Introduction

People who feel connected to one another often report being able to understand each other without words. Such descriptions highlight the importance of nonverbal coordination within communication and are consistent with the embodiment perspective, emphasizing that mental processes do not occur independently of bodily processes (Tschacher, 2020). The embodiment framework expands on our understanding of social interactions, as it becomes clear that communication is not limited to the verbal exchange of information, but is based on a reciprocal relationship between embodied agents (Fuchs & De Jaegher, 2009). For instance, subtle and unconscious coordination of bodily movements between interactants can be detected. Listeners, for example, can be observed to adjust their movements, however subtly, to the changes in direction, speed and intonation of utterances and movements of the speaker. In doing so, interactants relate to each other, synchronize their movements and develop a shared understanding, referred to as participatory sense-making (Fuchs & De Jaegher, 2009).

One way to empirically observe such interpersonal coordination is through synchronization phenomena (Bernieri & Rosenthal, 1991). Meta-analyses have shown that nonverbal synchrony, e.g. synchrony of body movements, is associated with prosocial attitudes and behaviors as well as social bonding and social cognition (Mogan et al., 2017; Vicaria & Dickens, 2016), though recent studies have shown associations with detrimental factors as well (for an overview, see Mayo & Gordon, 2020). Nonverbal synchronization processes range from the coordination of movement, posture and gestures to coordinated vocal frequencies or mutual patterns in heart rate, breathing and skin conductance (Tschacher et al., 2023; Wiltshire et al., 2020). All of these phenomena address the theme of interpersonal coordination, and various terms such as behavioral synchrony or interpersonal synchrony have been used (Paxton, 2015).

A specific form of nonverbal synchrony is movement synchrony. This is understood as the relatedness of movement dynamics over time (Ramseyer & Tschacher, 2011). Movement synchrony focuses on the temporal coordination and relatedness of movements irrespective of the specific type of movement or the body parts involved and is separate from mimicry, in which the same movement is implicitly imitated (Bernieri & Rosenthal, 1991). It encompasses any movements that coincide in temporal proximity and allows by the use of time-lag analyses for a differentiation between impulse-giving (leading) and impulse-following (pacing) movements (Ramseyer & Tschacher, 2011) in the dyad. The terms pacing and leading are usually understood from the counselors’ perspective, i.e. pacing means that the counselor follows the client’s impulses.

The development of highly automated methods to extract body movements from video recordings, such as Motion Energy Analysis (MEA, Ramseyer & Tschacher, 2011), led to an uptick in the study of movement synchrony. MEA is a software-based method that quantifies motion energy from video recordings by analyzing changes in pixels across consecutive frames. This method allows for the analysis of specific regions within the video, whether delineated by individuals or specific body areas. As a result, it generates quantitative time series (one for each defined region), reflecting the amount of movement change within these specified regions. These time series data can be utilized to compute movement synchrony between individuals or specific body regions by employing a synchrony algorithm. This computational approach allows for the identification of common patterns within the movement data, subsequently interpreted as movement synchrony.

In recent years, a promising line of research has emerged that investigates movement synchrony in psychotherapy. While there are rare cases where no correlation has been found (Paulick et al., 2018), the majority of studies to date have shown that higher levels of therapist-client synchrony are associated with meaningful variables such as experiencing higher self-efficacy, better clinical outcomes and especially higher relationship quality (Altmann et al., 2020; Ramseyer & Tschacher, 2011; Wiltshire et al., 2020). A recent systematic review of 17 studies found a neutral to mildly positive association between movement synchrony and either proximal (i.e. within-session) or distal therapy outcomes (Atzil-Slonim et al., 2023). Therefore, movement synchrony is considered an indicator of the therapeutic alliance (Koole & Tschacher, 2016; Tickle-Degnen & Rosenthal, 1990) and alliance, in turn, is a significant and robust factor of therapy outcome (Flückiger et al., 2012, 2018). The central theme underlying various conceptualizations of the therapeutic alliance is the trustful collaboration between therapists and clients (Flückiger et al., 2018; Luborsky et al., 1996). Often, this also emphasizes the alignment of client and therapist towards common goals (Bordin, 1979; Luborsky et al., 1996).

Recent research can explore the relationship between the therapeutic alliance and movement synchrony at an even more fine-grained level by considering pacing and leading dynamics. Studies have provided evidence that higher therapist pacing in the early phase of psychotherapy is associated with fewer dropouts (Schoenherr et al., 2019b) and contributes to a successful therapy progress (Ramseyer & Tschacher, 2011). Later in the treatment process, however, therapist pacing has been associated with higher symptom levels and interpersonal problems (Altmann et al., 2020) which contradicts the findings of Ramseyer and Tschacher (2011). Altmann et al. (2020) suggested that early pacing in psychotherapy may play an important role in establishing a therapeutic alliance, but heightened pacing at a later stage may lead the therapist to become involved in the patient’s maladaptive interpersonal patterns. Prinz et al. (2021b) assessed leading and pacing in terms of synchrony in physiological patterns (electrodermal activity) and found a relationship between therapist leading and clients’ emotional experience. While the assessment of pacing and leading dynamics is of great interest within the research on synchrony in psychotherapy, the existing evidence so far is weak and contradictory.

