Graduate School of Psychology
RESEARCH MASTER’S PSYCHOLOGY THESIS
Status (1st draft or revision): Final version Date: 2 / 03 /2016
1. WHO AND WHERE
Student
Name : Rogier Verhoef
Student ID number : 5807506
Address : Marnixstraat 75C
Postal code and residence : 1015VD Amsterdam Telephone number : 0614173401
Email address : rogier.verhoef1985@gmail.com
Supervisor(s)
Within ResMas (obligatory) : dr. Arnoud Arntz Specialisation : Clinical Psychology Second assessor : dr. Nexhmedin Morina Research center / location : University of Amsterdam
Number of credits (1 ec = 28 hrs) : 35 EC Ethics committee reference code : 2015-CP-4154
http://ce.psy-uva.nl/
Is a schema-congruent interpretation bias in Borderline Personality Disorder dependent on both the valence and ambiguity of events?
Rogier Verhoef
Abstract
Cognitive Models of Borderline Personality Disorder (BPD) propose that a schema-congruent interpretation bias plays an important role in BPD-symptomatology. This study (N = 47) examined interpretational biases in BPD (n = 18), Avoidant Personality Disorder (AVPD, n = 13) and non-patients (n = 16) using film clips where participants had to interpret ambiguous- and non-ambiguous situations differing in valence from a first person perspective. The film clips covered two themes (job interviews and blind-dates). After each film clip participants wrote down how they would rate themselves and the
event/other based on the scenario. Subsequently, participants rated five possible interpretations of the scenario (BPD specific, AVPD specific, OCPD specific, neutral, positive) on a 100 mm Visual Analog Scale (VAS). Results demonstrated BPD patients show more malevolent views of others, hold more self negative views and cope with more pessimism and criticism than non-patients in ambiguous and neutral situations. In addition, BPD patients evaluated others as more malevolent and criticized others more than AVPD patients on ambiguous and neutral job interviews and ambiguous negative blind-dates. BPD
specific results were most evident within the open response format and most pronounced on job interviews. Mediation analyses showed BPD specific belief fully mediate the relationship between BPD and BPD closed response interpretations. In contrast, no relationship was found between BPD specific beliefs and BPD specific open responses. In sum, this study seems to show moderate evidence for a schema-congruent interpretation bias in BPD patients in ambiguous situations. Limitations and possible directions for future research are discussed.
Introduction
Cognitive models of Borderline Personality Disorder (BPD) propose that BPD-symptomatology derives from maladaptive schemas. Schemas are defined as overarching mental representations of oneself, others, and the world (Beck, 1976). These schemas need to be seen as templates for the perception, encoding, storage and retrieval of information (Beck, Freeman, & Davis, 2004). In other words, these schemas form the glasses through which individuals perceive the world. Therefore maladaptive schemas generally distort information processing. One specific information processing bias is interpretation bias. Interpretation bias is the tendency to interpret ambiguous situations consistent with one’s beliefs. Maladaptive schemas in BPD are thought to be characterized by views of the self as being powerless, deprived and victimized whereas others are being perceived as rejecting, abusing and abandoning (Arntz, Weertman & Salet, 2011). Since cognitive models propose that information processing is guided by schemas it seems plausible that information processing in BPD is biased congruent with BPD specific schema content. In other words, maladaptive schemas in BPD should guide the interpretation of ambiguous events consistent with BPD beliefs about the self and others.
Research into interpretational biases in BPD is still in its infancy. So far only few studies
investigated the role of interpretation bias in BPD. Barnow et al. (2008) found that BPD patients evaluate characters in neutral clips more negative than non-patients. A study of Domes et al. (2008) demonstrated that BPD patients interpret mild emotional faces more malevolent than non-patients. In addition, Meyer, Pilkonis & Beevers (2004) showed BPD features are associated with tendencies to judge neutral faces as more rejecting than non-patients. Although prior studies show evidence for a negative interpretation bias in BPD they do not show whether interpretations of BPD patients are congruent with BPD schema-content. To infer the latter the exact content of interpretation bias in BPD patients has to be studied and compared with interpretation bias in other Personality Disorders (PD) such as patients with Avoidant Personality disorder (AVPD) which schema-content has theoretically minimal overlap with BPD. AVPD beliefs of the self are characterized by inadequacy, incompetence, weakness, and being undesirable and others are viewed as better people, being capable, selfconfident, competent, and potentially critical toward them (Arntz, Weertman & Salet, 2011; Beck et al., 2004). Furthermore, as opposed to BPD where in in case of imagined misfortune others are being blamed and criticized, AVPD patients blame themselves and cope with avoidance and compliance (Bamelis, Renner, Heidkamp & Arntz, 2011; Beck et al., 2004). A study of Arntz, Weertman & Salet (2011) directly compared BPD and AVPD patients on the
interpretation of ambiguous mild to moderately negative vignettes. BPD patients evaluated themselves as more deprived and victimized, others as more malevolent and showed higher levels of criticizing others compared to AVPD patients and non-patients. Hence, BPD patients evaluated vignettes congruent with BPD specific schema-content. In sum, studies so far seem to show evidence for a schema-congruent interpretation bias in BPD patients within ambiguous negative situations.
An important limitation of previous studies is that they did not investigate interpretations of BPD patients in both ambiguous- and non-ambiguous situations differing in valence. The prior is important to further specify how maladaptive schemas in BPD distort interpretations of events. One line of thinking could be that maladaptive schemas lead to a general negativity bias. Thus, maladaptive schemas in BPD patients result in schema-congruent interpretations across situations irrespective of valence and ambiguity. However, a study of Arntz and Ten Haaf (2012) showed evidence against this general negativity bias hypothesis. Within this study BPD patients evaluated mental health trainees based on a short telephone conversation where the trainee acted in a clear positive, clear negative or neutral manner. Results showed that BPD patients only showed more negativity in the negative condition than Cluster-C patients and non-patients. However, an important limitation of this study is that the neutral condition was less neutral than intended and had a positive emotional effect. Therefore the findings within the neutral condition need to be interpreted with caution. Nonetheless, the lack of more negative interpretations within the neutral and positive condition seem to rule out the general negativity bias hypothesis.
One explanation for the findings in previous studies could be that the effect of schemas on interpretations is dependent on both the valence and ambiguity of the event. Since within ambiguous situations the valence of the event is less clear it seems plausible that individuals address pre-established beliefs about the self and others to evaluate the event. In contrast, non-ambiguous situations are more clear in valence and consequently individuals have less need to rely on schemas to interpret these situations. Hence, within ambiguous situations maladaptive schemas in BPD distort interpretations congruent with BPD schema-content. Furthermore, it seems plausible that maladaptive schemas distort interpretations when the content of the event is related to specific schema-content. Since maladaptive schemas in BPD are characterized by negative views of the self and others it could be that they distort interpretations within situations of negative valence. Thus, within ambiguous and non-ambiguous situations of negative valence maladaptive schemas are triggered by the specific content of the event and therefore result in BPD schema-congruent interpretations. The prior line of thinking would propose that BPD patients show more BPD schema-congruent interpretations than AVPD patients and non-patients in situations of negative valence and ambiguous situations. An alternative hypothesis could be that BPD schema-congruent interpretations are exclusively steered by either the valence or ambiguity of the event. In other words, BPD patients show schema-congruent interpretations in either ambiguous situations irrespective of valence or situations of negative valence irrespective of ambiguity (see table 1 for models of BPD interpretations based on valence/ambiguity of events). In sum, a research design where BPD
patients need to interpret situations differing in ambiguity and valence would yield valuable information about how maladaptive schemas distort interpretations in BPD.
Therefore the current study examined interpretational biases in BPD patients, AVPD and non-patients using ambiguous and non-ambiguous film clips differing in valence that participants interpreted from a first person perspective. Furthermore, the events occurring in the film clips covered two domains, namely the domain of social performances (job interviews) and intimate relationships (blind-dates). AVPD was chosen to differentiate between schema-congruent interpretations in BPD and AVPD. It was
expected that BPD patients evaluated themselves as more victimized and deprived and others as more malevolent in ambiguous situations and situations of negative valence compared to AVPD patients and non-patients. Lastly, it was expected that BPD specific beliefs fully mediate the relationship between BPD and BPD specific interpretations.
Table 1: Models of Type of Interpretations of BPD patients compared to AVPD patients and non-patients within situations differing in ambiguity and valence.
