive Episodes During the Follow-Up Interval Interview data at T2 revealed that 32 of the participants including all CDs, PDs, and 1 ND (T1 group status) in our sample had suffered one or several MDEs during the follow-up interval and 1 individual belonging to the CD group had become chronically depressed (i.e., had no period of remission over the 9-year interval). The remaining 14 participants who were all NDs had not been depressed. We chose hierarchical multiple regression analysis to explore if the joint ability of the T1 history of MDE (two or more recurrent episodes, a single past episode, and no past episode), the T1 DAS, and each of the individual T1 Rorschach variables would add incremental validity to a model for predicting number of new MDEs over the follow-up interval. T1 history of MDE was selected as Solomon et al. (2000) and Beevers et al. (2007) showed that past depression is a predictor of future depression. Because research has indicated that DAS scores might be influenced by a depressed mood state (e.g., Zuroff et al., 1999), we decided to statistically control for depression severity at T1 (AdjBDI) by including this variable into the exploratory regression analyses. As shown in Table 4, AdjBDI was entered into Block 1, explaining 6.4% of the variance in number of new MDEs during the follow-up. After entry of T1 history of MDE, the T1 DAS, and each of the T1 Rorschach variables separately into Blocks 2, 3, and 4, respectively, the total variance explained by the model as a whole was 44%. T1 history of MDE explained a significant additional 23.1% of the variance in number of new MDEs during the follow-up, the DAS explained another nonsignificant additional 2.4% of the variance, and T1 MOR, which was the only Rorschach variable that added significantly, explained another significant additional 12.1% of the variance after controlling for depression severity. As shown in Table 4, T1 history of MDE (β = .390, p < .001) and T1 MOR (β = .374, p < .01) recorded very similar magnitudes.

THE RORSCHACH AND THE DAS: A FOLLOW-UP STUDY TABLE 4.—Hierarchical multiple regression analyses using the T1 BDI, T1 history of MDE, T1 DAS, and T1 Rorschach variable MOR as predictors of number of new MDEs over the follow-up interval (T1–T2). Model Summary at Each Block Entered

R2

R2

R

R

1 2 3 4

T1 BDIa T1 history of MDE T1 DAS T1 MOR

.064 .295 .319 .440

.064 .231∗∗∗ .024 .121∗

.254 .543 .565 .663

.254 .289 .022 .098

Variable

Variable Coefficients on Final Block B SE B β

Block


T1 BDIa T1 history of MDE T1 DAS T1 MOR

.027 .273 .009 .081

–.001 .800 .013 .411

–.005 .390 .168 .374

t –.039 2.934∗∗ 1.405 2.975∗∗

Note. N = 46. T1 = Time 1; T2 = Time 2; BDI = Beck Depression Inventory; MDE = major depressive episode; DAS = Dysfunctional Attitude Scale; MOR = morbid content. a Adjusted BDI–I to BDI–II. ∗ p < .05. ∗ ∗ p < .01. ∗∗∗ p < .001.

Prediction of the Rorschach and the DAS Outcome Scores at T2 We then conducted a series of hierarchical multiple regression analyses to assess the ability of T2 history of MDE to predict the Rorschach and the DAS outcome scores at T2 after controlling for depressed mood state by including depression severity at T2 (BDI–II) in the model. As shown in Table 5, T2 history of MDE explained a significant percentage of the variance in the Rorschach variables WSum6, X+%, X–%, MQ–, and MOR that varied between 9.7% (X–%) and 22.4% (MQ–), whereas for the DAS and the Rorschach variables C, SumY, AG, and EBPer the contribution of T2 history of MDE was nonsignificant.

TABLE 5.—Hierarchical multiple regression analyses using T2 history of MDE as a predictor of each T2 Rorschach and DAS outcome measures after controlling for T2 BDI–II.

7

Change and Stability in Mean Scores From T1 to T2 on the Rorschach Variables and the DAS Descriptive statistics for the CD, PD, and ND on the Rorschach variables and the DAS at T1 and T2 are presented in Table 6. We used the general linear model and conducted a series of mixed between–within subjects ANOVA (Pallant, 2007; Tabachnick & Fidell, 2007) to analyze change and stability in mean scores from T1 to T2 by assessing the impact of diagnostic group status (CD, PD, and ND diagnosed at T1) on the participants’ scores on the selected Rorschach variables and the DAS across two time periods (T1 and T2). We controlled for AdjBDI at T1 and BDI–II at T2. We calculated the differences between the estimated marginal means when controlling for T1 BDI and T2 BDI–II and the raw means for the CDs, PDs, and NDs at T1 and T2. In all cases, the standardized mean differences (Cohen’s d) were smaller than .2. As presented in Table 7, there was a significant interaction between diagnostic group status and time on seven of the nine Rorschach variables (WSum6, X+%, X–%, MQ–, SumY, MOR, and EBPer). The values of p varied from < .001 to .040 and partial eta-squared varied from .139 to .490 and were within the large and moderate range. For these Rorschach variables, PDs’ means had substantially changed from T1 to T2 in the direction of revealing more psychopathology and being more equal to CDs’ means, whereas the means for CDs and NDs at T2 as at T1 were within the psychopathological range for CDs and within the normal range for NDs. Figure 1 demonstrates the changes in estimated mean scores from T1 to T2 on WSum6 for CDs, PDs, and NDs, respectively; Wilks’s Lambda = .822, F(2, 45) = 4.654, p = .015, partial eta-squared = .178. Parallel patterns of significant change in estimated means were found for X+%, X–%, MQ–, SumY, MOR, and EBPer. For C and AG, the interaction between diagnostic group status and time was not significant (p was > .05 and partial eta-squared was .042 and .021, respectively, and within the small range). Concerning the DAS, there was also a significant interaction between diagnostic group status and time. The pattern of change from T1 to T2 was, however, different. Figure 2 presents

