Despite conflicting research findings underlying the debate of factors contributing to psychotherapy outcome, therapeutic alliance appears to be a key component of any intervention (Beutler 2004). A widely accepted definition of therapeutic alliance is agreement on therapy goals, agreement of therapy tasks, and therapeutic bond (Bordin 1979). In fact, research findings have identified therapeutic alliance as being important in the process of therapeutic growth for clients and as a main predictor of therapeutic outcomes (Castonguay and Beutler 2006; Rogers 1951; Shedler 2010). Thus, one means of improving client outcomes is to improve therapeutic alliance. To this end, researchers have examined ways to facilitate the training of therapeutic alliance, including examining empathy and mindfulness training (e.g., Thomas and Otis 2010).

Mindfulness represents a sophisticated technique for directing attention and facilitating present awareness and can be broadly defined as honing attention and emotional processes. In other words, it is learning what we pay attention to current internal and external stimuli and how long we pay attention to it with a nonjudgmental attitude (Brown and Ryan 2003; Kabat-Zinn 2005; Linehan 1993). Mindfulness fosters acceptance of the present moment through the intentional emotional and cognitive evaluations while resisting natural urges that may occur, such as distraction or resistance. Research has found a relationship between the practice of mindfulness and positive outcomes, including increases in emotional intelligence, insight, wisdom, self-awareness, empathy, and attention to and clarification of feelings (Baer et al. 2004; Brown and Ryan 2003; Goldstein 2002; Sweet and Johnson 1990). Other research findings posit engagement in mindfulness results in decreases of stress, anxiety, anger, and depression (Astin 1997; Kutz et al. 1985; Oxman 1995; Shapiro et al. 1998). Furthermore, research has demonstrated that the practice of mindfulness is negatively correlated with clinician burnout and compassion fatigue, as well as positively correlated with clinicians’ well-being (Shapiro et al. 2007; Thomas and Otis 2010). The act of mindfulness achieves these aforementioned outcomes through re-evaluating cognitive appraisals in a meta-cognitive, reflective process—which in turn, allows for the attention, awareness, and acceptance of emotions (Campbell and Christopher 2012; Cigolla and Brown 2011).

The conceptualization of empathy in a therapeutic context, as defined by Rogers (1946), is an accurate perception of others’ internal frame of reference as well as experienced emotions as if one were the person. In other words, it is the ability to walk in another’s shoes through suspending judgment and biases (Greason and Cashwell 2009). Empathy has long been regarded as a fundamental tenet of therapeutic alliance (Cozolino 2014; Duncan et al. 2010; Lambert and Barley 2002). Consistent with this notion, a meta-analysis identified empathy as a moderate predictor of treatment outcome and an essential ingredient of change in the therapeutic process (Greenberg et al. 2001). Aiken (2006) found that clinicians who practiced mindfulness were more capable of developing empathy for their clients through being more present with their clients’ suffering as well as assisting their clients with expressing their body sensations and feelings. Since mindfulness is thought to cultivate awareness as well as acceptance of the self and others (Campbell and Christopher 2012; Kabat-Zinn 2005; Linehan 1993), and awareness is related to empathy (Aiken 2006; Bruce et al. 2010), which is thought to be a key component to the development of therapeutic alliance (Greenberg et al. 2001), researchers have examined the relationship between mindfulness and therapeutic alliance.

The current definition of therapeutic alliance, also called working alliance, is thought to consist of three aspects: agreement on therapy goals, agreement of therapy tasks, and therapeutic bond (Bordin 1979), all of which are key components of therapeutic interventions regardless of approach (Beutler 2004). Mindfulness can be instrumental in developing and maintaining a strong therapeutic alliance. For example, the nature of therapy itself presents inherent challenges and opportunities for things to go awry within therapeutic alliance such as reacting negatively to clients (Dunn et al. 2013; Safran and Kraus 2014). Clinicians who practice mindfulness in general not only reduce the likelihood of reacting to the client in a way that would endanger the therapeutic relationship but increase their ability to nurture and foster successful relationships with their clients by being more attentive and responsive (Boswell et al. 2010; Davis and Hayes 2011; McCollum and Gehart 2010). Mindfulness has also been found to be an important pre-treatment variable correlated both with psychotherapy outcomes and ratings of therapeutic alliance (Ryan et al. 2012). Finally, empathy has been found to be cultivated through various means in order to foster therapeutic alliance. Researchers have hypothesized that one such means is through the practice of mindfulness (Bruce et al. 2010; Hopkins and Proeve 2013; Keane 2014; Schomaker 2013).

