Exploration of the Patient-Therapist Relationship in Psychotherapy

Research supports the value of psychotherapy. However, when it comes to understanding which technical ingredients are most important in different treatments, our knowledge is limited. Researchers from many domains hold the position that nonspecific common factors, such as patient expectancy, alliance, and therapist empathy, account for almost all change in psychotherapy. In this article the empirical evidence for the effects of one of the specific techniques in dynamic psychotherapy, transference work, is summarized.

In the dictionary from the American Psychoanalytic Association, Psychoanalytic Terms and Concepts (1), transference is defined as “the displacement of patterns of feelings, thoughts, and behaviour, originally experienced in relation to significant figures during childhood, onto a person involved in a current interpersonal relationship.” Not only clinical theory but also empirical evidence indicates that transference phenomena do exist in all relationships (2) and all schools of psychotherapy (3).

In psychoanalysis and psychodynamic psychotherapy, exploration and interpretation of the content of transference, transference work, is seen as central to the therapeutic process. Influential theorists (4–6) have argued that interventions focused on conflicts and interpersonal patterns in the patients’ relationships outside therapy, extratransference work, often invite intellectualized speculation. Transference interventions, on the other hand, are thought to be affectively immediate and compelling when they are accurate and may illuminate the “true” nature of problems in the patient’s relationships outside of therapy (4–9). Particularly for patients with more severe interpersonal difficulties, there is a strong probability that their problems will be activated during treatment, in the form of difficulties with collaboration, avoidance, dependency, and a higher dropout rate (2–7). Therefore, working with the patient-therapist relationship may be a productive approach to enhancing self-awareness and understanding with regard to interpersonal difficulties. The ubiquity of transference phenomena has implications for therapy regardless of orientation (2, 3).

There is an enormous theoretical and clinical literature on the transference concept and transference phenomena. However, in this literature it is difficult to find consensus on exactly what transference is, what a transference interpretation sounds like, and when such interpretations are effective (4, 5, 7–13). The notion of transference as pure enactments of early (family) relationships has been a matter of much debate in the psychodynamic literature (14–18). In recent decades relational theorists have focused more on transference as co-constructed by patient and therapist (14–17). Patients may also elicit certain responses from the therapist that fit their interpersonal expectations or fears (countertransference reactions) (12, 17, 19). Clinical theory emphasizes that the therapist is not in the privileged position of “determining the truth,” especially about causal origins of the patient’s experience and personality functioning (5–8, 12–14, 19). There is substantial continuity in personality functioning from childhood to adulthood (20, 21), but this may to a large extent be based on genetic or temperamental factors. Research in behavior genetics indicates that the heritability of borderline personality disorder, for example, may be in the range of 0.40 to 0.60. The estimated shared-in-families (familial) environmental effects are generally small (22). However, research in epigenetics (23), neurobiology (24, 25), social neuroscience (26), cognitive psychology (2, 18), and longitudinal epidemiological and clinical studies (21, 25, 27–30) has documented long-term effects of childhood environmental factors, although parsing out genetic effects on life events and parenting styles is difficult. It is also relatively well documented that genetic and biological factors can be activated, impeded, or transformed throughout life by a multitude of environmental factors and interpersonal experiences (2, 18, 31, 32), including psychotherapy (33–36).

Empirical studies of transference patterns taken together with developments in cognitive neuroscience (2, 37–40) suggest that patients’ characteristic patterns of thinking and relating to others show an overlap between relationships outside therapy and the therapeutic relationship. Patients often show several or even many transference reactions over the course of treatment. The real characteristics of the therapist inevitably influence the patient’s reactions and perceptions of the therapist (2, 3, 37–40).

