Introduction

Much attention in psychotherapy research is devoted to therapeutic alliance (Horvath, 2018, Prusiński, 2022a, with increasing evidence that it is a factor enhancing treatment outcomes (Prusiński, 2022b). The large body of research on alliance includes the issue of patient–therapist consensus on its quality. Studies showed that convergence in the perception of alliance was moderate (Tyron et al., 2007) and changed during treatment (Atzil-Słonim et al., 2015).

In some measure, given the different roles played by the patient and the psychotherapist in a dyad and their different professional, educational, and life experience, the non-coincidence of their therapeutic alliance evaluations seems to be a normal phenomenon (Bachelor, 2013). It is, however, worth knowing how large these differences are, what their direction is, and what factors, if any, determine the discrepancy in alliance ratings. In view of the divergent findings reported in the literature (Hartmann et al., 2015), these issues were addressed in the study presented below. Knowledge of how the alliance is experienced by their patient allows the psychotherapist to properly manage the process, ensuring that cooperation is effective and that the patient does not abandon treatment.

Patient–Therapist Convergence

Research in social psychology indicates that convergence—the process of two people becoming increasingly similar in their emotions, attitudes, perceptions, and experience—is common not only in informal relationships (Coyne et al., 2018). It is adaptive and strengthens attachment, making relationships more satisfying and stable. Convergence continues throughout the lifetime and is present also in the institutional dimension of people’s functioning, during the performance of many social roles (Mikulincer & Shaver, 2018). Psychotherapy is a special kind of social exchange in which a close patient–therapist relationship develops. With the growing recognition of the relational model of psychotherapy (as involving two individuals), it is also becoming common to view the therapeutic relationship as a system of dynamic and interdependent interactions (Nissen-Lie et al., 2021) in which both the patient and the therapist shape the treatment process and in which both individuals have opinions about what is happening between them. Due to the asymmetry inevitable in the dyad, they may perceive their alliance in different ways (Markin et al., 2014).

Some studies showed that it was therapists who underrated the relationship in the dyad, and the correlation between patient- and therapist-rated alliance was only moderate (Hartmann et al., 2015). The findings reported by Hartmann et al. (2015),who explored only a narrow fragment of the relationship (i.e., patient–therapist bond), are in line with an earlier study by Tryon et al. (2007), who also found that therapists significantly underrated the alliance with their patients. Discrepancies were larger in shorter therapies and in the case of treatment for mild and moderate psychopathology (compared to severe psychopathology).

Bachelor (2013), who examined patients with relational, anxiety, and personality disorders, also reported significant, though medium and low, correlations between patients’ and psychotherapists’ ratings. Marmarosh and Kivlighan (2012) found that patient–therapist consensus on alliance quality was higher at the beginning of treatment than in later stages. In their study, however, patients were not divided according to disorder, and the sample included psychodynamic, humanistic, and CBT psychotherapists.

A more recent study, with repeated evaluations of therapeutic alliance between patients suffering from emotional or anxiety disorders and CBT therapists, yielded evidence of therapists’ ratings being systematically lower than patients’ ratings. The discrepancy was more frequently found when disorder symptoms intensified, but the pattern was the opposite of the previously found one, with perspectives converging over the course of psychotherapy (Atzil-Słonim et al., 2015). This pattern was confirmed by Laws et al. (2017), whose research on patients with depressive disorders indicated increasing consensus on alliance quality during treatment. Theoretically, this may reflect the ongoing negotiations in the therapeutic relationship, which gradually becomes more harmonious and comes to be perceived similarly regardless of whether it thrives or breaks down. Cirasola et al. (2021) report ambiguous results on the issue in question. Patients’ and therapists’ ratings coincided when a two-dimensional structure of alliance was assumed and diverged for a one-factor structure.

