r The Association for Family Therapy 2008.

Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. Journal of Family Therapy (2008) 30: 194–214 0163-4445 (print); 1467-6427 (online)

Observing the therapeutic alliance in family therapy: associations with participants’ perceptions and therapeutic outcomes

´n Valentı Escudero,a Myrna L. Friedlander,b Nuria Varelac and Alberto Abascald
Positive and negative alliance-related behaviours of thirty-seven families seen in brief family therapy were rated from videotapes using the System for Observing Family Therapy Alliances (Friedlander et al., 2006b). Positive associations were found between in-session behaviour and participants’ perceptions of the alliance and improvement so far both early (session 3) and later in therapy (session 6). Binary logistic regression showed that successful outcomes (defined as consensus by therapist and all family members on general improvement and reduced problem severity) were significantly predicted by positive individual behaviour (Engagement in the Therapeutic Process, Emotional Connection with the Therapist, Safety within the Therapeutic System) in session 3 and productive withinfamily collaboration (Shared Sense of Purpose within the Family) in session 6. Shared Sense of Purpose was the alliance indicator most consistently associated with clients’ and therapists’ perceptions of therapeutic progress; moreover, it was the only alliance indicator to improve significantly over time in treatment.

In the past two decades, psychotherapy researchers have accumulated evidence indicating that the quality of the working alliance predicts therapeutic change across a variety of treatment modalities (Horvath and Bedi, 2002; Horvath and Symonds, 1991). This empirical link, which underscores the crucial importance of the therapeutic relationship, is pertinent to virtually all approaches to family treatment.
´a, ´ Departamento de Psicologı Facultad de Ciencias de la Educacion, Universidad de ˜ ˜ ˜ La Coruna, Campus de Elvina, 15071 La Coruna, Spain. E-mail: vescudero@udc.es. b Department of Educational and Counseling Psychology, University at Albany, State University of New York. c ˜ Department of Health Sciences, University of La Coruna, Spain. d ˜ Department of Psychology, University of La Coruna, Spain.
r 2008 The Authors. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice

Observing the therapeutic alliance


However, despite the importance of the relationship to practitioners and increasing theoretical interest in the role of common factors in systemic treatments (Sexton, 2007; Simon, 2007; Sprenkle and Blow, 2004, 2007), only a handful of studies have specifically tested the association between the working alliance and therapeutic change in couple and family therapy. None the less, these studies clearly demonstrate the centrality of the alliance in successful conjoint therapy across a diversity of problems and contexts – family therapy for substance-abusing adolescents (Shelef et al., 2005), couples groups for partner abuse (Brown and O’Leary, 2000), structured group training for marital skills (Bourgeois et al., 1990), integrative problem-centred therapy for couples (Knobloch-Fedders et al., 2004, 2007), functional family therapy for delinquent youths (Robbins et al., 2003), emotion-focused couple therapy (Johnson and Talitman, 1997), home-based family therapy (Johnson et al., 2002), brief couple (Symonds and Horvath, 2004) and family (Beck et al., 2006) therapy ‘as usual’, and family psycho-education for schizophrenia (Smerud and Rosenfarb, 2007). The convergence of these findings and the importance placed on the therapeutic relationship by clinicians has prompted renewed interest in the quality of the alliance in clinical practice and recognition of the need to provide therapists in training with specific strategies for fostering and maintaining strong alliances with their clients (Crits-Christoph et al., 2006b). The general purpose of the present study was to assess the relationship of the working alliance to therapeutic progress in brief family therapy. Given the complexity of the therapeutic process in conjoint treatment, it is reasonable to conjecture that different trajectories of alliance development could be associated with successful outcomes. In individual therapy, early alliances seem to be predictive of final outcomes (Constantino et al., 2002), but there is also evidence for different patterns of alliance development (Kivlighan and Shaughnessy, 2000). In family therapy, alliances can be strengthened or broken when additional family members come to a session for the first time or when powerful family secrets are revealed; events such as these can be positive turning points for some families or the catalyst for others to drop out (Friedlander et al., 2006a). Because couple and family therapy involves circular interactions among multiple individuals, the ‘expanded therapeutic alliance’ is a common factor that is unique to conjoint therapy (Sprenkle and Blow, 2004, p. 124). Family therapists are challenged to discover how to foster alliances with multiple clients whose working capacities,
r 2008 The Authors. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice

and (2) family members’ alliances with each other are of crucial importance (Beck et al..196 Valentin Escudero et al.. the study of the interrelations of alliance. clients need to come to some agreement with each other about the goals and value of therapy and to feel safe enough to explore painful issues and conflicts without fear of retaliation when the session ends (Friedlander et al. Although assessment of the alliance early in therapy reliably predicts post-treatment change. Thus strong alliances necessitate the establishment of a safe atmosphere where hostile family interchanges are curtailed. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . 2006a). 2003. 2006). observation of behaviour is optimal – indeed.. Horvath. in other fields of psychological research. 2006). 2006). Gaston and colleagues (1991). A recent re-examination of the alliance literature in the broader context of psychotherapy research has raised concerns about future studies on common factors in general and on the therapeutic alliance in particular (e. 2007. We reasoned that knowledge of which in-session behaviours correspond with clients’ and therapists’ private views on the therapeutic process as it unfolds would inform our understanding of how alliance development relates to treatment r 2008 The Authors... Thompson et al. The systemic complexity of this process is underscored by research showing that (1) a joint assessment of family members’ individual alliances with the therapist is a more sensitive predictor of outcome than any one client’s alliance with the therapist considered in isolation (Knobloch-Fedders et al.g.. perceived progress and final outcome is not only logical but indispensable (Castonguay et al. Castonguay et al. motivations. 2004). causality has not been clearly established (Crits-Christoph et al. 2004.. Aside from relating to a therapist. developmental needs and problem definitions are likely to differ. Robbins et al. yet most alliance research has relied solely on self-report. 2006. supporting the inference that the alliance is a causal factor. confidentiality limits are clarified. 2007). The therapeutic alliance is an interpersonal phenomenon as well as an intrapersonal process (Horvath. One intriguing question has to do with causality in the association between alliances and ongoing improvement. the complementarity of observation and selfreport has been widely acknowledged. found no association between alliance and outcome when prior change was taken into account. 2006. for example. To study the interpersonal aspects. the alliance is complicated by the fact that every client observes every other client’s relationship with the therapist. For this reason.. Friedlander et al. in press b. Symonds and Horvath.. At a basic level. and differences in family members’ roles and expectations are elucidated. 2006a).