We all bring something to a new encounter (we transfer), but since adult interpersonal functioning and behavior are determined by a multitude of genetic, biological, and interpersonal factors, I find it useful to adopt a broad definition of transference(s) and transference interpretation (5, 15). As such, transference(s) is defined as the patient's patterns of feelings, thoughts, perceptions, and behavior that emerge within the therapeutic relationship and reflect aspects of the patient's personality functioning (regardless of the developmental origin of these patterns). I use the terms “transference work” and “transference interventions” instead of “transference interpretation,” to emphasize a broader definition. Transference work is any therapist intervention that points out, refers to, wonders about, or explains the patient’s experience of the therapist and the ongoing patient-therapist interaction (4). There are important real aspects to the patient’s experience (2–4), but the targets of transference interventions are often problematic aspects of the patient-therapist relationship, such as biased cognitions and inappropriate or strong affects (2, 4, 7, 9, 13).

Transference work can be operationalized as five categories of interventions with explicit reference to the therapist (R. Ulberg, S. Amlo, P. Høglend, unpublished scales, 2012):

These categories of transference work combine the constructivist or relational view of transference (12, 14, 15) with the traditional construct of transference in relation to historical (genetic) origins (1, 8, 15, 16). Using data from 12 studies, I found (41) that that transference work interventions can be operationalized and reliably rated by trained judges from audiotaped sessions. Data from these studies indicate that low levels of transference interventions seem to be on average 0–2 interventions per session, a moderate level is 3 or 4, and a high level is 5 or 6 (41).

More than 30 studies have been published that attempt to provide empirical evidence on the effects of transference work in dynamic psychotherapy. The studies are too heterogeneous to be analyzed by means of meta-analytic technique, so a brief summary of the studies will follow. The focus will be on treatment outcome rather than in-treatment effects, since the impact on clinical practice is more speculative when outcome is not part of the empirical investigation.

Naturalistic Studies

One of the earliest studies to examine transference interpretation was the Menninger project. Kernberg and colleagues reported that patients with low ego strength did best in dynamic psychotherapy focusing on transference interpretation, compared with psychoanalysis or supportive psychotherapy. Patients with high ego strength did equally well in the different treatments (42). Four early studies addressed, as a separate category, interpretations that linked the patient’s feelings toward the therapist with the patient’s feelings toward his or her parents (genetic transference interpretations). These studies reported positive correlations between the frequency of this type of interpretation and nonsymptom outcome, such as psychodynamic and interpersonal functioning (43–46). A high number of correlations between other categories of therapist interventions and many different outcome measures were computed in each of these studies. Thus, the findings are at best tentative.

Uncategorized Transference Interventions

Later correlational studies have not distinguished between different categories of transference interventions. Eight studies reported that a higher frequency of transference interventions was associated with less favorable outcome (47–54). This led researchers and empirically informed clinicians to view transference work as ineffective or even harmful (55). These studies suggest that a low frequency of transference interventions might be useful, whereas a higher frequency may lead to negative treatment effects. Only two studies have reported no association between transference work and outcome (56, 57). A limitation in these studies is that different categories of transference interventions, similar to what I proposed earlier, were not identified. Another methodological problem is the issue of therapist responsiveness (58). Findings from several studies indicate that therapists increase their use of transference interventions with more difficult patients in the context of a negative or hostile therapeutic process (48, 52, 53, 55). This may partly explain the inverse correlations between frequency of transference work and outcome.

Therapeutic Immediacy

One recent study found that greater early treatment focus on the patient-therapist relationship, termed “therapeutic immediacy,” was positively associated with improved interpersonal functioning, especially among patients with more severe interpersonal problems (59). The average level of immediacy interventions in this study was found to be low to moderate. Conversely, Hill and colleagues reported no significant association between therapist immediacy and posttreatment interpersonal functioning (60). Therapist immediacy seems to involve more therapist strategic self-disclosure of emotions felt during the session (countertransference disclosure) than traditional transference work, as demonstrated in the following statements (59):

A summary of the naturalistic studies of transference work and outcome is shown in Table 1.

Experimental Studies

An early study by my colleagues and me, in which the amount of transference work was experimentally manipulated, indicated that therapy with a low frequency of transference interventions (average, 1 per session) had a more favorable outcome than therapy with a high frequency of these interventions (average, 6 per session), suggesting a causal negative effect of frequent use of transference interventions (61). The negative effect of frequent transference interventions tended to be stronger among patients with personality disorders (62).

