Psychotherapy Research, January 2011; 21(1): 1626

NARRATIVE AND PSYCHOTHERAPY

Narrative and emotion integration in psychotherapy: Investigating the relationship between autobiographical memory specificity and expressed emotional arousal in brief emotion-focused and client-centred treatments of depression

TALI ZWEIG BORITZ, LYNNE ANGUS, GEORGES MONETTE, LAURIE HOLLIS-WALKER, & SERINE WARWAR
Department of Psychology, York University, Toronto, Ontario, Canada
(Received 24 May 2009; revised 1 June 2010; accepted 10 June 2010)

Abstract Clinically depressed individuals have consistently been shown to demonstrate a bias for overgeneral autobiographical memory (ABM) disclosure, a strategy used to protect against the access of intense, primary emotions that may accompany specific memories. The present study examined how ABM specificity in client narratives was related to expressed emotional arousal in brief emotion-focused and client-centred psychotherapy for depression. Emotion episodes identified in two early-, two middle-, and two late-therapy transcripts drawn from 34 clients from the York I Depression Study were rated for degree of ABM specificity and expressed emotional arousal. A hierarchical linear modelling analysis demonstrated that greater ABM specificity was associated with higher expressed emotional arousal for clients who were no longer depressed at therapy termination.

Keywords: depression; emotion in therapy; experiential/existential/humanistic psychotherapy; process research; narrative; autobiographical memory

In psychotherapy, the process of representing lived experience as told stories establishes a shared context of meaning between client and therapist and helps therapists to better understand clients complaints and modes of adjustment (Angus & McLeod, 2004). Client disclosures of vivid, specific narrative accounts of emotionally salient life events is central to how therapists identify core conflictual interpersonal patterns (Luborsky, Barber, & Digeur, 1992) and develop an empathic understanding of clients key concerns (Angus, Lewin, Bouffard, & RotondiTrevisan, 2004) during therapy sessions. Research on autobiographical memory (ABM) has demonstrated that the specificity of disclosed memories differs for depressed versus nondepressed samples (Williams et al., 2007). Compared with nondepressed controls, clinically depressed subjects

demonstrate a bias for overgeneral, nonspecific ABM disclosure and show difficulties accessing specific, single-event ABM narratives. This is important because the inability to access and integrate specific episodic ABM has also been associated with reduced self-coherence, increased rumination and worry, impaired social problem solving, and a reduced capacity to imagine future events (Conway & Pleydell-Pearce, 2000). To assess and systematically identify ABMs for clinical and research applications, Singer and Moffitt (1992a) developed a classification of ABM subtypes. Single-event ABMs involve recall of a singular or episodic event that is more likely to evoke experience-near, sensory-perceptual imagery, and, importantly, affective responses (e.g., One Sunday we went for a walk in the park, and my father told me

This article is based on Tali Zweig Boritzs masters thesis completed at York University. Correspondence concerning this article should be addressed to Tali Zweig Boritz, Department of Psychology, York University, 4700 Keele Street, Toronto, Ontario M3J 1P3, Canada. Email: tboritz@yorku.ca ISSN 1050-3307 print/ISSN 1468-4381 online # 2011 Society for Psychotherapy Research DOI: 10.1080/10503307.2010.504240

Autobiographical memory and emotional arousal that he was leaving my mother). In contrast, summary memory, or overgeneral ABM, refers to ABMs that are nonspecific in content and that represent a compilation of multiple events over a series of occasions. There are two types of summary memory: generic and extended. Generic ABM involves the repetition of a category of events with a lack of discrete connection to a particular moment in time. A defining characteristic of generic ABMs is a blending of separate events into a schematic representation (e.g., Every Sunday my father would take me for a walk in the park). Extended ABMs represent imageless and detail-barren memories that occur over sweeping time spans (e.g., One summer my father and I took many walks in the park). Recently, Boritz, Angus, Monette, and HollisWalker (2008) examined ABM narrative specificity as it occurred in therapy transcripts of 34 clients undergoing brief emotion-focused therapy (EFT) and client-centred treatment (CCT) for depression in the York I Depression Study (Greenberg & Watson, 1998). Their findings demonstrated that ABM specificity significantly increased over the course of therapy, suggesting that the overgeneral ABM bias in depression may be subject to change as the result of treatment. However, increases in ABM specificity were not related to outcome. Reflecting on this finding, Boritz et al. (2008) contended that it may be that ABM specificity augments other process variables, such as client-expressed emotional arousal, that directly impact outcome in brief therapy for depression. Williams et al. (2007) argued that the increased visual and experiential imagery of specific ABM is more likely to evoke the original emotional content of a remembered event, thus providing a rich narrative context within which to explore salient emotions during the therapy hour. In a similar vein, Greenberg and Angus (2004) suggested that the disclosure of personally salient, specific ABMs is of particular importance for effective EFT for depression because emotions can only be understood*and have personal meaning for clients* when they are organized within a narrative framework that identifies what is felt, about whom, in relation to a specific need or issue. As such, the differentiation of client emotional experiences in the context of salient personal stories may be a key intervention strategy for client narrative change and beneficial outcomes in psychotherapy. However, few studies to date have examined the relationship between client ABM specificity and expressed emotions in actual therapy sessions. The purpose of the present study is to address this gap in the psychotherapy research literature and investigate the relationship between ABM narrative specificity and expressed emotional