If nonverbal coordination patterns in the therapeutic dyad are related to specific therapeutic interventions, the choice of the intervention may have an influence on therapeutic progress, but research on synchronization processes in different therapeutic interventions is scarce (Altmann et al., 2020; Wiltshire et al., 2020). A few studies have examined movement synchrony at the macro level in different types of therapy such as cognitive-behavioral therapy (CBT) or psychodynamic therapy (PDT; e.g. Schoenherr et al., 2019b). Most previous studies have examined the initial 15 min of a session and taken this as an index of synchrony for the entire session (e.g. Paulick et al., 2018; Ramseyer & Tschacher, 2011).

However, there are hardly any studies examining movement synchrony at the micro level, i.e. how it changes within sessions and how it is related to different therapeutic strategies or interventions. Tschacher et al. (2014) compared movement synchrony in different forms of interaction (including playful cooperation, serious cooperation and competition) in an experimental setting, which might be a blueprint of different therapeutic interventions. They found higher levels of synchrony, when subjects were cooperating playfully or competing.

A widely used intervention in systemic approaches is solution-focused questioning, which originated in solution-focused brief therapy (SFBT) (Bradley et al., 2010; Trepper, 2012). In SFBT, solutions are developed in conversation with clients by focusing on their desired futures, skills and past successes instead of analyzing problems and their origins (de Shazer et al., 1986). The solution-focused approach offers a particular set of questions (Trepper et al., 2011). These questions are intended to direct attention and consequently generate new ideas and further helpful ways of experiencing (Neipp et al., 2021).

To develop a focus, which allows for the exploration of individual possibilities, therapists often need to direct clients’ attention from a problem-focused perspective towards potential goals and solutions. Micro-analyses of therapeutic processes led by Insoo Kim Berg, one of the founders of SFBT (Choi, 2019) revealed that the therapist initiated almost all of the solution talks in the conversations. As for initiating problem talks, the ratio was balanced between therapist and client.

Various efficacy studies have demonstrated the favorable outcomes associated with a solution-focused therapeutic approach (Gingerich & Peterson, 2013; Kim, 2008). Grant (2012) was the first to directly compare the impact of solution-focused and problem-focused questioning. In an online coaching format for students, solution-focused questions led to increases in positive affect, decreases in negative affect, increases in self-efficacy, perceived closeness to the goal and more actively generated steps towards a solution compared to problem-focused questions. This study has since been replicated multiple times (Braunstein & Grant, 2016; Grant & Gerrard, 2020; Neipp et al., 2016; Theeboom, Beersma & van Vianen, 2016).

Prinz et al. (2021a) investigated movement synchrony in connection with therapeutic approaches by exploring its relationship with Grawe’s general mechanisms of change in 423 therapy sessions. Two of these mechanisms are resource activation and problem actuation, which can be regarded as very similar to solution-focused and problem-focused questioning. These mechanisms were rated retrospectively. Prinz et al. (2021a) hypothesized that synchrony and resource activation would be positively correlated due to their joint and co-creative nature, as well as their association with positive emotions (Gassmann & Grawe, 2006), and that problem actuation would be negatively correlated due to their belief that patients would be more preoccupied with themselves in this position. However, their results did not support these hypotheses. In this study, we will investigate the relationship between resource activation/problem actuation and movement synchrony by addressing this question within the paradigm of solution focused therapy introduced by Grant (2012, see above). Our approach is experimental and manualized, focusing on a specific intervention. We use a within-session design by dividing our sessions into three distinct parts, in which the degree of manualization is varied. According to Grawe, therapeutic alliance is another central mechanism of change (Grawe et al., 2001). Prinz et al. (2021a) did not include alliance as a variable, but since alliance and movement synchrony are closely related (Koole & Tschacher, 2016), we included it in our study design.

In this study, we aim to address the following questions: What are the differences in movement synchrony and therapeutic alliance between problem-focused and solution-focused therapy-like sessions? How do these differences vary across different parts of the interaction?

Since solution-focused conversations and movement synchrony are both related to positive emotional activation and self-efficacy (Grant, 2012), this might allow for the hypothesis that a solution-focused interaction would be associated with higher synchrony than a problem-focused interaction. Conversely however, evidence suggests that emotional activation precedes synchrony rather than vice versa (Tschacher et al., 2014), and given the lack of support for this hypothesis by Prinz et al. (2021a), we will take an exploratory approach by testing the undirected hypothesis that movement synchrony will differ in problem-focused and solution-focused interactions (H1).