Transference-Focused Psychotherapy

Two randomized controlled trials have examined the efficacy of transference-focused psychotherapy, an outpatient treatment designed for patients with personality disorders, particularly borderline personality disorder. The focus is on understanding how patients relate to their therapists, with the idea that this may improve the patient’s relationships outside of therapy (6, 7). In one clinical trial, 90 patients (93% female) were randomly assigned to transference-focused psychotherapy, dialectical behavior therapy, and psychodynamic supportive psychotherapy. There were no significant differences among the three treatments on the primary outcome variables (63). However, transference-focused psychotherapy was superior with regard to nonsymptom outcome variables, such as metacognitive and social-cognitive capacities (attachment patterns and reflective function) (64). Another study of transference-focused therapy included 104 female patients with borderline personality disorder. The patients were randomly assigned to transference-focused therapy and treatment by experienced therapists in the community. The authors reported that transference-focused therapy was superior in reducing borderline symptoms and interpersonal problems but not anxiety and depression (65). The comparison groups in the two studies did not focus on transference work, although no formal assessment of treatment integrity was reported.

Following is an illustrative example of transference-focused therapy (F. Yeomans, E. Caligor, personal communication, Dec. 11, 2013):

Mentalization-Based Treatment

Two studies of outpatient mentalization-based treatment, which has a strong individual therapy component in addition to group therapy, are included in this overview. In mentalization-based treatment, transference phenomena in the here and now are explored, but interpretation of repetitive interpersonal patterns is deemphasized. Strategic use of self-disclosure is encouraged (13, 66–68). In a study of 134 adult outpatients (80% female) with borderline personality disorders plus other comorbid personality disorders, patients were randomly allocated to 18 months of mentalization-based treatment or to structured clinical management, an individual supportive and problem-oriented therapy, which also includes group sessions. Mentalization-based treatment was superior with regard to both symptoms and personality functioning (68). In a study of 80 adolescents (85% female) with self-harm and comorbid depression, patients were randomly assigned to 1 year of mentalization-based treatment or to enhanced treatment as usual. Mentalization-based therapy was more effective in reducing self-harm and depression. It is interesting that both improved mentalization (reflective function) and reduced attachment avoidance explained (mediated) the treatment effects (69).

The following is an excerpt from session 25 of a patient’s treatment (67):

Alliance-Focused Treatment

Alliance-focused treatment is a form of brief relational therapy based on principles from relational psychoanalysis (14). Its primary task is to track ruptures in the therapeutic alliance and encourage the patient, together with the therapist, to explore what is going on in the therapeutic relationship. Alliance-focused therapy also includes more therapist self-disclosure, and therapists are specifically encouraged to acknowledge responsibility for the therapist's own contribution to alliance ruptures and misunderstandings (70). In a study conducted by Muran and colleagues, 128 patients with cluster C personality disorders were randomly assigned to 30 sessions of brief relational therapy, short-term dynamic psychotherapy, and cognitive-behavioral therapy. The three treatments were equally effective, but brief relational therapy had fewer dropouts (71). In a pilot study, Crits-Christoph and colleagues (72) found support for training therapists in a 16-session alliance-fostering therapy. The treatment focus was on agreeing on goals, identifying alliance ruptures, and exploring underlying feelings in the patient’s relationship to the therapist. The training produced small, nonsignificant improvements in symptoms but larger improvements in quality of life. Safran et al. provide the following example (73):

There are several limitations in these experimental studies. Their generalizability to everyday clinical practice may be limited. There are a number of other technique variables and process variables beyond the ones measured in these studies that may influence the effects of transference work. Furthermore, the studies did not report data on the frequency of transference interventions. Also, different categories of transference interventions were not identified. The comparison treatments in these studies are bona fide psychotherapies, but researcher allegiance effects cannot be ruled out .

A summary of the experimental studies of transference work and outcome is shown in Table 2.