17

arousal*observable, in-session verbal and nonverbal behaviour reflecting the intensity of emotional experience*in the context of early-, middle-, and late-therapy session transcripts drawn from the York I Depression Study (Greenberg & Watson, 1998). As well, this study seeks to determine whether the relationship between ABM narrative specificity and expressed emotional arousal is related to outcome at therapy termination. Previous processoutcome investigations of EFT and CCT for depression have focused almost exclusively on the significant contributions of expressed emotional arousal (Missirlian, Toukmanian, Warwar, & Greenberg, 2005; Warwar, 2003) to overall treatment outcomes in the context of the York I Depression Study (Greenberg & Watson, 1998). For these studies, researchers segmented early, middle, and late sessions into discrete emotion episodes (EEs; Greenberg & Korman, 1993; Korman, 1991), identifying therapy session narratives in which the client describes an emotion in response to a situation or event, real or imagined. In an examination of the relationship between expressed emotional arousal and outcome, Warwar (2003) predicted that higher levels of client emotional arousal would be related to better outcomes in a depressed clinical sample (N 032) engaged in psychotherapy. Using a linear regression approach, Warwar found that higher levels of midtherapy modal emotional arousal were related to better outcome on the Beck Depression Inventory (BDI; Beck, Steer, & Garbin, 1988; Beck, Ward, Mendelsohn, Mock, & Erbaugh, 1961) for both EFT and CCT dyads. Similarly, Missirlian et al. (2005) found that higher modal emotional arousal in the middle stage of therapy was related to overall therapeutic improvement in level of self-esteem on the Rosenberg Self-Esteem Scale (Rosenberg, 1965) for both EFT and CCT dyads at termination. Additionally, higher modal emotional arousal in middle and late stages of therapy were related to significant reductions in interpersonal dysfunction, as assessed by the Inventory of Interpersonal Problems (Horowitz, Rosenberg, Baer, Uren o, & Villasen or, 1988). Although Warwar (2003) and Missirlian et al. (2005) provide important information about how expressed emotional arousal operates in CCT and EFT for depression, the generalization of research findings to clinical practice is constrained by three methodological limitations. First, although both studies point to the importance of heightened expressed emotional arousal in therapeutic outcome (e.g., decreased symptoms, increased self-esteem, greater interpersonal problem solving), they provide little information about which other processes (e.g.,

18

T. Z. Boritz et al. is important to therapeutic outcome (nondepressed vs. depressed at therapy termination) above and beyond either variable alone. Method

narrative, interpersonal) occurring in EEs are most likely to be associated with heightened expressed emotional arousal. The second limitation concerns the type of statistical analyses conducted: Warwar (2003) used a linear regression analysis and Missirlian et al. (2005) used hierarchical regression analyses. Given the longitudinal, nested, and unbalanced structure of the data in the York I Depression Study, hierarchical linear modelling (HLM) is presently considered to be a more appropriate statistical approach for capturing the richness and complexity of psychotherapy process research. Finally, both studies focused on modal expressed emotional arousal as representative of client emotional arousal occurring throughout treatment. Although modal expressed emotional arousal indeed captures clients overall level of emotional arousal, peak expressed emotional arousal captures the most intense moments of expressed emotion within EEs and may be a better measure of client emotional processing. The present study examines ABM specificity in relation to both modal and peak expressed emotional arousal and attempts to replicate previous expressed emotional arousal findings using HLM analyses. In summary, although both cognitive-experimental (Williams et al., 2007) and psychotherapy (Greenberg & Angus, 2004; Singer & Salovey, 1993) researchers suggest that clients disclosures of specific ABM narratives may be an important starting point for accessing and expressing emotions in effective therapeutic treatments of depression, no comprehensive study to date has empirically investigated the relationship between ABM specificity and expressed emotional arousal. The present study is the first in the field of psychotherapy research to assess the relationship between ABM narrative specificity and expressed emotional arousal at early, middle, and late stages of therapy for depressed clients undergoing brief EFT and CCT and to investigate the relationship between ABM narrative specificity and expressed emotional arousal to overall therapeutic outcomes. On the basis of the expectation that clients disclosures of specific (single-event) ABMs may be an important precondition for heightened client emotional engagement during EFT and CCT sessions (Greenberg & Angus, 2004), we hypothesize that within EEs higher proportions of specific ABMs will be related to higher levels of expressed emotional arousal. Furthermore, because previous studies have demonstrated an association between expressed emotional arousal and outcome in brief experiential psychotherapy (Missirlian et al., 2005; Warwar, 2003), we predict that the relationship between ABM narrative specificity and expressed emotional arousal

Participants Clients. The sample consisted of 34 individuals (25 women, nine men) who participated in the York I Depression Study (Greenberg & Watson, 1998). All participants met criteria for major depressive disorder on the Structured Clinical Interview for the DSM-III-R (SCID; Spitzer, Williams, Gibbon, & First, 1989) and had BDI (Beck et al., 1988) scores of 16 or higher. Fourteen clients (41%) had at least one SCID-diagnosed Axis II personality disorder. Seventeen were assigned to CCT and 17 to EFT. Clients had a mean age of 39.64 years (SD 011.97). Thirty clients were White, three were Asian, and one was Latino. Therapists and therapist training. Eight female and three male therapists participated in the study. One was a psychiatrist, four had a PhD in clinical psychology, and six were advanced doctoral students in clinical psychology. Ten of the 11 therapists had at least 2 years of client-centred training and an average of approximately 5.5 years of therapy experience; the remaining therapist was trained initially in cognitivebehavioural therapy but received 12 weeks of client-centred training and supervision and was judged as competent to administer the treatment. The therapists prior training and experience in the use of the active experiential therapy tasks ranged from 12 weeks of training to a number of years of experience (M 01.5 years; Greenberg & Watson, 1998). All therapists received 24 weeks of additional training for the study based on manuals devised for this project (Greenberg, Rice, & Elliott, 1993; Greenberg, Rice, & Watson, 1994). They received training in CCT for 8 weeks, systematic evocative unfolding for 6 weeks, two-chair work for 6 weeks, and empty-chair work for 4 weeks. Training was provided by the originators of the manuals of these interventions. The therapists were supervised on one pilot client before the start of the project. During the first half of treatment with their pilot client, therapists implemented CCT, and during the second half they added the active experiential interventions. Each therapist served as his or her own control by seeing an equal number of clients in each of the two modalities.