Convergence in Alliance Ratings and Treatment Outcome

Scholars also investigate patient–therapist convergence in alliance perceptions and its effect on treatment outcomes, interpreting it as a dyadic or relational factor unrelated to the patient’s or the therapist’s individual characteristics (Manne et al., 2012). There are few studies on this issue, however. It is believed that differences in perceived alliance quality, emerging at a certain stage of treatment or gradually over time, may reflect the lack of mutual clarity and understanding of the goals of therapy or low perceived strength of bond, which can hinder the desired change in the patient (Eubanks-Carter et al., 2010). By contrast, convergence in alliance perception may attest to a mutual understanding of the therapy and to the natural progress of the therapeutic relationship, potentially translating into better treatment outcomes. In certain groups, such as patients with depressive disorders, the relational factor is fundamental for improvement in treatment (Constantino et al., 2008). Partial support for convergence in alliance ratings over time being significantly related to treatment outcome can be found in the study by Laws et al. (2017), with a sample of depressive patients.

Discrepancy as a Consequence of Measurement Technique

Another important issue is the way of measuring alliance. The question asked is what period alliance rating should cover (Nissen-Lie et al., 2021). The relationship is dynamic, which means it is influenced by the patient’s and the therapist’s internal representations and earlier experiences concerning the other person. Consequently, it is not a permanent feature of the dyad but changes through constant negotiations regarding the rupture and repair of the relationship (Safran & Muran, 2006). For this reason, an investigation of changes in the convergence of alliance ratings over time is a more valuable test of the hypothesis about consensus on alliance quality than an investigation of this convergence at specific time points.

It is essential that the therapist make an evaluation of the alliance itself rather than the patient’s perspective on and experience of the alliance (Hartmann et al., 2015). This is precisely what the therapist is asked to do in the Working Alliance Inventory (WAI; Horvath, 2018; Prusiński, 2021). The use of external evaluators’ ratings seems inadvisable. Interrater agreement is low, and the ratings are a function of the evaluators’ social roles, failing to offer direct access to patient and therapist perspectives (Halfon et al., 2019). An answer to the difficulties in assessing the actual condition of the alliance is combined measurement (Prusiński, 2021), with two sides assessing their mutual relationship. This possibility is offered by the WAI (Horvath, 2018), which yields an overall score that is more accurate, being a weighted outcome of two ratings. As noted by Muran et al. (2019), when exploring the dynamics of the therapeutic relationship, it seems necessary to focus not only on the processes taking place in the client or the therapist (the intrapersonal level) but also on those taking place between them (the interpersonal level).

The Structure of Alliance

The alliance in psychotherapy was operationalized based on Bordin’s (1979) pantheoretical model. While the therapeutic relationship is understood in a broad sense, as the conditions that the therapist is supposed to create for the patient and as the totality of mutual patient–therapist interactions, links, and relations involved in treatment, which includes the categories of counseling, guiding, or directing the patient (Czabała, 2013), alliance is a narrower and more precise concept. In its conceptualization that Bordin refers to as working alliance, emphasis is placed on agreement and cooperation in the patient–therapist dyad. The patient accepts and actively participates in the intervention agreed on together. Alliance is understood as a triad of goals (the cognitive component: participation in agreement on the important goals of the therapy), tasks (the behavioral component: agreement on important tasks necessary to accomplish the objectives of change), and bonds (the affective component: the strength of the ties in the dyad). Bordin’s model, adopted in the present study and highlighting the need for cooperation and mutual agreement on the psychotherapy treatment, is the most suitable one for research on the discrepancy between the patient’s and the psychotherapist’s perspectives on their relationship.

The Present Study

The key problem addressed in this empirical study was patient–therapist discrepancy in alliance ratings. The literature is ambiguous on this issue, though it more often indicates such a discrepancy and therapists’ tendency to underrate the alliance (Atzil-Słonim et al., 2015; Hartmann et al., 2015; Nissen-Lie et al., 2021). The present study tested convergence in alliance ratings at specific time points and the degree of change in consensus on alliance quality over the course of psychotherapy. It also investigated the link between convergence in alliance ratings and treatment success, hypothesizing a positive relationship between progressive convergence and better outcomes (Eubanks-Carter et al., 2010; Laws et al., 2017).

