How therapists' interpersonal behaviour is perceived by their patients and close others: A longitudinal and cross-situational study.

Due to their predictive abilities, therapist interpersonal behaviour is of great relevance for psychotherapy. However, there is a lack of knowledge about its stability inside but also outside of the therapy room within and between therapists. The current study investigates interpersonal behaviour of trainee therapists (N = 20) as perceived by four patients each suffering from generalized anxiety disorder and three closely related persons of every therapist (close others). Investigating repeated measures, four patients per therapist completed the Impact Message Inventory (IMI; Kiesler, 1987) three times over the course of their cognitive behavioural therapy. Furthermore, the IMI was completed by three close others at one assessment time. Therapist interpersonal behaviour was perceived as more friendly and less submissive when evaluated by close others compared to patients. Using a multilevel approach, our results indicate that therapists' interpersonal behaviour was perceived considerably stable across patients and over the course of treatment, and there is considerable uniformity of the IMI evaluations in respect to the particular subscales within and between therapists. Our results highlight the potential similarities of observer-based habitual therapists' interpersonal behaviour inside and outside of the therapy room.

Therapists' interpersonal behaviour is differently perceived by patients compared to close others.

Therapists' interpersonal behaviour is perceived as highly stable across patients and over the course of treatment.

A very low between‐therapist variability of therapists' interpersonal behaviour as perceived by their patients emerged in this study, pointing to an unexpected high homogeneity of interpersonal behaviour across the investigated therapists.

INTRODUCTION

“… therapists were people before they were professionals, …” noted Wolf, Goldfried and Muran (2017, p. 175) in respect of therapists' negative interpersonal responses, opening up to the question of the origin of therapists' behaviours shown in treatment. Until today, it is still mostly unknown whether therapist habitual interpersonal behaviour is impacted more by a therapist's personal characteristics or professional attitudes and roles. Furthermore, there is a lack of knowledge about the stability of therapists' interpersonal behaviour within and across life domains; therefore, the question arises if therapists show comparable patterns when perceived by their patients in comparison to the therapists' close others. Current literature shows that therapist effects explain about 5%–8% (Johns et al., 2019) of treatment outcomes and that some therapists are about 10 times more effective than others (Okiishi et al., 2003). However, only a few factors have been found to explain these effects such as therapist personal burdens (Nissen‐Lie et al., 2013), therapist occupational burnout (Delgadillo et al., 2018), use of maladaptive coping strategies and self‐doubt (Nissen‐Lie et al., 2017), and therapists' current alcohol‐related and/or financial stress (Xiao et al., 2017). Besides, therapists' perceived social support and degree of comfort with attachment (Dunkle & Friedlander, 1996) are associated with better therapy outcomes. Recent reviews emphasize in particular on therapists' interpersonal variables (Heinonen & Nissen‐Lie, 2020) next to several interaction effects (Lingiardi et al., 2018). The alliance is one such interaction between process and outcomes, where in an analysis the therapists' contribution to the alliance was found to be correlated to outcome but not the patients' contribution (Baldwin et al., 2007; Del Re et al., 2021). Nevertheless, the personal characteristics and/or professional skills that enable some therapists to establish better alliances and foster patient outcome have to be further studied (e.g., Lingiardi et al., 2018).