Autobiographical memory and emotional arousal Raters. Different raters were used for the various procedures. Two clinical psychology advanced doctoral students*a White female in her late 20s and a White female in her mid-40s*identified EEs. Two clinical psychology graduate students, both White women in their mid-20s, were responsible for expressed emotional arousal ratings. Three raters* a White female advanced doctoral student in her early 30s, a White female clinical psychology graduate student in her early 20s, and a White male clinical psychology graduate student in his early 20s*identified narrative sequences within EEs. Two clinical psychology graduate students, both White women in their mid-40s, identified ABM subtypes within external narrative sequences. Treatments Participants were randomly assigned to CCT or EFT. Treatment consisted of 15 to 20 one-hour weekly sessions (M 017.6 sessions). All sessions were both audio- and videotaped. CCT (Rogers, 1957, 1961) emphasizes three necessary conditions: unconditional positive regard, empathy, and congruence. Participating therapists were trained in CCT using manuals developed specifically for this purpose (Greenberg, Rice, & Watson, 1994, as cited in Greenberg & Watson, 1998; Rice & Greenberg, 1990). EFT (Greenberg et al., 1993) adopts clientcentred therapeutic conditions (i.e., unconditional positive regard, empathy, and genuineness) and integrates experiential interventions, such as gestalt techniques (e.g., empty- or two-chair dialogue, focusing, and systematic evocative unfolding), based on the presence of specific client markers. The main objective of EFT is the evocation and restructuring of maladaptive emotional schemes that are perceived to be the source of distress (Greenberg et al., 1993). Measures EEs (Greenberg et al., 1993; Korman, 1991). These are in-session segments in which a client expresses having experienced an emotional response or demonstrates an action tendency in relation to a real or an imagined situation (Korman, 1991). EEs comprise two main components: the situation (e.g., being criticized by a superior) and the emotional response (e.g., I felt stupid) or an action tendency associated with the emotional response (e.g., I was so embarrassed that I left work early). An EE is identified according to thematic content related to the emotional response, beginning when the emotional response is expressed in the therapy transcript and ending when a new emotional response emerges.

Client Emotional Arousal Scale III (CEASIII). This scale, adapted from Daldrup, Beutler, Engle, and Greenberg (1988), as cited in Warwar & Greenberg, 1999), measures the presence, intensity, and function of anger in focused expressive psychotherapy. The scale was revised by Machado (1992); Machado, Beutler, & Greenberg, 1999) to measure the intensity of six primary emotions (love, anger, fear, joy, surprise, sadness) and added vocal quality (Rice, Koke, Greenberg, & Wagstaff, 1979). The resulting 7-point Likert scale measures the overall intensity of emotional categories (pain/hurt, sadness, hopelessness/helplessness, loneliness, anger, resentment, contempt/disgust, fear/anxiety, love, joy/ excitement, contentment/calm/relief, shame/guilt, pride/self-confidence, anger and sadness at same time, pride and anger together, surprise/shock, and other), which are coded when clients acknowledge having an emotion or demonstrate an action tendency in response to an emotion. Scores of 1 to 3 are considered to represent lower expressed emotional arousal, while scores of 4 to 7 indicate higher expressed emotional intensity levels. This scale has demonstrated interrater reliabilities ranging from .75 to .81 (Warwar & Greenberg, 1999). Narrative processes coding system (NPCS; Angus, Levitt, & Hardtke, 1999). This two-step system allows researchers to reliably subdivide therapy transcripts into narrative process subtypes (Angus, Levitt, & Hardtke, 1996). External narrative sequences were of sole interest for this study because they capture client discourse focused on the description of personal life events and provide specific criteria for the identification of ABM narratives embedded within EEs. The NPCS has demonstrated interrater agreement for identifying narrative process sequences (range 08388%, Cohens k 0.75; Angus et al., 1999). ABM specificity. This was assessed using Singer and Moffitts (1992b) scoring manual for memory narrative subtypes (i.e., single event, generic, extended), which was adapted for use with psychotherapy transcripts (Angus et al., 1996). Included in the refinements of this method for psychotherapy transcripts was the addition of an initial step in the coding procedure to determine whether the ABM in question met criteria for a personal memory related to self (Brewer, 1996). Additionally, the category of not ABM was further subdivided into three subtypes: not autobiographical (about someone or something other than the client); not a memory (autobiographical but not a memory, e.g., semantic information conjecture or future plans); and too short to code (less than four lines of client and

