An Integrative Complexity Analysis of Cognitive Behaviour Therapy Sessions for Borderline Personality Disorder

513

The British Psychological Psychology and Psychotherapy: Theory, Research and Practice (2007), 80, 513–523 Society q 2007 The British Psychological Society www.bpsjournals.co.uk

An integrative complexity analysis of cognitive behaviour therapy sessions for borderline personality disorder

Kate Davidson1,2*, Steven Livingstone1, Katherine McArthur1, Lindsay Dickson2 and Andrew Gumley1,2 1Department of Psychological Medicine, University of Glasgow, UK 2Glasgow Institute of Psychosocial Interventions (GIPSI), NHS Greater Glasgow, UK

Objectives. Integrative complexity (IC), a measure of cognitive style, was used to analyse discourse in Cognitive Behaviour Therapy (CBT) sessions from patients with borderline personality disorder treated in the BOSCOT trial. It was predicted that patients’ level of integrative complexity would be positively associated with the outcome of therapy. That is, an increase in patients’ level of integrative complexity would be associated with good outcome. We also predicted that therapists would also show an increase in the level of complexity associated with their patient’s increase in integrative complexity and good outcome. Design. Ten patients who received CBT were categorized according to the outcome, good (N ¼ 5) and poor (N ¼ 5), using an algorithm that incorporated the number of suicide attempts and magnitude of change in severity of depression during therapy. Method. For each patient and their therapist, an early and a late therapy session were transcribed and coded for integrative complexity (IC) (N ¼ 20 sessions transcribed). IC scores for patients and therapists were compared across early and late therapy sessions and for good and poor outcomes of therapy. Results. The majority of discourse was at the lower levels of IC. Higher levels of IC at baseline were related to depression and anxiety. Good outcome was not associated with a change in the level of IC between earlier and later CBT sessions. Therapists, however, showed an increase in IC when patient’s outcome was poor. In addition, an increase in patient’s IC was associated with improvement in social functioning. Conclusions. Therapists may overcompensate for patient’s poor outcome by giving more complex explanations to patients. Higher complexity does not necessarily lead to better outcomes.

* Correspondence should be addressed to Professor Kate Davidson, Department of Psychological Medicine, University of Glasgow, Gartnavel Royal Hospital, 1055 Great Western Road, Glasgow, G12 0XH, UK (e-mail: [email protected]).

514 Kate Davidson et al.

Randomized controlled trials focusing on the differential effects of one or more treatments on standardized clinical outcomes tell us what works best, but only a little on the process of change for individual patients undergoing psychological therapy. The way individuals think about complex issues underlies and in some cases defines specific mental disorders, and is reflected in characteristic ways of coping. Increasing cognitive complexity, either implicitly or explicitly, might be considered to be one of the main aims of talking therapies. Cognitive behavioural therapy (CBT) aims to increase the ability to think about problems from different perspectives and by doing so, decrease the emotional and behavioural impact of problems. Integrative complexity represents a cognitive style that reflects characteristics of information processing occurring at a particular time or when dealing with a particular issue (Suedfeld & Pennebaker, 1997). Integrative complexity consists of two components: (a) differentiation, the recognition of more than one perspective on a problem or situation and (b) integration, the recognition of relations among differentiated components or perspectives (Suedfeld & Bluck, 1993). Low integrative complexity is marked by rigidity, all-or-none judgements, a desire for rapid closure and minimal uncertainty, and an intolerance of other points of view (Suedfeld, 1985). Higher levels of complexity are characterized by recognition of the validity, relevance or legitimacy of diverse approaches to an issue and awareness that solutions may be based on consideration of how these diverse approaches can be compared or combined. In addition, high levels of complexity are thought to indicate that the individual has invested more time, attention, mental effort, information search, processing and a careful thought into understanding the situation and drawing conclusions or making decisions. Low levels of integrative complexity have been related to acting out behaviours such as aggression (Bruch, McCann, & Harvey, 1991) and deliberate self-harm (Patsiokas, Clum, & Luscomb, 1979). The relationship of integrative complexity to therapy process and progress might elucidate the characteristic style of thinking and cognitive change in individuals with borderline personality disorder undergoing CBT. Being able to describe and measure this relationship would help us in developing more effective therapies for this group of patients. Using integrative complexity as a tool to analyse the content of therapy transcripts may highlight helpful and unhelpful processes in the therapy and potentially identify effective change techniques in CBT (Llewelyn & Hardy, 2001). Hitherto, the application of integrative complexity has been largely to non-clinical areas such as politics and social history. Higher levels of complexity in letters written by famous people have been associated with negative as opposed to positive life circumstances (Suedfeld & Bluck, 1993) but in contrast, lower levels of complexity were expressed in speeches (given by American Psychiatric Association presidents) during years when USA was engaged in war. The authors speculated that this apparent contradiction could be accounted for by ‘positive challenge’-type stress in the person’s own life, leading to an increase in complexity, which is different from the more impersonal nature of war situations. Suedfeld and Pennebaker (1997) examined the relationship between the recall of unpleasant and neutral memories, level of complexity and health outcomes in students. Those in the negative events group wrote essays that were significantly higher in complexity, implying the allocation of a more cognitive effort to the narrative. Those who wrote about negative events also showed a significant correlation between complexity and improvement in well-being, measured biochemi- cally as an increase in immunity. They concluded that confronting negative memories through talking or writing counteracts the negative effects of holding back one’s Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Integrative complexity and borderline personality disorder 515