Assessments of therapist abilities such as therapist interpersonal skills have been found to predict psychotherapy outcomes of patients up to several years (Anderson et al., 2015, 2016, 2009; Schöttke et al., 2015). Those investigations of therapist interpersonal behaviour refer to observer ratings in standardized assessment situations. However, there is a lack of literature concerning the interpersonal behaviour therapists habitually express while conducting psychotherapy and its stability. Especially, literature about patients' perceptions of their therapists' interpersonal behaviour is very limited. One of the most prominent preexisting theoretical frameworks to understand observer‐based interpersonal behaviour in general is rooted in Kiesler's (1979) interpersonal communication theory: the basis for his theory is the interpersonal circumplex model (Freedman et al., 1951; Wiggins, 1979) with the two central dimensions of affiliation and control (for an overview of names for these dimension, see Horowitz et al., 2006); where the affiliation dimension contains the opposing poles of hostility and friendliness and the control dimension contains the opposing poles of dominance and submissiveness. According to Kiesler, interpersonal behaviour of an individual (‘sender’) can be assessed by the perceived experience evoked in his or her counterpart (‘receiver’; see Schmidt et al., 1999). Thereby, it is assumed that first, the sender is not entirely aware of his interpersonal style of communication and especially, its influence on his counterpart; second, the interpersonal style of a sender's communication is experienced similarly across different receivers; third, receivers are able to detect their own internal response to the messages of a sender. Following Kiesler (1979, 1983, 1987, 1996), it is a promising approach to measure a sender's interpersonal behaviour by the receiver's internal response. For this purpose, Impact Message Inventory (IMI; Perkins et al., 1979) was developed, tested and translated into various languages (e.g., Caspar et al., 2016).

Whereas Kiesler's theory and the IMI measure primarily were developed to better understand patients' interpersonal behaviour, some researchers expected that the interpersonal behaviour of the therapists may be relevant as well (Goldfried & Davison, 1994; Kiesler, 1979; Wolf et al., 2017). For example, in the Gloria Interviews, the interpersonal behaviour of three prominent therapists could be well differentiated using IMI assessment (Schmidt et al., 1999), demonstrating a certain stability of an individual therapist's interpersonal behaviour. In line, a study investigating the material of one of the previously mentioned assessments found a high internal consistency (Cronbach's α = 0.94) of therapists' reactions across different patients (Munder et al., 2019). Furthermore, it is widely assumed that therapists' reactions are influenced by private experiences, indicating an influence of personal experiences into the interpersonal behaviour in therapy (e.g., Delgadillo et al., 2018; Gelso & Hayes, 2007; Rek et al., 2018). Conversely, there is as well reason to assume that a therapist's interpersonal behaviour may fluctuate between different patients over the course of treatment. Accordingly, Carson (1969) for example proposed ‘symptom‐free individuals’ to be able to flexibly and appropriately vary across interaction partners, enabling nonpatients such as therapists to adapt to their vis‐à‐vis. Furthermore, there are several treatment approaches that postulate a flexible responsiveness of therapists' interpersonal behaviour (e.g., Caspar, 2019).

Taken together, whereas the investigation of therapists' interpersonal characteristics recently got a particular focus in psychotherapy research, there is still little knowledge about the potential variability or stability of the therapists' habitual interpersonal behaviour as perceived by their patients and their close others.

Current study

The main aim of the current study was to investigate the therapists' interpersonal behaviour in their professional as well as in their private life (see Figure 1). More specifically, therapists' interpersonal behaviour was evaluated, each by four patients and three therapists' close others using IMI measure. First, we tested the variability of the therapists' interpersonal behaviour across all evaluators (Question; Q1a), and we contrasted the patients' evaluations with those of the close others (Q1b). Within the subsample investigating repeated patients' IMI assessments (three patients' assessments during therapy, four patients per therapist), we analysed the between‐patient differences (Q2a) and the within‐patient variability over the three assessment times (Q2b).

FIGURE 1

Illustration of the study design and corresponding research questions 1 and 2

METHODS

Design

This study was a subproject of a larger randomized controlled trial (RCT; study protocol: Flückiger et al., 2018). Cognitive behavioural therapy in a 16‐session format was provided for patients suffering from generalized anxiety disorder (GAD). The therapy was delivered according to the most recent Mastery Your Anxiety and Worry package (Zinbarg et al., 2006), which was developed to treat GAD and entailed progressive relaxation or mindfulness, cognitive restructuring, behavioural experiments as well as exposure (for more information of the particular trial and the major outcomes, please see Flückiger et al., 2021). Eighty patients who were randomly assigned to 20 therapists were invited to evaluate the IMIs of their therapists at Sessions 5, 10 and 15. In addition, three therapist's close others completed the IMIs as well (see Figure 1). In the patient sample, 72 out of 80 (90%) IMIs returned at Session 5, 70 (87.5%) at Session 10 and 60 (75%) at Session 15. In the close other sample, 51 out of 60 (86%) IMIs retuned. Overall, for 14 (75%) therapists, there was no missing data (i.e., 7 IMI evaluations per therapist). Data for this subproject were collected from January 2017 until January 2020.

