Empathy in psychotherapy: subjective ratings versus remote biosensing of interpersonal heart rate synchrony as outcome predictors
=================================================================================================================================

* Clara C. Gernert
* Peter Falkai
* Christine M. Falter-Wagner

## Abstract

In addition to understanding empathy in an affective and cognitive dimension, the physiological domain plays a crucial role, especially in the emotional dynamics of interpersonal interactions during psychotherapy. Within the complex bio-psycho-social system of cognitive behavioural therapy language, cognition, emotion and physiological states of both, client and therapist, intertwine through interaction dynamics. The current study aimed to explore interpersonal physiological dynamics during psychotherapy sessions as an objective biomarker for predicting therapy outcome. In a follow-up assessment design, involving 25 client-psychotherapist dyads, wearable sensors monitored individual’s heart rate, while video cameras recorded movement behaviour during regular cognitive behavioural therapy sessions. Post-session reports and symptom questionnaires were collected from both, clients and therapists, after each session. Results showed that synchrony in head movement and heart rate emerged during psychotherapy sessions. Notably, heart rate synchrony from the initial session predicted changes in patients’ self-rated global severity index over time. The objective predictor, heart rate synchrony, emerged as particularly robust, surpassing patients’ subjective ratings of affiliation in explaining a higher variance of the therapy outcome variable. These findings highlight the potential shown by remote biomarker sensing of interpersonal dynamics for the prediction of psychotherapeutic effectiveness.

Keywords
*   psychotherapy research
*   heart rate
*   interpersonal synchrony
*   objective biomarkers
*   therapy outcome
*   cognitive behavioural therapy

## Introduction

The concept of empathy has been extensively studied across various disciplines over the last century. Initially, Vischer introduced the term “Einfühlung” (“feeling into”) in 1873, describing the process of projecting feelings onto other people and objects by perception1. Lipps expanded on this concept, emphasizing the psychological perspective of understanding others by inner imitation (Lipps cited in2). Titchener later translated “Einfühlung” into “empathy”2. Ferenczi introduced the concept of empathy into the field of psychotherapy. He stated that empathic understanding serves as a positive alternative to the traditional interpretation of resistance in psychoanalytic therapy, particularly in trauma treatment3. He emphasised the therapist’s role in demonstrating responsiveness to the patient’s needs. Ferenczi used the term “elasticity”, later also known as “relaxation” (“Nachgiebigkeit”), to describe a necessary ability of the analyst. This ability creates space for the analysand’s feelings and thoughts, fostering the therapeutic alliance and facilitating progress3. Similarly, Rogers and Kohut emphasized the importance of empathy within their respective therapeutic approaches, viewing it as essential for facilitating therapeutic change4,5. However, the definition of empathy is complex and non-consensual across different disciplines. In the current study, we adopted the definition of empathy proposed by Husserl6 and Singer et al.7: Empathy is a dynamic process involving experiencing and understanding the subjective feelings and thoughts of the other while being aware of the distinct individuality6,7,8. This dynamic process, i.e., empathy, involves reciprocity and mutual awareness, leading to a triadic structure involving the empathizer, the empathised, and both forming a unique we-identity6. Research has shown that therapist’s empathy was a predictor for therapy outcome, surpassing the influence of specific interventions or therapist qualities9,10. The more empathetically the therapist is able to respond to the client’s needs, the more likely it is that the client experiences being understood and validated11. Client- perceived empathy has been associated with a decrease in negative self-treatment and symptom levels, as well as with changes in the client’s attachment style and interpersonal interaction12. Empathy, regarded as a relationship variable, plays a role in facilitating the development of a positive therapeutic alliance13. Therapeutic alliance operationalised as a “collaborative and effective bond between therapist and patient” (p.438 in14), has been linked to predicting clinical outcome15 and therapeutic success16. Evidence exists that observing others in distress can elicit emotional arousal and autonomic responses in observers17,18. Beyond measuring the physiological state and dynamics of a single person, theories came up suggesting interpersonal coordination might be more indicative of capturing empathy, which is to be understood interpersonally by definition19,20. Bernieri and Rosenthal defined interpersonal dynamics, also known as interpersonal synchrony (IPS) as21: “the degree to which the behaviors in an interaction are non-random, patterned or synchronized in both timing [and] form.” Evidence exists that IPS is not only limited to non-verbal behaviour, such as movement synchrony22, but also emerges in speech23, physiology7,24,25,26,27,28, and brain activity29,30. IPS is supposed to be fundamental in fostering social relationships and cohesion by showing associations with rapport31,32, affiliation33, successful cooperation25,34, prosocial commitment35, and empathy20,36. From a developmental perspective, Feldman proposed that biobehavioral synchrony is a core mechanism for acquiring resilience37. Over the last decades there has been growing interest in exploring IPS also in the field of psychotherapy38. While research on movement synchrony in psychotherapy focused on therapy outcome, synchrony in physiology and other modalities was mostly linked to therapeutic alliance38. Already in the 1960s it was suggested that empathy leads to mirroring of the autonomic nervous system between individuals39,40. Physiological IPS in psychotherapy was explored using variables like heart rate dynamics27,41, respiration27, and skin conductance20,26. Cardiac activity is a well-established biomarker for affective and cognitive states, as well as a parameter of sympathetic and parasympathetic arousal42. Heart rate (HR) activity has emerged as a more sensitive marker for empathic responses compared to skin conductance43. Additionally, the technique of heart rate variability biofeedback (HRVB), focusing on single-person cardiac activity, has shown promise in alleviating depressive symptoms44. Despite this, few studies have examined the impact of interpersonal HR synchrony, primarily focusing on its correlation with subjective ratings such as therapeutic alliance27 or empathy41, but not on therapy outcome. While empathy and therapeutic alliance are often subjectively assessed, with the danger of introducing inconsistencies and biases, current technology could offer the perspective of finding objective and quantifiable biomarkers to predict therapy outcome on the basis of interpersonal dynamics.