19

20

T. Z. Boritz et al. necessary to identify the high-arousal sessions to adequately test the hypothesis. Preparation of transcripts for analysis. There were three steps involved in the preparation of the transcripts for analysis. EEs were identified by two advanced doctoral students in clinical psychology. Raters independently segmented transcripts into EEs, each of which was demarcated by identifying the situation and the corresponding emotion or action tendency. Reliability of EE sampling was based on raters agreement on both the situation and emotion for each EE as well as agreement of the location of the EE within approximately a half page of a transcript (Cohens k 0.79; Warwar, 2003). The kappa represented average agreement between the pair of raters. Expressed emotional arousal was rated for each EE. Two trained graduate students used the CEAS-III (Warwar & Greenberg, 1999) under the supervision of registered psychologists. Raters were unaware of outcome. For the purpose of this study, only peak expressed emotional arousal scores were utilized. Cohens k for peak arousal ratings was .78 (Warwar, 2003). Narrative sequences were identified in the context of EEs. Three trained graduate students, unaware of outcome, used the NPCS to identify narrative process modes within the context of EEs and demonstrated good interrater agreement (Cohens k 0.88.95). Interrater agreement was determined based on an average between pairs of raters. External narrative sequences were of exclusive interest for further analysis. Once external segments were identified, they were subsequently coded for ABM specificity by two trained graduate students. According to both the NPCS manual (Angus et al., 1999) and the ABM coding system (Singer & Moffitt, 1992b), ABMs were identified as one of four mutually exclusive types: single event, generic, extended, and combination. Whenever the combination category was encountered (e.g., single event and generic), a decision was made based on the prevalence of one category over the other in the combination. Using these criteria, the trained raters achieved high interrater reliability (Cohens k 0.77.88; Hollis-Walker, 2005). The kappa represented average agreement between the pair of raters. Once all ABM subtypes were identified, the proportion for each subtype was calculated at the EE level (i.e., if there were two single-event ABMs and two generic ABMs within an EE, the proportion of single-event ABMs was 50% and the proportion of generic ABMS was 50%). For the purpose of the present study, generic and extended ABMs were

therapist dialogue). The original manual was determined to achieve an average interrater agreement of 93% (Cohens k 0 .78; Singer & Moffitt, 1992a). BDI (Beck et al., 1961). This 21-item inventory is used to assess depression. The BDI has high internal consistency and correlates highly with other self-report measures of depression and with clinician ratings of depression (r 0.60.90; Beck et al., 1988). Testretest reliability has been reported to be .65 (Ogles, Lambert, & Sawyer, 1995).

Procedure Treatment adherence. All therapists received manual-based training and were monitored through audio- and videotapes for adherence to treatment before and during therapy sessions. A full description of treatment adherence procedures and results is provided in Greenberg and Watson (1998). Both CCT and EFT required the provision of empathy for adherence to treatment protocols. Therapists in both conditions were rated as sufficiently high on average tape-rated empathy, and the two treatments were not significantly different on rated empathy. One case in each condition was eliminated because the treatment failed to adhere to the manual. Transcript selection. Two sessions from the early, middle, and late stages of therapy were selected for each of the 34 clients involved in the York I Depression Study. The initial session was excluded in all cases based on the assumption that it did not involve the delivery of therapy but rather focused on discussion of the presenting problem and the establishment of therapeutic alliance (Bordin, 1994; Horvath, 2001). The second and third sessions were selected to represent the early sessions. Late sessions were the two sessions immediately preceding the final therapy session. The middle sessions were chosen from among those occurring anytime after the third session and before the final three sessions of therapy. The term middle was chosen to characterize a working stage of therapy that comes after the alliance is developed and before the last few sessions of therapy, in which clients are usually engaged in more reflective processes (Horvath & Bedi, 2002). The selection process of middle sessions follows previous studies investigating emotional arousal (see Missirlian et al., 2005; Warwar, 2003). These sessions were selected to reflect the presence of expressed client emotional arousal as an aspect of the theory being tested that it is the processing of aroused emotion that provides the foundation for important change. Therefore, it is

Autobiographical memory and emotional arousal combined to represent summary memory or overgeneral memory. Herein, the term overgeneral ABM is used to represent instances where generic and extended ABM have been combined for analyses. For the HLM analyses, the proportion of ABM subtype at the EE level represented ABM specificity. Outcome categorization. To categorize therapeutic outcome in a statistically reliable way, we used the clinical significance groups for the BDI identified by Seggar, Lambert, and Hansen (2002). Their research has previously identified a threesample normative continuum, and cutoff points were established for the BDI: asymptomatic (AS) refers to those who exist in the community without depressive symptomatology; community symptomatic (C) refers to the functional yet normally depressed community; and clinically symptomatic (CS) refers to those who are actively experiencing psychological distress to the point of seeking or being required to seek treatment. In the York I Depression Study (Greenberg & Watson, 1998), all clients in the sample began treatment in the CS group (see Table I for BDI pre- and posttherapy descriptives). Given the previous research on ABM in depression, which links the phenomenon of overgeneral ABM with clinically symptomatic depression, the present study sought to target CS clients at therapy termination rather than examining outcome based on clinically significant change. Thus, Seggar et al.s (2002) cutoff scores were applied to the sample to determine two outcome groups: nondepressed at therapy termination (clients in the AS and C groups; BDI B14.29) and depressed at therapy termination (clients in the CS group; BDI ]14.29). On the basis of Seggar et al.s criteria, 28 clients in our
Table I. Mean Pre- and Post-therapy BDI Scores by Outcome and Therapeutic Modality

21

sample were categorized as nondepressed and six as unchanged (i.e., depressed) at therapy termination. Hierarchical Linear Modeling and the York I Depression Study An HLM approach is similar to a multiple regression model in that both calculate estimates of parameters and their standard errors. However, whereas the standard regression has a single error term, the HLM has error terms for each random effect included in the model. In HLM, regression coefficients are referred to as fixed effects and the error terms as random effects (Raudenbush & Bryk, 2002). The HLM method simultaneously conducts both between- and within-subjects analyses, which allows for a fine-tuned examination of the complex nature of the data gathered for psychotherapy process research (see Gibbons et al., 1993). The York I Depression Study data are longitudinal with a nested, multilevel structure. There are four nested levels of random effects: dyads, sessions within dyads, EEs within sessions, and narrative sequences within EEs. The random levels used for a particular analysis depend on the level of the response variable. For example, a response variable measured at the EE level would require three random levels in the analysis. The data are considered unbalanced because the number of observations varies across and within random levels. Hierarchical linear regressions were performed with the linear mixed-effects (lme) modeling function in the nonlinear mixed-effects (nlme) modeling package (Pinheiro, Bates, DebRoy, & Sarkar, 2007) using the R statistical language (R Development Core Team, 2007). Specific hypotheses were tested with Wald tests using estimated regression coefficients and their estimated variances. Results

Therapy group CCT Total Nondepressed Unchanged EFT Total Nondepressed Unchanged Combined Total Nondepressed Unchanged

Pre-therapy

Post-therapy

17 14 3 17 14 3 34 28 6

22.90 22.50 23.30 26.20 25.40 27.00 24.55 23.95 25.15

12.35 7.40 17.30 11.85 6.40 17.30 12.10 6.90 17.30

Note. CCT, client-centered therapy; EFT, emotion-focused therapy.