First Experimental Study of Transference Work (FEST)

FEST was conducted by my colleagues and me and is, to our knowledge, the only dismantling randomized clinical trial specifically designed to investigate long-term effects of transference work in dynamic psychotherapy. We randomly assigned 100 patients to 1 year of dynamic psychotherapy with a low to moderate level of transference work or to the same type of therapy without transference work. The same therapists administered both treatments after extensive training. Treatment integrity was documented with ratings of more than 450 full sessions. The only component that differed between the two treatments was use of a low to moderate frequency of transference work interventions. Thus, the design makes it possible to study causal effects of transference work.

There was no overall effect of transference work. However, patients with a low quality of object relations benefited significantly more from therapy with transference work compared to therapy without transference work (74). This effect was sustained during a 3-year follow-up period (75). Patients with mature relationships and greater psychological resources benefited equally well from both treatments. Furthermore, female patients responded significantly better than men to therapy with transference work (77, 78). Among the 46 patients with one or more personality disorders, 17 of 23 patients (74%) no longer met diagnostic criteria for any personality disorder in the transference group, versus 10 of 23 patients (43%) in the comparison group (76). The dropout rate was 0% in the transference group and 22% in the comparison group. Patients who did not receive transference work had about four times more additional mental health specialist treatment during the 3-year follow-up period, compared with patients who received transference work. All the therapists in this study had extensive experience and were specifically trained to deliver the two treatments, which limits generalizability to ordinary clinical practice.

The long-term effect of transference work among patients with low-quality object relations was mediated (explained) by increased gain of insight during therapy (79). Several studies suggest that changes in insight or self-understanding are specific to dynamic psychotherapy and are not associated with other treatments, such as cognitive-behavioral therapy or antidepressant medication (80–82). FEST extended this work by linking the use of specific techniques to gains in insight and subsequent improvement in interpersonal functioning (79). These findings are consistent with the clinical theory that insight may be a specific mechanism of change in dynamic therapy. It should be noted, however, that the association between insight and outcome cannot be experimentally controlled. The true causal mechanism of change could be some unknown variable correlated with insight. This is an inevitable limitation, to date, in mediator studies.

The following vignette, from session 29 of treatment, illustrates how work within the transference may promote that insight (S. Amlo, R. Ulberg, P. Høglend, 2012):

My colleagues and I have also explored the long-term effects of transference work in the context of therapeutic alliance and patients’ quality of object relations (83). Alliance was a positive predictor of outcome, independent of transference work. However, the specific effect of transference work on psychodynamic functioning was most positive for patients with low quality of object relations within the context of a weak therapeutic alliance. For patients with more mature object relations and high alliance, we observed a negative effect of transference work. These findings contrast with what is maintained in mainstream clinical theory, that only patients with a solid therapeutic alliance gain greater benefit from transference work. One way of understanding the findings is that more disturbed patients have trouble establishing a trusting therapeutic relationship. A thoughtful exploration of transference issues and alliance ruptures may disprove negative interpersonal expectations and strengthen alliance, self-understanding, and ultimately outcome. On the other hand, when treatment is going well for high-functioning patients, a focus on the therapeutic relationship can be awkward or seem like an indulgence for the therapist.

My colleagues and I also reported that therapist parental countertransference interacted with patients’ personality disorder pathology in predicting the long-term outcome of transference work (84). When therapists' parental feelings were more intense, the effect of transference work was very positive for patients with high levels of personality disorder pathology, but the effect was negative for patients with low levels of personality disorder pathology. One might speculate that a more parental role is experienced as intrusive, infantilizing, or unnecessary by patients with low levels of personality pathology, whereas for patients with more severe personality pathology a parental role in the therapist provides a source of support, nurturing, and empathic validation that is felt as missing in other relationships.

The complex moderation effects reported in the preceding may help in our understanding of some of the individual variability in response to transference work. However, until the findings are replicated in future studies with larger study groups, they must be considered preliminary.