Although studies have shown a relationship between mindfulness, empathy, and therapeutic alliance in a pairwise context, the nature of the three-way relationship has not been conducted. One previous study found that greater amounts of dispositional mindfulness were correlated with higher amounts of attention and empathy, which in turn influenced counseling self-efficacy (Greason and Cashwell 2009). However, the lack of similar research in this area justifies the need for further investigation into mechanisms of mindfulness within a clinical context in service of potentially improving therapeutic alliance.

The current study used a mediation model to examine if mindfulness affects therapeutic alliance because of its positive influence on empathy. We hypothesized that empathy would mediate the positive association between clinician mindfulness and therapeutic alliance that has previously been found to exist (Cozolino 2014; Duncan et al. 2010; Lambert and Barley 2002).

Method

Participants

Participants (n = 96) consisted primarily of doctorate-level licensed psychologists who are currently working in a clinician role. Of the 91% of participants who disclosed their degree earned (n = 88), 81% reported earning a Ph.D. (n = 78), 8% reported earning a Psy.D. (n = 8), and 2% reported earning an M.A. (n = 2). Participants ranged in age from 26 to 78 years old (M = 50.34; SD = 14.17 years) with 68% of the participants reporting as female (n = 65) and 31% (n = 30) male. With respect to ethnicity, 90% of the participants were White/Caucasian (n = 87), 4% were of Hispanic/Latino/Spanish decent (n = 4), 2% were Asian American (n = 2), 1% were Black/African-American (n = 1), and 2% reported themselves as other (n = 2).

In addition to demographics, participants were asked about their training experiences. With respect to model of training program, 79% reported scientist-practitioner (n = 76), 10% reported practitioner-scholar (n = 10), 7% reported clinical-scientist (n = 7), and 4% categorized their training program as other (n = 3). When asked about theoretical orientation, 60% reported their orientation as cognitive/behavioral (n = 57), 13% reported integrative/eclectic (n = 14), 11% reported systemic (n = 7), 5% reported interpersonal (n = 5), 2% reported client-centered/humanistic/existential (n = 2), 1% reported psychodynamic/psychoanalytic (n = 1), and 7% reported their orientation as other (n = 7). Finally, participants were asked if they have had any previous mindfulness training. Of the total sample (N = 96), 67% reported having previous mindfulness training (n = 64).

Procedure

Purposive and snowball sampling were used in this study. Purposive sampling was used through recruiting doctorate and masters-level clinicians who are currently working in a clinician role through professional psychology listservs, such as APA Divisions. Snowball sampling was used by asking participants identified on listservs to forward the survey link to colleagues prior to participating in the study. Participants were directed to an online survey administered through Qualtrics survey software, where they provided informed consent and completed measures of mindfulness, empathy, and therapeutic alliance. No compensation was offered to participants.

Measures

Mindfulness

Clinician mindfulness was assessed using the Five Facet Mindfulness Questionnaire (FFMQ; Baer et al. 2006). The FFMQ is a 39-item self-report questionnaire assessing five aspects of mindfulness through five subscales: Observing, Describing, Acting with Awareness, Non-Judging of Inner Experience, and Non-Reactivity to Inner Experience. Respondents rated the extent to which each item is true for them using a 5-point Likert-type scale ranging from 1 (never or very rarely true) to 5 (very often or always true). Higher scores reflect higher amounts of mindfulness. The FFMQ has demonstrated good construct validity with the Acceptance and Action Questionnaire-II (Curtiss and Klemanski 2014). In the current study, only the total score for FFMQ was included in analyses and demonstrated excellent internal consistency for the total score (α = .93).

