Cognitive Behavioural Therapy (CBT) is the form of treatment that has the gathered the most empirical support for its efficacy to date. Its efficacy in treating a variety of adult mental disorders has been well documented (see Butler, Chapman, Forman, & Beck, 2006). In fact, 94% of “well-established” empirically supported treatments (see Hunsley, Dobson, Johnston, & Mikail, 1999) are cognitive behavioural therapies. CBT is based on the premise that psychopathology is caused by dysfunctional thinking patterns, that in turn cause emotional and behavioral difficulties (Beck, Emery, & Greenberg, 1985). Thus, treatment is designed to modify dysfunctional cognitions and maladaptive behaviours, and to alter the manner in which people interpret and act on negative emotions. Treatment is delivered via a psychoeducational approach, in which the therapist begins by providing a rationale for CBT to the client and explains the relationship between cognition, affect, and behaviour (Hamilton & Dobson, 2002). Treatment is focused on solving current problems, is structured and directive, and is typically brief and time-limited (Beck, Emery, & Greenberg, 1985).
Increasing demands for cost-effective and empirically supported treatments have heightened interest in CBT. Moreover, the increased concerns of delivering the best possible treatment to consumers has made empirically supported treatments not only a reality, but also a professional and ethical responsibility. Yet, despite evidence of treatment efficacy, the precise mechanisms of change in CBT are not well-documented (Addis & Jacobson, 2000). However, with the efficacy of cognitive behavioural therapies now relatively well established, investigators have begun to consider the “active ingredients” of such approaches (Neimeyer & Feixas, 1990).
Moreover, evidence is accumulating regarding the specific and non-specific factors that influence outcome in CBT. Specific factors encompass therapist factors (e.g., techniques), client factors (e.g., homework adherence), and their interaction. Non-specific factors include those that are common across psychotherapies and not just specific to CBT, such as the therapeutic alliance and the resulting empirical data are now starting to shed light on a question that has beleaguered clinicians for quite some time. That is, is a good therapeutic relationship necessary for positive therapeutic outcome? (Hamilton & Dobson, 2002).
The Therapeutic Alliance
One difference among therapists that has been examined in its relation to outcome is a therapist’s ability to form a positive working relationship, or therapeutic alliance, with his/her clients. The therapeutic alliance, defined as the attachment and collaboration between the client and therapist (Bordin, 1979), has been identified as an instrumental factor related to positive therapeutic change. The terms therapeutic alliance, working alliance, helping alliance, and therapeutic empathy have all been used synonymously to denote specific parts of the alliance or the alliance as a whole (Horvath & Luborsky, 1993). The alliance has been one of the most widely examined variables in therapy process research, and continues to be so. In fact, in 1999 the APA Division of Psychotherapy Task Force was commissioned to identify, operationalize, and disseminate information on empirically supported therapy relationships (ESRs) (Norcross,
2002). Moreover, Grencavage and Norcross (1990) reviewed 50 published studies to determine what commonalities are shared among diverse psychotherapies. The largest common factor was the development of a therapeutic alliance between the client and therapist, cited in 56% of the studies as a common element across different psychotherapies. Thus, the alliance appears to be an important and necessary transtheoretical construct. In general, the positive contribution of the alliance towards outcome has been found across treatment modalities, using both client and therapist ratings (for an extensive review, see Horvath & Bedi, 2002).
History of Concept
The therapeutic alliance’s effect on outcome was first emphasized in psychodynamic therapy. Freud (1958) believed that an analyst’s positive and supportive attitude towards a client would allow him/her to develop a healthy attachment towards the therapist. Furthermore, he differentiated between this healthy attachment and an unhealthy or neurotic attachment, known as transference (Freud, 1958). Greenson (1965) later developed the concept of transference and coined the term working alliance to denote the client’s healthy and mature identification to the analyst. Zetel (1956) also developed the distinction between transference and alliance. She noted that a healthy relationship between a client and therapist allows the client to step back and receive the therapist’s interpretations regarding healthy and unhealthy attachments in therapy.