Testing for the relationship between singleevent ABM and peak expressed emotional arousal. Mean proportions of single-event ABM, levels of peak expressed emotional arousal, and number of EEs by stage of therapy, outcome group, and treatment type are presented in Table II. In order to determine the relationship between specific ABM and expressed emotional arousal, a hierarchical linear regression model was constructed using arousal as the dependent variable and proportion of singleevent (specific) versus overgeneral ABM as the explanatory or independent variable, with random intercepts for dyads and sessions within dyads. This model included proportion of single-event ABMs

22

T. Z. Boritz et al. Findings from this analysis demonstrated a significant overall effect of single-event ABM on peak emotional arousal, F(2, 744) 03.821, p 0.022. There was also evidence of a Stage )Outcome interaction effect on the relationship between peak expressed emotional arousal and single-event ABM, F(2, 145) 05.285, p 0.006. Specifically, among nondepressed clients, there was a significantly positive relationship between peak expressed emotional arousal and specific ABMs, t(744) 02.396, p 0.016 (95% confidence interval [CI]: 0.0230.233), whereas there was a nonsignificant negative relationship between peak expressed emotional arousal and specific ABMs among unchanged clients, t(744) 0 (1.379, p 0.168 (95% CI: (0.438 to 0.076; see Figure I). There was no evidence of a relationship between peak expressed emotional arousal and outcome, F(3, 32) 02.319, p 0.094. A Wald test conducted to determine whether there were differences in the trajectories of peak expressed emotional arousal over stages of therapy for nondepressed versus unchanged clients, controlling for single-event ABM, was significant, F(3, 30) 03.549, p 0.026. There was also evidence of significant changes in peak expressed emotional arousal by stage of therapy for both nondepressed, F(2, 145) 09.457, p 0.000, and depressed, F(2, 145) 0 4.441, p 0.013, clients. Among nondepressed clients, there were significant increases in peak expressed arousal from early to late stages, t(145) 04.076, p 0.00008 (95% CI: 0.2090.603), and middle to late stages, t(145) 0 3.753, p 0.00025 (95% CI: 0.1610.520), of therapy. There was a nonsignificant increase in peak expressed arousal among nondepressed clients from early to middle stages of therapy, t(145) 00.798, p 0 .425 (95% CI: 0.096 to 0.227). Unchanged clients evidenced a significant decrease in peak expressed arousal from early to middle stages of therapy, t(145) 0(2.962, p 0.003 (95% CI: (0.895 to ( 0.178). Among unchanged clients, there was a nonsignificant increase in peak expressed emotional arousal from middle to late stages of therapy, t(145) 01.544, p 0.124 (95% CI: (0.106 to 0.867), and a nonsignificant decrease in peak expressed emotional among unchanged clients from early to late stages of therapy, t(145) 0 (0.661, p 0.509 (95% CI: (0.624 to 0.311). Testing for the relationship between singleevent ABM and modal expressed emotional arousal. Tests of the model were replicated using modal expressed arousal. A test of the overall effect of single-event ABM on modal arousal was not significant, F(2, 744) 01.559, p 0.210. There was also no evidence of a relationship between modal

Table II. Mean Proportions of Single-Event ABM, Level of Peak Arousal, and Number of Emotion Episodes by Stage of Therapy, Outcome, and Therapeutic Modality Stage of therapy Therapy group N Early Middle Late Range

Mean proportions of single-event ABM CCT Total Nondepressed Unchanged EFT Total Nondepressed Unchanged Combined Total Nondepressed Unchanged CCT Total Nondepressed Unchanged EFT Total Nondepressed Unchanged Combined Total Nondepressed Unchanged CCT Total Nondepressed Unchanged EFT Total Nondepressed Unchanged Combined Total Nondepressed Unchanged 17 14 3 17 14 3 34 28 6 0.42 0.37 0.46 0.33 0.35 0.30 0.38 0.36 0.38 0.35 0.40 0.30 0.33 0.35 0.31 0.34 0.38 0.31 0.51 0.49 0.53 0.48 0.40 0.55 0.50 0.45 0.54 01 01 .14.69 01 01 .11.69 01 01 .11.69

Mean number of emotion episodes 17 14 3 17 14 3 34 28 6 25.90 29.79 22.00 20.75 22.50 19.00 23.33 26.15 20.50 23.74 30.14 17.33 24.10 25.86 22.33 23.92 28.00 19.83 24.15 26.29 22.00 19.55 23.43 15.67 21.85 24.86 18.84 1152 1452 1126 1242 842 1227 1152 852 1127

Mean level of peak arousal 17 14 3 17 14 3 34 28 6 3.36 3.37 3.35 3.44 3.47 3.40 3.40 3.42 3.34 3.13 3.29 2.97 3.56 3.71 3.41 3.35 3.50 3.19 3.55 3.64 3.45 3.45 3.75 3.14 3.50 3.70 3.30 2.694.33 2.694.33 2.973.87 2.584.41 2.844.41 2.584.14 2.584.41 2.694.41 2.584.14

Note. ABM, autobiographical memory; CCT, client-centered therapy; EFT, emotion-focused therapy.

within sessions; proportion of single-event ABMs within EEs; average proportions of single-event ABM within dyad, stage, and outcome grouping (nondepressed vs. unchanged); and interaction terms for outcome with all the other variables. These averages act as contextual variables (Allison, 2005). This type of model allows for the study of the relationship between single-event ABM and expressed arousal within EE, adjusted for possible confounding effects of the variables that vary between session and between dyads.