Since movement synchrony is closely related to the therapeutic alliance, the second hypothesis is that therapeutic alliance will differ in solution-focused and problem-focused interactions (H2). Since no previous study has addressed this question, we also take an exploratory approach by means of a two-tailed testing.

Furthermore, we hypothesize that counselors will show more leading behavior in solution-focused interactions (H3). This assumption is based on the findings of Choi (2019) that a solution focused approach was linguistically almost always initiated by the therapist. We expect this pattern to also be reflected in bodily interaction, since mental or linguistic and bodily processes are interdependent. Therefore we assume that the shifting of perspectives by the therapist may be also associated with changes in movement behaviour. The assumption is also supported by the findings on pacing and leading in therapeutic dyads, which suggest that synchrony led by the therapist is particularly associated with a change in behavior towards the desired outcome (Altmann et al., 2020).

Methods

Design

In a crossover design, participants were invited to attend two different counseling sessions, one with a solution-focus and one with a problem-focus. The order of these two sessions was randomized and for each session the participant met with a different counselor.

Participants and Counselors

We recruited participants via public announcements, social media postings and mailing lists. Inclusion criteria were age of 18 years or older and fluency in German. Exclusion criteria were motor limitations that might prevent movement synchrony with the counterpart and symptoms of COVID-19 infection, as established by an intake questionnaire. In addition, counselors were instructed to cancel the session if they knew the participant from another context. Based on the expectation of a relevant, medium-sized effect (d = 0.5) for the target comparison, we conducted a power analysis (d = 0.5; 1-β = 0.80, α = 0.05; 2t, dependent means) which indicated a requirement of at least 34 participants. Subsequently, 34 participants were recruited, of whom 30 identified as female, three as male and one as other. Their ages ranged from 19 to 74 years (M = 26.3, SD = 10.3). In terms of maximum educational level, twelve participants (35%) had a high school diploma and 22 (65%) had a university degree.

Seven counselors were recruited, six of whom were from a group of psychotherapists in training at the Systemic Institute of the Medical Center of the University of Freiburg, Germany and were taking part in a 5-year training program in systemic psychotherapy to obtain official licensing. Six of the seven counselors had a master’s degree in psychology as well as prior experience in counseling. Of these six therapists (all female), three had previously completed various further training courses in the systemic field and one had completed training in Gestalt therapy. Two of the therapists had not completed any therapeutic or counseling training prior to training at the systemic institute, but had already gained experience in clinical and counseling work through internships as well as prior experience in counseling. One counselor (male) was recruited from the Freiburg Family Therapy Association. He had already completed training in systemic counseling and had prior counseling experience.

Systemic therapy is an approach that is characterized by its foundation in systems theory, resulting in a focus on social interactions. Clients are understood in their social nature and seen in their relationships with others. The systemic approach offers models to explain complex, dynamic interactions in social systems. Solution-focused approaches are seen as part of systemic therapy (Friedlander et al., 2021).

The ethics committee of the Medical Center of the University of Freiburg, Germany approved the study. All participants gave written informed consent.

Procedures

Interested participants reached out to the experimenter via email to schedule two counseling sessions, with an average interval of seven days between them. The experimenter checked on the immune status of the participants, since the counseling sessions had to take place without a face mask during the COVID-19 pandemic. Participants had to either show proof of COVID-19 immunity through recovery or vaccination, provide a negative test result from an official rapid testing center, or arrive 20 min early to perform a rapid antigen test at the study site. Rapid tests were provided.

The participants were told the cover story that the study would assess how different approaches of the counselor would affect their perceptions of the client-counselor relationship. Neither counselors nor participants were informed about the assessment of movement synchrony. Participants were instructed to select two different personal problems prior to the counseling sessions, one for each session, that were frustrating and yet unresolved, but manageable and comfortable to talk about in the context of the study.

Upon arrival, participants were asked to fill in a questionnaire to assess the risk of an undetected COVID-19 infection. Next, they self-tested if necessary and were then guided to the interaction lab where the sessions took place.

The lab was equipped with two solid armchairs in fixed positions placed at an angle of 90°. Two small action cameras were mounted unobtrusively on the wall directly opposite each chair. Additionally, the lab was illuminated by four large, permanently installed LED panels, which provided a constant, semi-warm white light.

Upon arriving in the lab, participants were seated, informed about the study procedure and gave written informed consent. The video recording was started and the counselor was brought into the room and also sat down. The experimenter briefly introduced the counselor to the participant, wished well, turned a 20-minute hourglass and left the room.

The following counseling session lasted approximately 20 min and comprised of three parts, which are detailed below. After its completion, the counselor bid farewell to the participant and left. The experimenter then returned to administer several questionnaires. Upon questionnaire completion, the video recording was stopped and participants were asked about their emotional well-being. In the event that a session was perceived as distressing, participants were given the opportunity for thorough support and assistance. This was never the case. Subsequently, participants were debriefed concerning the intended movement analysis of the video recordings.