Four hypotheses were formulated:

H1: There is significant divergence in alliance ratings (i.e., the divergence between the patient’s and the psychotherapist’s perceptions of their relationship at specific time points during the therapeutic process); more precisely, H2: Therapists underrate alliance at specific time points during the therapeutic process;

H3: The discrepancy between alliance ratings decreases over the course of psychotherapy, resulting in a convergence of perspectives (i.e., in more similar perceptions of the relationship in the final phase of psychotherapy);

H4: The convergence of therapeutic alliance ratings over the course of psychotherapy is greater in dyads including patients with better treatment outcomes.

Method

Participants

The longitudinal study included 20 patient–psychotherapist dyads. The sample was composed of one psychotherapist, who was a person other than the researcher in this study, and patients undergoing individual outpatient treatment in private practice.

The patients, assessed by a psychiatrist or psychologist in accordance with ICD-10, suffered from adjustment disorders (F43.2). Patients diagnosed with psychosis, substance or alcohol addictions, personality disorders, dementia, or cognitive disorders were excluded. Women (n = 14) constituted 70% of the sample, and men (n = 6) constituted 30%. Patients were 19 to 54 years old (M = 38.60, SD = 11.71). Most of them had secondary (50%) or higher education (45%); 75% lived in cities with more than 500,000 residents, 70% were married or in relationships.

The psychotherapist, with a master’s degree in psychology and complete postgraduate specialization in psychotherapy, was 37 years old and worked in the systemic modality. He had 7 years of experience in the profession.

The treatment administered to the patients was systemic psychotherapy. Its length varied across dyads, ranging from 13 to 17 months. Four patients were simultaneously treated with pharmacotherapy, which was abandoned after 10 months. The treatment included a standard consultation phase during the first two sessions, in which the therapist conducted a clinical interview to learn about the symptoms, set the goals and the first tasks, and discussed the formal aspects of the treatment plan with the patient.

Sessions, usually lasting 50 min (M = 49.0, SD = 2.05), were held once a week. The length of treatment ranged from 50 to 59 sessions (M = 53.2, SD = 2.46). The therapy included three review sessions devoted to summing up the achievements, discussing the difficulties, and setting further goals. These were Sessions 12, 18, and 38, so that review did not coincide with the measurements performed as part of the study.

Measurement of Variables

Therapeutic Alliance

Therapeutic alliance quality was assessed using the full form of the WAI (Horvath, 2018) as adapted into Polish (Prusiński, 2021). The WAI has two versions, completed by the patient (WAI-PA) and by the psychotherapist (WAI-PT). Each version consists of 36 analogous items operationalizing the construct of working alliance, which the participant rates on a 7-point Likert scale as true or untrue about cooperation in the patient–psychotherapist dyad being evaluated. Measurement reliability was αWAI-PA = 0.95 and αWAI-PT = 0.98. CFA showed that measurement using the WAI was valid (Prusiński, 2021).

Treatment Outcome: Well-Being

Treatment outcome was operationalized as self-reported change in patients’ psychological well-being (Prusiński, 2022b) and assessed using a short version of the Psychological Well-Being Scale (PWBS; Karaś & Cieciuch, 2017). The PWBS consists of 18 items and measures overall long-term psychological well-being. Respondents rate the items on a 6-point scale from 1 = strongly disagree to 6 = strongly agree. Measurement reliability for the total score is αPWBS = 0.83. The PWBS has been found to have good and very good measurement validity (Karaś & Cieciuch, 2017).

Demographic Data and The Formal and Contextual Aspects of The Psychotherapy Process

An elaborate demographic survey (DS) was administered to control for demographic variables (sex, age, education) and variables relevant to the subject matter of the study (psychotherapist’s modality, psychotherapist’s work experience, the number of sessions held, the frequency of meetings, and session length). Two versions of the survey questionnaire were prepared: for the patient and for the psychotherapist. The latter included questions about the formal aspects of psychotherapy and variables concerning its context.

