by Joye C. Anestis
It's been a couple of weeks since I last tackled the topic of common factors of therapy (also known as nonspecific factors), so I thought it was time to bring them back up. Common factors are often the crux of arguments for therapeutic equivalence, the theory that all forms of psychotherapy are equally effective (often known as the dodo bird hypothesis). In this series, I have been evaluating the validity of this argument by examining the available scientific evidence for common factors. Previous posts have concluded that catharsis without problem-solving is hurtful (in fact, expressive-experiential therapies, which involve a great deal of emotional release, are considered potentially harmful psychotherapies) and the research on insight is too sparse to draw any firm conclusions. Today I address the most fervently researched common factor: the therapeutic alliance.
The therapeutic alliance refers to "the degree to which the therapy dyad is engaged in collaborative, purposeful work" [Hatcher & Barends (2006; p. 293) description of Bordin's (1979) description]. The therapeutic relationship and alliance are not synonymous; however, alliance can influence the quality of the relationship between client and therapist. In a relationship in which the alliance is strong, clients will feel safe and comfortable to discuss their concerns and engage in any therapeutic tasks required of them, and the therapist will be open, accepting, nonjudgmental, and professional. Obviously, developing a good alliance is a goal for most therapists, but is the alliance necessary to induce client change? It is perhaps intuitive to assume that a good therapeutic alliance would be related to better client outcome. And in this instance, research supports our common sense notions. In general, research finds that the better the alliance/relationship, the greater the client change (e.g., Horvath & Bedi, 2002; Horvath & Symonds, 1991; Kazdin et al., 2005; Martin et al., 2000). Alliance also seems to be related to engagement and retention in treatment (e.g., Meier et al., 2005).
But there are many caveats to this research on therapeutic alliance (see DeRubeis et al., 2005, for a more thorough discussion of this). In most alliance studies, the timeline between alliance and outcome has not been established. Alliance and outcome are measured at the end of treatment, and not before, during, and after. Because of this, it is plausible that client improvement occurs first and that increases the quality of the alliance, and not the other way around (Kazdin, 2005). Both alliance and outcome can be difficult constructs to measure effectively, and studies vary in how they quantify these variables. Finally, client and/or therapist factors may mediate the relationship between alliance and outcome. In other words, client and/or therapist characteristics may be third variables which explain the relationship between alliance and outcome. Very few studies have investigated this possibility; however, one recent study found evidence that therapist variability, and not client characteristics, may be most important in developing alliance (Baldwin, Wampold, & Imel, 2007).
I can't conclude that the therapeutic alliance functions in an important way across types of therapies. Many schools of psychotherapy (e.g., client-centered therapy) argue that therapeutic relationship factors are the primary source of change. Others (e.g., behavior therapy) acknowledge that a good alliance is necessary in order to motivate clients to engage in the more active aspects of treatment, but alliance itself is not the primary mechanism of change. Additionally, it is still unknown if alliance causes improvement or improvement causes alliance, or if client & therapist variables mediate the alliance-outcome relationship. In fact, although the alliance consistently predicts outcome, alliance only accounts for a small amount of the variance in predicting outcome. For example, a 2002 meta-analysis noted that the average alliance-outcome correlation is .21, indicating that alliance accounts for only 5% of the variance in outcome (Horvath & Bedi, 2002). This means that alliance explains only a small portion of client outcome. Think of it this way: consider the relationship between mood and test performance. Perhaps you were in a bad mood when you took your Abnormal Psychology final. What part of your overall test score could be accounted for by your bad mood? Wouldn't things such as hours spent studying and class attendance account for a larger portion of your performance? In this example, mood is not irrelevant but other factors are more influential. Similarly, although alliance is certainly important for a number of reasons (e.g., client motivation & retention), it is not sufficient by itself to produce client outcomes. Current research is trying to further understand the multitude of mechanisms which result in client change. The alliance is necessary, but it is not sufficient.