The current proof-of-concept study focussed on measuring interpersonal HR synchrony in naturalistic settings of cognitive behavioural therapy (CBT). HR was continuously monitored using wearable sensors during therapeutic interaction between psychiatric patients and their respective therapists. Body and head movement were captured using a stationary video camera. Following each CBT session, patients and therapists rated their subjective experience of therapeutic alliance and progress by completing a post-session report. Additionally, patients reported about symptoms and symptom severity using standardized questionnaires. HR synchrony was analysed as a potential predictor of therapy outcome, defined herein as patient-reported change in symptom severity between the initial and a follow-up session. Firstly, we aimed to confirm that IPS emerged during dyadic CBT sessions. While we assessed both physiological and movement IPS, the further analysis in the current study focused on physiological IPS, with movement IPS falling outside its scope. Secondly, we aimed to investigate whether HR synchrony is related to patients’ or therapists’ post-session report ratings of therapeutic alliance. Thirdly, we aimed to analyse whether physiological synchrony or subjective ratings predict therapy outcome and whether such a relationship interacts with patients’ initial symptom severity.

## Methods

### Participants

The study comprised 25 patient-therapist dyads. Patients were recruited from the Clinic for Psychiatry and Psychotherapy (LMU University Hospital, Munich), encompassing both inpatient and outpatient departments. Participants with cardiac disease or implants, acute substance abuse, high-dose beta- blockers, acute neurological disorders or severe skin defects were excluded from participation. All participants had either normal or corrected-to-normal vision. The therapists involved were not part of the research team and had no engagement in planning the study or analysing the datasets. The follow- up group consisted of 20 dyads, each participating in two CBT sessions, while five dyads participated only once. Six out of the 20 dyads in the follow-up group were eliminated from physiological data analysis due to insufficient signal quality found in the HR time series of at least one member within each dyad. *Kubios HRV Scientific* software45 was used to automatically detect noise in the blood volume pressure (BVP) signal. Dyads with over 25% corrected beats in at least one of the dyad members’ time series were excluded from further analysis. Regarding the outcome variable of patients’ symptom change over time, only dyads that underwent assessment twice and provided fully answered questionnaires from both sessions were included in the analysis. One dyad of the follow-up group did not complete all questionnaires and was consequently excluded. Therefore, the analysis included 13 dyads whose CBT sessions (n = 26) were at least two weeks apart. The follow-up group comprised a mixed clinical patient sample (69.2% female, 30.8% male, Mage = 32.3 ± 14.2). Each patient had at least one (n = 11) or two (n = 2) diagnoses of a mental disorder according to ICD-10, Chapter V(F)46. The follow-up group included 11 different therapists (72.7% females, 27.3% males, Mage = 30.5 ± 7.7), with two therapists participating with two different patients. A full demographic summary is shown in Tables 1-3 (see the supplementary materials). All participants provided written informed consent, and the study received IRB approval from the LMU Medical Faculty Ethics Board (19-170).

### Study design

The study was conducted within a designated therapy room at the Clinic for Psychiatry and Psychotherapy (LMU University Hospital, Munich). The experimental setup involved the use of a research-grade wristband sensor, provided for both patient and therapist. These sensors continuously captured physiological parameters non-invasively. To mitigate movement artefacts, the wearable device was worn on the non-dominant hand. For capturing dyad members’ body and head movements during the interaction, a fixed video camera recorded the entire CBT session (Mduration= 48 ± 4.2 min). Throughout the session, patients and therapists were seated on chairs, facing each other at a 60-degree angle. Language and spoken content were neither recorded nor analysed. At the beginning of each CBT session, therapists were instructed to set a timestamp with the wristband to facilitate temporal synchronization of both datasets. Subsequently, the dyad was left alone in the room for the entire CBT session with no specific instructions regarding spoken content. Following the CBT session, both patients and therapists completed a post-session report questionnaire and patients additionally provided ratings of their somatic and psychological symptom burden. A minimum of two weeks passed before recording a follow-up session, maintaining the same study setup.

### Materials

#### Movement data

The CBT sessions were recorded using a standard video camera with a frame rate of 25 frames per second (fps). The entire session, excluding the initial five minutes, was used for assessing movement IPS, using the software *Motion Energy Analysis V3.10* (MEA)47. MEA is a frame-differencing method quantifying grayscale pixel changes in pre-defined regions of interest (ROI) (for more details see48). In the current study, two distinct ROIs were defined, focusing on head and upper body movement.

#### Heart rate

The *Empatica* wristband sensor E4 (Empatica, Milan) was used to monitor participants’ HR throughout the CBT sessions. This sensor uses photoplethysmography to measure the blood volume pulse (BVP), serving as a parameter for cardiac output, with a sampling frequency of 64 Hz. Alongside BVP data, *Empatica* provides a HR time series for each dataset, containing average HR values sampled at a frequency of 1 Hz. These values are computed in 10-second intervals based on BVP data. The signal quality of participants’ BVP time series was assessed using *Kubios HRV Scientific V4.0.0* software 45,49. Standard parameters were used for decoding raw BVP data. An overview of these parameters is provided in Table 5 (see the supplementary material). Subsequently, *Kubios HRV Scientific V4.0.0*49 automatic beat correction algorithm was applied, identifying artefacts by analysing differences between successive RR (dRR) interval time series (for more details, see49). To ensure data integrity, the initial five minutes of each session were excluded from any analysis to account for significant motion artefacts and allow dyads to acclimate to the setting. Following this, the entire psychotherapy session was considered for data quality check and analysis. HR time series with a beat correction rate exceeding 25%, determined by *Kubios HRV Scientific V4.0.0*45, were excluded from further analysis. The analysed signal was HR, measured in beats per minute (bpm). A summary of the HR statistics can be found in Table 4 and Figure 1 (see both in the supplementary material).