Empathy Engagement

The degree to which a participant engages in empathy was measured through the Interpersonal Reactivity Index (IRI; Davis 1983). The IRI is a widely used questionnaire assessing empathy due to the fact that it measures both cognitive and affective components of empathy, whereas other scales only assess either the emotional component (e.g., Emotional Empathetic Tendency Scale; Mehrabian and Epstein 1972) or the cognitive component (e.g., Hogan Empathy Scale; Hogan 1969). The IRI has 28 self-report items with four subscales: Empathetic Concern (EC; tendency to experience feelings of others), Personal Distress (PD; tendency to experience distress when observing other’s distress), Fantasy (FS; tendency to act or identify with fictional characters in hypothetical situations), and Perspective Taking (PT; ability to adopt another’s outlook). Respondents rate each item using a 5-point Likert-type scale ranging from 0 (does not describe me well) to 4 (describes me very well). A mean score is derived from the IRI total scale. Higher scores reflect higher degree of empathy engagement. The IRI has demonstrated convergent validity with prosocial behaviors (Sze et al. 2012). In the current study, the IRI total scale score was not included in analyses. Rather, the four subscales were included in analyses and all demonstrated acceptable internal consistency. The internal consistency was acceptable for the subscale of Empathic Concern (α = .73) and good for the subscales of Perspective Taking (α = .81), Fantasy (α = .78), and Personal Distress (α = .80).

Working Alliance

The quality of therapeutic alliance was assessed through the Working Alliance Inventory, Therapist version (WAI-T; Horvath and Greenberg 1989). To control for accuracy and consistency of recalling clients, participants were asked to think of their most recent client when completing the measure. The WAI is a 36-item self-report questionnaire measuring aspects of therapeutic alliance as conceptualized by Bordin (1979) via the following subscales: Agreement on Therapy Goals, Agreement of Therapy Tasks, and The Therapeutic Bond. Respondents rate each statement using a 7-point Likert-type scale ranging from 1 (never) to 7 (always). A mean score is derived from the WAI-T total scale. Higher scores reflect a greater alliance from the perspective of the clinician. The total WAI-T has shown moderate to strong correlations with other questionnaires measuring therapeutic alliance, specifically the California Psychotherapy Alliance Scale (r = .51, p < .001) and the Pennsylvania Helping Alliance Scale (r = .44, p < .01; Fenton et al. 2001). In the current study, the total score for the WAI-T was included in analyses and demonstrated excellent internal consistency for the total score (α = .95).

Data Analyses

Twenty-two cases from the original data set (N = 118) were removed due to response incompleteness above 15% on a given measure. Missing values were then predicted for 10 cases missing less than 5% of data on a given measure from the remaining data file (N = 96) utilizing the expectation-maximization (E-M) approach. The E-M approach employs algorithms to predict expected values for missing data by estimating parameters from completed data (Moon 1996). Upon treating missing cases, a correlation analysis was conducted for all the study variables. Table 1 illustrates the means, standard deviations, and intercorrelations for these variables. The correlation table shows a clear relationship between mindfulness, empathy, and therapeutic alliance, thus enabling further statistical testing to be conducted on these variables.

Table 1 Means, standard deviations, and intercorrelations of study variables (N = 96)
Full size table

Four mediation analyses were used to test the hypotheses of the study using the Hayes PROCESS bootstrapping analysis software macro (Hayes 2012). In each model, the independent variable was mindfulness and was measured by the FFMQ (Baer et al. 2006). The outcome variable was the quality of therapeutic alliance and was measured with the WAI-T (Horvath and Greenberg 1989). Total scores were used for both the WAI-T and FFMQ since the subscales of each comprise the constructs of therapeutic alliance and mindfulness, respectively. As the exact role of empathy within the context of mindfulness and therapeutic alliance is unknown, to help shed light on the exact role of empathy between mindfulness and therapeutic alliance, each facet of empathy measured by the IRI (Davis 1983) was examined as a potential mediating variable in four separate simple mediation models. Finally, since a majority of the sample (67%, n = 64) reported previous mindfulness training, this was controlled for in the mediation analysis. Demographic variables were also controlled for (e.g., age and gender).