g. Behaviours are clustered within four dimensions. Safety within the Therapeutic System and Shared Sense of Purpose within the Family. we used r 2008 The Authors. (2) therapeutic progress at two points in time. which mirrors how family therapists judge the quality of the alliance from clients’ in-session behaviours. In the present study. 2006a. a sense of solidarity in relation to the therapy (‘we’re in this together’) and valuing their time with each other in therapy. and (3) treatment outcome as assessed by clients and therapists in their final session. and Shared Sense of Purpose refers to the level of productive family collaboration on therapy goals and tasks. in this study we examined the degree to which behavioural indicators of the alliance in conjoint family sessions predict participants’ self-reported perceptions of (1) the alliance. Thus. 2006b)..Observing the therapeutic alliance 197 gains. we used a recently developed measure. p. zero-order correlations showed that some aspects of client behaviour as rated on the SOFTA-o were associated with clients’ and therapists’ post-session alliance perceptions and ratings of improvement and session quality. Friedlander et al. The other two SOFTA dimensions. In that study.. (Friedlander et al. trained judges note the frequency. Viewing live sessions or videotapes. with somewhat more experienced therapists. 1994) aspect of the alliance is defined in the SOFTA-o as: family members seeing themselves as working collaboratively in therapy to improve family relations and achieve family goals. This within-system (Pinsof.. reflect the unique aspects of couple and family work. To study the alliance. we had a larger. Spanish sample of clients and therapists. tasks and emotional bonds. intensity and clinical context of specific positive and negative behaviours (e. two of which (Engagement in the Therapeutic Process and Emotional Connection with the Therapist) reflect Bordin’s (1979) classic conceptualization of the working alliance in terms of establishing therapeutic goals. Safety refers to the individual’s comfort level in interacting with family members in psychotherapy. essentially. ‘Client reveals a secret or something other family members didn’t know’ or ‘Family members try to align with the therapist against each other’) to rate the strength of the within-family alliance and each family member’s individual alliance with the therapist. 2006b). a felt unity within the family in relation to the therapy. 2006a. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . the System for Observing Family Therapy Alliances (SOFTA-o. 126) The present study replicates and extends an investigation of twentythree US families that was used in the development of the SOFTA-o (Friedlander et al. Simply put.

By identifying the kinds of interpersonal behaviour that predict clients’ and therapists’ private perceptions of the process and progress of therapy. coping with risky adolescent behaviour). More importantly.198 Valentin Escudero et al. ranging from a simple suggestion or a summary of the session to a structured homework assignment. Families were seen by six white therapists (three women.8 years (range 13–72). and the others had Master’s degrees in psychology. Although sixteen of the thirty-seven cases were couples. chronic illness.20). Two therapists had Ph.e. three men). presenting problems included intrafamily conflict. strategic and solution-focused. Thirty-seven families (N 5 82 individuals) who attended at least three conjoint sessions and whose members completed all self-report data were included in the final sample. we hoped to inform clinical practice as well as the training and supervision of family therapists. M age at intake 5 37. Families were assigned to therapists based on scheduling availability. All participants were white Spaniards. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . we tested the degree to which in-session behaviour could discriminate between families that improved substantially and those that did not. separation. Therapy was not time-limited. communication and intimacy.s. their therapy tended to be family focused (e. with interventions derived from various family therapy orientations. six other families dropped out before that session. which he or she then delivered to the family at the end of the session.g. with a range of four to twenty years of clinical experience (M 5 11. Method Participants The study was conducted in Spain at a university training clinic for family therapists. and we assessed clients’ retrospective views of session impact as well as their immediate (i. including structural. parenting problems. aged 30 to 50 (M 5 40. In general. With the therapist in the consulting room.D. the therapist would take a break after fifty minutes to plan a concluding or final intervention with the observing team. Five of these families completed therapy successfully before session 6. one to five team members observed the session on closed-circuit television or through a one-way mirror.4 years). Generally. although families were told at the outset r 2008 The Authors. adjustment to divorce. An integrative systems model of brief therapy was used. These concluding interventions were varied. post-session) perceptions. multiple regression analyses. or parental death.