Participants

Therapists

Trainee therapists were recruited via announcements of the principal investigator in therapist trainings. Twenty trainee therapists agreed to participate in the superordinate project. Inclusion criteria were: (a) a Master's degree in psychology and (b) being registered in an integrative cognitive‐behavioural psychotherapy‐training programme. Of this sample, 18 (90%) were female with an average age of 31.9 years (SD = 6.9, range 28–56). The trainee therapists had been working as therapists for 1.9 years (SD = 1.4, range 0–5), and their prior clinical experience was on average 49 completed therapies (SD = 67, range 0–240).

Patients

Patients were recruited via public announcements and adverts on mailing lists. Inclusion criteria were (a) diagnosis of GAD assessed with DSM 5, (b) being 18 years old or older, (c) informed consent and (d) speaking German. Exclusion criteria were (a) suicidal tendency as indicated by a score of 2 or higher on the corresponding item of the Beck Depression Inventory, (b) medication for current bipolar or psychotic disorder, or (c) current psychotherapy from another therapist. Comorbidities as well as prescribed medications for mood disorders did not lead to exclusion from this study. Eighty patients (75% female) met inclusion criteria and their average age was 31 years (SD = 9.5, range 21–67)

Close others

Close others of the therapists were recruited by asking the therapists to hand the questionnaires to three self‐selected close others such as family members, partners and close friends. To guarantee anonymity, only age and gender were assessed. Close others of the therapists were on average 36.7 years old (SD = 13.1, range 21–66) and 53% female.

Measures

Impact Message Inventory

Impact Message Inventory (IMI; Perkins et al., 1979) is an indirect measurement of the interpersonal behaviour of a target subject in which the interpersonal behaviour is rated by their interaction partners (raters). Usually, the target subjects are patients and the raters are their close others. However, in the current study, therapists are the target subjects and they are rated by their patients and close others. Raters evaluate their emotional, cognitive, and behavioural experience in reaction to the target subject on a 4‐point Likert‐style scale from ‘not at all’ (1) up to ‘very much so’ (4). The IMI is based on interpersonal theories enabling investigators to draw interpersonal styles by arranging interpersonal behaviour along the interpersonal circle (Schmidt et al., 1999) with the two distinct dimensions, control (submissiveness vs. dominance) and affiliation (hostility vs. friendliness). Fingerle (1998) translated the shortened version into German (IMI‐RD; Fingerle, 1998). For the IMI‐RD, alpha coefficients ranged from 0.68 up to 0.86 (Caspar et al., 2016).

For the present study purposes, patients rated their therapist's habitual interpersonal behaviour in therapy sessions. To take this particular situation into account, the IMI‐RD was adapted. Items that were unsuitable for therapy sessions such as “When I am with this person, he/she makes me feel that… I should tell him/her to stand up for himself” were not considered, and a total of 20 items were retained. To ensure comparability, this adapted version of the IMI was used for all participants. Psychometric properties of this shortened scale were as follows: Cronbach's α ranged from α = 0.68 (hostile) to α = 0.77 (friendly‐submissive) with a mean alpha of α = 0.74 for the subscales. The Kaiser–Meyer–Olkin (KMO) index returns an overall MSA = 0.81 which is considered as meritorious (Kaiser, 1974) and indicates a given suitability of data for factor analysis. The factor analysis yielded a comparative fit index (CFI) of 0.81 and a root mean square residual (RMSR) of 0.089 indicating an adequate model fit. In the current study, we decided to focus on the four main scales (dominant, submissive, hostile, friendly). We tested if therapists differ in their interpersonal behaviour in comparison to a patient population collected in a Swiss university outpatient centre: therapists are perceived as being significantly more dominant (t = −5.56, df = 311, p < 0.001), more friendly (t = −3.82, df = 317, p < 0.001), and less submissive (t = 11.98, df = 310, p < 0.001) than the outpatients and they are seen as similarly hostile (t = 1.44, df = 317, p = 0.152).

