Psychotherapy

Volume 38/Winter 2001/Number 4

EMPATHY
LESLIE S. GREENBERG1 York University JEANNE C. WATSON University of Toronto Empathy is defined, and its measurement is discussed. We then present the results of a new meta-analysis of the relation between measures of empathy and psychotherapy outcome from three perspectives (client, therapist, and observer). Variables that mediate this relationship also are discussed. The processes of change that empathy is posited to facilitate, as well as the different roles and forms that empathy may take in therapy are discussed. Results indicate that clients' and observers' perceptions that therapists understand their clients' internal experiences relate to outcome. This suggests it is important that therapists make efforts to understand their clients, and that this understanding be demonstrated through responses that address the needs of the client, as the client perceives them, on an ongoing basis.
In this brief article we seek to consider the role of empathy in psychotherapy. We first define it and briefly discuss measurement issues. We pres-

ROBERT ELLIOTT University of Toledo Arthur C. Bohart Saybrook Graduate School and Research Institute
ent the results of a meta-analysis of the effects of empathy on psychotherapy outcome. We then discuss the processes of change that empathy is posited to facilitate, as well as the different roles and forms that empathy may take in therapy. Definition and Measurement Following Carl Rogers (1980) we define empathy as follows:
[Empathy is] the therapist's sensitive ability and willingness to understand the client's thoughts, feelings, and struggles from the client's point of view. [It is] this ability to see completely through the client's eyes, to adopt his frame of reference, (p. 8 5 ) . . . It means entering the private perceptual world of the other . . . being sensitive, moment by moment, to the changing felt meanings which flow in this other person. . . . It means sensing meanings of which he or she is scarcely aware, (p. 142)

Empathy is best understood as a complex construct consisting of a variety of experiences and acts used in different ways by therapists of different orientations for different purposes. The clearest operational definition of empathy is BarrettLennard's (1981) delineation of the three different components and perspectives of empathy: that of the therapist's experience ("empathic resonance"), the observers' view ("expressed empathy"), and the client's experience ("received empathy"), organized into a sequential process model. Types of Measures Observer-rated empathy. Scales of this type ask the rater to decide if the content of the therapist's response detracts from the client's response, is interchangeable with it, or adds to or carries it forward by responding to feeling (e.g., Truax & Carkhuff, 1967). More recent empathy measures are based on broader understandings of forms of empathic responding (Elliott et al.,

1 First authorship is shared by all four authors. Correspondence regarding this article should be addressed to Leslie S. Greenberg, Department of Psychology, York University, 4700 Keele St., Toronto, Ont. M3J 1P3. E-mail: Igrnberg y orku. ca

380

Empathy
1982) but have not been used in psychotherapy outcome research. Client ratings. The most widely used clientrated measure is the empathy scale of the BarrettLennard Relationship Inventory (BLRI; BarrettLennard, 1962). Rogers (1957) hypothesized that clients' perceptions of therapists' facilitative conditions (positive regard, empathy, and congruence) predicted therapeutic outcome. Accordingly, the BLRI, which measures clients' perceptions, is the closest operational definition of Rogers's hypothesis. In earlier reviews, Barrett-Lennard (1981) and Gurman (1977) both concluded that clientperceived empathy predicted outcome better than observer- or therapist-rated empathy. Therapist ratings. There are therapist selfrating scales on the BLRI. Earlier reviews (Barrett-Lennard, 1981; Gurman, 1977) found that therapist-rated empathy neither predicted outcome nor correlated with client-rated or observerrated empathy. Intercorrelations between different types of empathy measures have generally been weak. Taperated measures were reported to correlate only moderately with client-perceived empathy, with a mean of .28, over 17 studies that were reviewed (Gurman, 1977). Considering what the different instruments are supposed to be measuring, however, it is not surprising that different measures of this complex construct do not correlate highly. Research Review We conducted a meta-analysis of the relation of empathy to psychotherapy outcome. The studies included in the analysis, their sample characteristics, the method of calculating the effect size, coding procedures, and inclusion criteria can be found in our chapter (Bohart, Elliott, Greenberg, & Watson, in press). The resulting sample consisted of 47 studies, encompassing 190 separate tests of the empathy-outcome association and a total of 3,026 clients. Study dates ranged from 1961 to 2000, with 68% carried out before 1980. The average study involved 64 clients (range = 8 to 320) seen for slightly fewer than 30 sessions (range = 3 to 228). Typically, the studies involved mixed, eclectic, or unknown types of therapy (70%); individual treatment (78%); "mixed neurotic" samples (47%; what today would include primarily affective and anxiety disorders); recent Ph.D. or M.D. therapists (38%); and posttreatment assessment of outcome (60%). Studies contributed between 1 and 42 separate effects (mean = 4.0). Of the 190 separate effects examined, the largest number used client measures (38%; about half of these the BLRI) or observer measures (33%, the vast majority variations of the Truax-Carkhuff scales); therapist measures were used for 14% of the effects (about half of these the BLRI). "Therapy to date" was the most common unit for measuring empathy, accounting for 64% of effects. For effect sizes, parametric or nonparametric correlations were used if available. Otherwise, the following conventions were used to estimate r. First, if there was a significance level, the equivalent t value was used and converted to r. If the result was nonsignificant, but there was enough information to calculate a t and then convert, this was done. If there was no other information, and the effect was nonsignificant, r was set at 0. If the authors indicated a "nonsignificant trend" but did not report a correlation, the trend was estimated by assigning an effect size (ES) half the size of a significant r. Overall relation between therapist empathy and outcome. The single best summary value of this relation is probably the study-level, weighted, unbiased r of .32, a medium effect size suggesting that empathy as a whole accounted for almost 10% of outcome variance. Uncorrected and corrected values were around .20 for effect-level analyses (see Bohart et al., in press, for tables and more details of analyses). The effect size of .32 is of the same order of magnitude (slightly larger than) as previous analyses of the relationship between therapeutic alliance and outcome (i.e., Horvath & Symonds, 1991; Martin, Garske, & Davis, 2000). Overall, empathy accounts for as much and probably more outcome variance than does specific intervention (compare to Wampold's [2001] estimate of 1 to 8% for interventions). Theoretical orientation was not found to be a moderator variable. Type of empathy measure, however, was a moderator variable. The perspective of the measure made a difference. Client measures predicted outcome the best (mean uncorrected r = .25), followed closely by observerrated measures (.23) and therapist measures (. 18); each of these mean effects was significantly greatly than zero (p < .001). Clients' feelings of being understood and observer ratings (and to a lesser extent, therapist impressions) appear to carry significant weight as far as outcome goes.