video-recording is the general practice in the clinic. are clustered within four dimensions: (1) Engagement in the Therapeutic Process (e. ‘Client shares a joke or light-hearted moment with the therapist’).1). raters make global ratings on each SOFTA dimension by taking into account the frequency. the rating must fall between À 1 and 11. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . intensity and context of the observed behaviours within the respective dimensions. The guidelines specify. negative as well as positive. That is. a second ten-session period was possible if mutually agreed upon. Prior research with the measure demonstrated known groups. that a rating of 0 ( 5 unremarkable or neutral) must be assigned when no alliancerelated behaviours are observed in a given dimension. 2006b. the family unit is rated on Shared Purpose. Next. A training manual provides operational definitions for each item and dimension. judges may assign ratings between 11 and 13.6 (SD 5 3. each family member receives a rating on Engagement. 2003) as a transtheoretical tool for both research and practice. (2) Emotional Connection with the Therapist (e. The average number of sessions for this sample was 7. Across five r 2008 The Authors. along with specific rating guidelines. on a continuum from extremely problematic ( À 3) to extremely strong (13). Friedlander et al.. trained raters first record the occurrence of each behaviour on a checklist. Escudero and Friedlander. rewinding the tape as necessary. with both positive and negative behaviours. ratings must be between À 1 and À 3. ‘Client implies or states that therapy is a safe place’). 2006. 2007). Emotional Connection and Safety. The forty-four client behaviours. ‘Family members offer to compromise’). 2006a. (3) Safety within the Therapeutic System (e. when only negative behaviours are observed.Observing the therapeutic alliance 199 that their progress would be reviewed after ten sessions. Instruments System for Observing Family Therapy Alliances.. ‘Client complies with therapist’s requests for enactments’). Treatment decisions were not constrained by the research.g. and (4) Shared Sense of Purpose within the Family (e. The SOFTA-o (Friedlander et al. 2006b) was created simultaneously in English and ´ ´ Spanish (Sistema de Observacion de la Alianza Terapeutica en Inter´ vencion Familiar. factorial. When only positive behaviours are observed. From videotapes. Smerud and Rosenfarb. valence.g. for example.. after viewing the session.g. predictive and concurrent validity with samples of outpatient couples and families in Spain and North America (Beck et al.g.

a brief self-report measure of the working alliance. was administered immediately following sessions 3 and 6 to both therapists and clients.’s (2006b) development studies for the SOFTA-o.0001 (clients) and r 5. we were able to assess clients’ retrospective perspective on the previous session that was not contaminated by their immediate post-session feelings. Although designed for individual psychotherapy...62. retrospective question asked clients immediately before each session about the usefulness of the previous session (‘How useful was the last session?’) on a 5-point scale from 1 5 not at all to 5 5 very much. The revised Helping Alliance Questionnaire (HAq-II. . Therapeutic progress.85.95 (Friedlander et al. 12 5 much better. 2006b). 1996). items are rated on a 1 (strongly disagree) to 5 (strongly agree) scale. Helping Alliance Questionnaire. 1976).66. the HAq-II was chosen for its psychometric properties and its use in one of Friedlander et al. A second post-session question assessed the global impact of the session (‘After this session I feel . just prior to the subsequent sessions. . One post-session question asked clients and therapists for ‘an estimate of improvement so far’ on a scale from 1 5 not at all to 5 5 very much.g.001 (therapists). po. Luborsky et al. po. 1991). A fourth. test-retest reliability over three therapy sessions (r 5. Luborsky et al. where À 2 5 much worse. 1983).200 Valentin Escudero et al.90). By asking about a session’s usefulness after a meaningful time interval (usually one to two weeks). r 2008 The Authors. the internal consistency reliability was a 5 . 0 5 the same. Third. clients and therapists rated the current status of the target problem on a 1 (worse than ever) to 10 (absolutely perfect) scale.72 to . and higher scores (range 19 to 114) reflect a stronger working alliance..’) on a 5-point scale. 11 5 better. the therapist and I find a way to work on my problems together’) assess various aspects of the therapeutic alliance (Luborsky. In the present sample. this item was taken from the original Penn Helping Alliance Questionnaire (Luborsky et al. for clients. The HAq-II’s nineteen items (e. Self-reported perceptions of therapeutic progress were assessed in four ways – immediately after every session (beginning with session 2) and. reported high internal consistency (a 5 . studies.78). À 1 5 worse. the correlation between sessions 3 and 6 was r 5. ‘In most sessions. and convergent validity with the California Psychotherapy Alliance Scale (Gaston and Marmar. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . the SOFTA-o authors reported average intraclass correlations ranging from .

Outcome determination To classify families as ‘improved’ or ‘not improved’ in terms of final outcome.e.Observing the therapeutic alliance 201 Responses to the improvement item were significantly associated across time for both clients and therapists (r 5.6 sessions (SD 5 2.3). 4 or 5 on improvement. Therapist and all clients in the family rated both (1) improvement and (2) problem status at the high end of the respective scales (i. (in press a).36 and .005). Therapists also provided informed consent. 6 to 10 on problem status) and (3) all family members’ problem status estimates showed a minimum 2-point gain (signifying a reduction in severity) over their initial evaluations of the problem. All but one family agreed to participate and signed a written consent form in the presence of the therapist. this procedure provided important selfreport data from families that dropped out of treatment as well as those that terminated by mutual agreement with the therapist. At the end of sessions 3 and 6 all family members age 131 completed the Haq-II in addition to the therapeutic progress measures they had to complete every session. No incentive was offered for participation. Procedure All families who requested treatment at the clinic during the period of data collection were asked by a research assistant to participate in a ‘naturalistic study about the process of family therapy’.39. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice .39. both pso. Based on these criteria. these families completed an average of 6. Although we did not use standardized instruments to assess outcome.8 per cent) were classified as not improved. we considered each family member’s estimate of (1) improvement so far and (2) target problem status after the last session attended.2 per cent) were classified as improved. the mean number of sessions for this group was 8. 2006a) was used for the training and allowed r 2008 The Authors..01. Three judges independently rated the videotapes of sessions 3 and 6. The improved families met the following three criteria.2 (SD 5 3. The remaining fourteen cases (37. The e-SOFTA software application (Friedlander et al. Judges were trained for twenty hours using practice tapes that were not included in the research sample. r 5. A similar result for clients was recently reported by Friedlander et al. po.4). twenty-three families (62.