Classical (Genetic) Transference Interpretations

In earlier psychoanalytic theory, transference-parent linking interpretations were seen as a necessary cornerstone of technique. However, a number of studies providing data on “dosage” levels of transference work (43–46, 59, 60, 85–88) indicate that transference-parent linking interpretations are used very seldom, if at all. Over time the use of classical linking interpretations seems to have fallen out of fashion, whereas use of self-disclosure seems to have gained in popularity. The phenomena of transference work in different studies may be more similar in practice than different clinical theories indicate. The clinical vignettes also illustrate the many similarities. Furthermore, the vignettes illustrate how transference phenomena should be explored in a spirit of inquisitive curiosity, inviting the patient to speculate together with the therapist.

Transference Work in Nondynamic Treatments

Gelso and Bhatia (3) reviewed 16 studies on nondynamic therapies. They described how transference influences nondynamic treatments and also often interferes with the treatment process in a negative way. Nondynamic therapists are able to assess transference phenomena when asked about it. However, research that examines how transferences of different types may be responded to in nondynamic therapies is very limited.

Castonguay and colleagues developed integrative cognitive therapy, which integrates a focus on alliance ruptures into cognitive therapy. In a pilot study, the results of integrative cognitive therapy compared favorably to previous findings from cognitive therapy (89). In a small randomized clinical trial, comparing integrative cognitive therapy to cognitive therapy, the integrative treatment produced greater improvement in depression and global symptoms (90).

In a large-scale study of the cognitive-behavioral analysis system of psychotherapy for patients with severe chronic depression, 40 psychotherapists reported on the relative degree to which they emphasized each of five listed intervention categories. These categories were cognitive interventions, behavioral interventions, agreement on goals, exploration of the patient-therapist relationship, and focus on developmental origins of symptoms. The only positive predictor of outcome was the overall degree of emphasis placed on discussing the patient-therapist relationship (91).

Functional analytic psychotherapy is a contemporary cognitive-behavioral therapy that focuses on how the patient’s interpersonal problems are played out in the therapy relationship (clinically relevant in-session behavior). In a pilot study (92), greater use of the patient-therapist relationship was associated with more favorable outcomes.

Future Research

Ideally, we need research that links specific transference interventions to outcome while controlling for other variables. Treatment component studies should be prioritized, especially with regard to the value of transference-parent linking interpretations and of therapist self-disclosure. If we can isolate specific components that are most effective, the search for mechanisms (mediators) may be narrowed. The study of mechanisms, such as insight, reflective function, self-assertion, or social avoidance, may improve basic knowledge.

Intensive process research is needed before we can understand the full impact of transference interventions and how they work. The presence or absence of such interventions, or their frequency, is not sufficient. The timing and quality of specific interventions are difficult to evaluate but may be crucial to how well the intervention is received by the patient (88). Interventions that precede transference work, in addition to those that follow, may carry weight as well. We also need studies of moderators and other contextual variables such as alliance (83) and therapist factors such as countertransference reactions (84), which may influence the effectiveness of transference interventions (84). More research on therapies that integrate transference work into other psychotherapy modalities is urgently needed.

Social processes are a candidate research domain in the Research Domain Criteria (RDoC) framework, based on the National Institute of Mental Health Strategic Plan (93). Social relationships are critical because interpersonal manifestations may be the most central observation of psychopathology, and the influence of poor social relationships on risk for mortality is comparable to, or even more important than, well-established risk factors, such as smoking (94). Research on the components, processes, and mechanisms of transference-based psychotherapies may improve basic knowledge about psychological aspects of interpersonal functioning. In addition, methods from social neuroscience (24, 26, 95), brain imaging (33–35, 40, 96–99), and genetics (23, 30, 36) may help in integrating psychological and biological explanations of interpersonal behavior, in order to ultimately improve treatment of disordered relationships.

Conclusions

The findings summarized in this article strongly suggest that transference interventions are active ingredients. Almost all the studies reviewed have shown significant associations between transference work and interpersonal and psychodynamic change. The average between-groups effect size in the experimental studies listed was 0.8 (large) and favored transference-based treatments.