![Figure 1](http://medrxiv.org/http://medrxiv.stage.highwire.org/content/medrxiv/early/2024/08/30/2024.08.29.24312787/F1.medium.gif)

[Figure 1](http://medrxiv.org/content/early/2024/08/30/2024.08.29.24312787/F1)

Figure 1 
Heart rate synchrony predicts change in patients’ global severity index. The linear trendline and its 95% confidence interval (shaded area) illustrate the prediction of change in patients’ global severity index (ΔGSI). HR synchrony was calculated using the peak-picking method for a lagged-windowed cross-correlation analysis (see *Methods*). Positive HR synchrony values denote in-phasic synchrony during a CBT session, while negative values indicate anti-phasic synchrony. Positive ΔGSI values indicate a reduction in patients’ initial GSI over time, while negative ΔGSI values indicate an increase in GSI. A ΔGSI value of zero reflects no change in patients’ GSI ratings over time. This objective predictor explained 61% of the variance in ΔGSI.

#### Self-report questionnaires

After the CBT session, patient and therapist completed a specific post-session report (*BPSR-P 2000/ BPSR-T 2000)*50. The questionnaire captures the patient’s subjective rating on eight subscales and the therapist’s rating on eleven subscales. Items are scored on a seven-point scale, where higher scores reflect a more positive agreement to the question. As we were specifically interested in the therapeutic relationship and the interactional perspective, two subscales for each group were selected for further analysis. *Therapeutic alliance* and *affiliation* were considered for the group of patients, and *therapeutic alliance* and *progress* were selected for the group of therapists (see Table 6 in the supplementary material). The selection of subscales is similar to previous studies (e.g., see27).

Patients’ depressive symptomatology was assessed using the German version of the *Beck Depression Inventory – Second Edition* (BDI-II)51. The BDI-II consists of 21 questions referring to the last 14 days, where each answer is scored with a value of 0 to 3. Higher total sum scores indicate a more severe depressive symptomatology. Patients’ overall symptom severity was evaluated using the German version of the *Brief Symptom Inventory* (BSI)52. The BSI measures psychological distress and clinically relevant somatic symptoms experienced over the past seven days on a four-point scale. The questionnaire consists of 53 items covering the following nine dimensions: somatization, obsession- compulsion, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism. *Global Severity Index* (GSI) and *Positive Symptom Distress Index* (PSDI) are two key metrics from the BSI. While the GSI score is a strong indicator of the client’s overall psychological distress experience, the PSDI provides information about the intensity level of positively rated dimensions. In our follow-up sample, BDI, GSI and PSDI scores were assessed after both sessions. For analysing therapy outcome, we were particularly interested in their change over time (i.e., ΔGSI and ΔPSDI). Positive ΔGSI values indicate a reduction of patients’ global symptom severity over time, while negative ΔGSI values indicate an increase. The same interpretation applies to ΔPSDI. For descriptive statistics see Tables 7-10 (see the supplementary material).

### Data Analysis

Data pre-processing and analysis of interpersonal physiological and movement synchrony were conducted in *RStudio 2023.09.0*53 using the statistical programming language *R V4.3.1*54.

#### Movement synchrony

To address the non-stationarity inherent in movement data and account for time delays between individuals’ responses, a lagged windowed cross-correlation analysis was employed for both ROIs. The analysis was conducted using the *rMEA* package55. The time series were cross-correlated using absolute values for non-overlapping window segments of 1-minute duration and a time lag of 5 seconds. These parameters follow previously used settings established in research on movement synchrony in psychotherapy22,47. A peak-picking algorithm was applied, where for each window the maximum absolute cross-correlation value across all different lags was detected. Absolute cross-correlation peaks were Fisher Z-transformed and aggregated by calculating the mean of all peaks per ROI, resulting in two movement IPS scores per session and dyad (i.e., head and body synchrony). In a control analysis, we compared movement IPS values between original and surrogate dyads. Surrogate dyads were created by data shuffling56. Each patient’s head and body movement time series were shuffled randomly to create a new time series, using 10 permutations. The new time series was paired with the unchanged time series from the belonging therapist. Surrogate dyads’ levels of movement synchronization were calculated using parameters identical to those used for the original dyads by applying a lagged windowed cross-correlation analysis. An average value out of all surrogate synchrony indices per ROI and per dyad was calculated.

#### Heart rate synchrony

HR time series were first screened for suspicious samples according to two criteria57: (i) samples exceeding 180 bpm or falling below 45 bpm were replaced, deeming such values as unrealistic within the current setting, and (ii) samples deviating by more than 25% from the value 1sec before were omitted. Such changes in HR within a brief time are considered unrealistic for HR data58. To account for the non-stationarity of the HR signal, the time series was split into non-overlapping segments of 4 minutes. For quantifying HR synchrony, a lagged window cross-correlation analysis with non-absolute values was calculated for each segment, using non-overlapping windows of 1-minute duration and a lag size of 30 seconds. Non-absolute cross-correlation values were used to differentiate in-phase from anti- phase HR synchrony59. Anti-phase HR synchrony can be understood by the example of one person’s HR is constantly high while the other one’s HR is constantly low. For each window within a segment, the maximum absolute cross-correlation value was detected across all different lags, referred to as the *peak cross-correlation*. Subsequently, the original sign of the non-absolute cross-correlation value was assigned to the absolute peak. Following this, all peaks per segment were averaged, and ultimately, the overall mean of all mean peaks was computed per dyad. We therefore obtained a measure of the strength of synchrony for each dyad. As for movement synchrony, HR synchrony scores from original dyads were compared to scores from surrogate dyads. Surrogate time series were generated by randomly shuffling the sequence of both dyad members’ time series, whereas means, distribution and trends of the time series are preserved. Surrogate synchrony values were calculated by applying a lagged window cross-correlation analysis with the same parameters as for original dyads.