thoughts or feelings. Such confrontation leads to improvement in mood, subjective feelings of happiness and a variety of health-related physiological and behavioural beneﬁts. In the light of these ﬁndings, it might be expected that in a clinical population, more favourable clinical outcomes might be associated with greater increase in the level of integrative complexity or having higher baseline levels of this ability at the outset of therapy. A randomized controlled trial of cognitive behaviour therapy (CBT) for borderline personality disorder was conducted by Davidson et al. (2006b). For 106 patients, followed up for 2 years, CBT was shown to be an effective therapy in reducing suicidal acts, dysfunctional beliefs, state anxiety and positive distress symptoms (Davidson et al., 2006a). As part of the above study, therapist competence in CBT was assessed by rating a random sample of audiotapes of the patient sessions. All therapists treating patients with CBT scored at or above a recognized level of competence on the Cognitive Therapy Rating Scale (CTRS). The present study aimed to examine integrative complexity in therapy transcripts of CBT sessions for patients with borderline personality disorder who had poor and good outcome. We predicted that:

(1) A patient’s level of integrative complexity at the start of therapy will be associated with scores on baseline measures of psychopathology. (2) When compared with patients with poor outcome in CBT, those with good outcome will show a greater degree of change in the level of integrative complexity between earlier and later CBT sessions. (3) Changes in the level of integrative complexity will be associated with changes in the level of psychopathology from baseline to the end of therapy period (at 12 months). For therapists, we predicted that

(4) An increase in therapist’s level of integrative complexity from earlier to later sessions of therapy will be related to good as opposed to poor outcome (as in prediction 2).

Method Ethical approval for the study was granted by the relevant NHS Ethics committees. Patients gave consent to have their therapy sessions recorded for the BOSCOT trial and analysed for purposes of the research. All identiﬁers were removed from the transcripts.

Algorithm for selecting patients with good and poor outcomes An a priori algorithm was used to select and categorize patients who received CBT according to outcome (see Figure 1). All patients who had CBT during the trial were entered into the algorithm and ﬁrst ranked in terms of the number of suicide attempts, then if this provided no further discrimination such that no suicide attempts had been made, a second ranking was made in terms of the magnitude of change on the Beck Depression Inventory (BDI) (Beck, Steer, & Brown, 1996). We then selected the extremes of these groups of patients with good and poor outcomes and listened to the tape-recordings of CBT sessions for clarity and ability to be transcribed. Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

516 Kate Davidson et al.

Figure 1. Algorithm for selecting patients with good and poor outcomes.