Situation (ambiguity/valence) Ambiguous Non Ambiguous Schema-Congruent
Interpretations steered by: Positive Mixed Negative Positive Neutral Negative
1. Valence & Ambiguity of
events BPD BPD BPD Healthy Healthy BPD
2. Valence of events Healthy BPD BPD Healthy Healthy BPD
3. Ambiguity of events BPD BPD BPD Healthy Healthy Healthy
4. Negativity Bias across
events BPD BPD BPD BPD BPD BPD
BPD = BPD patients respond with more BPD schema-congruent interpretations than AVPD and non-patients. Healthy = BPD patients respond with interpretations similar to non-patients.
Methods Participants
Participants within BPD (n =18) and AVPD (n=13) groups were recruited through multiple mental health care institutes specialized in PDs (“De Viersprong”, “Pro Persona” and “Arkin NPI”). Since BPD and AVPD patients are most often females and to rule out gender effects it was decided to only include females in this study, resulting in the most large and homogenous sample as feasible. Most BPD and AVPD patients included did not or only recently started treatment to make sure treatment had a minimal effect on outcomes. However, due to a strict timeframe of the study several BPD (n=2) and AVPD patients (n=3) were in the final phase of their treatment course. The non-patients group (n=16) was recruited from the general population through online advertisements. Sample sizes were calculated through power analyses based on effects found by a similar study of Arntz, Weertman & Salet (2011). This study found a large effect size (Cohen’s d = 2.7) when comparing BPD vs non-patients and a moderate effect size (Cohen’s d = 0.5) when BPD and AVPD patients were compared.
Participants were screened on Axis 1 and Axis 2 disorders with the Structured Clinical Interview for DSM-IV Axis 1 and 2 disorders SCID-I; First, Spitzer, Gibbon & Williams, 1996; Groenestijn van, Akkerhuis, Kupka, Schneider & Nolen, 1999; SCID-II; First, Gibbon, Williams & Benjamin, 1997; Weertman, Arntz & Kerkhof, 2000). To be included in the BPD or AVPD group, participants had to qualify for a primary BPD- or AVPD diagnosis and not suffer from bipolar disorder, dissociative identity disorders, psychotic disorders, clinical detoxification, attention-deficit/hyperactivity disorder and
immediate suicide risk. Other comorbid Axis-1 disorders and Axis-2 disorders/traits were allowed. To be included in the non-patients group, participants had to disqualify for any Axis-1 and Axis-2 disorder.
Table 2 presents demographics and diagnostics. Groups were similar on educational level but differed in age; AVPD patients were older than BPD patients and non-patients. BPD and AVPD groups did not differ in the number of present comorbid Axis-I disorders, Axis-II disorders, anxiety disorders, mood disorders, eating disorders and medication use. However, BPD patients showed more comorbid substance disorders than AVPD patients. There were no participants showing comorbid BPD and AVPD. Since AVPD patients were significantly older than BPD patients and non-patients and age showed a moderate effect, age was included as covariate within the analyses of the closed- and open responses. Table 2: Demographics and diagnostics of the three groups defined in descriptives (age) and frequencies (education level, axis-I and axis-II disorders and medication).
Demographics/Diagnostics Group
BPD ( n = 18) AVPD (n = 13) Non-patients ( n = 16)
Agea Mean = 26.89 (SD = 7.21) Mean = 34.29 (SD = 10.55) Mean = 26.25 (SD = 7.94) Educational Levelb Lower Level n = 3 (16.7%) n = 3 (23.1%) n = 3 (18.8%) Middle Level n = 8 (44.4%) n = 6 (46.2%) n = 3 (18.8%) High Level n = 7 (38.9%) n = 4 (30.8%) n =10 (62.5%) Axis-I disordersc Anxiety disorder d n = 11 (61.1%) n = 4 (30.8%) n = 0 (0%) Mood disorder e n = 6 (33.3%) n = 7 (53.8%) n = 0 (0%) Eating disorder f n = 7 (38.8%) n = 4 (30.8%) n = 0 (0%) Substance disorder g n = 5 (27.7%) n = 0 (0%) n = 0 (0%) Axis-II disordersh BPD n = 18 (100%) n = 0 (0%) n = 0 (0%) Avoidant PD n = 2 (11.1%) n = 10 (76.9%) n = 0 (0%) Dependent PD n = 0 (0%) n = 3 (23.1%) n = 0 (0%) OCPD n = 1 (5.5%) n = 1 (7.7%) n = 0 (0%) Medicationi Yes n = 15 (83.3%) n = 0 (0%) n = 0 (0%) No n = 3 (16.7%) n = 13 (100%) n = 16 (100%) a
Kruskal-Wallis χ2 [2] = 7.56, p = .023, r = 0.40. Pairwise comparisons showed that AVPD patients were significantly older than
BPD patients (U =61.00, z [29]= - 2.25, p = .025, r = -0.40) and nonpatients (U= 47.50, z [27]= - 2.48, p = .013, r = -0.56). b Kruskal-Wallis χ2 [2] = 2.04, p = .360, r = 0.21. c U =89.00, z [29]= -1.16, p = .247, r = -0.21. d U =95.00, z [29]= -1.02, p = .308, r = -0.18. e U = 93.00, z [29]= -1.12, p = .261, r = -0.20. f U = 107.50, z [29]= -0.46, p = .646, r = -0.08.
g U = 84.50, z [29]= -2.04, p = .041, r = -0.37. h U = 92.00, z [29]= -0.94, p = .311, r = -0.17. I U = 97.50, z [29]= -1.52, p = .128, r = -0.27. Materials Interpretation Task
The Interpretation Task (IT) consisted of 12 film clips (30-45 seconds) that participants needed to interpret from a first person perspective. Events occurring in the film clips covered two PD relevant domains, namely the domain of social performances (job interviews) and intimate relationships (blind- dates). Within each clip an actor (potential employer/blind-date partner) is talking into the camera (to the participant). Participants were asked to imagine that they were at a job interview or blind-date and that their potential employer or blind-date partner was directly talking to them. This to make it more vivid to participants that they were in a real life interaction. It is important to specify that it concerns a first person perspective and not an observer perspective (third person perspective). This because the central focus of cognitive therapies in clinical practice is that events are directly happening to the person him/herself. Each domain (job interviews and blind-dates) was represented by six clips that differed in valence and ambiguity (see table 1; distribution film clips on valence, ambiguity and domain). Half of the film clips concerned job interviews where a potential employer had something to say to the participant about the job vacancy in relationship with the participant’s qualifications. Participants were told they already applied for this job by writing a letter and sending their Curriculum Vitea (C.V.). The other half of the film clips concerned blind-dates where the blind date-partner had something to say to the participant. Participants were told they already interacted with the date-partner by email prior to this blind date.
After each scenario participants were asked to interpret these film clips within an open- and closed response format. Within the open response format, participants wrote down how they would evaluate themselves and the event/other based on the film clip. Participants were asked to write down the first things that came to mind. Within the closed response format, participants were asked to rate five possible interpretations of the scenario on how much this interpretation represented what they would think based on the presented film clip on a 100 mm Visual Analog Scales (VAS). The five interpretations consisted of a BPD specific interpretation, AVPD specific interpretation, OCPD specific interpretation, neutral interpretation and positive interpretation. In addition, at the end of the IT participants rated how vividly they could imagine that the event was happening directly to herself on a 100 mm Visual Analog Scales (VAS). This to assess whether participants were able to vividly interpret scenarios as a real-life interaction from a first person perspective. Film clips were presented in two different orders based on valence and ambiguity and randomized on type of domain. The order of the clips based on valence or ambiguity was presented as defined in table 3 and in reversed order. It was decided to not counterbalance the order of the film clips due to the larger sample size it would require. Instead the two fixed orders were equally often used in BPD group and non-patients group. Due to the uneven sample size this was not possible for the AVPD group.
Table 3: Distribution Film Clips on Valence, Ambiguity and Domain
Film Clip Valence Ambiguity Type of Domain
Nr 1: Ambiguous Positive scenario Positive Ambiguous Social Performance Nr 2: Neutral Scenario Neutral Ambiguous Social Performance Nr 3: Non-ambiguous Negative scenario Negative Non-ambiguous Social Performance Nr 4: Non-ambiguous Positive scenario Positive Non-ambiguous Social Performance Nr 5: Mixed Scenario (positive/negative) Mixed Ambiguous Social Performance Nr 6: Ambiguous Negative scenario Negative Ambiguous Social Performance Nr 7: Ambiguous Positive scenario Positive Ambiguous Intimate Relationships Nr 8: Neutral Scenario Neutral Ambiguous Intimate Relationships Nr 9: Non-ambiguous Negative scenario Negative Non-ambiguous Intimate Relationships Nr 10: Non-ambiguous Positive scenario Positive Non-ambiguous Intimate Relationships Nr 11: Mixed Scenario (positive/negative) Mixed Ambiguous Intimate Relationships Nr 12: Ambiguous Negative scenario Negative Ambiguous Intimate Relationships
Measures
SCID-I and SCID-II
The Dutch version of the Structured Clinical Interview for DSM-IV Axis-I disorders (SCID-I, First, Spitzer, & Williams, 1997; Dutch translation, Groenestijn et al., 1999) and the Dutch version of the Structured Clinical Interview for DSM-IV Axis-II disorders (SCID-II, First, Spitzer, Gibbon et al., 1997; Dutch translation, Weertman et al., 2000) were used to assess the presence of axis-I disorders and Axis-II disorders.