All seven counselors participated in problem- and solution-focused sessions. The total number of sessions per counselor ranged from three to 25.

Intervention

Counselors were provided with a written interview guide for the session procedure (see Online Resources 1 and 2). The sessions consisted of three distinct parts: (1) problem description (PD), (2) standardized intervention (IV-S) and (3) free intervention (IV-F). The combination of all three parts constitutes the total session (TOT).

In the PD part, the counselor briefly explained the procedure of the session, their roles and actions and the study context in a standardized way. Besides providing necessary information, this part was important for establishing a working relationship or alliance between the counselor and the participant. Next, the counselor asked the participant to describe the problem they had chosen for the session with a standardized question. This articulation of the problem by the participant was given the most time during the PD part. If the description was unclear or incomplete, the counselor was allowed to ask an additional question or express what they had understood up to two times. At the end of this part, the counselor paraphrased the problem presented and asked whether they had understood it correctly.

In the following IV-S part, the counselor asked seven standardized questions about the problem. There were two versions. In the solution-focused session, these questions were designed to explore potential solutions, in the problem-focused session, they were designed to explore the problem in more detail. In order to establish an authentic interaction, the counselors were allowed to react to the answers of the participant up to two times (e.g. “Are there any further thoughts?”). Furthermore, they were allowed to make slight changes in the wording of the questions to fit the specific context. However, the sequence of the seven questions had to remain unchanged. The questions were designed on the basis of Grant (2012).

Next, the counselor started with the IV-F part. The aim was to continue with the respective session (solution-focused or problem-focused) about the presented problem in a less restrictive way to allow for more spontaneous and authentic interaction sequences than the previous standardized intervention parts. The interview guide suggested some questions, but the counselor had the freedom to use them or ask others. The only constraint was that they had to follow the specific session condition (solution-focused or problem-focused).

Measures

Movement Synchrony

In order to ensure consistent and reliable motion energy representation of the interactants’ movements, we maintained a fixed camera position and consistent artificial lighting conditions throughout the experimental sessions. We employed two compact action cameras (Insta360 One R 1”) to record the videos, with a resolution of 4352 × 2488 and a frame rate of 30 frames per second (for full video settings see Online Resource 3). Subsequently, we utilized the software Blackmagic DaVinci Resolve to crop the two videos, synchronize them using their audio tracks and ultimately combine them into a split-screen video with dimensions of 1920 × 1080.

The temporal boundaries of the video were determined by specific events. The video commenced when the experimenter exited the room and it concluded at the last moment before either interactant exhibited any movement towards standing up. The IV-S part commenced with the counselor posing the first standardized question, while the IV-F part began with the first question asked by the counselor following the completion of the seven standardized questions from the IV-S part.

Motion Energy Analysis (MEA) is both the name of the software and the underlying algorithm, which is used to extract motion energy data representing the amount of visual change between each two adjacent video frames (Ramseyer, 2020b; Ramseyer & Tschacher, 2011). We consistently applied a minimal motion detection threshold of 13 based on Ramseyer (2020b) to avoid artifacts generated by subtle changes in light or noise of the camera sensor, for example. In the split screen videos, two regions of interests (ROI) were defined. One included the full body of the participant and one the full body of the counselor. To find all areas in the video affected by body movement in the video image, movement overview images of the full session were generatedFootnote 1. The ROIs were then delineated in such a way that all movements of one person occurred within the ROI while excluding any possible overlap with movements of the other person. As a result, two synchronized time series, with 30 data points per second for each person in the session, were obtained.

Movement synchrony was computed using the package rMEA (Kleinbub & Ramseyer, 2021) for the statistical programming language R (R Core Team, 2022). As a pre-processing step in rMEA, all MEA time series were scaled by their standard deviation using the MEAscale function. No smoothing was performed. Each of the three defined parts plus the total video (PD, IV-S, IV-F, TOT) of a session is partitioned into windows of 60s length (bandwidth and step size). Each window contains two time series with 60 × 30 = 1800 data points. In order to perform windowed cross-lagged correlation (WCLC) of these two time series, they were shifted against each other frame-by-frame up to time lags of +/- 5 s (lag size). For each lag including 0 a Pearson-correlation of the two time series was computed. This resulted in 301 correlations per window. These correlations were then Fisher-z transformed and next converted into absolute values. In the next step, these 301 correlations were averaged into a single r-value and are then also averaged over all windows of one part. Thus, we arrived at one z-value for each of the parts of each session.

For examining leading and pacing behavior in the interaction, we applied the same windowed cross-lagged correlation technique as described above. Instead of averaging all 301 correlation coefficients, we split the data into two sets: 150 lags where the counselor’s time series preceded that of the participant (leading) and 150 lags where the counselor’s data followed the participant’s data (pacing). Subsequently, we computed separate average correlation coefficients for these lag sets.