Results

When examining the subscale of personal distress, which is the tendency to experience distress when observing other’s distress, the total statistical model accounted for a significant portion of the variance in therapeutic alliance (R2 = .34, p < .001). The a path from clinician mindfulness to personal distress (B[SE] = − 5.56[1.08], β = − .53, p < .001, sr2 = .22) was statistically significant with a medium effect size (see Fig. 1). The b path from personal distress to therapeutic alliance was statistically significant (B[SE] = − .04[.01], β = − .33, p < .001, sr2 = .01]) with a small effect size. Bias-corrected bootstrap results for the indirect (ab) of clinician mindfulness on therapeutic alliance through personal distress revealed a statistically significant ab effect (ab[SE] = .24[.11], 95% CIs [.09, .51], z = 2.28, p = .02). The total medium association of clinician mindfulness and therapeutic alliance (c path; B[SE] = .62[.14], β = .47, p < .001, sr2 = .18) was reduced when the mediational variable was accounted for in the model with a small, statistically significant relationship remaining (B[SE] = .41[.15], β = .29, p = .001, sr2 = .05). Thus, the subscale of personal distress partially mediated the relationship between clinician mindfulness and therapeutic alliance.

Fig. 1
figure 1

Mediating or intervening role of personal distress empathy in the relationships between trait mindfulness and therapeutic alliance. Depicted are the standardized regression coefficients for each path of the mediation model. Gender, age, and previous mindfulness training were included as covariates. *p < .05; **p < .01; ***p < .001

Full size image

When examining the subscale of perspective taking, which is the ability to adopt another’s outlook, the total statistical model accounted for a significant portion of the variance in therapeutic alliance (R2 = .25, p < .001). The a path from clinician mindfulness to perspective taking (B[SE] = 3.83[1.03], β = .41, p < .001, sr2 = .13) was statistically significant with a medium effect size. The b path from perspective taking to therapeutic alliance was not statistically significant (B[SE] = − .00[.01], β = − .02, p = .783, sr2 = .00]) with a trivial effect size. Bias-corrected bootstrap results for the indirect (ab) of clinician mindfulness on therapeutic alliance through perspective taking revealed a non-statistically significant ab effect (ab[SE] = − .01[.05], 95% CIs [− .09, .11], z = − .16, p = .21). The total medium association of clinician mindfulness and therapeutic alliance (c path; B[SE] = .62[.14], β = .47, p < .001, sr2 = .18) was maintained when the mediational variable was accounted for in the model with a medium, statistically significant relationship remaining (B[SE] = .64[.15], β = .45, p < .001, sr2 = .15). Thus, the subscale of perspective taking did not mediate the relationship between clinician mindfulness and therapeutic alliance.

When examining the subscale of empathetic concern, which is the tendency to experience feelings of others, the total statistical model accounted for a significant portion of the variance in therapeutic alliance (R2 = .28, p < .001). The a path from clinician mindfulness to empathetic concern (B[SE] = 1.33[.93], β = .16, p = .156, sr2 = .01) was not statistically significant with a small effect size. The b path from empathetic concern to therapeutic alliance was not statistically significant (B[SE] = .02[.02], β = .14, p = .150, sr2 = .01]) with a small effect size. Bias-corrected bootstrap results for the indirect (ab) of clinician mindfulness on therapeutic alliance through empathetic concern revealed a non-statistically significant ab effect (ab[SE] = .03[.04], CIs [− .01, .16], z = .81, p = .75). The total medium association of clinician mindfulness and therapeutic alliance (c path; B[SE] = .62[.14], β = .47, p < .001, sr2 = .18) was maintained when the mediational variable was accounted for in the model with a medium, statistically significant relationship remaining (B[SE] = .63[.14], β = .45, p < .001, sr2 = .17). Thus, the subscale of empathetic concern did not mediate the relationship between clinician mindfulness and therapeutic alliance.