Predictor Outcome Measures

R

T2 history of MDE R

T2 WSum6 T2 X+% T2 X–% T2 MQ– T2 C T2 SumY T2 AG T2 MOR T2 EBPer

Rorschach variables .488 .534 .530 .635 .425 .365 .307 .688 .430

.143 .177 .101 .212 .005 .015 .000 .137 .039

T2 DAS

Self-report variable .515

.018

β .383∗ –.439∗∗ .344∗ .524∗∗∗ .070 .114 –.014 .456∗∗∗ .198 .149

Note. N = 46. T2 = Time 2; MDE = major depressive episode; DAS = Dysfunctional Attitude Scale; BDI = Beck Depression Inventory; WSum6 = weighted sum of cognitive special scores; X+% = conventional form; X–% = distorted form; MQ– = human movement with minus quality; C = pure color response; SumY = all diffuse shading responses; AG = aggressive movement; MOR = morbid content; EBPer = EB Pervasive. ∗ p < .05. ∗ ∗ p < .01. ∗∗∗ p < .001.

FIGURE 1.—Illogical and incoherent thinking (WSum6). CD = clinically depressed; PD = previously depressed; ND = never depressed.


8

TABLE 6.—Raw mean scores and standard deviations for the CD, PD, and ND on the DAS and the Rorschach variables at T1 and T2. CDa

PDb

T1



DAS BDId

T2

NDc

T1

M

SD

M

SD

17.79 .48 .29 .99 1.14 .64 1.36 5.43 2.68

13.29 .14 .08 .68 .95 .63 1.99 3.59 2.87

13.71 .59 .22 1.57 1.00 1.29 1.21 4.43 2.37

11.55 .12 .07 1.10 .96 1.77 1.03 4.09 1.75

114.57 11.29

25.36 7.88

128.29 15.71

25.53 7.81

M

T2 SD

M

Rorschach variables 4.61 8.76 .11 .58 .06 .21 .72 1.18 .87 .59 .87 1.35 1.30 .94 1.58 2.82 1.31 2.02 Self-report variables 113.35 22.46 106.65 11.41 7.15 8.00 4.47 .68 .19 .65 .41 .53 .76 1.88 .62

T1 SD

T2

M

SD

M

SD

10.54 .10 .04 .88 1.28 1.32 1.25 2.10 1.47

5.20 .63 .19 .07 .07 .33 .47 1.40 .41

5.05 .17 .09 .26 .27 .49 1.06 .82 .85

1.22 .79 .10 .00 .27 .07 .40 .20 .33

2.63 .05 .05 .00 .59 .26 .63 .41 1.29

24.82 5.96

108.40 4.67

24.49 9.33

99.87 4.20

20.75 3.23

Note. CD = clinically depressed; PD = previously depressed; ND = never depressed; DAS = Dysfunctional Attitude Scale; T1 = Time 1; T2 = Time 2; WSum6 = weighted sum of cognitive special scores; X+% = conventional form; X–% = distorted form; MQ– = human movement with minus quality; C = pure color response; SumY = all diffuse shading responses; AG = aggressive movement; MOR = morbid content; EBPer = EB Pervasive; BDI = Beck Depression Inventory. a n = 14. bn = 17. cn = 15. dThe AdjBDI at T1 and the BDI–II at T2.

the changes in estimated means on DAS for CDs, PDs, and NDs, respectively; Wilks’s Lambda = .857, F(2,45) = 3.599, p = .04, partial eta-squared = .143 within the large range. CDs’ means were significantly increased, indicating more dysfunctional attitudes at T2 than at T1; the partial eta-squared was in the large range; and PDs’ and NDs’ means were within the normal range at both assessment points.

Relative Stability for the Rorschach Variables and the DAS Between T1 and T2 To examine the relative stability of the outcome variables from T1 to T2, we used test–retest Pearson product–moment correlation coefficient. As DAS scores might be influenced by a depressed mood state (e.g., Zuroff et al., 1999), we calculated zero-order and partial correlations controlling for depression

severity both at T1 (AdjBDI) and T2 (BDI–II) for all the outcome measures. We conducted two series of analyses; one with the full sample (N = 46) and one with the vulnerable sample of individuals with single and recurrent MDE at T1 (n = 31). The results are presented in Table 8. For the full sample there were moderate to strong positive, partial test–retest correlations for the DAS scale and the Rorschach variables WSum6, X–%, MQ, C, SumY, AG, and MOR, with r ranging from .34 (SumY) to .67 (MOR). The test–retest partial correlation for X+% and EBPer was nonsignificant. An inspection of the test–retest zero-order correlations ranging from .29 (X+%) to .70 (MOR) showed almost the same pattern of significant and nonsignificant test–retest correlations, except that the zero-order correlations for X+% and EBPer were significant. The analyses with the vulnerable sample revealed a very similar pattern of significant and nonsignificant correlations. Thus, controlling for

TABLE 7.—Mixed between–within subjects analysis of variance interaction effects for T1 defined CD, PD, and ND on the Rorschach and DAS Variables at T1 and T2 after controlling for T1 BDI a and T2 BDI–II. Variable

Wilks’s Lambda


.822 .658 .608 .852 .958 .860 .979 .782 .861

DAS

.857

F(2, 45)

Significance

Rorschach variables 4.654 .015 11.195