Therapeutic Alliance

The Helping Alliance Questionnaire (HAQ, Alexander & Luborsky, 1986) is a well-established scale to assess the therapeutic alliance in psychotherapy. It is translated and validated in German (Nübling et al., 2017) and has demonstrated good psychometric properties. We only applied the 6-item subscale “therapeutic relationship” of the German version (Bassler et al., 1995). Items are scored on a 6-point Likert scale (1 = I strongly feel it is not true, 6 = I strongly feel it is true). Participants completed the HAQ at the end of both sessions.

Adherence

All videos and all parts were reviewed for adherence to the interview guide and group-specific focus. Sequences in which counselors asked questions of the opposite strategy were flagged.

Camera Effect and Research Suspicion

At the end of each session, participants were asked three questions about a potential camera effect on a six-point Likert scale. At the end of the second session, irrespective of the condition, they were additionally asked two questions about their assumptions about the aim of the study.

Statistical Analysis

To assess meaningful movement synchrony and control for coincidental synchrony, we used a procedure called surrogate synchrony (SUSY) developed by Ramseyer and Tschacher (2011). This involved initially segmenting the time series data into 60-second windows, a process already undertaken in the calculation of real synchrony. For the pseudosynchrony analysis or SUSY, these pre-established windows were recombined through a shuffling process to create hypothetical interactions between participants and counselors that never really took place. The resulting over 4000 time series combinations, excluding actual interaction pairs, were subjected to synchrony computation, as previously described. Subsequently, the pseudosynchrony values derived from this process were independently compared with real synchrony values by an independent t-test. This was done separately for all defined session parts. We computed the effect size Cohen’s d for movement synchrony by subtracting the mean pseudosynchrony value from the mean real synchrony value divided by the standard deviation. In instances of significant variations in variances between groups, we used the Welch–Satterthwaite equation to estimate variances instead of the standard pooled variance.

To assess H1 (i.e. solution-focused and problem-focused interactions differ in movement synchrony) we computed a 2 × 3 repeated measures ANOVA with the repeated factors part (three levels: PD, IV-S, IV-F) and condition (two levels: solution-focused vs. problem-focused). Here, we expected a significant part × condition interaction. The structure of our data showed some nesting. The counselors conducted several sessions; each participant had two sessions with two different counselors. In order to account for this dependency in data (i.e., participants were nested in specific pairs of counselors), we conducted multilevel linear modelling and included the pair of counselors as additional level in the model. However, this approach did not improve the model fit. Thus, we remained with the 2 × 3 ANOVA. We performed post-hoc t-tests for dependent data. Outlier analyses for this and the following analysis were performed according to Tukey’s (1977) fences. Results with excluded outliers are reported only if their exclusion altered the significance of the results.

To assess H2 (i.e. solution-focused and problem-focused interactions differ in therapeutic alliance), we computed a simple t-test for dependent data specifically assessing HAQ scores across conditions. To assess H3 (i.e. counselors show more leading in solution-focused group interactions than in problem-focused interactions), we employed two distinct analyses to capture nuanced aspects of counselor-participant coordination. First, we measured leading synchrony, which informs us about the extent of successful movement synchrony initiation by the counselor. For the purposes of this study, we use the term absolute leading synchrony to describe leading synchrony as described above, in order to distinguish it from relative leading synchrony, which will be described below. For this analysis we computed a 2 × 3 repeated measures ANOVA with the repeated factors condition (two levels: solution-focused vs. problem-focused) and part (three levels: PD, IV-S, IV-F) for the dependent variable absolute leading synchrony. We then performed post-hoc t-tests for dependent data.

On the other hand, relative leading synchrony sheds light on the dynamics between the counselor and participant, offering insights into which party - counselor or participant - successfully initiates movement synchrony more frequently during a session. To calculate relative leading synchrony, we divided leading scores by pacing scores for each interaction part. These quotients were then logarithmized to create equal scaling between different values and to center the data around zero. Next, we computed the same 2 × 3 repeated measures ANOVA for the dependent variable relative leading synchrony.

In all statistical analyses, we employed the zABS value to represent synchrony, aligning with established practices in earlier works (Kleinbub & Ramseyer, 2020; Ramseyer & Tschacher, 2011). When presenting descriptive statistics, we transformed zABS scores back to Pearson’s r, maintaining the irreversible use of absolute values.

Results

The average duration of the sessions was 19:58 min (SD = 02:57, range 13:00–27:56). The average duration of the PD part was 04:12 min (SD = 02:13, range 01:35–15:29), of the IV-S part 09:01 min (SD = 03:00, range 03:12–17:06) and of the IV-F part 06:45 min (SD = 02:58, range 0:00–14:24). Three sessions had no IV-F part due to time constraints (i.e. the duration of previous parts). Counselors conducted an average of 9.7 sessions (SD = 7.3, range 3–25).