More recent process research within short-term psychodynamic therapy has supported the importance of the alliance. Suh, O’Malley, Strupp, and Johnson (1989) found that the alliance formed within the first three sessions was predictive of post-treatment outcome. On the other hand, the clients of therapists who were not initially warm and empathic evidenced early deterioration in treatment from which recovery was difficult.
The importance of the therapeutic alliance was further underscored with the advent of client-centred therapy. Rogers (1957) asserted that a therapist’s warm and empathic stance, congruency, and ability to accept a client unconditionally were necessary and sufficient conditions for positive therapeutic outcome. Early research findings indicated that therapists who possessed these qualities were more successful in securing better outcomes for their clients (Rogers, Gendlin, Kiesler, & Truax, 1967). However, the positive relationship between Rogers’ therapeutic conditions and outcome has not been found across treatment modalities (Orlinsky & Howard, 1986). Furthermore, Rogers only examined the therapist’s contribution to the alliance.
Some reviews, however (e.g., Gelso & Carter, 1985), concluded that Rogers’ therapist-offered conditions only account for part of the complex and, at times, interactive factors that contribute to positive outcome (Horvath & Symonds, 1991).
The positive role of alliance on outcome has also been found within behavioural therapy (e.g., Ryan & Gizynski, 1971; Keijsers, Schaap, Hoogduin, & Peters, 1991). The role that the alliance plays is conceptualized somewhat differently in behaviour therapy. The therapeutic relationship is regarded as an important factor that increases the likelihood that the client will listen to the therapist and allow for the implementation of behavioural techniques (Sweet, 1984).
Thus, the more effective a therapist is at fostering a positive therapeutic alliance, the greater the probability that he/she will be successful in changing a client’s behaviour. As an illustration,
Morris and Suckerman (1974) found that therapist warmth was significantly related to outcome in systematic desensitization for snake-phobic clients. Significant effects between alliance and outcome have also been found in exposure and response prevention treatment for Obsessive
Compulsive Disorder (Hoogduin, deHaan, & Schaap, 1989), in vivo exposure treatment for agoraphobia (Williams & Chambless, 1990), behaviour therapy for problem drinking (Miller et al., 1980) and behavioural treatment for marital therapy (Holtzworth-Munroe, Jacobson, DeKlyen, & Whisman, 1989).
Meta-Analytic Reviews of Alliance and Outcome
In a meta-analytic review of 40 studies, Orlinsky and Howard (1986) reported that therapeutic empathy was significantly correlated with outcome. Furthermore, in a review of 24 studies examining alliance and outcome, Horvath and Symonds (1991) found an average effect size of r = 0.26 across treatments. In more recent reviews, the effect size for alliance was found to be r = 0.21 across 89 studies (Norcross, 2001a) and r = 0.22 in another review that measured the alliance across 79 studies (Martin, Garske, & Davis, 2000). Although most studies have found a positive relationship between alliance and outcome across treatments (e.g., Ackerman & Hilsenroth, 2003; Horvath & Symonds, 1991; Krupnick et al., 1996), some have not (e.g, Krupnick et al., 1994).
Qualification on General Results
Some studies have shown that the therapeutic relationship is correlated more highly with client outcome than are specified treatment interventions (Lambert 8c Barley, 2001; Norcross,
2001a). However, whether the alliance is a curative factor, per se, currently remains unknown.
Gaston (1990) provided three hypotheses regarding the role that the alliance might play across therapies: the alliance could be directly and, thus, causally related to outcome, it may be a mediator of outcome, or it may interact with other variables to effect positive change. For example, Anthony, Ogrodniczuk, Piper, and McCallum (2003) found that the therapeutic alliance was a mediator of clients’ expectancy of improvement and outcome in short-term individual therapy. Specifically, they found that the alliance directly accounted for one-third of the effect of expectancy on outcome.