#### Statistical Analysis

For the control analysis of body movement and HR synchrony, independent t-tests were applied to assess the significance between original and surrogate dyads. The Shapiro-Wilk test confirmed normally distributed synchrony scores for both groups in both conditions (p > .05). Head movement synchrony, exhibiting a deviation from normal distribution for the original data in the Shapiro-Wilk test (p < .05), was evaluated using a Mann Whitney U-test. HR synchrony measurements exceeding the level of a random effect were included in the following three main analyses, to (i) investigate whether interpersonal physiological synchrony in CBT sessions is associated with patients’ or therapists’ post- session report of alliance, and (ii) analyse whether physiological synchrony or subjective ratings of therapeutic alliance are related to therapy outcome, and (iii) test whether such a relationship is bound to interaction with patients’ initial symptom severity. For the first part, a correlation analysis was performed using patients’ and therapists’ rating scores, as well as HR synchrony indices. All parameters were measured after or in the first session of the follow-up group. For the second part, two follow-up outcome variables were added to the correlation analysis (i.e., ΔGSI, ΔPSDI). ΔGSI as an index of change in patients’ psychological distress experience and ΔPSDI as an index of the change in patients’ symptom intensity over time. Variables showing significant correlations with these outcome variables were used as predictors in a simple linear regression model. All linear regression models met the necessary statistical requirements, including independence of errors and homoscedasticity. Additionally, a stepwise regression analysis was conducted to examine whether subjective ratings or interpersonal HR synchrony scores were a better predictor for ΔPSDI. For the third analysis, assessing the presence of an interaction effect, we examined both the main effect and two-way interaction effects of HR synchrony measures and the initial symptom burden (i.e., PSDI_S1) as predictors for the outcome variable (i.e., ΔPSDI). Both variables, HR synchrony and PSDI\_S1 were centred first to facilitate a clearer interpretation of the interaction effects. The variance inflation factor was calculated to control for multicollinearity.

## Results

### Synchrony versus pseudo-synchrony

Body movement IPS scores from original dyads (N = 12, M = 0.220, SD = 0.019) did not significantly differ from surrogate dyads’ body movement synchrony scores (N = 12, M = 0.209, SD = 0.017), t(22)= -1.476, *p* >.05). Original dyads showed significantly higher head movement synchrony scores (N = 12, M = 0.230, SD = 0.035) compared to surrogate dyads (N = 12, M = 0.205, SD = 0.010), *U*=34.0, p = .03). HR synchrony values from original dyads (N = 13, *M* = -0.215, *SD* = .27) significantly exceeded surrogate synchrony values (N = 13, *M* = -0.016, *SD* = .079), t(24) = 2.549, *p* < .05).

### Interpersonal physiological synchrony and post-session ratings

HR synchrony, measured in the first session of the follow-up group, showed a significant positive correlation with patients’ post-session rating scores of affiliation (*r* = .736, *p* = .004). Even after applying the Bonferroni correction (*p*B1 = 0.007), this correlation retained statistical significance. Therapists’ assessments of therapeutic progress showed a significant positive association with HR synchrony scores, where the therapist was leading (*r* = .608, *p* = .027). However, this correlation lost significance after Bonferroni correction (*p*B1 = 0.007). Notably, no significant associations were observed between patients’ subjective ratings and therapists’ post-session evaluations (p > .05). A comprehensive summary is provided in Table 11 (see the supplementary material).

### Prediction of symptom change

Both therapists’ ratings of progress and patients’ ratings of affiliation showed a positive association with ΔGSI (p < .05). However, these associations lost significance after applying the Bonferroni correction (*p*B2 = 0.006). Furthermore, HR synchrony with the therapist leading was positively correlated with ΔGSI (r=0.781, p = .002), remaining significant after Bonferroni correction (*p*B2 = 0.006). A simple linear regression model indicated that this physiological synchrony measurement significantly predicted ΔGSI, (*R*2 = .610, *F*(1,11) = 17.22, *p* = .002, β = 0.792). HR synchrony as an objective predictor explained 61% of the variance in ΔGSI. As depicted in Fig. 1, the positive slope shows higher in-phasic HR synchrony with the therapist leading was associated with a greater reduction of patients’ overall psychological distress experience over time. Additionally, a strong anti-phasic HR synchrony with the therapist leading was associated with an increase in patients’ GSI over time.

Patients’ ratings of affiliation showed a significant positive association with ΔPSDI (*rho* = .741, *p* = .004), as did HR synchrony (rho=0.841, p < 0.001). Both correlations remained significant after Bonferroni correction (*p*B2 = 0.006). A simple linear regression model revealed that patients’ ratings of affiliation on its own significantly predicted ΔPSDI (*R*2 = .432, *F*(1,11) = 8.362, *p* = .015, β = 0.047). Higher rating scores of affiliation were associated with a decrease in symptom intensity levels, and vice versa. HR synchrony on its own also significantly predicted ΔPSDI (*R*2 = .459, *F*(1,11) = 9.339, *p* = .011, β =0.756), showing that higher in-phasic HR synchrony was associated with a decrease of PSDI over time and high anti-phasic HR synchrony was associated with an increase of patients’ PSDI scores in the follow-up session.

However, combining both objective and subjective predictors in a stepwise regression model, HR synchrony was the sole predictor left explaining 46% of variance in ΔPSDI and there was no significant proportion of variance explained by subjective ratings over and above HR synchrony. Fig. 2 gives an impression of both linear models.

![Figure 2](http://medrxiv.org/http://medrxiv.stage.highwire.org/content/medrxiv/early/2024/08/30/2024.08.29.24312787/F2.medium.gif)

[Figure 2](http://medrxiv.org/content/early/2024/08/30/2024.08.29.24312787/F2)

Figure 2 
Predicting therapy outcome by subjective and objective parameters. **A**) Change in patients’ symptom intensity levels (ΔPSDI) were predicted by patients’ post-session ratings of affiliation, assessed after the first measured session of the follow-up group. This subjective predictor explained 43% of the variance in ΔPSD. **B**) ΔPSDI was also predicted by interpersonal heart rate (HR) synchrony calculated from the first measured session of the follow-up group, using the peak-picking method applied to a lagged-windowed cross-correlation analysis (see *Methods*). Positive HR synchrony values indicate in-phasic synchrony during CBT sessions, while negative values indicate anti-phasic synchrony. This objective predictor explained 46% of the variance in ΔPSDI. The plots in (A) and (B) depict the linear trendline in red, along with its 95% confidence interval (shaded area). Positive ΔPSDI values indicate a reduction of patients’ initial symptom intensity (PSDI) over time, while negative values denote an increase. A value of zero reflects no change between both measured sessions. Comparing both predictors from (A) and (B) in a stepwise regression model, only HR synchrony remained as the significant predictor for ΔPSDI, with no additional significant variance explained by subjective ratings beyond HR synchrony.