96 (Shared Sense of Purpose). or with any of the therapeutic progress measures. a statistically significant change. A randomly selected sample of sessions (25 per cent) was assessed for interrater reliability. Mean intraclass correlations were . to a positive mean rating at session 6. Observed alliance behaviour Figure 1 depicts clients’ alliance-related behaviour in sessions 3 and 6.005) r 2008 The Authors. these demographic variables were not associated with clients’ in-session behaviour. nor were any of the correlations with therapists’ years of clinical experience. Results As preliminary analyses. Three of the four SOFTA-o dimensions showed positive and stable mean ratings at both points. the judges to view and rate the final sample of videotapes from their individual computers. Ratings of improvement so far were also consistently positive and somewhat higher at session 6 for both clients and therapists.202 Valentin Escudero et al. .73 (Engagement).84 (Safety). None of these bivariate correlations was statistically significant. Taken together. . po. The e-SOFTA provides a direct comparison of two or more judges’ ratings. which became significantly more productive over time. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . In other words. indicating generally problematic within-family alliances. Significant correlations were found between clients’ and therapists’ alliance perceptions only after session 3 (r 5. Shared Purpose went from a negative mean rating in session 3.31. with slightly higher scores at session 6.76 (Emotional Connection). Self-reported perceptions: alliance and therapeutic progress Table 1 shows that clients and therapists reported generally strong alliances on the HAq-II at both time periods. with participants’ perceptions of the alliance. for the twenty-six families that completed at least six sessions. qualitative impressions may be recorded and compared to aid consensus negotiations for discrepant ratings. correlations were computed between therapist gender and all study variables. Notably. these results indicate that individual alliance behaviour tended to be moderately strong and consistent. . and the primary shift was in the within-system alliance.

5 0. Note: ENGAGE 5 engagement in the therapeutic process. improvement scores were significantly r 2008 The Authors.0 0. session 6 N 5 54.72 1. Moreover. session usefulness (scores range from 1 to 5) was rated immediately prior to the subsequent sessions.12 0.91 1.54 SD 10.04 1. rs 5 .37 2. Improvement (scores range from 1 to 5) and session impact (scores range from À 2 to 12) were rated immediately after sessions 3 and 6.56 3. Clients’ post-session scores correlated significantly with their retrospective session usefulness scores. however. SSP 5 shared sense of purpose within the family. both pso.13 3.05.5 SOFTA-o ratings 1. CONNECT 5emotional connection with the therapist.08 1.59 0.0 −0. where 0 5 unremarkable or neutral. SOFTA-o dimension ratings can range from À 3 (extremely problematic to 13 (extremely strong).73 99. scores range from 19 to 114). 1996. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice .79 1.95 Note: session 3 N 5 82.79 0.78 M 98. n po..0001.41 2.84 Session 6 SD 8.23 2.48 0.5 203 3 session ENGAGE CONNECT 6 SAFETY SSP* Figure 1 Changes in observed alliance ratings (SOFTA-o) from session 3 to session 6.00 3.49 96.05 6.07 0.Observing the therapeutic alliance 1. SAFETY 5 safety within the therapeutic system. TABLE 1 Descriptive statistics on self-report measures Session 3 M Clients’ perceptions HAq-II Improvement so far Session impact Session usefulness Therapists’ perceptions HAq-II Improvement so far 96. HAq-II 5 Helping Alliance Questionnaire ([28]Luborsky et al. indicating more congruent perspectives on the quality of the alliance early in treatment.85 6.61 (session 6).49 (session 3) and .

session 6 r 5.001.53).18. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . except that Engagement did not reach statistical significance.38. 46) 5 6. r 2008 The Authors. and with both Shared Purpose and Emotional Connection at session 6.17.204 Valentin Escudero et al. Different from clients. therapists’ perceptions of improvement were also correlated with Engagement and Emotional Connection at session 3.001.49. 76) 5 3. Observed alliance behaviours as predictors Bivariate correlations between each of the four SOFTA-o dimension ratings and clients’ and therapists’ self-reported alliance (Haq-II) and improvement-so-far estimates are shown in Table 2. A series of regression analysis was subsequently performed to assess the predictive validity of the SOFTA model.41. po.62] and Safety [b 5 . In terms of therapeutic progress. as described below. associated with both immediate impact (session 3 r 5.12]. R2 5 . R2 5 . R2 5 .80.38. At session 3. adj. R2 5 . Shared Purpose was significantly correlated with therapist-rated improvement at sessions 3 and 6. Clients’ alliance scores were significantly correlated with each of the four SOFTA dimensions at session 3 and with Emotional Connection and Safety at session 6. po. This result means that clients whose behaviour reflected more comfort in the session and a stronger bond with the therapist tended to view the alliance more favourably in session 6.80. 76) 5 3. the beta weights for this equation indicate that Emotional Connection [b 5 .02. adj. The first series of multiple regression analyses was conducted with the four SOFTA-o dimension ratings as predictors and total HAq-II scores (perceived alliance) as the criterion variable.57) and retrospective usefulness (session 3 r 5. session 6 r 5. the observed alliance ratings significantly predicted clients’ alliance perceptions [F(4. In other words.01. SOFTA-o ratings also significantly predicted clients’ HAq-II scores at session 6 [F(4. client-rated improvement was significantly correlated with Shared Purpose at session 3. all indices of therapeutic progress were positively interrelated. Therapists’ alliance perceptions were significantly predicted from clients’ alliance behaviour at session 3 [F(4.40] were unique predictors. t 5 2. t 5 2. Alliance perceptions. suggesting that clients’ perceptions of session helpfulness were retained over time and played a role in perceived improvement. po.34. all pso. Therapists’ perceptions of alliance showed a similar pattern.32].