Data analysis

A multilevel modelling approach was used to analyse the nested data (Raudenbush & Bryk, 2002). First, we investigated the variability of therapist's interpersonal behaviour (Q1a) and tested whether therapists' professional interpersonal behaviour is perceived differently by their patients in comparison with therapists' private interpersonal behaviour by three close others based on the four above‐mentioned scales (Q1b). These comparisons were tested using a multilevel model where the group association (i.e., patients vs. close others) was added as predictor at Level 1 and therapists as grouping variable at Level 2.

In order to test whether therapists' interpersonal behaviour varies between patients (Q2a) and within patients over time (Q2b), a hierarchical multilevel model was performed for every scale with fixed intercept and fixed slope. Five predictors were integrated into the model: time, patient and therapist as well as their interactions time * patient and time * therapist. Thereby, time as the repeated measure t 1–t 3 was nested in patient at Level 2 which was nested in therapists at Level 3.

We used R statistical software for data preparation and statistical analyses (R Development Core Team, 2014). In order to evaluate the psychometric properties, Cronbach's α was calculated with the package psych (Revelle, 2018), and the package lavaan (Rosseel, 2012) was used to compute factor analysis of the IMI. For hypotheses testing, multilevel models were performed using the package lmerTest (Kuznetsova et al., 2017).

RESULTS

First, we investigated the variability of therapists' interpersonal behaviour (Q1a). Overall, therapists' interpersonal behaviour tends to be perceived as friendly (M = 3.72, SD = 0.40, range = 1.8–4.0) moderately dominant (M = 2.99, SD = 0.47, range = 1.0–4.0) and somewhat hostile (M = 1.57, SD = 0.39, range = 1–2.8); they may seem somewhat submissive (M = 1.96, SD = 0.48, range = 1.0–3.3); the standard deviations were noticeably low among patients and close others and within patient at all three time points (a figure of every therapists' individual ratings can be found in supplementary material). Then, we compared patients' perceptions of therapists' interpersonal behaviour with the therapists' close others' perceptions (Q1b). Results revealed that patients compared to close others experience therapists as significantly less friendly (t = 3.8, df = 85, p < 0.001) and less submissive (t = 6.26, df = 19, p < 0.001), but no significant differences were found in the hostile and dominant scales (see Table 1 and Figure 2).

TABLE 1

Therapists' interpersonal behaviour perceived by close others compared to patients

	Close others	Patients	Difference
IMI scale	M (SD)	M (SD)	M diff (SE)	t (df)	p
Hostile	1.56 (0.38)	1.46 (0.37)	−0.10 (0.07)	1.47 (45.30)	=0.147
Dominant	2.99 (0.42)	3.06 (0.47)	0.07 (0.09)	0.71 (21.76)	=0.482
Friendly	3.73 (0.29)	3.51 (0.33)	−0.22 (0.06)	3.80 (84.66)	>.001***
Submissive	1.96 (0.52)	1.34 (0.36)	−.62 (0.10)	6.26 (19.06)	>.001***

Abbreviations: df, degrees of freedom; M, mean; SD, standard deviation; SE, standard error; t, test statistic of the linear mixed model.

p < 0.05.

p < 0.01.

p < 0.001.