Given that patients could profit differently from therapy and show different associations with HR synchrony depending on initial symptom burden, we entered the latter as a predictor to the HR synchrony model. Indeed, there was a significant interaction effect between HR synchrony and patients’ initial symptom intensity level (PSDI_S1), measured in the first session (β = -1.119, p = 0.041), indicating that the relationship between HR synchrony and ΔPSDI differed depending on the level of the initial PSDI burden. However, the main effect of patients’ initial symptom intensity burden was insignificant (β = 0.154, p = 0.129). The interaction model significantly explained 70% of the variance in ΔPSDI (R² = 0.697, F(3, 9) = 6.886, p = 0.010) (see Fig. 3). The VIF values were all smaller than 1.7, suggesting that multicollinearity did not influence these results.

![Figure 3](http://medrxiv.org/http://medrxiv.stage.highwire.org/content/medrxiv/early/2024/08/30/2024.08.29.24312787/F3.medium.gif)

[Figure 3](http://medrxiv.org/content/early/2024/08/30/2024.08.29.24312787/F3)

Figure 3 
Interaction of initial symptom burden and heart rate synchrony in predicting therapy outcome. Prediction of change in patients’ symptom intensity level over time (ΔPSDI) based on the interaction effect of heart rate (HR) synchrony and patients’ initial symptom intensity level (PSDI_S1). HR synchrony is measured in the first session of the follow-up group, using a lagged-window cross-correlation combined with the peak-picking algorithm (see *Methods*). The black line represents the fitted regression line with a 95% confidence interval shaded in light grey. Higher PSDI\_S1 values, indicated by red-coloured dots, reflect greater symptom intensity levels in the first session, while lower values, represented by blue-coloured dots, suggest a less severe initial symptom burden. The interaction effect and the main effect of HR synchrony were significant, while the main effect of PSDI\_S1 was not. Overall, the interaction model explained 70% of the variance in ΔPSDI.

An overview of correlations between objective and subjective variables, as well as additional regression models and details of the interaction model can be found in Tables 12 – 15 (see the supplementary material).

## Discussion

Empathy is believed to foster the development of therapeutic alliance and is recognized as a robust predictor for therapeutic success60. However, the conventional subjective assessment of therapeutic alliance and empathy arguably overlooks the intricate intertwining of client and therapist not only through language, cognition, and emotions but also in their physiological states, suggesting a form of “embodied empathy”. In the current study, we aimed to address this gap by incorporating physiological data collection alongside subjective reports from therapists and patients in a follow-up assessment of CBT. Wearable sensors continuously measured heart rate dynamics, while questionnaires were used to capture subjective ratings of therapeutic alliance and patients’ symptomatology.

An initial sanity check revealed significantly higher levels of HR synchrony in real dyads compared to randomly generated surrogate dyads, highlighting that IPS existed as a non-random phenomenon within our recorded CBT sessions. This is in line with previous research reporting IPS in psychotherapy38. We found a positive association between overall HR synchrony and patients’ subjective ratings of affiliation. Additionally, we observed that HR synchrony, with the therapist having the lead, was positively correlated with therapists’ ratings of therapy progress. Both findings support prior research, linking physiological IPS to therapeutic alliance27. In contrast, we did not find a significant correlation between patients’ and therapists’ post-session ratings of therapeutic alliance. This discrepancy underscores the occurrence of divergent subjective experiences among dyad members, despite their shared interaction. Such variability is common not only within the therapeutic context but also in daily interactions, reflecting the aspect of individuality in patterns of cognition, experiences, motivation, beliefs, and personality structure. In psychotherapy, where patient’s experiences take precedence over those of the therapist, reciprocity as well as mutual responsiveness regarding each other’s experiences and emotions are therefore limited. Patients’ mental health conditions could serve as another confounding variable. More than half of the patients included in our follow-up group were clinically diagnosed with an affective disorder, primarily depression. Depressive symptoms such as low mood, loss of interest, low energy, or lack of pleasure (see ICD-1046) can also contribute to difficulties in interaction61. However, this insignificant finding underscores the limitations of the use of subjective ratings for predicting therapy success.

Given that both factors on their own, i.e. subjective ratings of affiliation by patients and HR synchrony, significantly predicted symptom change, we performed a stepwise regression to see, whether subjective ratings add any variance explanation over and above HR synchrony. Yet, HR synchrony was the only predictor left, indicating the higher variance explanation and no additional predictive value of subjective ratings. HR synchrony explained 46% of the variance in ΔPSDI. Specifically, greater in-phase HR synchrony was associated with a greater reduction in symptom intensity, whereas high anti-phase HR synchrony was linked to an increase in symptom intensity levels.

An important aspect to consider is the question whether patients profit differently depending on initial symptom burden. Hence, we discovered that the relationship between HR synchrony and ΔPSDI varied depending on the level of patient’s initial PSDI burden. This interaction model explained 70% of the variance in ΔPSDI. The negative beta coefficient suggests that the relationship between HR synchrony and ΔPSDI becomes more negative as the initial PSDI burden increases. This implies that low initial symptom intensity may be associated with more in-phase synchronous HR patterns, potentially contributing to symptom reduction over time. However, our findings report statistical associations and do not imply causation. Notably, the initial symptom intensity burden alone did not significantly predict ΔPSDI. This significant interaction model underscores the importance of considering both subjective and objective factors for therapy outcome prediction. While objective measures like HR synchrony offer valuable insights, our results also highlight the significance of subjective ratings by the patient, such as the initial symptom burden, as a moderating variable for optimizing prediction models for therapy outcome.

Previous research has described IPS as a process variable, with depressive patients exhibiting low interpersonal movement synchrony at the beginning but showing an increase in IPS synchrony over time62. Further research is necessary to analyse the causal relationship, including whether specific psychiatric symptoms are associated with decreased HR synchrony in interaction, whether this effect remains consistent across various interaction partners (i.e., therapists), and whether the level of HR synchrony increases in parallel with decreasing symptom burden over time.