Full adherence to the counselor guide was partially compromised in seven out of the 68 sessions, all of which were problem-focused. In these cases, the counselors either occasionally and inadvertently used wording that was closer to the alternative condition, or the session as a whole seemed less clearly confined to a single condition than others. We conducted sensitivity analyses by excluding these seven cases. Results for these analyses are reported only if their exclusion altered the significance of the results. Participants indicated a subtle camera effect and low study suspicion (for complete findings, refer to Online Resource 4).

Movement Synchrony

First, we assessed the general presence of synchrony by comparing real synchrony values to pseudosynchrony. For the total interaction, the mean real synchrony was r = .11 while the mean pseudosynchrony was r = .10. A finer evaluation of all different lags showed that for the total interaction, real synchrony was on average above pseudosynchrony for 100% of all possible lags between − 5s and + 5s (mean synchrony value of all values at specific lag) as seen in Fig. 1. Since variances between real synchrony and pseudosynchrony differed, F(67, 4488) = 1.756, p < .001, the Welch-Sattertwaithe approximation to the degrees of freedom was used to perform a t-test and to calculate the respective effect size, yielding t(68) = 5.36, p < .001, d = 0.74. Thus, there was statistically significant synchrony in all sessions (more real synchrony compared to pseudosynchrony). The finer details and comparisons for each part are provided in Table 1.

Fig. 1
figure 1

Lag-plot for both synchrony and pseudosynchrony of normalized correlations between lags. Note The lag-plot depicts the comparison between synchrony and pseudosynchrony, showcasing normalized correlations at various lags. The bold lines represent the average values for both synchrony and pseudosynchrony. Notably, the average synchrony z-values consistently outperform the corresponding average pseudosynchrony values. The thinner lines represent synchrony values for individual 60s windows, upon which the average synchrony line is based

Full size image
Table 1 One-tailed t-tests comparing actual synchrony to pseudosynchrony by part
Full size table

The descriptive statistics of movement synchrony across different conditions are provided in Table 2. The H1 analysis yielded a statistically significant main effect of condition on movement synchrony, F(1, 31) = 4.623, p = .04, partial η2 = 0.13 and a significant interaction between condition and part, F(2, 62) = 3.542, p = .04, partial η2 = 0.10, supporting H1. The post-hoc t-tests for each part revealed, that the movement synchrony was higher in solution-focused than in problem-focused interactions in part PD (p = .007). However, the movement synchrony did not differ in parts IV-S (p = .25) and IV-F (p = .90). Moreover, in the sensitivity analysis excluding the seven compromised sessions the interaction effect was no longer significant, F(2, 48) = 2.061, p = .14, while the main effect still was, F(1, 24) = 4.334, p = .048, partial η2 = 0.15.

Table 2 Descriptive statistics for movement synchrony by condition and part
Full size table

Unfortunately, it was not possible to conduct an exploratory assessment of a counselor effect using a repeated measures ANOVA with the factors condition and parts. This was due to the limited number of sessions (i.e. three and four) provided by two of the counselors.

Therapeutic Alliance

There was no difference in therapeutic alliance between conditions, t(33) = 0.13, p = .90, failing to support H2. We exploratively examined the relationship between alliance ratings and movement synchrony. However, the correlations between synchrony and the HAQ scores were not significant (p > .05) neither for the overall interactions nor for any of the individual parts, whether analyzed by condition or across both conditions. In spite of this, the sensitivity analyses excluding the seven compromised sessions revealed negative correlations between synchrony and therapeutic alliance for the problem-focused condition in overall synchrony (TOT; r = -.41, p = .03) as well as in part IV-S (r = -.43, p = .03), indicating that lower levels of synchrony were associated with higher levels of therapeutic alliance (see Online Resource 5 for complete table). The descriptive statistics for therapeutic alliance are presented in Table 3.

Table 3 Descriptive statistics for therapeutic alliance (HAQ Sum Scores)
Full size table

Differences in Leading

Absolute leading synchrony are synchrony values calculated specifically for the lags where the counselor’s movement precedes the participant’s movement. As with movement synchrony, absolute leading synchrony varied between conditions and across the different parts of the interaction (see Table 4). For absolute leading we found a statistically significant main effect of condition, F(1, 31) = 4.33, p = .05, partial η2 = 0.12 and a significant interaction part × condition, F(2, 62) = 6.592, p = .003, partial η2 = 0.18, supporting H3, which expected more leading behavior in solution-focused interactions. The post-hoc tests for each part revealed this to be true in part PD (p = .005), but not in parts IV-S (p = .15) and IV-F (p = .36). However, in the sensitivity analysis excluding the seven compromised sessions the main effect of condition was no longer significant, F(1, 24) = 3.527, p = .07, while the interaction effect still was, F(2, 48) = 3.534, p = .04. The 2 × 3 repeated-measurements ANOVA for relative leading as dependent variable revealed no significant effects (p-values ranged between p = .14 and p = .92).