PDBQ
The Personality Disorder Belief Questionnaire (PDBQ; Arntz et al., 2004) subscale of BPD was used to assess BPD beliefs about the self, the world and others. The PDBQ shows good psychometric properties (Arntz et al., 2004).
Development of Stimulus Material
For the development, validation and example of the stimulus materials (scenarios, closed- and open responses) see appendix 1 and 2.
Procedure
Data was collected in a laboratory room at the University of Amsterdam. Participants were explained that this study investigates how individuals feel/think at job interviews/blind-dates and that they need to interpret 12 film clips from a first person perspective within an open- and closed response format. Participants were asked to write down the first thing that came to mind. In addition, participants were told that there were no correct or incorrect answers and that their interpretation of the scenario was
the best possible answer. The study was approved by the UvA’s ethical committee and participants gave informed consent. The task took approximately 1.5 hour.
Statistical Analyses
Closed responses/belief ratings
The closed responses belief ratings were analyzed with a Mixed between-within subjects ANCOVA with as between factor Group and as within factors Ambiguity and Valence, the 5 closed response interpretations (BPD specific, AVPD specific, OCPD specific, neutral and positive) as dependent variables and age as covariate. It was decided to not include comorbid Axis-I disorders and comorbid traits of other PDs as covariates since the non-patients group showed no comorbidity and therefore including Axis-I and Axis-II comorbidity as covariates would obscure group effects. Mixed ANCOVA’s were followed by simple contrasts (independent t-tests/Welch’s t-tests; based on assumption of homogeneity variances) and Bonferroni-Holmes post-hoc tests to correct for multiple testing. Within the Bonferroni-Holmes procedure p-values are sorted from smallest to largest. The smallest p-value is tested against an alpha-level of α/m, where m is the number of tests. The next smallest p-value is tested against α/(m-1) and so on. The procedure is stopped when a p-value is non-significant. Thus, all subsequent p-values are also treated as non-significant. This was done for each hypothesis and contrast separately. Since omnibus F-tests are not optimal to test hypotheses aimed at specific group comparisons, contrasts (BPD vs. non-patients and BPD vs. AVPD) related to the hypotheses were defined a priori. Thus, contrasts related to the hypotheses were examined even when the omnibus F-test failed to reach significance and significant contrasts are reported. Furthermore, since this study aims to analyze group specific effects, only effects including Group were further examined. In addition, since PDs showed higher belief ratings in general than non-patients (see appendix 3) and this study aims to examine PD specific effects, also corrected belief ratings were calculated. This was done by subtracting each individual’s mean belief rating over all scenarios from the individual’s rating of each closed response interpretation within a scenario. Due to the large number of analyses within this study only corrected belief ratings are discussed within the result section. Results of the uncorrected belief ratings are presented within appendix 4.
Open responses
To analyse open responses two master thesis students, who were blind for group, collaboratively created themes covering the open responses based on 6 randomly selected participants of each group. Both raters classified these themes by mutual agreement into 28 categories. Next, both raters rated the open responses according to these 28 categories. Within each scenario each category was coded as 0 (no) or 1 (yes). To test the test-retest correlations of the scoring system Cohen’s Kappa’s were calculated for 20 randomly selected participants. Cohen’s Kappa’s were satisfying (ranging from 0.469 to 0.811 with a mean of 0.612, median of 0.644 and standard deviation of 0.11). Subsequently, the 28 categories were combined to create 10 high-order categories based on BPD, AVPD and healthy schema content representing
self-positive views (e.g. capable, adequate, desirable), self-negative views (e.g. inadequate, undesirable, inferior), other positive views (e.g. sympathetic, trustable, accepting, admiring), others being powerful views (e.g. capable, desirable, confident), other negative views in relation to self (e.g. not interested, keeping distance, critical, overpowering, rejecting), other malevolent views (e.g. not to be trusted, manipulating, arrogant, abusing) and coping strategies as optimism, pessimism, avoidance and being critical. These 10 high-order categories formed the dependent variables for the open responses.
Since the dependent variables were count variables and the distributions of the dependent variables were positively skewed Generalized Linear Mixed Models (GLMMs) were used to analyze the open responses. Within the GLMMs, Group, Ambiguity, Valence and Age were included as fixed factors. It was decided to not include comorbid Axis-I disorders and comorbid traits of other PDs as covariates since the non-patients group showed no comorbidity and therefore including Axis-I and Axis-II comorbidity as covariates would obscure group effects. To test the hypotheses contrasts were defined a priori (BPD vs. non-patients and BPD vs. AVPD) and significant contrasts related to the hypotheses were examined and reported even when the omnibus F-test failed to reach significance. Furthermore, since this study aims to analyze group specific effects, only effects including Group were further examined. To control for multiple testing the Bonferroni-Holmes procedure was used. In addition, since most dependent variables were count variables with many observations having value 0 and variances often larger than means a negative binominal distribution with a log-link function was applied to the GLMMs. However, several dependent variables showed only observations with the value 0 or 1 and therefore a binary logistic regression GLMM with a logit-link fuction was used to analyse these variables. Lastly, based on the study design it was decided to compare the fit of GLMMs with a compound symmetry covariance structure with the fit of GLMMs with a scaled identity covariance structure or diagonal covariance structure on the data. Comparing Akaike’s Information Criteria corrected (AICc) of the GLMMs
demonstrated that the covariance structure with a significant superior fit was dependent on the dependent variable of interest. Thus, measurements of some dependent variables in the GLMMs showed
heterogeneous variances and zero correlations and measurements of other dependent variables showed homogenous variances and correlations or homogenous variances and zero correlations. Therefore it was decided to apply the GLMM with a significant superior fit to the specific data of interest. In addition, when no significant difference in model fit was observed, the most parsimonious model was preferred (GLMM with a compound symmetry covariance structure).
Mediation analyses
To test whether the relationship between BPD and BPD schema-congruent interpretations is mediated by BPD specific beliefs a Preacher-Hayes Mediation Bootstrap Method (Preacher & Hayes, 2008) was used. Due to the different scaling of the closed and open response variables (100mm VAS scale vs count variables) the mediation analysis was run separately for the closed and open responses. The closed response BPD interpretation variable was based on the mean of the corrected BPD belief ratings across all scenarios. For the open responses the BPD schema-congruent interpretation variable was based
on the total score on Other Malevolent Views across all scenarios. The mediating variable is based on BPD specific beliefs as measured by the PDBQ.
Manipulation check
As manipulation check, participants were asked how vividly they could imagine that they were at a real-life job interview or blind-date on a 100 mm Visual Analog Scales (VAS). This to assess whether participants were able to vividly interpret the scenario as a real-life interaction from a first person perspective. It was decided to exclude participants with a score lower than 40. Results showed no participants needed to be excluded (see appendix 3). As an additional manipulation check of the ambiguity/valence of the film clips, uncorrected and corrected Positive belief ratings on job-interviews and blind-dates were examined. Results confirm the expectations about the valence/ambiguity of the film clips (see appendix 3). In addition, a Mixed ANOVA with Type of Order of the film clips as between subject variable, All Film Clips as within subject variable and uncorrected BPD belief ratings as dependent variable showed no interaction between Type of Order of the film clips and All Film Clips on uncorrected BPD belief ratings, F(1,33) = 1.23, p = .160.
Results
Closed responses
Job interviews; Corrected BPD belief ratings
Table 4. Main- and interaction effects regarding corrected BPD belief ratings on job interviews.
Type of effect F (df) P-value
Group 14.77 (2,43) < .001 Ambiguity 0.05 (1,43) = .829 Valence 8.14 (2,86) = .001 Group x Ambiguity 5.99 (2,43) = .005 Group x Valence 1.64 (4,86) = .172 Ambiguity x Valence 19.78 (2,86) < .001
Group x Ambiguity x Valence 8.99 (4,86) < .001
*Note significant effects are in bold.
Since the higher-order interaction Group x Ambiguity x Valence was significant and theoretically expected no lower-order omnibus effects were further examined. However, the significant Group x Ambiguity and Group x Ambiguity x Valence interactions suggested that the magnitude and/or direction of the Group x Ambiguity interaction was dependent on Valence. Therefore the Group x Ambiguity x Valence interaction was further examined and three plots are presented of the Group x Ambiguity x Valence interaction for each level of Valence.