While showing promising results with respect to the use of objective biosensing of interpersonal dynamics, this proof-of-concept study is limited due to relatively small sample size and exclusion of datasets with high artefact rates in physiological data. As we were interested in physiological IPS, low data quality from even one dyad member led to the exclusion of the whole dyad. Wearable sensors are usually more sensitive to motion artefacts, leading to lower data quality. While most physiological IPS research traditionally relied on classical devices like electrocardiograms for HR measurements, our study highlights the potential of wearable devices in capturing physiological dynamics during CBT. Those devices are less intrusive, which is important when aiming to capture naturalistic interaction, particularly in the setting of psychotherapy, where content is often quite intimate.

Additionally, the heterogeneity of our patient group might limit the generalizability of our findings. Nevertheless, among the participants in our study, over half presented with at least one diagnosis falling within the spectrum of affective disorders (ICD-10 F3 diagnosis46). This cohort is notably representative of patients commonly encountered in the setting of CBT. Given the frequent occurrence of comorbid diagnoses among psychiatric patients in clinical practice, our findings underscore the potential utility of physiological IPS as a transdiagnostic measure for therapy outcome. In our proof-of-concept study, our aim in terms of basic research was to explore the potential of using wearable sensors in CBT to capture interpersonal physiological dynamics. Future research is warranted to validate our findings in a larger sample size and conduct group-wise comparisons among individuals with various psychiatric disorders. Furthermore, the follow-up assessment period in our study was relatively short, spanning a minimum of two weeks, due to the recruitment setting in a psychiatric clinic characterized by high patient turnover. This duration was chosen as a compromise to ensure the feasibility of conducting a follow-up assessment, which was crucial for measuring therapy outcome, within the clinic’s constraints while still allowing for meaningful data collection. Nevertheless, we observed changes in symptom severity scores over time in the majority of patients. Due to the sample size, we could only run regression models with two predictor variables at the same time. By having a larger sample size, multivariate analysis should be calculated to control for confounders and interaction effects. However, additional analyses controlling for potential confounders (e.g., age and sex difference) confirmed the robustness of HR synchrony as a significant predictor for ΔPSDI (see Table 14 in the supplementary material).

In conclusion, our study demonstrates the value of utilizing wearable sensors in CBT settings. It emphasizes the predictive strength of HR synchrony as an objective biomarker for therapy outcome, surpassing subjective ratings of therapeutic alliance. Moreover, integrating the concept of embodied empathy into further research on evidence-based psychotherapy might have the potential to tailor therapy approaches and enhance patient-therapist matching, ultimately leading to improved treatment outcomes.

## Author contributions statement

All authors conceptualised the study. C.F.W. secured research funding. C.C.G. recruited participants and collected data. C.C.G. and C.F.W. developed the analysis plan. C.C.G. pre-processed and analysed the data. C.C.G and C.F.W. wrote the first draft of the manuscript. P.F. and C.F.W. supervised the study. All authors approved the manuscript.

## Data Availability Statement

The participants of this study did not give written consent for their data to be shared publicly, so due to the sensitive nature of the research supporting data is not available. Supplementary material is available in the online version.

## Additional Information

### Competing interests

The authors declare no competing interests.

### Funding

This work was supported by the FöFoLe-Programme 2018/2020 (Medical Faculty, LMU Munich, Promot 18/2018).

### Ethical Standards

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.

## Acknowledgements

We are thankful for participants’ commitment and contribution to our study.

*   Received August 29, 2024.
*   Revision received August 29, 2024.
*   Accepted August 30, 2024.


*   © 2024, Posted by Cold Spring Harbor Laboratory

The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission.

## References

1.  1.Vischer, R. On the Optical Sense of Form. in Empathy, Form and Space: Problems in German Aesthetics (eds. Mallgrave, H. F. & Ikonomou, E.), 1873–1893 (University of Chicago Press, 1994)
    
    

2.  2.Titchener, E. B. Lectures on the Experimental Psychology of the Thought Processes. (New York, The Macmillan Company, 1909).
    
    

3.  3.Ferenczi, S. The elasticity of psycho-analytic technique. in Final Contributions to the Problems and Methods of Psycho-Analysis Vol. 3 (ed. Balint, M.), 87–102 (New York: Bruner/Mazel, 1980).
    
    

4.  4.Kohut, H. Introspection, Empathy, and the Semicircle of Mental Health. International Journal of Psycho-Analysis 63, 395–408 (1982).
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=7152804&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

5.  5.Rogers, C. R. The Necessary and Sufficient Conditions of Therapeutic Personality Change. Journal of Consulting Psychology 21, 95–103 (1957).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/h0045357&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=13416422&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1957CBB1500001&link_type=ISI) 

6.  6.Husserl, E. Zur Phänomenologie der Intersubjektivität: Texte aus dem Nachlass. Zweiter Teil: 1921- 1928. (ed. Kern, I.) (Springer, 1972).
    
    

7.  7.Singer, T., Critchley, H. D. & Preuschoff, K. A common role of insula in feelings, empathy and uncertainty. Trends in Cognitive Science 13, 334–340 (2009).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.tics.2009.05.001&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=19643659&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000269411800005&link_type=ISI) 

8.  8.Zahavi, D. Second-Person Engagement, Self-Alienation, and Group-Identification. Topoi 38, 251–260 (2019).
    
    

9.  9.Bohart, A. C., Elliott, R., Greenberg, L. S. & Watson, J. C. Empathy. in Psychotherapy relationships that work (ed. Norcross J.) (New York: Oxford University Press, 2002).
    
    

10. 10.Luborsky, L., McLellan, A. T., Wopody, G. E., O’Brien, C. P. & Auerbach, A. Therapist success and its determinants. Arch. Gen. Psychiatry 42, 602–611 (1985).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1001/archpsyc.1985.01790290084010&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=4004503&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1985AJQ9500010&link_type=ISI) 

11. 11.Bohart, A. C. & Greenberg, L. S. Empathy and psychotherapy: An introductory overview. in Empathy reconsidered: New directions in psychotherapy (American Psychological Association, 1997).
    