Measurement Procedure and Schedule

To eliminate the potential effect of disorder type on alliance development, the group of patients was selected using non-random expert sampling (only adjustment disorders; Cirasola et al., 2021). The therapist selected was a specialist who treated mainly this type of disorders, which promised a greater number of psychotherapeutic processes for analysis and eliminated potential differences in treatment strategy that would have occurred if the study had included more therapists. The conditions of conducting the psychotherapeutic processes and the study itself were agreed on with the psychotherapist: a uniform method of treatment, understood as working in the same modality with every patient, controlling for potential bias factors (the dynamics of relationship formation, tendencies to develop habits and manifest one’s effectiveness, introducing diversity in interventions, controlling one’s directiveness during the process, attention to burnout) and eliminating them, especially through supervision. The rules of repeated measurement were set out, in accordance with the schedule. Each potential subject was first approached by the psychotherapist (qualification based on diagnosis) and then by the researcher (detailed discussion of the purpose and conditions of the study). The final patient inclusion decision was made by the researcher. The psychotherapist did not participate in data analysis; he did not know the contents of the hypotheses, and neither did the patient. Patients were recruited as they appeared in the psychotherapist’s office. All participants gave their informed consent, and the study was approved by the research ethics board at the author’s university. Measurements always took place after a completed session. Alliance was measured first, well-being assessment followed, and finally respondents completed the demographic survey. Participants received no remuneration for taking part in the study. Alliance was measured throughout the psychotherapy process, until its completion.

Therapeutic alliance was evaluated three times: in the initial, middle, and final phases of psychotherapy (Laws et al., 2017; Mellando et al., 2017), both by patients and by the psychotherapist. As in earlier studies (Hartmann et al., 2015), measurements were not performed during the consultation phase (Sessions 1–2), the review sessions (Sessions 12, 18, and 38), and the final phase of treatment (the last four sessions), when the approaching separation could have affected the evaluation of the process. Additionally, based on Haflon et al.’s (2019) recommendations, the measurement moments chosen (based on signals from the psychotherapist) were neither preceded nor followed by sessions at risk of negative or positive intensive experiences, a breakdown of the alliance, or its deep fluctuation—so that momentary intensification would not interfere with the natural long-term trajectory of the alliance. As a result, the numbers of sessions followed by a measurement were allowed to differ slightly across the processes examined. The procedure was as follows: Measurement I between Sessions 3 and 6 (M = 4.35, SD = 0.99), Measurement II between Sessions 24 and 31 (M = 26.45, SD = 2.16), and Measurement III between Sessions 46 and 55 (M = 49.2, SD = 2.46). Well-being was estimated in the initial and final phases of the psychotherapy process. Two measurements of well-being were performed, which accompanied the first and third measurements of alliance. Well-being was self-reported by the patients.

Results

Discrepancies in Alliance Ratings at Specific Moments of Psychotherapy

To test the first and second hypotheses (H1 and H2), it was necessary to analyze global alliance ratings from patients and psychotherapists. They served as the basis for tests of difference and two-dimensional correlation analyses. The ratings were compared and patient–psychotherapist discrepancies in ratings from three measurement moments were determined. The results of the analyses are presented in Table 1.

Table 1 Patient–therapist discrepancies in alliance quality rating at three measurement moments (longitudinal study, n = 20)
Full size table

The results of difference tests show a discrepancy between patient- and psychotherapist-rated therapeutic alliance at the first two measurement moments: early and medium alliance. In the final measurement, the discrepancy is present but not statistically significant. Therapists rated alliance quality lower than patients did.

The pattern of discrepancies emerging from difference tests is supported by correlations. Patients’ and psychotherapists’ alliance ratings are not significantly correlated until the third measurement. Correlations indicate moderate consensus on alliance quality only at the end of treatment. The results of analyses support the first and second hypotheses, though the convergence detected in the final phase of psychotherapy requires discussion.