FIGURE 2

Data distribution of ratings by close others and patients per IMI scale

Second, we investigated whether patients perceived their therapists' interpersonal behaviour differently (i.e., between‐patient differences; Q2a) and whether patients' perception of their therapists' interpersonal behaviour fluctuated over time (i.e., within‐patient variability; Q2b). Results indicated no significant differences between patients in any of the four scales at any point of time (see Table 2). These findings indicate that therapists' interpersonal behaviour was perceived similarly across patients. Finally, we tested whether patients' perceptions of their therapists' interpersonal behaviour varies over time (Qb2). Results did not show any significant main effect for therapist nor a significant main effect for time in any of the four scales. This indicates that patients experience therapists' interpersonal behaviour as highly stable over the course of treatment. But we found significant interactions of patient and Time3, indicating that for some patients the perception of their therapists' interpersonal behaviour changed from Time1 to Time3. Those patients perceived their therapist as significantly more hostile, significantly more dominant and significantly less friendly at Time3 compared to Time1. All three effects were small as indicated by effect sized below Cohen's ds = 0.5 (Cohen, 1998).

TABLE 2

Therapist interpersonal behaviour between and within patients

		Patients
IMI scale	Fixed effects	γ	SE	t	p	Cohen's d s
Hostile	Time 1	1.296	0.884	1.47
	Time 3diff	−1.098	0.989	1.11	= 0.27
	Patient	−0.001	0.001	0.54	= 0.57
	Therapist	0.001	0.007	0.26	= 0.79
	Patient * time 3	0.002	0.000	4.28	<0.001***	0.32
	Therapist * time	0.006	0.008	0.75	= 0.45
Dominant	Time 1	2.930	1.35	2.16
	Time 3 diff	0.967	0.953	0.91	= 0.36
	Patient	0.000	0.000	0.03	= 0.98
	Therapist	0.001	0.012	0.10	= 0.92
	Patient * time 3	0.002	0.000	2.6	= 0.004 **	0.21
	Therapist * time	0.006	0.009	0.67	= 0.49
Friendly	Time 1	3.577	0.971	3.68
	Time 3 diff	1.425	1.066	1.33	= 0.18
	Patient	0.000	0.000	1.18	= 0.24
	Therapist	−0.002	0.008	0.22	= 0.83
	Patient * time 3	−0.003	0.000	6.25	<0.001***	0.20
	Therapist * time	0.006	0.009	0.68	= 0.50
Submissive	Time 1	0.584	0.862	0.68
	Time 3 diff	−0.353	0.894	0.40	= 0.69
	Patient	0.000	0.000	0.07	= 0.94
	Therapist	0.007	0.008	0.91	= 0.37
	Patient * time 3	0.000	0.000	1.43	=0.15
	Therapist * time	0.002	0.008	0.28	= 0.83

Abbreviations: Cohen's d , effect size after Cohen (1992) with pooled standard deviation; N, number of participants; SE, standard error; t, test statistic of two‐level hierarchical model; γ, predictor.

p < .05.

p < .01.

p < .001.

DISCUSSION

The current study investigated therapist interpersonal behaviour evaluated with IMI by patients and close others. We found small but statistically significant differences of patients' and close others' perception of therapists' friendliness and submissiveness. Similar differences in therapists' interpersonal behaviour between personal and professional relationships emerged when therapists rated themselves: in a study conducted by Heinonen and Orlinsky (2013), therapists reported showing more warmth, nurturance, protection and intuition in personal than in professional relationships. Fincke, Möller and Taubner (2015) found that therapists indicated being more affiliated to and less controlling in personal compared to professional relationships. There are several differences between patients and close others that could have led to this result: first, close others are relatives of the therapists as most therapists reported having asked family members, partners and good friends, whereas the patients knew their therapists from their therapy sessions only. Therefore, close others could have rated the therapists towards social desirability. Second, close others were self‐selected by the therapists, whereas the patients were randomly assigned by the study protocol. Hence, it is possible that therapists chose only those persons of whom they assumed to be especially positively inclined towards themselves. However, there is little knowledge about whether these close relatives evaluate the therapists differently from other relatives that are not selected from the therapist. Third, the relationship qualities in close relations can be assumed to be reciprocal (Patterson et al., 1993) and driven by emotional needs (Heinonen & Orlinsky, 2013), whereas, within therapy, the therapist is assumed to be primarily focused on his or her patient (Norcross & Hill, 2002). Furthermore, the therapeutic context itself may impact the IMI evaluations. As therapists may have an active role in leading the process of change, they may be perceived less submissive. Moreover, as therapists may sometimes have to address uncomfortable, unpleasant or painful topics which were avoided by the patients, they may be perceived as less friendly and more dominant by their patients.