Autobiographical memory and emotional arousal

Non-Depressed
0.0 0.2 0.4 0.6

23

Unchanged
0.8 1.0

Early
3.8

Middle

Late

Peak Arousal

3.6

3.4

3.2

3.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Proportion of Single-Event ABM

Figure I. Relationship between peak expressed emotional arousal and proportions of single-event autobiographical memory (ABM) by stage of therapy and therapeutic outcome. Results demonstrate a signicant positive relationship between peak arousal and single-event ABM for nondepressed clients; a nonsignicant negative relationship between peak arousal and single-event ABM for unchanged clients; signicant differences in peak arousal trajectories for nondepressed versus unchanged clients; a signicant increase in peak arousal for nondepressed clients from early to late and middle to late stages of therapy; and a signicant decrease in peak arousal for unchanged clients from early to middle stages of therapy.

expressed emotional arousal and outcome, F(3, 32) 01.764, p 0.173. Testing for treatment differences in the relationship between single-event ABM and expressed emotional arousal. To test for possible treatment differences on the relationship between single-event ABM and expressed emotional arousal, treatment type was added to the previously described model. There was no evidence of an overall treatment effect on peak arousal, F(8, 28) 01.48, p 0.208, or of an overall treatment effect on modal arousal, F(12, 26) 01.016, p 0.462. Discussion The present study sought to empirically investigate the relationship of ABM specificity and expressed emotional arousal to therapeutic outcome (nondepressed vs. unchanged at therapy termination). Initial analyses determined that there was a significant positive relationship between ABM specificity and peak emotional arousal for the sample as a whole: At the EE level, greater proportions of singleevent ABM were significantly related to higher levels of peak expressed emotional arousal. Moreover, the slope of this relationship was consistent across all stages of therapy. These findings are consistent with the research literature that contends that the construction of detailed descriptions of specific ABM narratives requires the evocation of visual and/or experiential imagery, which is more likely to evoke deeper emotional arousal (Moffitt, Singer, Nelligan,

Carlson, & Vyse, 1994; Raes, Hermans, de Decker, Eelen, & Williams, 2003; Singer & Moffitt, 1992a; Singer & Salovey, 1993; Wessel, Meeren, Peeters, Arntz, & Merckelbach, 2001; Williams & Broadbent, 1986; Williams, Stiles, & Shapiro, 1999). The picture became more complicated when the relationship between ABM specificity and expressed emotional arousal was examined in relation to therapeutic outcome. In support of assertions that clients capacities to disclose specific ABM with greater emotional intensity are useful for the successful application of active therapeutic interventions and productive therapy (e.g., Greenberg et al., 1993), it was hypothesized that higher proportions of specific ABMs in combination with higher levels of expressed emotional arousal would be related to better therapeutic outcome (movement from depressed to nondepressed status on the BDI at therapy termination). This finding was supported: For clients who were nondepressed at therapy termination, higher proportions of single-event ABMs were significantly related to higher levels of peak arousal across all stages of therapy. In contrast, among clients who were depressed at therapy termination, higher proportions of specific or single-event ABMs tended to be associated with lower levels of peak arousal, although this relationship was not significant. It is possible that the lack of significance associated with this finding is due to the relatively small sample size (n 06) of the unchanged outcome subgroup. Future research is necessary to determine whether this is indeed the case.

24

T. Z. Boritz et al. action tendencies, new personal meanings, and self-understandings. Future research is necessary to better understand what differentiates clients who demonstrate high narrative and emotion integration in psychotherapy from those who display low integration. In particular, it will be important to examine the contribution of therapeutic alliance in facilitating or impeding narrative and emotion integration processes. One limitation of the present study was that the sample consisted of unbalanced outcome groups, with a relatively small unchanged outcome subgroup. It is possible that a larger sample size would allow for more conclusive results regarding the relationship between ABM specificity and expressed emotional arousal in the unchanged group and differences between the outcome groups. Another limitation was the singular use of the BDI to determine outcome. It is possible that a more comprehensive assessment of outcome may have yielded different findings. The present study also did not include any measures of therapy follow-up; thus, we cannot be certain whether the immediate post-therapy depression levels were maintained. Future research including therapy follow-up measures would be helpful in determining whether shifts in narrative and emotion integration contributed to long-term and sustainable depression reduction. The findings from the present study suggest that within psychotherapy for depression the evocation and disclosure of specific ABMs is necessary but not sufficient for good therapeutic outcome. Rather, the integration of expressed emotional arousal within the context of specific ABMs may provide a narrative framework that encourages the generation of new meaning and increases the likelihood that new views of self, self-experiences, and ways of relating to others will be transferred to interpersonal contexts and domains outside of psychotherapy. This finding has important clinical implications because it may require determined monitoring of clients narrative and emotional content by therapists. This type of microprocessing is inherent in some types of therapy, such as EFT (Greenberg et al., 1993), where specific ABMs can serve as markers for interventions and as a means to facilitate therapeutic interventions that heighten emotional arousal. Indeed, it is the heightening of emotional arousal within the context of disclosures of specific ABMs that appears to be crucial for better therapeutic process over the course of therapy. That being said, in order to determine whether the pattern of results reflects change across psychotherapies or a pattern of change specific to experiential therapies, it will be important and necessary to replicate these findings for other therapeutic modalities that do not emphasize emotional