    

12. 12.Watson, J. C., Steckley, P. L. & McMullen, E. J. The role of empathy in promoting change. Psychotherapy Research 24, 286–298 (2014).
    
    

13. 13.Castonguay, L. G. & Beutler, L. E. Principles of therapeutic change: A task force on participants, relationships, and techniques factors. Journal of Clinical Psychology 62, 631–638 (2006).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jclp.20256&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=16538667&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000237766500001&link_type=ISI) 

14. 14.Martin, D. J., Garske, J. P. & Davis, M. K. Relation of the therapeutic alliance with outcome and other variables: A meta-analytic review. Journal of Consulting and Clinical Psychology 68, 438–450 (2000).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/0022-006X.68.3.438&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=10883561&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000087651500009&link_type=ISI) 

15. 15.Ardito, R. B. & Rabellino, D. Therapeutic alliance and outcome of psychotherapy: Historical excursus, measurements, and prospects for research. Frontiers in Psychology 2, 1–11 (2011).
    
    

16. 16.Horvath, A. O., Del Re, A. C., Flückiger, C. & Symonds, D. Alliance in Individual Psychotherapy. Psychotherapy 48, 9–16 (2011).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/a0022186&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21401269&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000288524500003&link_type=ISI) 

17. 17.Eisenberg, N. et al. Relation of sympathy and personal distress to prosocial behavior: a multimethod study. J. Pers. Soc. Psychol. 57, 55–66 (1989).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/0022-3514.57.1.55&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=2754604&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1989AC43800005&link_type=ISI) 

18. 18.Stotland, E. Exploratory Investigations of Empathy. Advances in Experimental Social Psychology 4, 271–314 (1969).
    
    

19. 19.Levenson, R. W. & Ruef, A. M. Empathy: A Physiological Substrate. Journal of Personality and Social Psychology 63, 234–246 (1992).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/0022-3514.63.2.234&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=1403614&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1992JH48100005&link_type=ISI) 

20. 20.Marci, C. D., Ham, J., Moran, E. & Orr, S. P. Physiologic correlates of perceived therapist empathy and social-emotional process during psychotherapy. Journal of Nervous and Mental Disease 195, 103–111 (2007).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/01.nmd.0000253731.71025.fc&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=17299296&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000244398300001&link_type=ISI) 

21. 21.Bernieri, F. J. & Rosenthal, R. Interpersonal Coordination: Behavior Matching and Interactional Synchrony. in Fundamentals of nonverbal behavior Vol. 3 (eds. Feldman, R. S. & Rimè, B.) Ch. 11 (Cambridge University Press, 1991).
    
    

22. 22.Zimmermann, R. et al. Movement Synchrony in the Psychotherapy of Adolescents With Borderline Personality Pathology – A Dyadic Trait Marker for Resilience? Frontiers in Psychology 12, (2021).
    
    

23. 23.Amiriparian, S. et al. Synchronization in Interpersonal Speech. Frontiers in Robotics and AI 6, (2019).
    
    

24. 24.Algumaei, M., Hettiarachchi, I., Veerabhadrappa, R. & Bhatti, A. Physiological Synchrony Predict Task Performance and Negative Emotional State during a Three-Member Collaborative Task. Sensors 23, (2023).
    
    

25. 25.Behrens, F. et al. Physiological synchrony is associated with cooperative success in real-life interactions. Scientific Reports 10, 1–9 (2020).
    
    

26. 26.Gernert, C. C., Nelson, A., Falkai, P. & Falter-Wagner, C. M. Synchrony in psychotherapy: High physiological positive concordance predicts symptom reduction and negative concordance predicts symptom aggravation. International Journal of Methods in Psychiatric Research 33, (2024).
    
    

27. 27.Tschacher, W. & Meier, D. Physiological synchrony in psychotherapy sessions. Psychotherapy Research, (2019).
    
    

28. 28.Palumbo, R., et al. Interpersonal Autonomic Physiology: A Systematic Review of the Literature. Personality and Social Psychology Review 21, 99–141 (2017).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1177/1088868316628405&link_type=DOI) 

29. 29.Dikker, S., Silbert, L. J., Hasson, U. & Zevin, J. D. On the same wavelength: Predictable language enhances speaker-listener brain-to-brain synchrony in posterior superior temporal gyrus. Journal of Neuroscience 34, 6267–6272 (2014).
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6Njoiam5ldXJvIjtzOjU6InJlc2lkIjtzOjEwOiIzNC8xOC82MjY3IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjQvMDgvMzAvMjAyNC4wOC4yOS4yNDMxMjc4Ny5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

30. 30.Mu, Y., Guo, C. & Han, S. Oxytocin enhances inter-brain synchrony during social coordination in male adults. Social Cognitive and Affective Neuroscience 11, 1882–1893 (2016).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/scan/nsw106&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=27510498&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

31. 31.Miles, L. K., Nind, L. K. & Macrae, C. N. The rhythm of rapport: Interpersonal synchrony and social perception. Journal of Experimental Social Psychology 45, 585–589 (2009).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jesp.2009.02.002&link_type=DOI) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000266065700019&link_type=ISI) 

32. 32.Vacharkulksemsuk, T. & Fredrickson, B. L. Strangers in sync: Achieving embodied rapport through shared movements. Journal of Experimental Social Psychology 48, 399–402 (2012).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jesp.2011.07.015&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22389521&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

33. 33.Hove, M. J. & Risen, J. L. It’s All in the Timing: Interpersonal Synchrony Increases Affiliation. Social Cognition 27, 949–961 (2009).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1521/soco.2009.27.6.949&link_type=DOI) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000272909000011&link_type=ISI) 

34. 34.Valdesolo, P. & DeSteno, D. Synchrony and the social tuning of compassion. Emotion 11, 262–266 (2011).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/a0021302&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21500895&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000289272000006&link_type=ISI) 

35. 35.Mogan, R., Fischer, R. & Bulbulia, J. A. To be in synchrony or not? A meta-analysis of synchrony’s effects on behavior, perception, cognition and affect. Journal of Experimental Social Psychology 72, 13–20, (2017).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jesp.2017.03.009&link_type=DOI) 

36. 36.Imel, Z. E. et al. The association of therapist empathy and synchrony in vocally encoded arousal. Journal of Counseling Psychology 61, 146–153 (2014).
    