In respect to the therapeutic context (within‐ and between patients), we found no substantial variation and significant differences in the perception of the therapists' habitual interpersonal behaviour. A similar result was obtained in an investigation with depressive patients, where therapist IMI change could not be assessed due to restricted variability (Coyne et al., 2018). In our study, only a small interaction effects with time emerged for a few patients, indicating a change in their perception of the interpersonal behaviour of their therapist. These patients rated their therapists as more hostile, more dominant and less friendly at the end of treatment compared to the beginning. One explanation could be that—as outlined above—therapists had to address unpleasant topics to the patients, which let the patients change their perception over the course of treatment. However, similar deteriorations with slightly less friendly and more hostile interpersonal behaviour have as well been documented in the Vanderbilt II study, where these effects were attributed to the investigated manualized psychodynamic training (Henry et al., 1993).

Most strikingly, however, the variances between all perceptions of the therapists were noticeably low. Hence, the results seem to speak for a relatively homogeneous perception of the interpersonal style for all therapists rather than an individual therapist's personal style or individual adaptation/responsiveness (see Supporting Information). We only can speculate about this unexpected high stability of perceptions of the therapists across patients and time. Reasons may lay in the therapists themselves, in the present study context as well as in the IMI assessment: first, therapists decided to become a mental health professional and they pursued this career for quite a while. Next, they decided for cognitive‐behavioural therapy postgraduate training, were selected by the training centres and agreed to participate in an RCT. Overall, this may have reduced the diversity of interpersonal styles in the professional psychotherapy context. Indeed, in the stereotype literature, it is assumed that people use stereotypes of professions as guidance to their vocational choice and that self‐concepts are positively correlated to the stereotype of their profession (Hollander & Parker, 1969). Furthermore, hiring decisions have been shown to be influenced by stereotypes (e.g., Nadler & Kufahl, 2014); that is, the training centres as well may have chosen their trainees based on congruencies with a psychotherapists' interpersonal stereotype. Additionally, stereotypes affect subsequent perceptions of people (Cohen, 1981). It is well known in the literature that memory‐based ratings are often relied on abstractions such as stereotypes (Srull & Wyer, 1989). Therefore, the therapist stereotype of patients and close others may as well have influenced IMI ratings. Partly in line with our finding of therapists being perceived as moderately dominant, friendly, not hostile and somewhat submissive, an investigation by Levy (1988) found the stereotype of a therapist was perceived more as leader than as a follower, warm, concerned with others and relaxed. Moreover, one could even argue in the sense of a déformation profesionnelle, a French term used to describe the effect of an (over‐) internalization of the professional role which leads to the usage of professional perspectives and practices in everyday lives (Rey, 2008). Indeed, an interview study investigating the effects of practice on the personal life of therapists found that over 70% of the investigated therapists perceived themselves to act therapeutically outside practice (Farber, 1983).

Second, the study context could have influenced the results. All therapy sessions were videotaped within a manualized cognitive‐behavioural therapy approach (Zinbarg et al., 2006). The videotaping could have led to ‘controlled’ and less spontaneous behaviours in respect to general therapeutic skills and cognitive behavioural interventions. However, close others' perceptions of therapists were similar as well without those constraints.