Discrepancies in Alliance Ratings Over the Course of Psychotherapy

For the purpose of checking if patient–therapist discrepancies in therapeutic alliance ratings decreased over the course of psychotherapy (H3), an additional, more complex longitudinal analysis of differences was performed, including all measurement moments simultaneously. Based on global alliance ratings from patients (WAI-PA) and psychotherapists (WAI-PT), mean discrepancies (MD) between these ratings and their standard deviations (SDD) were computed for each of the three measurement moments (I, II, III): Measurement I: MD(P–T) = 55.35, SDD(P–T) = 28.99; Measurement II: MD(P–T) = 38.20, SDD(P–T) = 29.72; Measurement III: MD(P–T) = 21.35, SDD(P–T) = 22.77. Next, on their basis, a one-way ANOVA for independent samples was performed.

The analyses yielded statistically significant differences between patient–psychotherapist discrepancies in alliance ratings during psychotherapy: F(2, 57) = 7.34, p < 0.01. Post hoc comparisons using the NIR test for all measurement moments revealed statistically significant differences between all discrepancies computed at the three measurement moments. Patient–psychotherapist discrepancies in alliance ratings decreased over the course of psychotherapy. This indicates a growing convergence of perspectives—increasingly similar perceptions of the condition of the relationship. The above analysis results support H3.

Discrepancies in Alliance Ratings and Treatment Outcome

The fourth hypothesis (H4), postulating a greater increase in alliance quality ratings among patients with better treatment outcomes, was tested in several stages. First, based on two measurements of well-being (initial and final) as the operationalization of treatment outcome and using the reliable change index (RCI = 12.18) to assess the clinical significance of psychotherapy outcomes, each patient was classified into one of two groups: more (n = 6) vs. less (n = 14) successful treatment (Fuertes et al., 2013). Next, for both groups, calculations were performed that yielded mean decrease (MDECREASE) in patient–therapist discrepancies in alliance quality ratings between the initial (Measurement I) and final (Measurement III) phases. Its values were MDECREASE = 32.33 (SDDECREASE = 22.32) for patients with more successful treatment and MDECREASE = 34.72 (SDDECREASE = 36.67) for those with less successful treatment. In accordance with H4, a decrease in patient–therapist discrepancies was expected. The discrepancies were found to decrease in both groups. Finally, through an analysis of intergroup differences, it was examined if the decreases in patient–therapist alliance rating discrepancies between the initial and final phases of psychotherapy differed significantly between the groups of patients with more and less successful treatment. The analysis revealed that these decreases were similar and that there were no statistically significant differences in the reported values of decrease, t(18) = 0.18, p = 0.44. Hypothesis H4 is therefore rejected.

Discussion

The analysis concerned the interesting phenomenon of patient–therapist convergence in therapeutic alliance ratings. This issue was investigated among patients suffering from adjustment disorders and, to a smaller extent due to the limitations of this study, in their systemic psychotherapists. Studies in the regime described above are not frequent, which makes the findings particularly valuable.

The psychotherapist and the patient differ in their evaluation of the therapeutic alliance. Considered at specific points in time, differences are present especially in the initial and middle phases of psychotherapy. The final phase is marked by similar patient and therapist ratings. Nevertheless, what should be noted when analyzing divergence in alliance ratings is the converging perspectives of the two sides of the dyad regarding alliance perception. There is some evidence suggesting that this convergence reflects the gradual strengthening of attachment accompanied by increasingly similar perceptions of experiences (Mikulincer & Shaver, 2018). The convergence seems to be characteristic not only of informal relationships; it is also present in conditions of professional assistance (Coyne et al., 2018).