The present study extended examination of expressed emotional arousal to include analyses of both modal and peak arousal scores. Whereas previous studies using modal expressed emotional arousal found that modal arousal was related to outcome, the present study did not find evidence of this for either modal or peak arousal. It is important to note, however, that there are differences in the definition of outcome between the present study and the previous research that found evidence of such a relationship. This is one possible explanation for the discrepant findings. Another possible explanation is that when the highly sensitive and finely tuned nature of HLM is applied, previously demonstrated relationships between expressed arousal and outcome are no longer indicated. That being said, this finding is theoretically consistent with the premise of the present study: that the relationship between expressed emotional arousal and ABM specificity is important to therapeutic outcome over and above either variable on its own. This was established for ABM specificity in Boritz et al. (2008), which demonstrated that ABM specificity alone was not related to therapeutic outcome. Now, with the present findings similarly suggesting this for expressed emotional arousal, it appears that neither narrative processes nor emotional processes are singularly related to therapeutic outcome; rather, it is their integration that is associated with successful psychotherapeutic treatment. These findings present a complex picture of how ABM narrative and expressed emotion work together in psychotherapy. They suggest that from the beginning of therapy there is a distinct relationship between narrative organization and emotional processing, which may influence the therapeutic process, as evidenced by a consistent slope between ABM specificity and peak expressed emotional arousal, over the course of therapy, for the clinical sample as a whole. Clients who were nondepressed at therapy termination evidenced a significant, positive relationship between heightened ABM specificity and expressed emotional arousal, whereas those who were depressed at therapy termination evidenced a nonsignificant negative relationship between these two factors. It is possible that for the unchanged subgroup the symbolization and expression of emotions evoked in the context of specific ABM disclosures is too intrapersonally and/ or interpersonally threatening to be sustained in the therapy relationship. Importantly, this emotional avoidance strategy undermines the possibility of schematically organizing and integrating destabilizing and distressing emotional experiences as coherent narratives that can be reflected on for the identification of primary adaptive emotions and

Autobiographical memory and emotional arousal processing (e.g., cognitivebehavioural therapy). Nevertheless, as the first in vivo study to investigate the relationship between ABM specificity and expressed emotional arousal in brief psychotherapeutic treatments for depression, the present study provides new information about the interplay of narrative and emotion and a greater understanding of how to evoke and deepen clients emotional experiences in psychotherapy. References
Allison, P. D. (2005). Fixed effects regression methods for longitudinal data using SAS. Cary, NC: SAS Institute Inc. Angus, L. E., Levitt, H., & Hardtke, K. K. (1996). Narrative process coding system: Training manual. Unpublished manuscript, York University, Toronto. Angus, L. E., Levitt, H., & Hardtke, K. K. (1999). The narrative processes coding system: Research applications and implications for psychotherapy practice. Journal of Clinical Psychology, 55(10), 12551270. Angus, L. E., Lewin, J., Bouffard, B., & Rotondi-Trevisan, D. (2004). Whats the story? Working with narrative in experiential psychotherapy. In L. E. Angus & J. McLeod (Eds.), The handbook of narrative and psychotherapy: Practice, theory, and research (pp. 87101). Thousand Oaks, CA: Sage. Angus, L. E., & McLeod, J. (2004). The handbook of narrative and psychotherapy: Practice, theory, and research. Thousand Oaks, CA: Sage. Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-ve years of evaluation. Clinical Psychology Review, 8, 77100. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561571. Bordin, E. (1994). Theory and research on the therapeutic working alliance: New directions. In A. Horvath & L. Greenberg (Eds.), The working alliance: Theory, research, and practice (pp. 1337). New York: Wiley. Boritz, T., Angus, L. E., Monette, G., & Hollis-Walker, L. (2008). An empirical analysis of autobiographical memory specicity subtypes in brief emotion-focused and client-centered treatments of depression. Psychotherapy Research, 18(5), 584593. Brewer, W. F. (1996). What is recollective memory? In D. C. Rubin (Ed.), Remembering our past: Studies in autobiographical memory (pp. 2190). Cambridge, UK: Cambridge University Press. Conway, M. A., & Pleydell-Pearce, C. W. (2000). The construction of autobiographical memories in the self-memory system. Psychological Review, 107, 261288. Daldrup, R. J., Beutler, L. E., Engle, D., & Greenberg, L. S. (1988). Focused expressive psychotherapy: Freeing the overcontrolled patient. New York: Guilford Press. Gibbons, R. D., Hedeker, D., Elkin, I., Waternaux, C., Kraemer, H. C., Greenhouse, J. B., et al. (1993). Some conceptual and statistical issues in analysis of longitudinal psychiatric data. Archives of General Psychiatry, 50, 739750. Greenberg, L. S., & Angus, L. E. (2004). The contributions of emotion processes to narrative change in psychotherapy: A dialectical constructivist approach. In L. E. Angus & J. McLeod (Eds.), The handbook of narrative and psychotherapy: Practice, theory, and research (pp. 331349). Thousand Oaks, CA: Sage. Greenberg, L. S., & Korman, L. (1993). Assimilating emotion into psychotherapy integration. Journal of Psychotherapy Integration, 3, 249265.