Table 5. Between Group Comparisons: Group x Ambiguity x Valence regarding corrected BPD belief ratings on job interviews.
Group x Ambiguity x Valence Comparisons Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs AVPD -24.53 (5.5) vs -25.87 (6.8) 0.15 = .881 0.06 AVPD vs non-patients -25.87 (6.8) vs -15.06 (5.8) -1.21 = .236 -0.47 Ambiguous Positive BPD vs non-patients 20.88 (5.0) vs -13.61 (5.3) 4.77 < .001 1.69 BPD vs AVPD 20.88 (5.0) vs 26.69 (6.2) -0.74 = .465 -0.27 AVPD vs non-patients 26.69 (6.2) vs -13.61 (5.3) 4.99 < .001 1.92 Neutral BPD vs non-patients 24.48 (5.9) vs -9.73 (6.3) 3.97 < .001 1.40 BPD vs AVPD 24.48 (5.9) vs 36.99 (7.4) -1.34 = .191 -0.50 AVPD vs non-patients 36.99 (7.4) vs -9.73 (6.3) 4.85 < .001 1.87 Mixed BPD vs non-patients 8.49 (5.5) vs -21.50 (5.9) 3.70 < .001 1.31 BPD vs AVPD 8.49 (5.5) vs -19.06 (6.9) 3.11 = .005 1.16 AVPD vs non-patients -19.06 (6.9) vs -21.50 (5.9) 0.26 = .789 0.10 Ambiguous Negative BPD vs non-patients 18.93 (5.8) vs -20.12 (6.2) 4.57 < .001 1.62 BPD vs AVPD 18.93 (5.8) vs 18.02 (7.3) 0.10 = .922 0.04 AVPD vs non-patients 18.02 (7.3) vs -20.12 (6.2) 4.00 < .001 1.54 Non-Ambiguous Negative BPD vs non-patients 31.29 (6.6) vs 22.06 (7.0) 0.96 = .346 0.34 BPD vs AVPD 31.29 (6.6) vs 25.60 (8.2) 0.55 = .590 0.20 AVPD vs non-patients 25.60 (8.2) vs 22.06 (7.0) 0.33 = .745 0.13 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 1. Group x Ambiguity x Valence (level 1; positive) regarding corrected BPD belief ratings on job interviews.
Figure 2. Group x Ambiguity x Valence (level 2; neutral/mixed) regarding corrected BPD belief ratings on job interviews.
Figure 3. Group x Ambiguity x Valence (level 3; negative) regarding corrected BPD belief ratings on job interviews. -30 -20 -10 0 10 20 30 Non-Ambiguous
Positive Ambiguous Positive
BPD AVPD Non-patients -40 -20 0 20 40 Neutral Mixed BPD AVPD Non-patients
Between group comparisons showed that BPD patients gave higher corrected BPD belief ratings than non-patients on the ambiguous positive, ambiguous negative, neutral and mixed job interview. BPD patients and non-patients showed no differences in corrected BPD belief ratings on the non-ambiguous positive and non-ambiguous negative job interview. In addition, AVPD patients gave higher corrected BPD belief ratings than non-patients on the ambiguous positive, ambiguous negative and neutral job interview. AVPD patients and non-patients showed no differences in corrected BPD belief ratings on the non-ambiguous positive, non-ambiguous negative and mixed job interview. In contrast, BPD and AVPD patients showed similar corrected BPD belief ratings on all job interviews except for the mixed job interview where BPD patients gave higher corrected BPD belief ratings. In addition, mean differences between BPD and AVPD patients on other job interviews were in both directions.
The three plots show a clear trend between groups. On the ambiguous negative and non-ambiguous positive job interview, BPD patients, AVPD patients and non-patients showed no differences in corrected BPD belief ratings. In contrast, on the ambiguous negative, neutral and ambiguous positive job interview, BPD and AVPD patients showed higher corrected BPD belief ratings than non-patients. In addition, on the mixed job interview BPD patients showed higher corrected BPD belief ratings than AVPD patients and non-patients whereas AVPD patients and non-patients showed similar corrected BPD belief ratings. In contrast, AVPD patients showed relatively high corrected BPD belief ratings compared to BPD patients on the neutral job interview.
Job interviews; Corrected AVPD belief ratings
Table 6. Main- and interaction effects regarding corrected AVPD belief ratings on job interviews.
Type of effect F (df) P-value
Group 26.75 (2,43) < .001 Ambiguity 4.96 (1,43) = .031 Valence 2.71 (2,86) = .072 Group x Ambiguity 0.36 (2,43) = .697 Group x Valence 0.57 (4,86) = .687 Ambiguity x Valence 0.90 (2,86) = .412
Group x Ambiguity x Valence 1.89 (4,86) = .120 *Note significant effects are in bold.
-30 -20 -10 0 10 20 30 40 Non-Ambiguous
Negative Ambiguous Negative
BPD AVPD Non-patients
Results showed a main effect of Group and Ambiguity. Since this study aims to analyse group effects, only the main effect of Group was further examined.
Table 7. Main effect of Group regarding corrected AVPD belief ratings
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients -5.11 (3.2) vs -22.51 (3.4) 3.77 < .001 1.33 BPD vs AVPD -5.11 (3.2) vs 16.55 (3.9) -4.34 < .001 -1.61 AVPD vs non-patients 16.55 (3.9) vs -22.51 (3.4) 7.58 < .001 2.92 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
The main effect of Group showed that AVPD patients gave higher corrected AVPD belief ratings than BPD patients and non-patients on job interviews. In addition, BPD patients gave higher fcorrected AVPD belief ratings than non-patients on job interviews.
Job Interviews; Corrected OCPD Belief ratings
Table 8. Main- and interaction effects regarding corrected OCPD belief ratings on job interviews.
Type of effect F (df) P-value
Group 1.39 (2,43) = .259 Ambiguity 5.68 (1,43) = .022 Valence 1.22 (1.749,75.214) (GG) = .298 Group x Ambiguity 3.28 (2,43) = .047 Group x Valence 4.10 (3.498, 75.214) (GG) = .007 Ambiguity x Valence 3.49 (2,86) = .035
Group x Ambiguity x Valence 1.01 (4,86) = .406 *Note significant effects are in bold.
Results showed a main effect of Ambiguity and interaction effects of Group x Ambiguity, Group x Valence and Ambiguity x Valence. Since this study aims to analyze Group effects, only the interactions including Group (Group x Ambiguity and Group x Valence) were further examined.
Table 9. Between Group Comparisons: Group x Ambiguity regarding corrected OCPD belief ratings.
Ambiguity Group Comparison Adjusted Mean (S.E.) t P-value Cohen’s d
Non-ambiguous BPD vs Non-patients 2.00 (3.2) vs 0.86 (3.5) 0.24 = .810 0.08 BPD vs AVPD 2.00 (3.2) vs 12.29 (4.0) -2.01 = .054 -0.75 AVPD vs non-patients 12.29 (4.0) vs 0.86 (3.5) 2.16 = .0391 0.83
Ambiguous BPD vs Non-patients 4.39 (2.8) vs -2.71 (3.0) 1.73 = .092 0.61
BPD vs AVPD 4.39 (2.8) vs -3.84 (3.5) 1.86 = .073 0.69 AVPD vs non-patients -3.84 (3.5) vs -2.71 (3.0) -0.25 = .806 -0.10 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 4. Group x Ambiguity regarding corrected OCPD belief ratings on job interviews.
Between group comparisons showed no significant differences in corrected OCPD belief ratings on ambiguous and ambiguous job interviews between BPD patients, AVPD patients and
non-patients. However, not a clear trend between groups. AVPD patients gave relatively high corrected OCPD belief ratings on non-ambiguous job interviews compared to BPD patients and non-patients. In contrast, BPD patients gave relatively high corrected OCPD belief ratings on ambiguous job interviews compared to AVPD patients and non-patients.
Table 10. Between Group Comparisons: Group x Valence regarding corrected OCPD belief ratings on job interviews.