The increase in the convergence of patient and therapist perspectives over the course of psychotherapy may indicate that the beginning of psychotherapeutic work is naturally associated with a lack of consensus on what the therapeutic relationship is and what it is like and reflects the initial lack of adjustment in the dyad at the cooperation level (Atzil-Slonim et al., 2015). As the psychotherapy progresses, alliance perceptions in the dyad become increasingly similar. Laws et al. (2017) referred to this phenomenon as increasing patient–therapist consensus. These results are also in line with Bordin’s (1979) pantheoretical model, according to which a well-functioning alliance is one where patients and therapists build a consensus on the goals and tasks in psychotherapy and, with time, strengthen their bond as well. The decrease in discrepancy may attest to the development of alliance over time.

The divergences observed and their direction pertain to alliance in psychotherapy for mild forms of psychopathology (adjustment disorders). Previous research suggested that patients with severe disorders, such as personality disorders (Cierpiałkowska & Frączek, 2017) or attachment disorders (Bachelor, 2013), had difficulties building a therapeutic relationship and that the alliances they build were marked by instability, criticism about the relationship, and a focus on the need for control rather than on building a genuine alliance. The present study showed that patients with milder psychopathology rated the alliance higher than their psychotherapist did. The severity of psychopathology may therefore be a significant factor affecting alliance quality reflected in the patient’s ratings of the alliance.

The phenomenon of psychotherapists underrating the alliance is only slightly supported by the results of the present research due to the participation of only one psychotherapist. It should be stressed, however, that the psychotherapist’s mode of work was strictly defined and controlled on many levels, which means the underrating cannot be solely a product of his bias. The literature offers two main explanations for psychotherapists underrating the alliance. One of them points to their vigilance and responsibility for the treatment process, which manifests itself in more critical and cautious evaluations (Atzil-Slonim et al., 2015). Sometimes, psychotherapists’ inadequate experience and immature technique is highlighted as a factor behind their sense of incompetence and the underestimation of their work (Fitzpatrick et al., 2005), which may be reflected in low alliance ratings. In this study, underrating cannot have been caused by the psychotherapist’s critical self-evaluation stemming from a lack of self-confidence and experience because the psychotherapist participating in the study had worked in the profession for many years.

The progressive patient–therapist convergence in alliance ratings during psychotherapy is not greater in patients with better treatment outcomes. This suggests the dissimilarity of therapeutic alliance perceptions is not essential for change (Eubanks-Carter et al., 2010). Patients with adjustment disorders suffer from states of tension, depression, and emotional disturbances that arise during adaptation to big changes in life and that usually impair social adjustment and effective functioning. In this diagnosis, relatively little attention is devoted to the relational or attachment factor, and the expected change, if present, may not stem from it. The situation is different in patients suffering from depressive (Constantino et al., 2008) or personality disorders (Cierpiałkowska & Frączek, 2017); in their case, apart from the patient’s or the therapist’s individual characteristics moderating treatment outcomes, there is a so-called dyad quality effect (Laws et al., 2017).

Limitations

Limitations of this empirical study should be acknowledged. The main ones among them are the small sample of patients, one psychotherapist, only three measurements to cover the entire psychotherapeutic process, and no control for demographic variables (e.g., sex, age) or independent variables from the context of participants’ life due to the time of psychotherapy sometimes reaching one and a half years. The characteristics of the sample restricted the range of statistics that could be used in analyses. One must also be careful with generalizations concerning other psychotherapists’ alliance ratings. Even the strictest research plan does not prevent the potential distortion of ratings because the reported dynamics of relationship building may be characteristic of a particular therapist—the one who took part in the study. The psychotherapist's directiveness or burnout may have influenced the therapeutic processes examined. The study also ignored an equally important factor, possibly of significance for the therapeutic alliance: pharmacotherapy. Patient–therapist convergence in perspectives was investigated in only one type of treatment administered for one type of disorder. Although the data collected paint a picture of a natural development of therapeutic alliance and treatment, reflecting actual psychotherapeutic practice, it does not seem legitimate to generalize the findings of this study to the psychotherapeutic processes of patients suffering from different disorders or to psychotherapists using different treatment strategies.