Therapy sessions With the patient’s consent, CBT sessions were recorded on TDK cassette tapes. Early and late sessions were selected for audibility and then sent for transcription to a person who was blind to the hypotheses of the study. The median session number transcribed for early in therapy tapes was 4 (inter-quartile range: 3–5.75; min–max 1–10); for late in therapy tapes, the median session number was 25 (inter-quartile range: 13.5–27.25; min–max 9–32). There was no signiﬁcant difference between patients who had good and poor outcomes in terms of the session number selected for transcribing and coding (earlier in therapy session number: good vs. poor: t ¼ 0:58, df ¼ 8, p ¼ :58, two-tailed; later in therapy session number: good vs. poor: t ¼ 0:26, df ¼ 8, p ¼ :80, two-tailed).

Coding for integrative complexity (IC) A Research Assistant (S.L.), blind to the category of patient outcome, coded the transcripts for integrative complexity. This was done according to The Coding Manual for Conceptual/Integrative Complexity (Baker-Brown et al., 2005). This manual has been designed to allow individuals to learn to code for IC while working independently, without having to attend workshops. Once the manual has been studied, test items are available from Professor Suedfeld’s on-line IC Workshop. Answers to test items are submitted on-line and then scored by a member of the IC Workshop’s staff. A person is considered to be a qualiﬁed coder once he/she has reached the level of .85 reliability with (other) expert coders, and the scorer (S.L.) reached a reliability of .90. In addition, a second qualiﬁed coder rated two transcripts (10% of the total) to assess study reliability and inter-rater reliability was found to be satisfactory (k ¼ :76; .82). Suedfeld on his homepage states that ‘as complexity looks at the structure of one’s thoughts, rather than the content, it is scorable from almost any verbal materials, including speech transcripts’ (Suedfeld, 2005). However, adapting the IC concept to a paradigm it was not originally designed for led to certain novel methodological problems. The percentage of material deemed unscorable in the transcripts was far larger than that has been noted in discourse analysis, where less than 5% was unscorable (Liht, Suedfeld, & Krawczyk, 2005). We found that a range between 45% and 80% was unscorable. Sentences were often begun by one party and then ended by the other, in a way that made it impossible to assign a score to either. In addition, questions are usually considered unscorable within the framework of IC scoring, but it was observed that Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Integrative complexity and borderline personality disorder 517

therapists in particular often asked what seemed to be rhetorical questions, which were otherwise high in IC. On consultation with the originator of IC, Professor Suedfeld, we agreed that where questions were complex and revealed what a person was thinking, they could be given an IC score. For the purposes of scoring, each utterance from the therapist and the patient is described as a turn, for example:

T ‘Did your trip back go okay?’¼ a ‘turn’ P ‘Yeah, no problems’¼ a ‘turn’

A score was assigned to every individual ‘turn’ taken during a session, for both the patient and the therapist. In this way, we were able to look at the minimum and maximum scores for both the therapist and the patient in the entire session.

Measures of psychopathology In the randomized controlled trial (BOSCOT) (Davidson et al., 2006b), patients completed a variety of measures of psychopathology at baseline and at six monthly time intervals up to and including 24 months. The present study includes data gathered from patientswho were randomized to CBT and who completed the following measures at baseline and at 12 months around the completion of the treatment phase. We selected a subset of baseline measures to reﬂect a range of psychopathology including symptoms and social functioning. The subset of measureswas selected a priori and before data analysis. Three measures were omitted as they were seen as less relevant to the hypothesis of this study. Symptom measures included theBrief Symptom Inventory (BSI), Positive Symptom Total (BSI-PST), derived from counting the number of items endorsed with a non-zero response (Derogatis & Melisaratos, 1983); Beck Depression Inventory-II (BDI) (Beck et al., 1996) and State-Trait Anxiety Inventory (STAI) (Spielberger, Gorsuch, & Lushene, 1970). The changes in social functioning are measured by the Social Functioning Questionnaire (SFQ) (Tyrer, 1990; Tyrer et al., 2004) and those in beliefs thought to be related to personality disorder are measured by the Schema Questionnaire (YSQ) (Young & Brown, 1990).