29n .38nn .48nn . .14 .32nn .34nn Session 6 .47nn Session 6 Session 3 Therapist report Improvement so far Session 6 .14 .09 .05.09 .02 .01 .29nn .31nn .08 .23n SOFTA-o Engagement Emotional connection Safety in the system Shared sense of purpose Note: n po.03 . nn po.33nn .13 .28nn .51nn Haq-II Session 3 .28nn .29nn .22n .53nn .01.18 .15 .17 .20n .TABLE 2 Observed alliance behaviour and self-reported perceptions: bivariate correlations Client report Haq-II Session 3 .16 .44nn Session 3 .10 Improvement so far Session 6 À .

R2 5 . In session 6.78. 73) 5 6. Therapist-rated improvement was significantly predicted from client behaviour at both time periods. higher ratings on Shared Purpose. t 5 3.01. similar to a t test for beta weights.22] and session 6 [F(4. R2 5 . were uniquely associated with greater improvement [b 5 . with client.and therapist-rated improvement as criterion variables. t 5 2. po. indicates which predictor variables.34. if any.19. t 5 3. uniquely contribute to this group discrimination. R2 5 . In session 3. adj.10].0001]. binary logistic regression may be used to predict the presence or absence of the criterion from a set of predictor variables.45. session 3 [F(4. Examination of the beta weights showed that Emotional Connection (b 5 .05]. clients tended to rate the family’s problems as more improved when they had a stronger bond with the therapist and when the within-system alliance was productive. adj. Treatment outcome Binary logistic regression with a backward selection method was used to identify which behavioural alliance indicators best predicted treatment outcome.98.23]. Whereas client-rated improvement was not significantly predicted by alliance behaviour in session 3.46.36. indicating that therapists tended to view the alliance more favourably when family members were collaborating productively with one another early in therapy [b 5 . The Wald statistic. po.01) were unique predictors. Shared Sense of Purpose was the only significant independent predictor. That is. adj. improvement was only uniquely predicted by clients’ Shared Sense of Purpose behaviour [b 5 . po.26. R2 5 . Results r 2008 The Authors. 47) 5 4. adj. A second set of regression analyses was conducted. t 5 2. improved versus not improved.0001. but despite the significant bivariate correlations (shown in Table 2) for Emotional Connection and Safety.87. po. and marginally higher ratings on Engagement.206 Valentin Escudero et al. 47) 5 4.35. R2 5 . Tests of the beta weights showed that. the regression analysis indicated that the set of SOFTA-o behaviours did not reach statistic significance at session 6.05) and Shared Purpose (b 5 . as defined earlier).01. When the criterion variable is categorical and dichotomous (in this case. t 5 2. R2 5 .21]. it was in session 6 [F(4.50. in session 3.94 po.24.01. R2 5 . Perceived improvement.15. po. po.27. The eight predictors entered backward into the logistic regression were the four SOFTA-o ratings at session 3 and at session 6.0001]. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice .29. R2 5 . po.

po. a 2 w (4) 5 10.02.63 Note: the criterion variable. Table 3 shows the beta weights and Wald statistics for the four critical variables in the prediction model. Horvath. showed a significant model composed of session 3 Engagement. In other words.90. This chi-square statistic evaluates goodness-of-fit by creating ordered groups of participants and then comparing the number actually in each group (i. not merely private perceptions (Hatcher and Barends. LRw2 5 .17]. a nonsignificant chi-square indicates that the data adequately conformed to the prediction model. Discussion There is a consensus in the psychotherapy literature that the therapeutic alliance is a salient common factor in all treatment. 2006. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice .63 0.59. LRw2 5 57. R2 5 . The logistic regression results further indicated that treatment outcomes for 70 per cent of the families in the data set were correctly predicted by a linear combination of these four variables.08 3.Observing the therapeutic alliance 207 TABLE 3 A predictive model of treatment outcome from observed alliance behaviour Binary logistic regression modela SOFTA-o predictors Engagement (session 3) Emotional connection (session 3) Safety in the system (session 3) Shared sense of purpose (session 6) b 0.05. 2006).93 0. Consistent with this recommendation. which signifies that the observed results did not differ significantly from the prediction model. indicating that the critical indicators for the improved cases were positive individual alliance behaviour early in therapy and strong withinfamily behaviour in session 6.02.11 0. po.57. Emotional Connection and Safety and session 6 Shared Sense of Purpose [w2(4) 5 10. ns.47 1. treatment outcome. the observed frequency) with the number predicted by the logistic regression model (the predicted frequency).05. was defined as a dichotomous categorical variable: improved (n 5 23) 5 1 versus not improved (n 5 14) 5 0.51 0. The Hosmer–Lemeshow (1989) goodness-of-fit statistic was w2(8) 5 2.e.52 Wald 1. we found that specific kinds of observable. in-session behaviours were predictive of clients’ and r 2008 The Authors. R2 5 .17.90. but the time has come for alliance research to focus on behaviour.

