Select Page

Therapists and clients: who brings what to the alliance?

So important is the therapeutic relationship to clients, it would appear, that they would be prepared to accept a treatment with a lower recovery rate to receive it.
Further, clients’ ratings of the alliance early on in therapy strongly predict subsequent improvement. But what do therapists and clients each bring to the therapeutic alliance, and who makes a bigger difference?

In the last blog, I outlined how the alliance, based on the work of Ed Bordin [i],  is comprised of three elements; agreement on a) the therapeutic goals and b) the tasks that make up therapy, as well as c) the bond between therapist and client. Further on, I outlined the critical contribution the alliance makes to therapy outcomes, showing that the alliance accounts for seven times more of the variance in therapy outcomes than the therapy model.

In this blog, I want to further underline the link between alliance and outcomes and look briefly at the relationship between alliance and client dropout. I’ll also explore the relative contributions that client and therapist make to the alliance-outcome link and how, as therapists, we vary in our ability to form alliances.

I’ll start, however, with a piece of research which outlines just how much clients value elements of the alliance.


Relationship or outcomes: what do clients value more?

There’s little doubt that given a choice, clients will choose an effective therapy delivered within a positive working relationship. The two are not mutually exclusive. As part of a research study into client preferences by Swift and Callaghan [ii], however, adult clients presenting for therapy at a university based psychology department in the US were asked to indicate how much less effective a therapy they would accept to ensure the presence of one of four common therapy factors. The factors in question were:

  1. A therapist with whom they could develop a good, positive working relationship
  2. A therapist described as warm, empathetic, and accepting
  3. A therapist who has completed many years of schooling and clinical experience
  4. A therapist who listens and allows you to do more of the talking in session

The image above is an example of the questions that were asked to determine client preferences. Starting from an intervention with a 70% recovery rate, clients indicated for each factor how much lower a recovery rate they would accept to ensure the presence of that factor. To quote the study:

‘It was found that clients were willing to receive a significantly less effective intervention by (a) 49% to ensure that their therapist was empathetic and accepting, (b) 38% to ensure that a therapeutic relationship could be developed, (c) 35% to ensure that sessions were client directed, and (d) 26% to ensure that their therapist was more experienced.’

It is notable that the factors which clients were prepared to make the greatest sacrifices for, namely a warm, empathetic, and accepting therapist, and a good, positive working relationship, are factors central to the therapeutic bond component of the alliance.

Early alliance strength predicts subsequent improvement rates

What is the relationship between the strength of the alliance and client outcomes? Does improvement in symptoms build a strong alliance, or does a strong alliance build better outcomes?

Research to date has provided evidence for both propositions, but a study published in 2014 by Falkenström et al [iii] has provided further strong support for the idea that a strong early alliance is a prerequisite for subsequent symptom improvement. The study, based on 719 clients seen by 69 therapists in Swedish primary care, measured the strength of the alliance at the third session using the Working Alliance Inventory (WAI), and clients’ levels of distress using the CORE Outcome Measure (CORE-OM) at each session. Among its key findings were the following:

  1. Greater improvement during the first three sessions predicted higher ratings of working alliance at session three.
  2. Higher symptom levels at the start of treatment predicted a poorer working alliance at session three.
  3. Working alliance, measured at session three, significantly predicted subsequent symptom change from across the rest of the treatment, even while controlling for prior symptom improvement

This study seems to lend support to what is known as a reciprocal influence model. In the authors’ words:

‘According to this model, the relationship between alliance and outcome is bi-directional; symptom change influences later alliance, and alliance influences later symptom change.’

Weaker alliances lead to higher rates of drop out

Given that the alliance is comprised of goal, task and bond elements, it would be not be surprising if deficits in any one of these three areas in some way undermined the client’s faith in either the proposed therapy approach, or the therapist him or herself.

While there has been relatively less research studying the link between alliance and dropout, than that looking at the alliance-outcome link, a meta-analysis from 2010 found evidence of a clear relationship between alliance and client dropout. The meta-analysis of 11 studies by Sharf et al [iv], found a moderately strong relationship between dropout and therapeutic alliance (d = .55), with clients with weaker alliances being more likely to drop out of therapy.


What do therapists and clients each contribute to the alliance?

Who, or indeed what, contributes to variations in the alliance that may be related to outcome? DeRubeis et al (2005) [v] suggested four possible sources of variance:  the therapist, the client, the interaction between the two, and prior symptom improvement.

A study by Scott Baldwin and colleagues published in 2007 sought to clarify this question. [vi]  Based on a study of 331 clients seen by 80 therapists, the study concluded that neither prior symptom change, the client, nor the interaction between therapist and client significantly contributed to variations in the alliance that were related to outcome. By contrast, the authors found that it was therapist variability in the alliance that was significantly correlated with outcome. Therapists who generally had higher (client rated) alliance ratings also had better outcomes than those therapists who had generally lower alliance ratings. Between therapists, the final outcome was related to the alliance, whereas within therapists, it was not.

The figure above illustrates this point. The measures used in the study were the client completed WAI and the Outcome Questionnaire 45 (OQ-45). Based on the study data, the prediction was that clients of a therapist one standard deviation above (or below) the mean WAI score for therapists would, on average, have an end of therapy measure score 5.8 points higher (or lower) than clients of a therapist with a WIA score on the average for therapists.

  1. Clients value the bond elements of the alliance, so much so that they may be willing to accept a less effective intervention to ensure that their therapist is empathetic and accepting, and that a therapeutic relationship can be developed.
  2. The two are not mutually exclusive, we can aim to build a strong alliance and good outcomes!
  3. The strength of the alliance early in therapy strongly predicts subsequent improvement (or indeed lack of it), and is also moderately predictive of whether clients will remain engaged or drop out.
  4. Neither prior symptom change, client, nor interaction between therapist and client significantly contribute to variations in the alliance that are related to outcome. It is therapist variability in the alliance that is most significantly linked with outcome.

What’s next?

If the alliance is so central to clients’ engagement in therapy, and to their outcomes, how can we be sure that the strength of our alliances (especially in the critical early stages) is sufficient? Also, given that it is therapist variability in the alliance that is most significantly linked to outcome, what do we need to be doing to ensure its robustness?

To answer these questions, in the next (and last) blog in this series I will be exploring:

  • How we can get an objective measure of the alliance.
  • What we should be doing more (and less) of, to ensure that we are routinely building strong alliances with our clients.


[i] Bordin E. 1979. The generalizability of the psychoanalytic concept of the working alliance. Psychotherapy: Theory, Research, and Practice, 16, 252–260.

[ii] Swift J & Callahan J. 2010. A Comparison of Client Preferences for Intervention Empirical Support Versus Common Therapy Variables. Journal of Clinical Psychology, Vol. 66(12), 1217—1231

[iii] Falkenström F et al. 2014. Working alliance predicts psychotherapy outcome even while controlling for prior symptom improvement. Psychotherapy Research, 2014 Vol. 24, No. 2, 146–159,

[iv] Sharf J et al. 2010. Dropout and therapeutic alliance: A meta-analysis of adult individual psychotherapy. Psychotherapy Theory, Research, Practice, Training, Vol. 47, No. 4, 637–645

[v] DeRubeis RJ et al. 2005. A Conceptual and Methodological Analysis of the Nonspecifics Argument. Clinical Psychology: Science and Practice. Vol 12, no. 2 Summer 2005

[vi] Baldwin et al. 2007. Untangling the Alliance–Outcome Correlation: Exploring the Relative Importance of Therapist and Patient Variability in the Alliance. Journal of Consulting and Clinical Psychology. Vol. 75, No. 6, 842–852