Finally, when examining the subscale of fantasy, which is the tendency to act or identify with fictional characters in hypothetical situations, the total statistical model accounted for a significant portion of the variance in therapeutic alliance (R2 = .25, p < .001). The a path from clinician mindfulness to fantasy (B[SE] = 1.43[1.33], β = .12, p = .284, sr2 = .01) was not statistically significant with a small effect size. The b path from fantasy to therapeutic alliance was not statistically significant (B[SE] = .01[.01], β = .06, p = .514, sr2 = .00]) with a trivial effect size. Bias-corrected bootstrap results for the indirect (ab) of clinician mindfulness on therapeutic alliance through fantasy revealed a non-statistically significant ab effect (ab[SE] = .01[.03], 95% CIs [− .01, .11], z = .42, p = .33). The total medium association of clinician mindfulness and therapeutic alliance (c path; B[SE] = .62[.14], β = .47, p < .001, sr2 = .18) was maintained when the mediational variable was accounted for in the model with a medium, statistically significant relationship remaining (B[SE] = .65[.14], β = .46, p < .001, sr2 = .17). Thus, the subscale of fantasy did not mediate the relationship between clinician mindfulness and therapeutic alliance. The mediation results for all four models are displayed in Table 2.

Table 2 Mediation of the effects of trait mindfulness on therapeutic alliance
Full size table

Discussion

Similar to previous literature (Bruce et al. 2010; Hopkins and Proeve 2013; Keane 2014; Schomaker 2013) and in line with our hypothesis, clinician-reported empathy mediated the association between mindfulness and therapeutic alliance. However, not all aspects of empathy measured served as mediators in the relationship between mindfulness and therapeutic alliance. Clinicians’ tendency to experience distress when observing other’s distress was inversely related to mindfulness and therapeutic alliance, and it partially mediated the relationship between these two variables. That is, clinicians who reported being more mindful were less likely to report experiencing negative emotions when their clients experience negative emotions, thus reportedly helping to strengthen therapeutic alliance.

Although our findings align with previous literature, this study is an initial attempt to shed light on the role empathy plays in the relationship between mindfulness and therapeutic alliance. As such, it is considered to be preliminary research exploring these relationships. As previously mentioned, it has been theorized that clinicians who practice mindfulness are less likely to react to clients in ways that might lead to a rupture within the therapeutic alliance due to experiencing less negative emotions as well as being more attentive and responsive (Astin 1997; Boswell et al. 2010; Davis and Hayes 2011; Dunn et al. 2013; Kutz et al. 1985; McCollum and Gehart 2010; Oxman 1995; Safran and Kraus 2014; Shapiro et al. 1998). The results of this study appear to align with the aforementioned notion that clinicians who exhibit higher levels of trait mindfulness are able to maintain a more effective therapeutic alliance. This is because mindfulness has the potential to reduce the likelihood that clinicians react in ways that endanger the therapeutic relationship, such as experiencing personal distress when the client expresses distress. Indeed, it may be difficult for clinicians to be present and effective with a client during difficult work if the clinician also becomes upset.

In the current study, not all components of empathy tested were found to play a role within the relationship of mindfulness and empathy: perspective taking, empathic concern, and fantasy. Previous research has established that clinicians who practice mindfulness are more present, attentive, responsive, have increased emotional intelligence, and more attune to and clarifying of feelings (Aiken 2006; Baer et al. 2004; Brown and Ryan 2003; Davis and Hayes 2011; McCollum and Gehart 2010; Sweet and Johnson 1990), all of which are encapsulated within the constructs of perspective taking, empathic concern, and fantasy. Thus, taken together, the current study may suggest that mindfulness is not enough to significantly increase the aforementioned aspects of empathy. Rather, mindfulness may actually reduce the likelihood of a clinician experiencing emotional contagion, which in turn, increases therapeutic alliance. However, other research (e.g., Celen 2014) identifies cognitive and affective aspects of empathy as a factor that improve therapeutic alliance. Thus, mindfulness, in fact, may not be the only way to avoid emotional contagion, but it appears to be an efficient means for doing so.