negotiating conflicts.. Shared Sense of Purpose within the Family was the sole SOFTA-o dimension to predict therapists’ reported improvement both times it was assessed. 2006a). some clients’ observed ratings of Safety changed dramatically depending on which members of the family attended the session. We surmise that the significant statistical change was clinically significant as well. the results highlight the central role of one specific.208 Valentin Escudero et al. demonstrating a feeling of safety) played a greater role early in therapy. in that the average rating rose from a negative value (signifying problematic within-system alliances) at session 3 to a positive value at session 6. the importance of the ‘expanded therapeutic alliance’ in conjoint therapy (Sprenkle and Blow. and taking risks with other family members. and it was also the only behavioural variable that improved significantly over time in treatment (see Figure 1). observable aspect of the alliance in understanding how family therapy works – the degree to which clients develop a strong sense of purpose about common goals and about the therapeutic context as a way to solve their family problems. connecting with the therapist. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . a measure designed for individual psychotherapy. r 2008 The Authors. which in the SOFTA-o is referred to as Shared Sense of Purpose within the Family. Unique to conjoint family therapy. Safety within the Therapeutic System. only conjoint family therapy requires clients to feel safe exploring painful issues. 2006). Overall. which is said to be a pre-condition for successful family work (Friedlander et al. 1996).. distinctively predicted therapists’ overall alliance perceptions. That is. emerged in our data as a unique contributor to clients’ self-reported views on the alliance. behaviours indicative of a strong shared sense of purpose were more important later on. whereas alliance-related behaviours of individual family members (engaging in the process. It goes without saying that the degree to which safety is possible for a given family member depends on many factors other than the therapist’s attitude or techniques. Although safety is essential in all therapy modalities. Notably. therapists’ privately held perceptions of the working alliance. In estimating family improvement. clients as well as therapists were attuned to within-family alliances. In this way. therapeutic progress and success achieved in brief family systems therapy.. An equally important aspect of conjoint treatment. despite our use of the HAq-II (Luborsky et al. 2004) was confirmed by the results of this study. in a series of case studies with four families (Beck et al. 1994). behavioural indicators of the within-system alliance (Pinsof. With respect to final treatment outcome.

’s findings in several ways. is reflected in positive behaviours such as asking the therapist personal questions. Emotional Connection to the Therapist. 1996). our study extends Friedlander et al. In that study as well as the present one. Moreover. retrospective session usefulness and outcome status in the last session attended. an aspect of conjoint treatment that was relatively overlooked by the early theorists in our field. Friedlander et al. The binary logistic regression analysis showed that positive individual alliance behaviour early on in therapy and positive withinalliance behaviour in session 6 permitted classification of 70 per cent of the sample’s outcomes as either improved or not improved. for a family to be classified as improved. mirroring the therapist’s body posture. In addition to having a Spanish sample that was more heterogeneous than the exclusively low-income. Simply put. these results have r 2008 The Authors. another unique contributor to clients’ alliance perceptions.. Tests of the beta weights identified which observed alliance dimensions were of unique importance to participants’ self-reported perceptions. (2006b). being reluctant to respond to the therapist and so on. In our sample. and in both studies clients’ Emotional Connection and Safety ratings were significant correlates with therapists’ perceptions of alliance strength. Although the measurement of outcome was not based on symptom reduction but on participants’ perceptions of improvement. the observable strength of the client/therapist bond was significantly associated with clients’ reported alliance and therapy progress in sessions 3 and 6. positive within-family interactions appear to be critical to successful family therapy. We assessed immediate session impact. However. That is. did not assess treatment outcomes in relation to SOFTA-o behaviour. The significant relationships between SOFTA-o alliance behaviour and HAq-II alliance perceptions replicate the results of Friedlander et al. associations between behaviour and perceptions were more robust later than earlier in therapy.Observing the therapeutic alliance 209 Importantly. For clinical practice. sharing a light-hearted moment with the therapist. negative behaviours include avoiding eye contact with the therapist. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . we used regression analysis to test the predictive validity of the SOFTA-o variables. at-risk US families sampled by Friedlander et al. This finding underscores how sensitive therapists have become to the importance of creating a strong therapeutic connection with families (Flaskas and Perlesz. the therapist and all clients needed to report good or very good ‘improvement so far’ and a notable reduction in target problem severity. we used a conservative determination of outcome.

Robbins and colleagues. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . Second. supervisors can encourage students. to pay attention to specific behavioural indicators of alliance. Our families were seen in relatively brief therapy (7. 2003) and Multidimensional Family Therapy (Robbins et al. Our results show that the time at which the alliance is observed makes a difference in the prediction of improvement and final outcome. If tape-recording is not available or not feasible. Although this method is consistent with some systemic ideas. despite the growing support for family therapy alliances. information can be fed back to therapists (and potentially the clients themselves) about these behaviours as therapy unfolds. found that although alliance pattern was a significant predictor of treatment retention in both Functional Family Therapy (Robbins et al. Finally. therapists can use a checklist after their sessions to note the kinds of individual and interpersonal behaviours their clients engaged in. there are at least two ways to use the observable behavioural indicators for therapeutic benefit. First.210 Valentin Escudero et al. the present results must be interpreted in the context of a specific cultural context and treatment approach. That is. The naturalistic nature of the present study brings with it some limitations that should not be ignored. and there was diversity in the subsystems participating in treatment. r 2008 The Authors... in the training of novice therapists. in reviewing their sessions. Moreover.6 sessions. 43 per cent of the sample were couples. 2006). future researchers should not overlook how the alliance interacts with specific characteristics of treatments. for example. future research should include measures of symptoms and family functioning. Not only length of therapy but also the approach to therapy could produce different alliance-improvement profiles. For example. Do in-session alliance behaviours play the same role in different and lengthier approaches to family therapy? In other words. the nature of the alliance patterns varied in these theoretically distinct approaches. on average) using a brief systemic approach with an observing team that provided input on therapeutic interventions and homework assignments. the indices of therapeutic progress and the outcome determination used in this study were derived solely on the basis of a consensus of participant perceptions. but the majority of their presenting problems related to parenting and other family difficulties rather than to specific marital issues. implications beyond the necessity for therapists to work diligently to develop a family’s shared sense of purpose. As is common in field studies. most of our families presented with multiple problems.