Results Data were analysed using parametric or non-parametric tests depending on the normality of the distribution of data. Generally, IC scores were not normally distributed and non-parametric tests were utilized.

Sample Table 1 gives a description of the sample of patients who had good and poor outcomes. No differences between good and poor outcome patients were found on age, gender or in the total number of CBT sessions received across a 12-month period with both groups having near the maximum 30 sessions recommended in the trial protocol as being feasible. (Age: t ¼ 0:85; df ¼ 8; p ¼ :42; two-tailed; Gender: x2 ¼ 0:48; df ¼ 1; p ¼ :49; total number of CBT sessions received: t ¼ 0:27; df ¼ 8; p ¼ :79). For those with good outcome, no patient had made a suicide attempt during the ﬁrst 12 months of the trial. Therefore, good outcome concerned only change in depression scores as measured by the BDI. The median change in depression scores was 25 points Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

518 Kate Davidson et al.

Table 1. Description of patients with borderline personality disorder with good and poor outcome

Suicide Magnitude of change* Total number of CBT Age Gender M:F attempts** on BDI scores sessions received Outcome (Mean SD) (N) (N, range) (Median, I-Q range) (Median, I-Q range)

Good 29.8 (4.4) 1:4 0 (0) 25 (22–33.5) 27 (14.5–30.5) Poor 34.8 (12.4) 2:3 5 (0–2) 6 (21–17.5) 26 (11–30.5)

* Positive scores indicates improvement. ** Suicide attempts over ﬁrst 12 months of trial. for those with good outcome. Of those allocated to poor outcome, one person had two suicide attempts and a further three patients had made one suicide attempt each. There were no further suicide attempts in those who had received CBT and as a result the patient ranked ﬁfth was determined by change in depression scores. The median change in depression scores was six points for this group.

Integrative complexity score categories for patients and therapists The total number of passages that received an IC score (coded) for patients with good outcome ðN ¼ 5Þ was 531 and 648 for those with poor outcome ðN ¼ 5Þ: There was no signiﬁcant difference between those who had good and poor outcomes in terms of the total number of scorable passages ðt ¼ 0:96; df ¼ 8; p ¼ :37Þ: The number of passages that did not receive an IC score, because they could not be scored due to the content, was 467 for those patients with good outcome and 729 for those with poor outcome. There was no signiﬁcant difference between good and poor outcome groups in terms of passages that could not be scored ðt ¼ 1:53; df ¼ 8; p ¼ :16Þ: For therapists, a similar pattern was evident – the total number of passages receiving an IC score for therapists whose patient had good outcome was 420 and 615 for poor outcome ðt ¼ 2:25; df ¼ 8; p ¼ :06Þ: The number of passages spoken by therapists which did not receive an IC score, because they could not be scored due to the content, was 591 for those patients with good outcome and 773 for those with poor outcome ðt ¼ 1:02; df ¼ 8; p ¼ :30Þ: Figure 2 illustrates that the majority of passages or turns were rated as level 1 in integrative complexity (IC). Therapists showed a greater range of IC categories (1–7) when compared with patients (1–4), but overall the pattern is similar for patients and their therapists indicating that the level of IC in the sample of therapy tapes was not high. There were no differences in the total number of each IC level for patients and therapists, taking all scores into account, regardless of the good or poor outcome or the stage of therapy. (IC Level 1: U ¼ 255; 190; N1 ¼ 845; N2 ¼ 604; ns; Level 2: U ¼ 12; 950; N1 ¼ 185; N2 ¼ 140; ns; Level 3: U ¼ 12; 445; N1 ¼ 131; N2 ¼ 190; ns; Level 4: U ¼ 513; N1 ¼ 18; N2 ¼ 57; ns).