25

Greenberg, L. S., Rice, L. N., & Elliott, R. (1993). Facilitating emotional change: The moment-by-moment process. New York: Guilford Press. Greenberg, L. S., Rice, L. N., & Watson, J. (1994). Manual for client-centered therapy. Unpublished manuscript, York University, Toronto. Greenberg, L. S., & Watson, J. (1998). Experiential therapy of depression: Differential effects of client-centered relationship conditions and process interventions. Psychotherapy Research, 8(2), 210224. Hollis-Walker, L. (2005). Emotional arousal and autobiographical memory specicity within emotional episodes in brief psychotherapy for depression. Unpublished masters thesis, York University, Toronto. Horowitz, L. M., Rosenberg, S. E., Baer, B. A., Uren o, G., & Villasen or, V. S. (1988). Inventory of Interpersonal Problems: Psychometric properties and clinical application. Journal of Consulting and Clinical Psychology, 56, 885892. Horvath, A. O. (2001). The alliance. Psychotherapy: Theory, Practice, Training, 38, 365372. Horvath, A. O., & Bedi, R. P. (2002). The alliance. In J. C. Norcross (Ed.), Psychotherapy relationships that work: Therapist contributions and responsiveness to patients (pp. 3769). New York: Oxford University Press. Korman, L. M. (1991). Emotion episodes. Unpublished masters thesis, York University, Toronto. Luborsky, L., Barber, J., & Digeur, L. (1992). The meanings of narratives told during psychotherapy: The fruits of a new observational unit. Psychotherapy Research, 2(4), 277291. Machado, P. P. P. (1992). Clients emotional arousal in therapy: Development of a rating scale. Unpublished manuscript, Psychotherapy Research Project, University of California, Santa Barbara. Machado, P. P. P., Beutler, L. E., & Greenberg, L. S. (1999). Emotion recognition in psychotherapy: Impact of therapist level of experience and emotional awareness. Journal of Clinical Psychology, 55(1), 3957. Missirlian, T., Toukmanian, D., Warwar, S., & Greenberg, L. (2005). Emotional arousal, client perceptual processing, and the working alliance in experiential psychotherapy for depression. Journal of Consulting and Clinical Psychology, 73(5), 861 871. Moftt, K. H., Singer, J. A., Nelligan, D. W., Carlson, M. A., & Vyse, S. A. (1994). Depression and memory narrative type. Journal of Abnormal Psychology, 103(3), 581583. Ogles, B. M., Lambert, M. J., & Sawyer, J. D. (1995). Clinical signicance of the National Institute of Mental Health Treatment of Depression Collaborative Research Program data. Journal of Consulting and Clinical Psychology, 63, 321326. Pinheiro, J., Bates, D., DebRoy, S., & Sarkar, D. (2007). nlme: Linear and nonlinear mixed effects models (R package version 3.183). Retrieved from http://cran.r-project.org/web/packages/ nlme/index.html. R Development Core Team. (2007). The R project for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved from http://www.R-project.org. Raes, F., Hermans, D., de Decker, A., Eelen, P., & Williams, J. M. G. (2003). Autobiographical memory specicity and affect regulation: An experimental approach. Emotion, 3(2), 201206. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage. Rice, L. N., & Greenberg, L. S. (1990). Fundamental dimensions in experiential therapy: New directions in research. In G. Lietaer, J. Rombauts & R. Van Balen (Eds.), Client-centered and experiential psychotherapy in the nineties (pp. 397414). Leuven, Belgium: Leuven University Press.

26

T. Z. Boritz et al.
Warwar, S. H. (2003). Relating emotional processes to outcome in experiential psychotherapy of depression. Unpublished doctoral dissertation, York University, Toronto. Warwar, S. H., & Greenberg, L. S. (1999). Client emotional arousal scale IIIR. Unpublished manual, York Psychotherapy Research Centre, Toronto. Wessel, I., Meeren, M., Peeters, F., Arntz, A., & Merckelbach, H. (2001). Correlates of autobiographical memory specicity: The role of depression, anxiety and childhood trauma. Behaviour Research and Therapy, 39, 409421. Williams, J. M. G., Barnhofer, T., Crane, C., Hermans, D., Raes, F., Watkins, E., et al. (2007). Autobiographical memory specicity and emotional disorder. Psychological Bulletin, 133(1), 122148. Williams, J. M. G., & Broadbent, K. (1986). Autobiographical memory in suicide attempters. Journal of Abnormal Psychology, 95, 144149. Williams, J. M. G., Stiles, W. B., & Shapiro, D. A. (1999). Cognitive mechanisms in the avoidance of painful and dangerous thoughts: Elaborating the assimilation model. Cognitive Therapy and Research, 23(3), 285306.

Rice, L. N., Koke, C. J., Greenberg, L. S., & Wagstaff, A. K. (1979). Manual for the client vocal quality classication system. Toronto: York University, Counselling and Development Centre. Rogers, C. R. (1957). The necessary and sufcient conditions of therapeutic personality change. Journal of Consulting Psychology, 21, 95103. Rogers, C. R. (1961). On becoming a person. Boston: Houghton Mifin. Rosenberg, M. (1965). The Self-Esteem Scale. Princeton, NJ: Princeton University Press. Seggar, L. B., Lambert, M. J., & Hansen, N. B. (2002). Assessing clinical signicance: Application to the Beck Depression Inventory. Behavior Therapy, 33, 253269 Singer, J. A., & Moftt, K. H. (1992a). An experimental investigation of specicity and generality in memory narratives. Imagination, Cognition. and Personality, 11(3), 233257 Singer, J. A., & Moftt, K. H. (1992b). A scoring manual for narrative memories. Unpublished manual, Department of Psychology, Connecticut College, New London. Singer, J. A., & Salovey, P. (1993). The remembered self: Emotion and memory in personality. New York: Free Press. Spitzer, R., Williams, J. M. G., Gibbon, M., & First, M. (1989). Structured clinical interview for DSM-III-R. Washington, DC: American Psychiatric Press, Inc.