several studies (Beck et al. To obtain a richer picture of alliance development over the course of therapy. Friedlander et al.. 2006b. 2006. or Shared Purpose. Qualitative and quantitative case studies can shed light on the correspondence between behaviour and clients’ privately held perceptions of what is taking place in therapy. and specific treatment variables. That is. Future research on therapists’ contributions to the alliance is also crucially important. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . however. to date. is a logical next step in advancing our knowledge of how family therapy works. 2007) have demonstrated the value of this instrument to detect distinctive and crucial aspects of the alliance in conjoint work with families. 2006. the SOFTA-o. For the present clients. clients tend to behave more congruently with their sense of how therapy is progressing.. in press b. the set of SOFTA-o ratings only predicted clients’ improvement estimates at session 6. like the therapeutic alliance. particularly for the families with successful outcomes. we need qualitative data and systematic case studies. suggesting that. the role played by their own behaviour seemed more important to their perceived improvement in later sessions than in earlier ones. This study raises several other questions about the alliance in family therapy. Naturally. reflects how clinicians process behavioural information to make global judgements. Emotional Connection.. we only observed clients’ alliance-related behaviour. Study of specific therapeutic techniques to improve or ‘repair’ (Pinsof. our participants did not show gains in alliance strength from session 3 to session 6 except with respect to intrafamilial interactions. and these remained stable. the therapist’s behaviour makes a difference in the level of alliance and the nature and extent of improvement each family achieves. In this investigation. Smerud and Rosenfarb. uniform process but rather a discontinuous one in which key therapeutic events play a central role. 1994) the alliance is the next step in this burgeoning r 2008 The Authors. the three individual dimensions of the alliance seem to have been favourably established early on in the therapy. Two multiple case studies (Beck et al.. over time. in press a) with the SOFTA-o found that alliance development is not a steady. The one employed in this study. On average. Safety. That is.Observing the therapeutic alliance 211 Research that considers the interaction of common factors. Friedlander et al. One question is how the alliance develops. Using a behaviour-based instrument to estimate the alliance is indispensable. We have no information on the degree to which therapists contributed (positively or negatively) to client Engagement.

(2000) Therapeutic alliance: predicting continuance and success in group treatment for spouse abuse. References Beck. success in conjoint family therapy. Sabourin. r 2008 The Authors. J. Acknowledgements This study was facilitated by a Spanish government research grant (PROFEXT-PR 2006-0213) awarded to the first author. S. L. Research. (2006b) Can therapists be trained to improve their alliances? A preliminary study of alliance-fostering psychotherapy. Connolly Gibbons. and Wright. Constantino. L.. 16: 268–281.. G. and Gallop. P. J. Crits-Christoph.. and O’Leary. M. Friedlander. Journal of Consulting and Clinical Psychology. D. Crits-Christoph. and Schut. (1990) Predictive validity of therapeutic alliance in group marital therapy. Bordin. and Hearon. Bourgeois. Boston. (2006a) Does the alliance cause good outcome? Recommendations for future research on the alliance.) Counseling Based on Process Research: Applying What we Know. C. Practice. J. Narducci. Psychotherapy. Training. Psychotherapy Research. It seems essential to understand how therapists can best bring about a strong shared sense of purpose to facilitate retention and. line of research. Psychotherapy: Theory. Tryon (ed. K. 43: 271–279. and Holtforth. M. Castonguay. R. Gibbons. 81–131. L. Brown.. The authors would like to thank John Carpentar and Mark Rivett for their support. 16: 252–260. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . J. P. M. S... S. 68: 340–345. 32: 355–368. ultimately. Training. Of particular interest is how therapists nurture a strong shared sense of purpose within the family. (1979) The generalizability of the psychoanalytic concept of the working alliance. D. would be an important focus for future study. Constantino.212 Valentin Escudero et al. Research. It is reasonable to expect that some kinds of presenting concerns – such as those that the family defines in strict win/lose terms – present more difficulty for intrafamilial alliances than others. M. P. G. M.. A. Consideration of type of problem. M. (2006) The working alliance: where are we and where should we go? Psychotherapy: Theory. K. B. and Escudero.. Castonguay. V. In G. Crits-Christoph. MA: Allyn and Bacon. E. Schamberger. Journal of Consulting and Clinical Psychology. M. Practice. B. L. pp. G. 58: 608–613. B. 43: 280–285. J.. with the interventions to reframe it systemically. (2002) The working alliance: a flagship for the ‘scientist-practitioner’ model in psychotherapy. (2006) Three perspectives of clients’ experiences of the therapeutic alliance: a discovery-oriented investigation. Journal of Marital and Family Therapy.