Early IC level and baseline measures of psychopathology Table 2 shows the correlation between patients’ level of IC score in early therapy session (median) and the following baseline measures: Brief Symptom Inventory (BSI-PST), Beck Depression Inventory-II (BDI), State-Trait Anxiety Inventory (STAI State), Social Functioning Questionnaire (SFQ) and Young’s Schema Questionnaire (YSQ). There Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Integrative complexity and borderline personality disorder 519

Figure 2. Frequency of IC score categories (range 1–7) for patients and therapists with good ðN ¼ 5Þ and poor outcomes ðN ¼ 5Þ (data from 20 transcripts, 2 per patient).

were signiﬁcant associations between baseline levels of anxiety and depression and early session IC scores for patients. No signiﬁcant associations were found between early session IC scores and dysfunctional schemas, and between brief symptom inventory and social functioning.

Table 2. Relationship between patients’ early session IC score (median for session ¼ 1) and baseline measures of psychopathology (Spearman’s rho)

IC score in early session

Measures at baseline Spearman’s rho Np

BDI 0.693 10 .03 STAI state 0.678 9 .05 BSI PST 20.213 10 .55 YSQ 0.462 10 .18 SFQ 0.141 10 .70

BDI, Beck Depression Inventory; YSQ, Young’s Schema Questionnaire; BSI, Brief Symptom Inventory (PST ¼ Positive Symptom Total); STAI, State-Trait Anxiety Index; SFQ, Social Functioning Questionnaire. Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

520 Kate Davidson et al.

Differences in IC between early and later sessions for patients and therapists for good and poor outcomes The difference in median IC scores for those with good and poor outcomes was assessed between later and earlier sessions for both therapists and patients separately (see Table 3). We anticipated that those patients with good outcome in CBT would show a greater degree of change in the level of integrative complexity between earlier and later CBT sessions. We did not conﬁrm this hypothesis for either patients or therapists. Contrary to our prediction, we found that when outcome for patients is poor, therapists displayed a signiﬁcantly higher level of integrative complexity in later sessions as compared with earlier sessions. This is better illustrated by the therapist mean IC score for early sessions: 1.74 (SD 0.26) vs. late session IC score: 2.04 (SD 0.27).

Table 3. Differences in Integrative Complexity (IC) score (median and inter-quartile range, min–max) (mean and SD also illustrated) between early and late sessions for patients and therapists with good and poor outcomes (Wilcoxon signed ranks tests) (one-tailed test)

Good outcome Poor outcome (N ¼ 5) (N ¼ 5)

Median Mean Wilcoxon z Median Mean Wilcoxon z (IQR) (SD) ( p) (IQR) (SD) ( p)

Patient Early session 1 (1–2) 1.35 (0.23) 0.314 (.38) 1 (1–2) 1.41 (0.14) 0.043 (0.48) Late session 1 (1–2) 1.52 (0.35) 1 (1–2) 1.43 (0.24) Therapist Early session 1 (1–3) 1.67 (0.36) 0.674 (.25) 1 (1–3) 1.74 (0.26) 3.274 (0.0005) Late session 1 (1–3) 1.84 (0.41) 1 (1–3) 2.04 (0.27)

Relationship between changes in IC score and changes in psychopathology measures We predicted that changes in the level of integrative complexity will be associated with changes in level of psychopathology from baseline to end of the therapy period (at 12 months). As can be seen in Table 4, there were no signiﬁcant associations between changes in mean IC scores from early to later sessions and changes on measures of depression (BDI), anxiety (STAI), dysfunctional beliefs (YSQ) and positive symptom total (BSI-PST). However, there was a highly signiﬁcant association between changes in social functioning (SFQ) and changes in IC scores between early and later sessions.