Third, IMI assessment is based on Kiesler's circumplex theory and its underlying assumptions (Kiesler, 1979, 1983, 1987, 1996). The first assumption presumes that the ‘sender’ is not aware of his influence on others. However, many would expect therapists to be aware of their influence and impact on others (Caspar, 2019; Fauth & Williams, 2005; Jennings & Skovholt, 1999; Stiles et al., 1998) and therefore may be responsive to the others' perceptions in respect to the therapists' preferences of how they would like to be perceived by the others. The second assumption is that the sender's communication is experienced similarly across different receivers. This assumption seems to be met by our results. The third assumption is that ‘receivers’ are able to detect their own internal response to a sender. However, the receivers in this study were patients and therefore the decoding of their therapists interpersonal behaviour may be distorted by these individuals (Caspar et al., 2016), e.g., the shared characteristics of patients that suffers from GAD. However, close others evaluated the therapists as well, and (even if not tested) it is most likely that these persons generally did not suffer from a GAD. Last but not least, the perceived behavioural uniformity in the IMI assessment could represent an evaluative outcome rather than a behavioural indicator. In other words, IMI evaluations could be a consequence of various behaviours of therapists to get a favoured picture of themselves in their patients, for example, via a therapeutic responsiveness. In any case, it is unlikely that all therapists are just so responsive to individual patients that eventually all patients come up with the same perception. Altogether, our results indicate an unexpected homogeneity in the therapists' interpersonal behaviour as experienced and perceived by their counterparts. Borrowing Wolf, Goldfried and Murans (2017) words, one could also state: therapists are those people that became professionals.

LIMITATIONS

This study has some limitations. First, this study may lack generalizability as the sample consists of only German speaking therapists in their early career. Second, part of the uniformity of this sample may be due to selection effects of the study itself. Therapists were enrolled in Swiss CBT training programmes and participated in the RCT. Furthermore, we had to adapt the questionnaire for patients as well as close others. Therefore, inappropriate items for one of each context were not considered. Differences in therapists' interpersonal behaviour may be found when assessed with more items and more distinguished questions. However, we did find small effects. On the one hand, a difference in friendliness and submissiveness between close others and patients emerged and on the other hand an interaction was found, indicating an increase in perceived hostility of some therapist over the course of treatment. Last but not least, in the current study, we used IMI assessment, which builds on the perception of interpersonal behaviour. However, the behaviours themselves that lead to a certain perception were neglected. Limitations notwithstanding, results are compelling as they provide preliminary information about potential variability of therapists' interpersonal characteristics across multiple evaluators. The results obtained showed a decrease in the perceived friendliness of the therapist for some patients, implying the possible occurrence of negative events or developments over the course of therapy. Furthermore, a uniformity‐like stability is indicated by the finding of an unexpected low variability of perceived interpersonal behaviour across all therapists, which implies an interplay of different unifying mechanisms such as selection, stereotypes and adaptations.

For future research, it would be promising to use a combination of assessments to understand interpersonal behaviour from different points of view and in order to replicate the results obtained in this preliminary investigation. Especially the investigation of possible moderators of patients' perceptions of their therapists could add essential information about therapist effects and may be an important link to process research. Furthermore, studies with other patient and therapist populations may provide estimates of the generalizability of the effects. Thereby, studies with more experienced therapists may help to get a more differentiated picture of therapists' habitual interpersonal behaviour. Last but not least, studies combining patient outcome with perceived interpersonal behaviour of their therapist may help to deepen the understanding of therapist effectiveness. However, such future direction would be particularly relevant for those samples where the IMI data indicates less within and between uniformity than in the present study.

DATA TRANSPARENCY STATEMENT

There is no prior manuscript that analysed this set of data. Furthermore, there is no manuscript submitted or in pipeline that is based on the present dataset.

FUNDING INFORMATION

This study was supported by the Swiss National Science Foundation (Grants: PP00P1_163702, PP00P1_190083; principal investigator: Christoph Flückiger).

Data S1. Supporting Information

Click here for additional data file.

Data S2. Supporting Information

Click here for additional data file.