Valence Group Comparison Adjusted Mean (S.E.) t P-value Cohen’s d
Positive BPD vs Non-patients 11.25 (4.4) vs 24.33 (4.7) -2.03 = .051 -0.72 BPD vs AVPD 11.25 (4.4) vs 11.32 (5.5) -0.00 = .993 0.00 AVPD vs Non-patients 11.32 (5.5) vs 24.33 (4.7) -1.80 = .082 -0.69 Neutral/Mixed BPD vs Non-patients 3.17 (4.0) vs -14.24 (4.3) 2.97 = .006 1.05 BPD vs AVPD 3.17 (4.0) vs 6.45 (5.0) -0.52 = .609 -0.19 AVPD vs Non-patients 6.45 (5.0) vs -14.24 (4.3) -3.16 = .004 -1.22 Negative BPD vs Non-patients -4.82 (3.7) vs -12.85 (4.0) 1.48 = .150 0.52 BPD vs AVPD -4.82 (3.7) vs -5.10 (4.6) 0.05 = .963 0.02 AVPD vs Non-patients -5.10 (4.6 ) vs 12.85 (4.0) -1.28 = .213 -0.49 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 5. Group x Valence regarding corrected OCPD belief ratings on job interviews. -5 0 5 10 15 Non-ambiguous Ambiguous BPD AVPD Non-patients
Between group comparisons showed that BPD and AVPD patients gave higher corrected OCPD belief ratings than non-patients on job interviews of neutral/mixed valence. On job interviews of positive and negative valence no significant differences in corrected OCPD belief ratings between groups were found.
Note a clear trend between groups. On job interviews of neutral/mixed valence, BPD patients and AVPD patients gave relatively high corrected OCPD belief ratings compared to non-patients whereas on job interviews of positive valence non-patients gave relatively high corrected OCPD belief ratings compared to BPD and AVPD patients. On job interviews of negative valence no clear trend was observed.
Job Interviews; Corrected Neutral Belief ratings
Table 11. Main- and interaction effects regarding corrected Neutral belief ratings on job interviews.
Type of effect F (df) P-value
Group 8.01 (2,43) = .001 Ambiguity 1.98 (1,43) = .167 Valence 1.52 (2,86) = .225 Group x Ambiguity 0.92 (2,43) = .405 Group x Valence 3.26 (4,86) = .015 Ambiguity x Valence 3.84 (2,86) = .025
Group x Ambiguity x Valence 0.98 (4,86) = .424 *Note significant effects are in bold.
Results showed a main effect of Group and interaction effects of Group x Valence and Ambiguity x Valence. Since this study aims to analyse group effects, only effects including Group were further examined. Furthermore, since Group was involved in the interaction effect of Group x Valence, the main effect of Group was qualified and therefore only the interaction of Group x Valence was further examined.
Table 12. Between Group Comparisons: Group x Valence regarding corrected Neutral belief ratings on job interviews.
Valence Group Comparison Adjusted Mean (S.E.) t P-value Cohen’s d
Positive BPD vs Non-patients 17.31 (5.1) vs 16.16 (5.5) 0.16 = .878 0.06 -20 -10 0 10 20 30
Positive Neutral/Mixed Negative
BPD AVPD Non-patients
BPD vs AVPD 17.31 (5.1) vs 6.60 (6.4) 1.33 = .194 0.49 AVPD vs Non-patients 6.60 (6.4) vs 16.16 (5.5) -1.15 = .260 -0.44 Neutral/Mixed BPD vs Non-patients 15.73 (4.5) vs 32.45 (4.8) -2.55 = .016 -0.90 BPD vs AVPD 15.73 (4.5) vs -6.22 (5.6) 3.09 = .004 1.13 AVPD vs Non-patients -6.22 (5.6) vs 32.45 (4.8) -5.28 < .001 -2.03 Negative BPD vs Non-patients 7.86 (4.5) vs 23.55 (4.8) -2.38 = .0242 -0.84 BPD vs AVPD 7.86 (4.5) vs 10.94 (5.6) -0.43 = .670 -0.16 AVPD vs Non-patients 10.94 (5.6) vs 23.55 (4.8) -1.71 = .099 -0.66 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 6. Group x Valence regarding corrected Neutral belief ratings on job interviews.
Between group comparisons showed that non-patients gave higher corrected Neutral belief ratings than BPD and AVPD patients on job interviews of neutral and mixed valence. In addition, BPD patients gave higher corrected Neutral belief ratings than AVPD patients on the job interviews of neutral and mixed valence. On job interviews of positive and negative valence, BPD patients, AVPD patients and non-patients showed similar corrected Neutral belief ratings.
The plot shows a trend between groups. Non-patients patients gave relatively high corrected Neutral belief ratings on job interviews neutral/mixed and negative valence compared to BPD and AVPD patients. AVPD patients gave relatively low corrected Neutral belief ratings across job interviews of positive, negative and neutral/mixed valence.
Job Interviews; Corrected Positive Belief ratings
Table 13. Main- and interaction effects regarding corrected Positive belief ratings on job interviews.
Type of effect F (df) P-value
Group 26.61 (2,43) < .001 Ambiguity 3.05 (1,43) = .088 Valence 7.83 (2,86) = .001 Group x Ambiguity 5.51 (2,43) = .007 Group x Valence 2.34 (4,86) = .062 Ambiguity x Valence 5.06 (2,86) = .008
Group x Ambiguity x Valence 1.61 (4,86) = .180
2 Non-significant against a Bonferroni-Holmes corrected alpha level of 0.017
-10 0 10 20 30 40
Positive Neutral/Mixed Negative
BPD AVPD Non-patients
*Note significant effects are in bold.
Results showed main effects of Group and Valence and interaction effects of Group x Ambiguity and Ambiguity x Valence. Since this study aims to analyse group effects, only effects including Group were further examined. Furthermore, since Group was involved in the interaction effect of Group x Ambiguity, the main effect of Group was qualified and therefore only the interaction of Group x Ambiguity was further examined.
Table 14. Between Group Comparisons: Group x Ambiguity regarding corrected Positive belief ratings.
Ambiguity Group Comparison Adjusted Mean (S.E.) t P-value Cohen’s d
Non-ambiguous BPD vs Non-patients -5.54 (3.2) vs 5.37 (3.4) -2.35 = .025 -0.83 BPD vs AVPD -5.54 (3.2) vs -14.59 (4.0) 1.80 = .082 0.67 AVPD vs non-patients -14.59 (4.0) vs 5.37 (3.4) -3.85 < .001 -1.48 Ambiguous BPD vs Non-patients 3.53 (3.4) vs 29.26 (3.7) -5.13 < .001 -1.81 BPD vs AVPD 3.53 (3.4) vs -12.44 (4.3) 2.94 = .006 1.09 AVPD vs non-patients -12.44 (4.3) vs 29.26 (3.7) -7.44 < .001 -2.86 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 7. Group x Ambiguity regarding corrected Positive belief ratings on job interviews.
Between group comparisons showed non-patients gave higher corrected Positive belief ratings than BPD and AVPD patients on ambiguous job interviews. In addition, BPD patients gave higher corrected Positive belief ratings than AVPD patients on ambiguous job interviews. In contrast, non-patients gave higher corrected Positive belief ratings than BPD and AVPD non-patients on non-ambiguous job interviews, whereas BPD and AVPD patients showed similar corrected Positive belief ratings on non-ambiguous job interviews.
The plot shows a clear trend, non-patients gave higher corrected Positive belief ratings than BPD and AVPD patients on non-ambiguous and ambiguous job interviews. The difference in corrected Positive belief ratings between non-patients and the experimental groups was the largest in ambiguous situations.
Blind-dates; Corrected BPD Belief ratings
Table 15. Main- and interaction effects regarding corrected BPD belief ratings on blind-dates.
Type of effect F (df) P-value
-20 -10 0 10 20 30 40 Non-ambiguous Ambiguous BPD AVPD Non-patients
Group 4.70 (2,43) = .014 Ambiguity 0.77 (1,43) = .385 Valence 0.44 (2,86) = .645 Group x Ambiguity 1.55 (2,43) = .223 Group x Valence 3.12 (4,86) = .019 Ambiguity x Valence 3.90 (2,86) = .024
Group x Ambiguity x Valence 3.12 (4,86) = .019
*Note significant effects are in bold.
Since the higher-order interaction Group x Ambiguity x Valence was significant and theoretically expected no lower-order omnibus effects were further examined. However, the significant Group x Valence and Group x Ambiguity x Valence interactions suggested that the magnitude and/or direction of the Group x Valence interaction was dependent on Ambiguity. Therefore the Group x Ambiguity x Valence interaction was further examined and two plots are presented of the Group x Ambiguity x Valence interaction for each level of Ambiguity.
Table 16. Between Group Comparisons: Group x Ambiguity x Valence regarding corrected BPD belief ratings on blind-dates.