    

37. 37.Feldman, R. What Is Resilience: An Affiliative Neuroscience Approach. World Psychiatry 19, 132–150 (2020).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/wps.20729&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32394561&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

38. 38.Wiltshire, T. J., Philipsen, J. S., Trasmundi, S. B., Jensen, T. W. & Steffensen, S. V. Interpersonal Coordination Dynamics in Psychotherapy: A Systematic Review. Cognitive Therapy Research 44, 752–773 (2020).
    
    

39. 39.Ax, A. F. Goals and Methods of Psychophysiology. Psychophysiology 1, 8–25 (1964).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1469-8986.1964.tb02616.x&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=14201850&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

40. 40.Kaplan, H. B., & Bloom, S. W. The Use of Sociological and Social-Psychological Concepts in Physiological Research: A Review of Selected Experimental Studies. The Journal of Nervous and Mental Disease 131, 128–134 (1960).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/00005053-196008000-00006&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=13751281&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1960CHW6800006&link_type=ISI) 

41. 41.Kodama, K., Tanaka, S., Shimizu, D., Hori, K. & Matsui, H. Heart Rate Synchrony in Psychological Counseling: A Case Study. Psychology 9, 1858–1874 (2018).
    
    

42. 42.Coutinho, J. F., Silva, P. O. & Decety, J. Neurosciences, empathy, and healthy interpersonal relationships: Recent findings and implications for counseling psychology. Journal of Counseling Psychology 61, 541–548 (2014).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/cou0000021&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=25285714&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

43. 43.Oliveira-Silva, P. & Gonçalves, Ó. F. Responding empathically: A question of heart, not a question of skin. Applied Psychophysiology Biofeedback 36, 201–207 (2011).
    
    

44. 44.Pizzoli, S. F. M. et al. A meta-analysis on heart rate variability biofeedback and depressive symptoms. Sci. Rep. 11, (2021).
    
    

45. 45.Tarvainen, M. P., Lipponen, J. A., Niskanen J.-P. & Ranta-aho, P. O. Kubios HRV Software User’s Guide. [https://www.kubios.com/publications/](https://www.kubios.com/publications/) (2021)
    
    

46. 46.World Health Organization. *International Statistical Classification of Diseases and Related Health Problems*, *10th Revision (ICD-10)*. (1993).
    
    

47. 47.Ramseyer, F. & Tschacher, W. Nonverbal synchrony in psychotherapy: Coordinated body movement reflects relationship quality and outcome. Journal of Consulting and Clinical Psychology 79, 284–295 (2011).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/a0023419&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21639608&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 

48. 48.Ramseyer, F. Motion energy analysis (MEA): A primer on the assessment of motion from video. Journal of Counseling Psychology 67, 536–549 (2020).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/cou0000407&link_type=DOI) 

49. 49.Lipponen, J. A. & Tarvainen, M. P. A robust algorithm for heart rate variability time series artefact correction using novel beat classification. Journal of Medical Engineering and Technology 43, 173–181 (2019).
    
    

50. 50.Flückiger, C., Regli, D., Zwahlen, D., Hostettler, S. & Caspar, F. Der Berner Patienten- Und Therapeutenstundenbogen 2000: Ein Instrument zur Erfassung von Therapieprozessen. Zeitschrift für Klinisch Psychologie und Psychotherapie 39, 71–79 (2010).
    
    

51. 51.Hautzinger, M., Keller, F. & Kühner, C. BDI-II: Beck-Depressions-Inventar; Revision, 2. Auflage. (Pearson Assessment, 2009)
    
    

52. 52.Franke, G. H. BSI: Brief Symptom Inventory - Deutsche Version. (Beltz, 2000).
    
    

53. 53.RStudio Team. RStudio: Integrated Development Environment for R. Boston. [http://www.rstudio.com/](http://www.rstudio.com/) (2020).
    
    

54. 54. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. [https://www.r-project.org](https://www.r-project.org) (2021)
    
    

55. 55.Kleinbub, J. R. & Ramseyer, F. T. rMEA: An R package to assess nonverbal synchronization in motion energy analysis time-series. Psychother. Res. 31, 817–830. doi:10.1080/10503307.2020.1844334 (2021).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1080/10503307.2020.1844334&link_type=DOI) 

56. 56.Moulder, R. G., Boker, S. M., Ramseyer, F. & Tschacher, W. Determining synchrony between behavioral time series: An application of surrogate data generation for establishing falsifiable null- hypotheses. Psychological Methods 23, 757–773 (2018).
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/met0000172&link_type=DOI) 

57. 57.Stuldreher, I. V., van Erp, J. B. F. & Brouwer, A. M. Robustness of Physiological Synchrony in Wearable Electrodermal Activity and Heart Rate as a Measure of Attentional Engagement to Movie Clips. Sensors 23, (2023).
    
    

58. 58.Cheung, M. N. Detection of and Recovery from Errors in Cardiac Interbeat Intervals. Psychophysiology 18, 341–346 (1981).
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=7291452&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F08%2F30%2F2024.08.29.24312787.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1981LS16100019&link_type=ISI) 

59. 59.Fujiwara, K., Nomura, K. & Eto, M. Antiphase synchrony increases perceived entitativity and uniqueness: A joint hand-clapping task. Frontiers in Psychoogy. 14, (2023).
    
    

60. 60.Horvath, A. O. & Bedi, R. P. The alliance. in Psychotherapy relationships that work: Therapist contributions and responsiveness to patients. (ed. Norcross, J. C.) (Oxford University Press, 2002).
    
    

61. 61.Kupferberg, A. & Hasler, G. The social cost of depression: Investigating the impact of impaired social emotion regulation, social cognition, and interpersonal behavior on social functioning. Journal of Affective Disorders Reports 14, (2023).
    
    

62. 62.Paulick, J. et al. Diagnostic Features of Nonverbal Synchrony in Psychotherapy: Comparing Depression and Anxiety. Cognitive Therapy and Research 42, 539–551 (2018).