In terms of an interactive effect with other variables, Norcross (2001a) found that the alliance had effects that were dependent upon clients’ diagnoses. For example, for Generalized
Anxiety Disorder (GAD) and Obsessive Compulsive Disorder (OCD), the treatment modality effect size on outcome was much greater than the alliance’s, but for depression, the alliance was more powerful. In contrast, the alliance might actually be less important or even less beneficial for some diagnoses. For example, a negative correlation has been found between alliance and outcome for paranoid clients (Norcross, 2001a). Thus, the therapeutic relationship probably exerts a more direct causal impact on some disorders and in some treatment modalities than in others (Norcross, 2001b).
The relation of alliance to outcome has been found across differing perspectives including those of therapist, client, and external observers. However, the majority of findings indicate that alliance is the strongest predictor of outcome when it is assessed from the client’s perspective (Horvath & Luborsky, 1993; Orlinsky & Howard, 1986; Henry, Strupp, Schacht, & Gaston, 1994). Horvath and Symonds (1991) found that the effect size for the relationship between alliance and outcome was r = 0.21 when clients rated the alliance and r – 0.17 when therapists rated the alliance, and the difference between effect sizes was not statistically significant. Safran and Wallner (1991) found that alliance was strongly related to outcome in short-term cognitive therapy when assessed by clients. On the other hand, when therapists rated the quality of the alliance, a moderate relation to outcome was found. In contrast, Krupnick et al (1996) found that ratings of patients’ contribution to the alliance contributed significantly to outcome in CBT, however, ratings of therapists’ contribution to the alliance to outcome were not significantly related to outcome. Although empirical findings generally indicate that clients’ ratings of the therapeutic alliance are more likely to predict clinical recovery therapists ratings (Burns & Auerbach, 1996), it is impossible at this time to draw definitive conclusions about the alliance rater effects on outcome in CBT. The meta-analytic reviews that have examined rater effects on alliance and outcome thus far have not included many CBT studies in their data set.
For example, in a review of 90 studies published between 1976 and 2000 (Horvath & Bedi,
2002) included only 5 cognitive therapy studies. Horvath and Symonds (1991) included only 2 cognitive therapy studies in their review of the alliance across 20 studies. Thus, a meta-analysis examining the effects of moderator variables (including type of alliance rater), within CBT specifically, and with a larger sample size, is warranted at this time.
The role of the alliance in cognitive therapy has been, for the most part, largely misunderstood. CBT has historically been inaccurately conceptualized as a mechanistic and structured treatment, in which the therapeutic relationship is downplayed, or ignored. Although the alliance has not been as frequently examined in the cognitive behavioural psychotherapies as in other modalities of treatment, theoretically, its importance has always been underscored. One notable exception is Ellis (1962) who thought that the alliance is neither necessary nor sufficient to produce positive change, and could in fact be counterproductive.
A central principle of cognitive therapy is that therapy is a collaborative effort between therapist and patient (Beck et al., 1985). This principle emphasizes that cognitive therapy is based on a collaborative approach whereby the alliance between the client and therapist is used to help the client arrive at solutions to his/her problems. The most skilled cognitive therapists are those who establish a mutually respective relationship with their clients, in which they combine genuine, empathic understanding of their problems, with competence in CBT treatment delivery (Dobson & Shaw, 1993).
Beck et al. (1985) listed and discussed a series often principles upon which cognitive therapy is based. Principle three states: “A sound therapeutic relationship is a necessary condition for effective cognitive therapy” (pp.167). Beck et al. noted that, without a warm therapeutic relationship, the techniques and procedures of cognitive therapy are unlikely to work.
Thus, the relationship is the medium through which the “work” in cognitive therapy can take place. Cognitive therapists must, however, be able to clearly define the nature of the alliance. It should be warm and empathic such that the client feels safe. It must not be overly smothering, such that objective analysis by the therapist can occur, and such that the therapist does not become overly entangled in a client’s hopelessness (Dobson & Shaw, 1993).
There are no empirical data to suggest that CBT therapists are cold and mechanical
(Keijsers, Schaap, & Hoogduin, 2000). For example, Marmar, Gaston, Gallagher, and Thompson
(1989) investigated the effects of alliance on outcome in cognitive, behavioral, and brief psychodynamic therapy for depressed elderly patients and found a significant within treatment relation between alliance and outcome in cognitive and behavioural therapies only. Similarly, when compared with other treatments, higher or similar alliance ratings have been reported in CBT (Raue, Castonguay, & Goldfried, 1993; Raue, Goldfried, & Barkham, 1997; Salvio, Beutler, Wood, & Engle, 1992).