Group x Ambiguity x Valence Comparisons Adjusted Mean (S.E.) t P-value Cohen’s d
Non-Ambiguous Positive BPD vs non-patients -14.44 (5.0) vs -28.46 (5.3) 1.92 = .068 0.69 BPD vs AVPD -14.44 (5.0) vs 1.09 (6.2) -1.97 = .059 -0.73 AVPD vs non-patients 1.09 (6.2) vs -28.46 (5.3) 3.61 = .001 1.39 Ambiguous Positive BPD vs non-patients 2.56 (6.7) vs -25.09 (7.2) 2.81 = .008 0.99 BPD vs AVPD 2.56 (6.7) vs -5.04 (8.4) 0.71 = .481 0.26 AVPD vs non-patients 5.04 (8.4) vs -25.09 (7.2) 1.82 = .081 0.70 Neutral BPD vs non-patients 31.06 (7.0) vs 18.68 (7.5) 1.21 = .236 0.43 BPD vs AVPD 31.06 (7.0) vs 8.84 (8.7) 2.00 = .055 0.74 AVPD vs non-patients 8.84 (8.7) vs 18.68 (7.5) -0.86 = .397 -0.33 Mixed BPD vs non-patients 7.90 (7.1) vs -8.47 (7.6) 1.57 = .126 0.56 BPD vs AVPD 7.90 (7.1) vs 26.71 (8.9) -1.65 = .111 -0.61 AVPD vs non-patients 26.71 (8.9) vs -8.47 (7.6) 3.03 = .005 1.17 Ambiguous Negative BPD vs non-patients 34.62 (6.2) vs 11.84 (6.6) 2.52 = .0173 0.89
BPD vs AVPD 34.62 (6.2) vs 12.02 (7.7) 2.31 = .0284 0.86
AVPD vs non-patients 12.02 (7.7) vs 11.84 (6.6) 0.01 = .986 0.00 Non-Ambiguous Negative BPD vs non-patients 37.42 (6.9) vs 29.61 (7.3) 0.78 = .442 0.28 BPD vs AVPD 37.42 (6.9) vs 21.13 (8.6) 1.50 = .145 0.56 AVPD vs non-patients 21.13 (8.6) vs 29.61 (7.3) -0.76 = .456 -0.29 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
3
Non-significant against a Bonferroni-Holmes corrected alpha level of 0.01
Figure 8. Group x Ambiguity x Valence (level 1; non-ambiguous) regarding corrected BPD belief ratings on blind-dates.
Figure 9. Group x Ambiguity x Valence (level 2; ambiguous) regarding corrected BPD belief ratings on blind-dates.
Between group comparisons showed that BPD patients gave higher corrected BPD belief ratings than non-patients on the ambiguous positive blind-date. On other blind-dates no significant differences in corrected BPD belief ratings between BPD patients and non-patients were found. However, large mean differences were in favour of BPD patients on all blind-dates except for the non-ambiguous negative blind-date. In addition, AVPD patients gave higher corrected BPD belief ratings than non-patients on the non-ambiguous positive and mixed blind-date. On other blind-dates no significant differences in
corrected BPD belief ratings between AVPD patients and non-patients were found. In contrast, BPD and AVPD patients showed similar corrected BPD belief ratings across all blind-dates. However, note mean differences in corrected BPD belief ratings between BPD and AVPD patients were in both directions.
The two plots show a clear trend between groups. BPD patients gave relatively high corrected BPD belief ratings compared to non-patients on the ambiguous negative and ambiguous positive blind-date. In addition, BPD patients gave relatively high corrected BPD belief ratings compared to AVPD patients on blind-dates of negative valence and the neutral blind-date. In contrast, AVPD patients gave relatively high corrected BPD belief ratings compared to BPD patients and patients on the non-ambiguous positive and mixed blind-date.
Blind-dates; Corrected AVPD Belief ratings
Table 17. Main- and interaction effects regarding corrected AVPD belief ratings on blind-dates.
Type of effect F (df) P-value
-40 -30 -20 -10 0 10 20 30 40 50 Non-Ambiguous
Positive Neutral Non-AmbiguousNegative
BPD AVPD Non-patients -30 -20 -10 0 10 20 30 40
Ambiguous Positive Mixed Ambiguous Negative
BPD AVPD Non-patients
Group 7.55 (2,43) = .002 Ambiguity 0.42 (1,43) = .520 Valence 2.54 (2,86) = .085 Group x Ambiguity 0.10 (2,43) = .910 Group x Valence 1.37 (4,86) = .250 Ambiguity x Valence 2.45 (2,86) = .093
Group x Ambiguity x Valence 8.52 (4,86) < .001
*Note significant effects are in bold.
Results showed a main effect of Group and an interaction effect of Group x Ambiguity x Valence. Since Group was involved in the interaction effect of Group x Ambiguity x Valence, the main effect of Group was qualified and therefore only the interaction of Group x Ambiguity x Valence was further examined.
Table 18. Between Group Comparisons: Group x Ambiguity x Valence regarding corrected AVPD belief ratings on blind-dates.
Group x Ambiguity x Valence Comparisons Adjusted Mean (S.E.) t P-value Cohen’s d
Non-Ambiguous Positive BPD vs non-patients -17.29 (5.6) vs -26.74 (5.9) 1.16 = .254 0.41 BPD vs AVPD -17.29 (5.6) vs 5.07 (6.9) -2.54 = .0175 -0.94
AVPD vs non-patients 5.07 (6.9) vs -26.74 (5.9) 3.51 = .002 1.35 Ambiguous Positive BPD vs non-patients -24.63 (4.3) vs -20.23 (4.6) -0.70 = .490 -0.25 BPD vs AVPD -24.63 (4.3) vs -33.69 (5.4) 1.33 = .195 0.49 AVPD vs non-patients -33.69 (5.4) vs -20.23 (4.6) -1.91 = .066 -0.74 Neutral BPD vs non-patients -15.52 (4.5) vs -22.66 (4.8) 1.08 = .290 0.38 BPD vs AVPD -15.52 (4.5) vs -20.92 (5.6) 0.75 = .458 0.29 AVPD vs non-patients -20.92 (5.6) vs -22.66 (4.8) 0.23 = .816 0.09 Mixed BPD vs non-patients -11.39 (5.6) vs -19.90 (6.0) 1.04 = .306 0.37 BPD vs AVPD -11.39 (5.6) vs 18.49 (7.0) -3.38 = .002 -1.26 AVPD vs non-patients 18.49 (7.0) vs -19.90 (6.0) 4.21 < .001 1.62 Ambiguous Negative BPD vs non-patients -10.31 (5.2) vs -22.45 (5.5) 1.61 = .117 0.57 BPD vs AVPD -10.31 (5.2) vs 6.67 (6.4) -2.08 = .0476 -0.77
AVPD vs non-patients 6.67 (6.4) vs -22.45 (5.5) 3.46 = .002 1.33 Non-Ambiguous Negative BPD vs non-patients -25.25 (6.7) vs -17.24 (7.1) -0.82 = .421 -0.29 BPD vs AVPD -25.25 (6.7) vs 3.71 (8.4) -2.70 = .0127 -1.00
AVPD vs non-patients 3.71 (8.4) vs -17.24 (7.1) 1.90 = .069 0.73 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 10. Group x Ambiguity x Valence (level 1; non-ambiguous) regarding corrected AVPD belief ratings on blind-dates.
5
Non-significant against a Bonferroni-Holmes corrected alpha level of 0.01
6
Non-significant against a Bonferroni-Holmes corrected alpha level of 0.01
Figure 11. Group x Ambiguity x Valence (level 2; ambiguous) regarding corrected AVPD belief ratings on blind-dates.
Between group comparisons showed that AVPD patients gave higher corrected AVPD belief ratings than non-patients on the non-ambiguous positive, mixed and ambiguous negative blind-date. In addition, AVPD patients showed higher corrected AVPD belief ratings than BPD patients on the mixed blind-date. In contrast, BPD patients and AVPD patients showed similar corrected AVPD belief ratings on other blind-dates, however on the ambiguous negative, non-ambiguous negative and non-ambiguous positive blind-date large mean differences were in favour of AVPD. BPD patients and non-patients showed no differences in corrected AVPD belief ratings on blind-dates.
The two plots show a clear trend between groups. AVPD patients gave relatively high corrected AVPD belief ratings compared to BPD patients and non-patients on the non-ambiguous positive and non-ambiguous negative blind-date. No differences in corrected AVPD belief ratings between groups were found on the neutral blind-date. In contrast, AVPD patients gave relatively high corrected AVPD belief ratings compared to BPD patients and non-patients on the mixed and ambiguous negative blind-date. No differences in corrected AVPD belief ratings between groups were found on the ambiguous positive blind-date. Note that for AVPD patients, the effect of Valence on corrected AVPD belief ratings was dependent on Ambiguity.