D.. and Barends. and Lemeshow. Horvath. M. A. M. 1–12. L. 23: 135–152. (2006) The alliance in context: accomplishments. Journal of Marital and Family Therapy. L. 43: 292–299. Practice. (1989) Applied Logistic Regression. E. C. and Perlesz.. Research. (2006b) System for Observing Family Therapy Alliances: a tool for research and practice. Hatcher. (1996) The return of the therapeutic relationship in systemic therapy. P. B. J. and Heatherington. A. (1991) Manual for the California Psychotherapy Alliance Scales – CALPAS. A. 37–69. and Cragun. Unpublished manuscript. N... A. M. A. (1991) Relation between the working alliance and outcome in psychotherapy: a meta-analysis. y aplicaciones del instrumento [System for Observing Family Therapy Alliances (SOFTA). V. Norcross (ed. Psychotherapy: Theory. W. L.. S. Research. A. and Ketring. R. DC: American Psychological Association. O. Escudero. In J. and Martens. Marmar. Journal of Counseling Psychology. L. R. Gallagher. Cabero. and Friedlander. Journal of Counseling Psychology. L. D. L.. Wright. pp. Cross-cultural development. L. Canada. (in press a) How do therapists enhance family alliances? Sequential analyses of therapist ! client behavior in two contrasting cases. C. Journal of Counseling Psychology. Horvath. O. J. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . challenges. Friedlander. M. M. W. A. (2006) How a return to theory could help alliance research. C. Psychotherapy: Theory. E. A.. reliability. R. Department of Psychiatry.. and Shaughnessy. Friedlander. Training. 4: 362–371. Practice. Psychotherapy: Theory. and Thompson. London: Karnac Books. 25: 32–36. and Bedi. W. 38: 139–149. and Muniz de la Pena. O. A. O. In C. (2003) El sistema de observacion de la alianza ´ ´ terapeutica en intervencion familiar (SOATIF): Desarrollo trans-cultural.. Lambert. D. pp. Flaskas and A. Johnson. (2006a) Therapeutic Alliances in Couple and Family Therapy: An Empirically Informed Guide to Practice. 53: 214–225. C. McGill University: Montreal. P. and future directions. (2000) Patterns of working alliance development: a typology of client’s working ratings. (1991) Alliance prediction of outcome beyond in-treatment symptomatic change as psychotherapy process. (2002) The therapeutic alliance in home-based family therapy: is it predictive of outcome? Journal of Marital and Family Therapy. Escudero. Mosaico (Journal of the Spanish Federation of Family Therapy Associations). New York: Oxford University Press. Horvath.Observing the therapeutic alliance 213 ´ Escudero. S. C. (1997) Predictors of success in emotionally focused marital therapy. P. L. Flaskas. Perlesz (eds) The Therapeutic Relationship in Systemic Therapy. Psychotherapy Research. 43: 258–263. ˜ ˜ Friedlander. Practice. L.) Psychotherapy Relationships that Work: Therapist Contributions and Responsiveness to Patients. Research. Escudero. Heatherington.. Journal of Counseling Psychology. and instrument use]. V. and Marmar. V. (2002) The alliance. Johnson. Gaston. L. Friedlander. D. S. L. Training. New York: John Wiley & Sons. Lambert. and Symonds. r 2008 The Authors.. Hosmer. M. O. Kivlighan. R. V. E. C. 28: 93–102. M. L. Training. and Talitman. Horvath. Washington. Gaston. (in press b) A step toward disentangling the alliance/improvement cycle in family therapy. fiabilidad. 1: 104–113.

M. Journal of Nervous and Mental Disease. 17: 534–544. Knobloch-Fedders. Alexander. B. clinical. Siqueland. S.. S. J. L. Turner.. A. Margolis. W. S. (2004) Optimizing the alliance in couples therapy. (1994) An integrative systems perspective on the therapeutic alliance: theoretical.. Barber. pp. W. W. and Mann. Turner. Sexton. A. and Blow. Liddle. S. W. J. Franks. L.. (1976) Helping Alliances in Psychotherapy: The Groundwork for a Study of Their Relationship to its Outcome. A. 171: 480–491. Crits-Cristoph. Thompson.214 Valentin Escudero et al. O. and Flynn. D. H. P. Sprenkle. G. Journal of Family Therapy. S. G. J. D. 29: 39–55. Diamond. M. (2003) Alliance and dropout in family therapy for adolescents with behavior problems: individual and systemic effects.. 73: 689–698. In A. (2007) Treatment engagement: building therapeutic alliance in family-based treatment.. H. Sprenkle. M. 33: 245–257.. Pinsof. J. and Cohen. Greenberg (eds) The Working Alliance: Theory. Knobloch-Fedders. and Practice.. Luborsky. Robbins. G. W. and Blow. Manuscript under review. K. global rating method. 29: 104–108. Journal of Family Psychology. G. P. M. Alexander. J. Pinsof. 20: 108–116. (2007) The therapeutic alliance and family psychoeducation in the treatment of schizophrenia: a prospective changeprocess study. (2006) Adolescent and parent therapeutic alliances as predictors of dropout in multidimensional family therapy. (2007) The heart of the matter: continuing the conversation. A. Contemporary Family Therapy. and Kogan. Journal of Marital and Family Therapy. and Horvath. and Daley. (2007) Therapeutic alliance and treatment progress in couple psychotherapy. New York: Wiley & Sons. L.. and Mann. Journal of Family Psychology.. (2004) The formation of the therapeutic alliance in couple therapy. (1983) Two helping alliance methods for predicting outcomes of psychotherapy: a counting signs vs..M. S. J. H. S. Luborsky. L. M. Journal of Consulting and Clinical Psychology. M. H. Dakof. O. E. Alexander.. D. L. Luborsky. D. A. (2007) The role of the therapist as the bridge between common factors and therapeutic change: more complex than congruency with a worldview.. J. Journal of Psychotherapy Practice and Research. Johnson. Horvath and L. (2004) Common factors and our sacred models.. B. M. J.. 29: 100–103. r 2008 The Authors.. 29: 109–113. and Perez. (1996) The revised Helping Alliance Questionnaire (HAq-II). M. G. L. J. Research. Robbins. and Rosenfarb. I. 5: 260–271. Najavits. C.. L. A. Journal of Family Therapy. Pinsof. C. (2007) The therapist as a moderator and mediator in successful therapeutic change. 173–195. F. B.. Simon. Journal of Family Therapy. Shelef. L. Lantry. (2005) Adolescent and parent alliance and treatment outcome in multidimensional family therapy. and research implications. K. L. Symonds. New York: Brunner/Mazel. Family Process.. Journal compilation r 2008 The Association for Family Therapy and Systemic Practice . M. 30: 113–129. Journal of Marital and Family Therapy. F. 43: 443–455. S. Smerud. A. and Liddle. T. H. Family Process.. Bender. Diamond. M. 43: 435–442. P.