CBT differs from client-centred approaches in its conceptualization of alliance in one major way. Although the client-centered treatment frameworks endorse a positive alliance as a necessary and sufficient condition of therapy, CBT approaches typically maintain that a good therapeutic relationship is necessary but is not a sufficient condition for positive change (Burns & Auerbach, 1996). The alliance may serve as a precondition for the application of cognitive and behavioural change mechanisms. More recently in the CBT literature, in certain contexts, the alliance has been regarded as a possible change mechanism in and of itself (e.g., Norcross, 2002; Safran & Segal, 1990). Specifically, ruptures in the therapeutic alliance can be used as opportunities to examine a client’s maladaptive thinking patterns and beliefs. Safran and Segal (1990) have underscored the importance of using interpersonal processes in cognitive therapy, and have helped clients develop more adaptive cognitive schemas within the context of the therapeutic relationship. Safran and Wallner (1991) have also developed a variant of cognitive therapy that places specific emphasis on the therapeutic relationship as an essential variable related to positive change.
Although the aforementioned developments are interesting, the specific role that the alliance plays in effecting positive outcomes in CBT is still not clear. The magnitude of the alliance’s effect on outcome and the effect’s variability across different disorders is also unclear at this time. On the other hand, the alliance might allow for a client’s acceptance of a CBT therapist’s techniques, such as cognitive interpretations and restructuring. As such, the alliance might be reflective of positive outcome, rather than causally related. For example, Burns and Nolen-Hoeksema (1992) found, via a structural equation model, that depressed patients treated by CBT rated their therapists as more caring and empathic as they were improving. Thus, depression severity evoked a reciprocal effect on empathy, and the magnitude of the effect was moderate to large given that each 1 point increase on the Empathy Scale (ES; Persons & Burns,1985) used in the study led to a decrease of 1.37 points on The Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) at post-treatment. Similarly, Raue, Castonguay, and Goldfried (1993) found that clients who were more anxious/depressed were rated as having a lower level of alliance (by external observers) than clients who were less symptomatic. Raue, Goldfried, and Barkham (1997) found that alliance ratings were higher in “high impact” sessions of CBT than those for “low impact.” In addition, DeRubeis and Feeley (1990) found that alliance scores were predicted by, rather than predictive of, positive therapeutic change in CBT for depression. Thus, a positive alliance may also be an indicator that things are going well in therapy overall.
For the most part, the relation between alliance and outcome in CBT has been deemed to be moderate to large (Muran et al., 1995; Safran & Wallner, 1991; Wright & Davis, 1994). Nevertheless, some studies of CBT have not found a relationship between alliance and outcome.
For example, Wilson, Fairburn, Agras, Walsh, and Kraemer (2002) did not find a relationship between alliance and outcome for CBT for bulimia nervosa. However, the results were based on two single alliance scores for each patient, one gathered early in treatment, and the other at mid-treatment.
Thus, the effects of alliance were not assessed over time. Others have found an interactive effect between alliance and outcome with the point at which it was measured. For example, DeRubeis and Feeley (1990) found that alliance was related to outcome only later in treatment for cognitive therapy for depression, and not in the initial stages of therapy.
Although meta-analyses of the alliance literature have been conducted (Horvath &
Symonds, 1991; Martin et al., 2000), a review of the alliance’s effect specifically within CBT has not. Given that, overall, the alliance seems to be often moderately related to outcome, and given that its role within CBT has not been extensively empirically examined, a meta-analysis examining at the effects of alliance within CBT is necessary. Furthermore, given that most studies, including those within CBT, have found a positive relationship between alliance and outcome, a simple review of the literature is not sufficient. A more in depth-analysis (via moderator analyses) is required in order to ascertain better the exact mechanism by which the alliance exerts its effects on outcome within CBT.