Blind-dates; Corrected OCPD Belief ratings
Table 19. Main- and interaction effects regarding corrected Neutral belief ratings on blind-dates.
Type of effect F (df) P-value
Group 3.35 (2,43) = .045 Ambiguity 0.17 (1,43) = .679 -30 -25 -20 -15 -10 -5 0 5 10 Non-Ambiguous
Positive Neutral Non-AmbiguousNegative BPD AVPD Non-patients -40 -30 -20 -10 0 10 20 30
Ambiguous Positive Mixed Ambiguous Negative
BPD AVPD Non-patients
Valence 0.11 (2,86) = .898
Group x Ambiguity 0.81 (2,43) = .454
Group x Valence 1.96 (4,86) = .108
Ambiguity x Valence 3.44 (2,86) = .037
Group x Ambiguity x Valence 1.02 (4,86) = .401 *Note significant effects are in bold.
Results showed a main effect of Group and an interaction effect of Ambiguity x Valence. Since this study aims to analyse group effects, only the effect of Group was further examined.
Table 20. Main effect of Group regarding corrected OCPD belief ratings on blind-dates.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients -11.86 (2.6) vs -5.71 (2.8) -1.63 = .113 -0.58 BPD vs AVPD -11.86 (2.6) vs -16.77 (3.2) 1.20 = .239 0.45 AVPD vs non-patients -16.77 (3.2) vs -5.71 (2.8) -2.62 = .014 -1.00 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that AVPD patients gave higher corrected OCPD belief ratings than non-patients and similar corrected OCPD belief ratings as BPD patients on blind-dates. In addition, BPD patients and non-patients gave similar corrected OCPD belief ratings on blind-dates.
Blind-dates; Corrected Neutral Belief ratings
Table 21. Main- and interaction effects regarding corrected Neutral belief ratings on blind-dates.
Type of effect F (df) P-value
Group 10.76 (2,43) < .001 Ambiguity 0.76 (1,43) = .389 Valence 1.35 (2,86) = .266 Group x Ambiguity 0.04 (2,43) = .965 Group x Valence 0.93 (4,86) = .454 Ambiguity x Valence 4.44 (2,86) = .015
Group x Ambiguity x Valence 1.83 (4,86) = .131 *Note significant effects are in bold.
Results showed a main effect of Group and an interaction effect of Ambiguity x Valence. Since this study aims to analyse group effects, only the effect of Group was further examined.
Table 22. Main effect of Group regarding corrected Neutral belief ratings on blind-dates.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients -2.13 (2.5) vs 12.34 (2.6) -4.05 < .001 -1.43 BPD vs AVPD -2.13 (2.5) vs -3.50 (3.1) 0.35 = .726 0.13 AVPD vs non-patients -3.50 (3.1) vs 12.34 (2.6) -3.97 < .001 -1.53 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that non-patients gave higher corrected Neutral belief ratings than BPD and AVPD patients. In addition, BPD patients and AVPD patients gave similar corrected Neutral belief ratings on blind-dates.
Blind-dates; Corrected Positive belief ratings
Table 23. Main- and interaction effects regarding corrected Positive belief ratings on blind-dates.
Type of effect F (df) P-value
Group 11.77 (2,43) < .001 Ambiguity 0.81 (1,43) = .375 Valence 4.61 (2,86) = .013 Group x Ambiguity 1.60 (2,43) = .213 Group x Valence 0.28 (4,86) = .892 Ambiguity x Valence 6.02 (1.500,64.496) (GG) = .008
Group x Ambiguity x Valence 1.83 (3.000,64.496)(GG) = .761 *Note significant effects are in bold.
Results showed main effects of Group and Valence and an interaction effect of Ambiguity x Valence. Since this study aims to analyse group effects, only the effect of Group was further examined.
Table 24. Main effect of Group regarding corrected Positive belief ratings on blind-dates.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients -9.11 (2.5) vs 6.95 (2.6) -4.44 < .001 -1.57 BPD vs AVPD -9.11 (2.5) vs -8.51 (3.1) 0.15 = .879 0.06 AVPD vs non-patients -8.51 (3.1) vs 6.95 (2.6) -3.83 < .001 -1.47 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that non-patients gave higher corrected Positive belief ratings than BPD and AVPD patients. In addition, BPD patients and AVPD patients gave similar corrected Neutral belief ratings on blind-dates.
Open responses
Job interviews; Self-Positive Views
Table 25. Main- and interaction effects regarding Self-Positive Views on job interviews.
Type of effect F (df) P-value
Group 16.17 (2,57) .000 Ambiguity 8.85 (1,264) .003 Valence 6.58 (2,252) .002 Group x Ambiguity 0.91 (2,243) .404 Group x Valence 0.43 (4,205) .785 Ambiguity x Valence 3.53 (2,245) .031
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and compound symmetry covariance structure.
Results showed main effects of Group, Ambiguity and Valence and an interaction effect of Ambiguity x Valence. Since this study aims to analyse group effects, only the main effect of Group was further examined.
Table 26. Main effect of Group regarding Self-Positive Views on job interviews .
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients 0.07 (0.02) vs 0.27 (0.06) -3.60 = .001 -1.27 BPD vs AVPD 0.07 (0.02) vs 0.02 (0.01) 2.05 = .045 0.76 AVPD vs non-patients 0.02 (0.01) vs 0.27 (0.06) -4.40 < .000 -1.69 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that non-patients reported more Self-Positive Views than BPD and AVPD patients. In addition, BPD patients reported more Self-Positive Views than AVPD patients.
Job interviews; Self-Negative Views
Table 27. Main- and interaction effects regarding Self-Negative Views on job interviews.
Type of effect F (df) P-value
Group 9.02 (2,50) < .001 Ambiguity 11.90 (1,251) = .001 Valence 3.71 (2, 238) = .026 Group x Ambiguity 0.03 (2,267) = .976 Group x Valence 0.34 (4,177) = .848 Ambiguity x Valence 2.80 (2,241) = .063
Group x Ambiguity x Valence 0.31 (4,236) = .871
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and compound symmetry covariance structure.
Results showed main effect of Group, Ambiguity and Valence. No interactions were significant. Since this study aims to analyse group effects, only the main effect of Group was further examined.
Table 28. Main effect of Group regarding Self-Negative Views on job interviews.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients 0.17 (0.04) vs 0.02 (0.01) 3.98 < .001 1.41 BPD vs AVPD 0.17 (0.04) vs 0.52 (0.09) -3.66 = .001 -1.36 AVPD vs non-patients 0.52 (0.09) vs 0.02 (0.01) 5.64 < .001 2.17 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that AVPD patients reported more Self-Negative Views than BPD patients and non-patients. In addition, BPD patients reported more Self-Negative Views than nonpatients.
Job interviews; Other Positive Views
Table 29. Main- and interaction effects regarding Other Positive Views on job interviews.
Type of effect F (df) P-value
Group 6.28 (2,115) = .003 Ambiguity 11.74 (1,86) < .001 Valence 50.01 (2,76) < .001 Group x Ambiguity 0.73 (2,99) = .485 Group x Valence 1.98 (4,78) = .106 Ambiguity x Valence 13.10 (2,76) < .001
Group x Ambiguity x Valence 2.18 (4,73) = .080
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and diagonal covariance structure.
Results showed main effect of Group, Ambiguity and Valence and an interaction effect of Ambiguity x Valence. Since this study aims to analyse group effects, only the main effect of Group was further examined.
Table 30. Main effect of Group regarding Other Positive Views on job interviews.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients 0.63 (0.07) vs 0.85 (0.10) -3.14 = .002 -1.11 BPD vs AVPD 0.63 (0.07) vs 0.82 (0.10) -2.52 = .013 -0.94 AVPD vs non-patients 0.82 (0.10) vs 0.85 (0.10) 0.27 = .790 0.10 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that BPD patients reported less Other Positive views than AVPD patients and non-patients. In addition, AVPD patients and non-patients showed no differences in Other Positive views.
Job interviews; Other Powerful Views
Table 31. Main- and interaction effects regarding Other Powerful Views on job interviews.
Type of effect F (df) P-value
Group 9.61 (2,52) < .001 Ambiguity 7.19 (1,253) = .008 Valence 2.68 (2,240) = .070 Group x Ambiguity 1.41 (4,236) = .231 Group x Valence 0.76 (2,255) = .471 Ambiguity x Valence 0.14 (2,225) = .870
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and compound symmetry covariance structure.
Results showed main effect of Group and Ambiguity. No interactions were significant. Since this study aims to analyse group effects, only the main effect of Group was further examined.
Table 32. Main effect of Group regarding Other Powerful Views on job interviews.