Discussion We found that in transcripts of cognitive therapy for borderline personality disorder, the overall level of integrative complexity was low for both patients and therapists. The majority of discourse is at the lowest levels (1–4) with only therapists using the highest levels of complexity when discussing issues with patients (1–7). We found no differences between therapists and patients in terms of the frequency with which they use lower levels of complexity (though therapists alone used higher levels of complexity). Patients’ level of integrative complexity at the start of therapy was signiﬁcantly associated with their baseline scores on depression and anxiety but not with their Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Integrative complexity and borderline personality disorder 521

Table 4. Association between change in IC score between early and late sessions and change on measures of psychopathology between baseline and 12-month follow-up (Spearman’s rho)

Change in mean IC score (late session – early session)

Spearman’s rho pN

Change in BDI score 0.285 .43 10 Change in STAI state total 20.052 .90 9 Change in BSI PST 0.286 .42 10 Change in YSQ total 0.285 .43 10 Change in SFQ total 0.784 .007 10

* BDI, Beck Depression Inventory; YSQ, Young’s Schema Questionnaire; BSI, Brief Symptom Inventory (PST ¼ Positive Symptom Total); STAI, State-Trait Anxiety Index.

dysfunctional beliefs, symptom total or social functioning. Higher levels of IC were associated with higher levels of anxiety and depression indicating that patients may be more preoccupied with problems and attempting to provide more psychologically effortful explanations for these. Individuals who are good problem solvers may judge the importance of a particular problem in the context of other demands on their decision- making resources, and invest as much time as the problem warrants. Suedfeld and Pennebaker (1997) have stated that higher complexity does not necessarily produce appropriate solutions to problems and it is possible to think of complex arguments that in hindsight led to the wrong conclusions. We had predicted that patients with good outcome would show a greater degree of change in the level of integrative complexity between earlier and later CBT sessions. We did not confirm this. When outcome is good, patients and therapists showed no significant change in the level of complexity across sessions. We did, however, find an association between poor outcome and changes in the level of integrative complexity used by therapists between early and later sessions. Therapists appear to use higher levels of IC in later sessions when patients had poor outcome. This may reflect therapists having to compensate by working ‘harder’ at providing more psychological or sophisticated explanations to patients to account for their problems when the outcome appears to be poor. However, the scores of therapists and patients generally mirrored each other, demonstrating the reciprocal nature of interaction. Although not part of our original hypotheses, in a random sample of transcripts from five patients (50%) of whom two had good outcome and three had poor outcome, the therapist and the patient had the same score in 45% of ‘turns’. That is to say, if the therapist made a comment that scored ‘3’ for example, the patient replied with a comment that also scored ‘3’. Unscorable passages were disregarded, as is the convention with IC, but all scores of 1 or greater were included. Changes in the levels of complexity as the therapy progressed were not significantly related to changes in measures of symptoms and dysfunctional beliefs, but were significantly related to social functioning – an increase in complexity was associated with improvement in social functioning. It is possible that an increase in social functioning is accompanied by more social interaction and that this, in itself, brings about higher levels of integrative complexity as patients reflect on problems and interactions with others. Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

522 Kate Davidson et al.

The study has some limitations. Our sample is, by necessity in our case, highly selected. We looked at good and poor outcomes of cognitive therapy in a sample of 10 out of 38 people who consented to have therapy sessions audiotaped from a pool of 52 patients who received CBT in the BOSCOT trial. We selected those who had the best outcome and those who had the poorest outcome, according to the algorithm. This strategy may have lead to unknown sources of bias. In addition, not all tapes were sufﬁciently clear to be transcribed and we had to select earlier or later tapes for transcribing. Although IC has been used in the analyses of more structured discourse before (Liht et al., 2005), applying integrative complexity to therapeutic discourse is new. In a more natural discourse, we found that patients and therapists take frequent turns at talking, interrupt each other and use a questioning format to make statements. In addition, we noted that the majority of therapeutic discourse, when analysed by turn taking, is at the lowest level of IC. This is not surprising as natural discourse differs from a prepared, but then spoken, speech. It may be that IC is more appropriate for the analyses of the latter type of speech and that it does not reﬂect more natural conversational speech. With this in mind, it may be that low IC scores give the misleading impression that what goes on in therapy is not complex. Had we found a greater association between IC and outcomes, we might have concluded that CBT serves to increase the complexity of a patient’s thoughts regarding their problems. Whilst this may be true of CBT, our methodology could not assess this fully. Although it may be possible to apply IC successfully to therapeutic discourse, we may not have completely captured the essence of therapeutic dialogue by taking the ‘turn’ as the unit of analysis. Future studies may wish to extract a topic within a session as the unit of analysis of integrative complexity. Alternatively, researchers may wish to further develop the concept of integrative complexity to analyse therapy discourse.