Table 4 Descriptive statistics for leading by condition and part
Full size table

Discussion

This study aimed to assess the differences in movement synchrony and therapeutic alliance between two approaches to therapy-like interactions: problem-focused and solution-focused counseling. Session data were analyzed with a focus on differences in movement synchrony and therapeutic alliance, while examining the variation of these differences across different interaction parts.

We found a significant main effect for condition and movement synchrony, supporting H1. We also found a significant interaction between condition and part for movement synchrony. Post-hoc tests revealed that the problem-focused and solution-focused conditions differed during the initial problem description part, but not during the subsequent intervention parts. This interaction effect was no longer significant in a sensitivity analysis excluding cases with occasional protocol breaches. Our second hypothesis, which assumed a difference in the therapeutic alliance between groups, was not supported. Our third hypothesis, that we would see more leading behavior in the solution-focused sessions, was again supported only in the initial problem description part.

Since we expected to find differences in the group-specific intervention parts, we were surprised to find significant differences only in the initial problem description part. The counselor’s task in this relatively brief part was to establish rapport with the client and to elicit a comprehensive description of their respective problem, using the same standardized phrases and questions for both conditions. One possible explanation for these differences is that the counselors couldn’t be blinded and were thus, aware of the session-condition throughout. As the counselors were trained in systemic psychotherapy, which emphasizes the advantages of the solution-focused approach, they may have, consciously or unconsciously, preferred the solution-focused condition, adapted their movements and interaction patterns accordingly and thus influenced movement synchrony. Such a preference is known in the literature as allegiance, which refers to the therapist’s or researcher’s personal belief in the superiority of a particular psychotherapy treatment. Allegiance has been shown to significantly influence intervention effect sizes (Dragioti et al., 2015), change expectancy (Field et al., 2017) and therapeutic outcomes (Toska et al., 2010).

Similar to our study, Prinz et al. (2021a) examined the relationship between movement synchrony and therapeutic interventions. However, our study differs methodologically as well as in our findings. Our controlled experimental design enabled us to directly manipulate counseling approaches. It is important to note that while our study examined differences between counseling approaches across different dyads, Prinz et al. only found results within the same dyads but not between dyads. This discrepancy could potentially help to explain our results, as our study never had more than one session per dyad and without within-dyad comparisons for the intervention, we are unable to evaluate ideographic variations that could influence movement synchrony. Regarding interventions, Prinz et al. (2021a) observed a negative association specifically between synchrony and resource activation, which can be considered similar to our solution-focused approach, in their real-world therapy sessions. These conflicting findings highlight the complexity of the interaction between intervention techniques and movement synchrony, emphasizing the need for further investigation into the nuanced dynamics of therapeutic processes. Furthermore, methodological differences, such as different parameters for synchrony calculation and the combination of multiple interventions within sessions in Prinz et al. (2021a), underscore the importance of considering contextual factors when interpreting findings across studies. These disparate findings provide valuable insights into the multifaceted nature of therapeutic interactions and underscore the importance of ongoing dialogue within the field.

We did not find any relation between therapeutic alliance and the two different therapeutic strategies employed, nor did we find an association between therapeutic alliance and movement synchrony. So far there are only few studies assessing therapeutic alliance in relation to movement synchrony showing heterogeneous findings. Similar to our results also Ramseyer et al. (2020a) and Paulick et al. (2018) did not find an association for patients’ alliance. Moreover, one study showed a positive association for body synchrony but not for head synchrony (Ramseyer & Tschacher, 2014) and one study reported a positive association (Ramseyer & Tschacher, 2011) for patients’ ratings but not for therapists’ ratings. Based on this summary, the often-mentioned evidence of a positive relationship between alliance and movement synchrony must be called into question; the situation seems to be more complex. Thus, our failure to find an association between therapeutic alliance and movement synchrony may be either due to the limited statistical power of our study to detect very small effects, or to other moderators of this association that have not yet been identified. Alternatively, the experimental sessions may have been too brief to establish a robust alliance.

With respect to leading and pacing, the existing literature also provides contradictory findings, as pointed out in the introduction. There are only few studies that assess this specific dynamic and these studies show substantial variations with respect to patients, conditions and outcomes. In the present study we found indications that leading behavior differed depending on the experimental condition. However, since this was not part of our hypotheses, this finding must be regarded as exploratory and cannot be used to support any conclusions. Nevertheless, these findings may serve as a foundation for further research to contribute to a more comprehensive understanding of the underlying dynamics.