Group Adjusted Mean (S.E.) t P-value Cohen’s d
BPD vs non-patients 0.04 (0.02) vs 0.11 (0.04) -1.63 = .110 -0.58 BPD vs AVPD 0.04 (0.02) vs 0.37 (0.09) -2.72 = .009 -1.01 AVPD vs non-patients 0.37 (0.09) vs 0.11 (0.04) 3.60 < .001 1.39 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Between group comparisons showed that AVPD patients reported more Other Powerful views than AVPD patients and non-patients. In addition, BPD patients and non-patients reported similar Other Powerful views.
Job interviews; Other Negative to Self Views
Table 33. Main- and interaction effects regarding Other Negative to Self Views on job interviews.
Type of effect F (df) P-value
Group 0.00 (2,260) = 1.000 Ambiguity 0.00 (1,260) = .998 Valence 15.07 (2,250) < .001 Group x Ambiguity 0.00 (2,260) = 1.000 Group x Valence 1.64 (4,249) = .164 Ambiguity x Valence 4.61 (2,248) = .011
Group x Ambiguity x Valence 2.90 (4,246) = .022
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and compound symmetry covariance structure.
Results showed a main effect of Valence and interaction effects of Ambiguity x Valence and Group x Ambiguity x Valence. Since this study aims to analyze Group effects, only the Group x Ambiguity x Valence interaction was further examined.
Table 34. Between Group Comparisons: Group x Ambiguity x Valence regarding Other Negative to Self Views on job interviews.
Group x Ambiguity x Valence Comparisons Adjusted Mean (S.E.) t P-value Cohen’s d
Non-Ambiguous Positive BPD vs non-patients 0.00 (0.00) vs 0.00 (0.00) 0.00 = 1.000 0.00 BPD vs AVPD 0.00 (0.00) vs 0.00 (0.00) 0.00 = 1.000 0.00 AVPD vs non-patients 0.00 (0.00) vs 0.00 (0.00) 0.00 = 1.000 0.00 Ambiguous Positive BPD vs non-patients 0.32 (0.11) vs 0.18 (0.09) 1.03 = .306 0.36 BPD vs AVPD 0.32 (0.11) vs 0.41 (0.15) -0.46 = .646 -0.17 AVPD vs non-patients 0.41 (0.15) vs 0.18 (0.09) 1.31 = .193 0.50
Neutral BPD vs non-patients 1.19 (0.21) vs 0.54 (0.15) 1.50 = .0138 0.53 BPD vs AVPD 1.19 (0.21) vs 1.73 (0.32) -1.39 = .165 -0.52 AVPD vs non-patients 1.73 (0.32) vs 0.54 (0.15) 3.33 = .001 1.28 Mixed BPD vs non-patients 0.87 (0.18 ) vs 0.36 (0.12) 2.31 = .0219 0.82 BPD vs AVPD 0.87 (0.18 ) vs 0.17 (0.10) 3.41 = .001 1.27 AVPD vs non-patients 0.17 (0.10) vs 0.36 (0.12) -1.26 = .210 -0.49 Ambiguous Negative BPD vs non-patients 1.35 (0.23) vs 0.97 (0.20) 1.29 = .200 0.46 BPD vs AVPD 1.35 (0.23) vs 1.81 (0.33) -1.14 = .254 -0.42 AVPD vs non-patients 1.81 (0.33) vs 0.97 (0.20) 2.18 = .03010 0.84
Non-Ambiguous Negative BPD vs non-patients 1.57 (0.24) vs 1.45 (0.25) 0.35 = .726 0.12 BPD vs AVPD 1.57 (0.24) vs 1.48 (0.30) 0.22 = .824 0.08 AVPD vs non-patients 1.48 (0.30) vs 1.45 (0.25) 0.09 = .929 0.03 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 12. Group x Ambiguity x Valence (level 1; non-ambiguous) regarding Other Negative to Self Views ratings on job interviews.
Figure 13. Group x Ambiguity x Valence (level 2; ambiguous) Other Negative to Self Views ratings on job interviews.
Between group comparisons showed AVPD patient reported more Other Negative to Self Views than non-patients on the neutral job interview. In addition, BPD patients reported more Other Negative
8
Non-significant against a Bonferroni-Holmes corrected alpha level of 0.008
9
Non-significant against a Bonferroni-Holmes corrected alpha level of 0.008
10 Non-significant against a Bonferroni-Holmes corrected alpha level of 0.017
0 0,5 1 1,5 2 Non-ambiguous
Positive Neutral Non-ambiguousNegative
BPD AVPD Non-patients 0 0,5 1 1,5 2 Ambiguous
Positive Mixed AmbiguousNegative
BPD AVPD Non-patients
to Self Views than AVPD patients on the mixed job interview. BPD patients and non-patients showed no differences in Other Negative to Self Views on job interviews.
The two plots show a clear trend between groups. AVPD patients reported relatively many Other Negative to Self Views on the ambiguous negative and neutral job interview compared to BPD patients and non-patients. In contrast, BPD patients reported relatively many Other Negative to Self Views on the mixed job interview compared to AVPD patients and patients. On the ambiguous negative, non-ambiguous positive and non-ambiguous positive no clear differences in Other Negative to Self Views between groups were observed.
Job interviews; Other Malevolent Views
Table 35. Main- and interaction effects regarding Other Negative to Self Views on job interviews.
Type of effect F (df) P-value
Group 0.50 (2,263) = .610 Ambiguity 0.01 (1,263) = .921 Valence 0.41 (2,263) = .663 Group x Ambiguity 0.16 (2,263) = .850 Group x Valence 0.17 (4,263) = .956 Ambiguity x Valence 0.79 (2,263) = .456
Group x Ambiguity x Valence 0.17 (4,263) = .954
*Note significant effects are in bold. The GLMM applied used a negative binomial distribution with a log-link function and scaled identity covariance structure.
Since it was hypothesised that BPD patients show more Other Malevolent Views than AVPD patients and patients on ambiguous job interviews and job interviews of negative valence, non-significant terms were not omitted from the model and pairwise comparisons (BPD vs AVPD and BPD vs non-patients) were examined on each level of Ambiguity and Valence.
Table 36. Between Group Comparisons: Group x Ambiguity x Valence regarding Other Malevolent Views on job interviews.
Group x Ambiguity x Valence Comparisons Adjusted Mean (S.E.) t P-value Cohen’s d
Non-Ambiguous Positive BPD vs non-patients 0.00 (0.00) vs 0.00 (0.00) 0.07 = 0.947 0.00 BPD vs AVPD 0.00 (0.00) vs 0.00 (0.00) 0.06 = 0.950 0.00 Ambiguous Positive BPD vs non-patients 0.59 (0.14) vs 0.06 (0.05) 3.58 < .001 1.27 BPD vs AVPD 0.59 (0.14) vs 0.08 (0.07) 3.26 = .001 1.21 Neutral BPD vs non-patients 0.59 (0.14) vs 0.12 (0.07) 3.04 = .003 1.07 BPD vs AVPD 0.59 (0.14)vs 0.08 (0.07) 3.26 = .001 1.21 Mixed BPD vs non-patients 0.43 (0.12) vs 0.06 (0.05) 2.88 = .004 1.02 BPD vs AVPD 0.43 (0.12) vs 0.00 (0.00) 3.58 < .001 1.33 Ambiguous Negative BPD vs non-patients 0.48 (0.12) vs 0.00 (0.00) 3.80 < .001 1.34 BPD vs AVPD 0.48 (0.12 vs 0.25 (0.11) 1.36 = .175 0.51
Non-Ambiguous Negative BPD vs non-patients 1.07 (0.19) vs 0.84 (0.18) 0.92 = .358 0.33 BPD vs AVPD 1.07 (0.19) vs 0.67 (0.19) 1.50 = .136 0.56 *Note significant effects are in bold. Adjusted means are estimated at age = 28.89 (mean age).
Figure 14. Group x Ambiguity x Valence (level 1; positive) regarding Other Malevolent Views on job interviews.
Figure 15. Group x Ambiguity x Valence (level 2; neutral/mixed) regarding Other Malevolent Views on job interviews.
Figure 16. Group x Ambiguity x Valence (level 3; negative) regarding Other Malevolent Views on job interviews.
Between group comparisons showed BPD patients reported more Other Malevolent Views than non-patients on the ambiguous positive, neutral, mixed and ambiguous negative job interview. In addition,
0 0 0 0 0 1 1 1
Non-Ambiguous Positive Ambiguous Positive
BPD AVPD Non-patients 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 Neutral Mixed BPD AVPD Non-patients 0 0,2 0,4 0,6 0,8 1 1,2 Non-Ambiguous
Negative Ambiguous Negative
BPD AVPD Non-patients