References Baker-Brown, G., Ballard, E. J., Bluck, S., de Vries, B., Suedfeld, P., & Tetlock, P. E. (downloaded 25.04.05). www.psych.ubc.ca/, psuedfeld/MANUAL.pdf. Beck, A. T., Steer, R. A., & Brown, G. K. (1996) Manual for the Beck Depression Inventory -II. Psychological Corporation, San Antonia, Texas. Bruch, M. A., McCann, M., & Harvey, C. (1991). Type A behavior and processing of social conﬂict information. Journal of Research in Personality, 25(4), 434–444. Davidson, K., Norrie, J., Tyrer, P., Gumley, A., Tata, P., Murray, H., et al. (2006a). The effectiveness of cognitive behaviour therapy for borderline personality disorder: Results from the BOSCOT trial. Journal of Personality Disorders, 20, 450–465. Davidson, K. M., Tyrer, P., Gumley, A., Tata, P., Norrie, J., Palmer, S., et al. (2006b). Rationale, description, and sample characteristics of a randomised controlled trial of cognitive therapy for borderline personality disorder: The BOSCOT study. Journal of Personality Disorders, 20(5), 431–449. Derogatis, L. R., & Melisaratos, N. (1983). The brief symptom inventory: An introductory report. Psychological Medicine, 13, 595–605. Liht, J., Suedfeld, P., & Krawczyk, A. (2005). Integrative complexity in face-to-face negotiations between the Chiapas Guerrillas and the Mexican Government. Political Psychology, 26, 543–552. Llewelyn, S., & Hardy, G. (2001). Process research in understanding and applying psychological therapies. British Journal of Clinical Psychology, 40(1), 1–21. Copyright © The British Psychological Society Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Integrative complexity and borderline personality disorder 523

Patsiokas, A. T., Clum, G. A., & Luscomb, R. L. (1979). Cognitive characteristics of suicide attempters. Journal of Consulting and Clinical Psychology, 47(3), 478–484. Spielberger, C., Gorsuch, R., & Lushene, R. (1970). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists Press. Suedfeld, P.(1985). APA presidential addresses: The relation of integrative complexity to historical, professional and personal factors. Journal of Personality and Social Psychology, 49(6), 1643–1651. Suedfeld, P. (downloaded 25.04.05). www.psych.ubc.ca/, psuedfeld/index2.html. Suedfeld, P., & Bluck, S. (1993). Changes in integrative complexity accompanying signiﬁcant life events: Historical evidence. Journal of Personality and Social Psychology, 64(1), 124–130. Suedfeld, P., & Pennebaker, J. (1997). Health outcomes and cognitive aspects of recalled negative life events. Psychosomatic Medicine, 59(2), 172–177. Tyrer, P. (1990). Social functioning questionnaire. Personality disorder and social functioning. In D. F. Peck & C. M. Shapiro (Eds.), Measuring human problems: A practical guide (pp. 136–137). Chichester: Wiley. Tyrer, P., Nur, U., Crawford, M., Karlsen, S., McLean, C., Rao, B., Johnson, T., et al. (2005) Social Functioning Questionnaire: A rapid and robust measure of perceived functioning. International Journal of Social Psychiatry, 51, 265–275. Young, J. E., & Brown, G. (1990) Young Schema Questionnaire. In Young J. E. (1999) Cognitive therapy for personality disorders: A schema-focused approach. Third edition. Professional Resource Exchange, Inc.

Received 16 March 2006; revised version received 22 December 2006