On the other hand, a potential reason for the other aspects of empathy not playing a role may be the use of the IRI. Although there are other measures of empathy, the IRI was chosen because it assesses both the cognitive and emotional components of empathy and is, thus, widely used. However, empathy can be measured through other methods. For example, the Therapist Empathy Scale (TES; Decker et al. 2014) measures empathy within a clinical context by way of an external observer. It could be that had empathy been assessed other ways, the findings of this study would fully align with theory. As previously mentioned, this is an initial attempt to provide preliminary support the role of empathy within the relationship between mindfulness and therapeutic alliance. As such, this study provides a theoretical justification for future research to be conducted assessing the ecological value of these findings, which may include use of the TES.

The aim of this study was to achieve an in-depth understanding of the role empathy has within the relationship of mindfulness and therapeutic alliance. Although the results of this study provide a preliminary understanding of the type of empathy fostered through mindfulness, it is entirely possible that other variables not assessed may also explain the positive relationship between clinician mindfulness and therapeutic alliance. In addition to empathy, Rogers (1957) proposed that acceptance and genuineness are also mechanisms of therapeutic change. It is possible that acceptance and genuineness also play a role in the relationship between clinician mindfulness and therapeutic alliance. Thus, this is an area worthy of future research. Another potential area of future research is investigating other ways in which the personal practice of mindfulness is beneficial for clinicians, which was controlled for in the current study. Mindfulness training is a reflective process that can foster awareness and acceptance of the self and others, and has been found to yield an improved relationship with the self and increased clinical skills (Campbell and Christopher 2012; Keane 2014; Shapiro et al. 2007). Future research should explore how mindfulness fosters self-growth for the clinician, and in turn, if this is perhaps another vehicle by which the capacity to be therapeutic is increased.

Limitations

Noteworthy are the limitations of this study. First, the findings of this study are based on the self-report of clinicians. The constructs of mindfulness, empathy, and therapeutic alliance are limited in their accuracy when measured via self-report. Thus, the findings of this study are limited to the confines of self-report. Incorporating additional or alternative methods may have shed further light on the relationship between mindfulness, empathy, and therapeutic alliance (e.g., interviews, observation). Second, this study employed the use of snowball sampling by asking clinician to forward the survey link to colleagues. In employing this approach, variability is introduced to the intended sampling frame. Third, this study was cross-sectional, limiting our ability to draw causal inferences regarding mediation. Fourth, the sample size of this study is a limitation. Additional findings may have been yielded if a larger sample size was acquired. A fifth limitation was the potential for participants to guess the study hypothesis, which could create the possibility for demand characteristics. Finally, another limitation of this study was not asking participants to provide client demographics when thinking of their most recent client during the WAI-T. Thus, important clinical context is missing such as client diagnosis, duration of treatment, or setting of treatment (e.g., inpatient versus outpatient). It is possible that the clinical setting may impact clinician mindfulness, which may affect empathy, and in turn, therapeutic alliance.

Notwithstanding the aforementioned limitations, the results of this study provide preliminary support for the theoretical links between mindfulness and empathy and its positive benefit on therapeutic alliance. As these results provide preliminary support, they should be interpreted cautiously. That being said, these findings offer perspective into just how important mindfulness may be to the therapeutic relationship because of its effect on empathy. Indeed, empathy is regarded as a requisite for a positive therapeutic alliance (Cozolino 2014; Duncan et al. 2010; Lambert and Barley 2002). The next step in this line of research would be to conduct ecological research which may involve the use of experimental designs and data collection beyond self-report measures.