Our sensitivity analyses yielded several changes in the results. Excluding the seven cases with guideline deviations resulted in the interaction effect of H1 and the main effect of H3 no longer reaching significance, while also resulting in the emergence of a significant interaction in H2. This suggests that the omitted sessions may have exerted a considerable influence on our findings. It is difficult to determine whether the characteristics present in these sessions acted as confounders or enriched our samples. On the one hand, excluding these sessions may have increased the homogeneity of our groups and thus produced more robust results. On the other hand, it’s possible that removing these sessions resulted in the loss of valuable data that could have contributed to more representative results. In addition, the disappearance of significant results may be due to a decrease in test power.

Our study offers several strengths for advancing the understanding of movement synchrony in therapeutic interactions. Firstly, we have shifted this type of research to a more controlled experimental setting. This allowed us to directly compare two different types of counseling, which to our knowledge has not been done before. Another strength of our approach is that it allows us to divide the session into several shorter segments that can be assigned to conditions. Thereby we compare for the first time structured interventions on a within-session basis. Such an approach allows for a more fine-grained understanding of the role of movement synchrony in psychotherapy sessions.

Although our manualized and experimental approach, involving participants rather than patients, has inherent limitations regarding ecological validity, the counseling sessions were structured to resemble authentic therapeutic encounters. This enhances the fidelity of our study to real-world contexts. The sessions consisted of an initial problem description part, followed by standardized and free intervention parts. This design allowed for the assessment of synchrony between different phases of the session, facilitating a detailed analysis of movement patterns throughout the therapeutic process, while the experimental design ensured a clear differentiation between intervention strategies. The inclusion of a free intervention was intended to serve as a test against the standardized intervention, as the latter’s standardization could potentially affect movement synchrony. However, although the emulation of authentic therapeutic interactions is carefully structured, it may not fully capture the subtleties of real-world counseling dynamics.

Our study is subject to several further limitations. One is that the video-based method used operates in a two-dimensional space, neglecting the three-dimensional nature of real-world interactions and resulting in the loss of significant motion information. While first progress has been made, e.g. by computing motion energy on extracted body keypoints (Fujiwara & Yokomitsu, 2021), it remains limited to two dimensions. Future research is needed to explore options for cost-effective 3D analysis (e.g. VideoPose3D, Pavllo et al., 2019).

Second, we used the rMEA package to obtain synchrony scores, which computes an overall mean of all calculated synchrony values (from different lags and windows) per session. Although this approach is common, it has faced criticism for neglecting several aspects, such as the transitions between different states of coordination (García & Di Paolo, 2018), the frequency of synchronous events (Schoenherr et al., 2019a) and the reasons for interactants to enter and interrupt synchronous behavior (Mayo & Gordon, 2020). In addition, this approach (rMEA) uses the absolute value of the cross-correlation, rather than separating between positive and negative correlations. This aspect has been discussed recently (cf. Coutinho et al., 2019; Tschacher & Meier, 2020) as it treats not only in-phase but also anti-phase behavior as synchrony events. However, the majority of the studies in the literature using MEA have used absolute values and thus this approach allows for the respective comparisons with the existing literature. Future studies should consider alternative approaches, such as peak-picking algorithms, frequency assessment (Altmann, 2011) or autoregressive models such as mlVAR (Epskamp et al., 2018). Additionally, the use of commonly chosen values for bandwidth, step size and lag size, which to our knowledge have not been validated empirically, may have impacted the synchrony results (Moulder et al., 2018; Schoenherr et al., 2019a). As outlined by Schoenherr et al. (2019a), our approach aligns with the so called WCLCS subtype, which examins strength rather than frequency in windows cross-lagged correlation analysis. Schoenherr et al. advocate for smaller window sizes (3–5 s) as the optimal configuration to detect synchrony. However, they acknowledge that larger windows, such as ours (60 s), may effectively evaluate overall synchrony strength across interactions, emphasizing the importance of meticulous parameter selection. One potential issue with larger window sizes and a non-overlapping step-size is that the evaluated time series may not be entirely evaluated for synchrony, since there are usually remnants at the end of each time series that are truncated because they are not large enough for a new window. If the tail of the series is of particular interest, e.g. to evaluate turn-taking, our synchrony evaluation approach may not be the most suitable for the task.

Another limitation of our study is the variability in the number of sessions conducted by different counselors. Although we ensured that each participant interacted with different counselors during both of their sessions, the difference in the number of sessions conducted by individual counselors may introduce a potential source of variation. For instance, two counselors contributed only three and four sessions, respectively, while another conducted 25. This fact prohibited the analysis of a potential counselor effect due to the limited number of sessions of some counselors.

In conclusion, our study provides valuable insights into movement synchrony in two different therapeutic approaches. It shows that different therapeutic strategies affect the non-verbal components of interaction. On the other hand, we found no evidence that either a problem-focused or a solution-focused approach had a significant impact on the therapeutic alliance as rated by the client. Overall, our study contributes to the growing body of literature on the importance of dyadic variables in therapeutic processes, particularly movement synchrony. Further research is needed to better understand the complex interplay between these dyadic variables and their impact on therapeutic outcomes.