381

Leslie S. Greenberg et al.

The most interesting and unexpected finding was a negative correlation between therapist level of experience and empathy-outcome. That is, larger effects were obtained for more junior therapists. Empathy may have a relatively small association with outcome for more experienced therapists. There are at least two possible reasons for this finding. First, inexperienced therapists may vary more in empathy, while the smaller correlation for experienced therapists may reflect a restriction of range or ceiling effect. Alternatively, experienced therapists may have developed additional skills such as effective problem solving, so that clients are more likely to forgive empathic misattunements. Another finding of note was that empathy appears to predict improvement in nonspecific measures of outcome (e.g., global improvement, client satisfaction) better than it does for more specific, problem-focused outcome measures. The question of causalitywhether empathy causes therapeutic outcome or is a correlate of itcannot be answered definitively from the above analyses. A number of studies lend some support for the causal role of empathy using structural equation modeling (Burns & Nolen-Hoeksema, 1992), supervisors' ratings of empathy (Miller, Taylor, & West, 1980), and measures of therapists' facilitative interpersonal skills, including accurate empathy, before therapy (Anderson, 1999). Further evidence comes from studies that show that client-centered therapy (CCT) is effective (Elliott, 2001). The mean ES of pre-post differences in CCT is .97 (n = 44 studies). For controlled effects (versus wait-list and no-treatment controls) the mean ES is .80 (n = 13). Usually one cannot attribute causal efficacy to a component of a therapy (e.g., interpretation in psychoanalysis) because the whole therapy is effective. However, although other relational elements are present in client-centered therapy, the vast majority of the time is occupied with therapeutic work in the form of empathy (e.g., Bradley & Brady, 1990). None of the above evidence definitively establishes a causal role for empathy. In contrast to Truax and Carkhuff's (1967) contention that empathy is primarily therapist-determined, other studies have indicated that therapists' levels of empathy vary as a function of the client (e.g., Beutler, Johnson, Neville, & Workman, 1972). Further research is needed to clarify this relationship. Nevertheless, empathy is probably better conceived of as a mutually created climate variable, rather than as a variable unilaterally "provided" by the therapist. Limitations of the Research Reviewed Watson (2001) has discussed problems with the research on empathy. These entail (a) the questionable validity of some of the outcome measures (e.g., client satisfaction); (b) lack of appropriate, sensitive outcome measures; and (c) restricted range of predictor variables. This was particularly a problem in the Mitchell, Truax, Bozarth, and Krauft (1973) study, for instance, where most of the therapists scored below the minimum considered to be sufficient to be effective, and outcome was only modest to moderate in the study. It is not surprising that significant correlations were not found. Further, the research is plagued by confounds among variations in time of assessment, experience of raters, and sampling methods; reliance on obsolete diagnostic categories; and incomplete reporting of methods and results. How Does Empathy Lead to Good Outcome? Four factors have been identified theoretically as potential mediators between empathy and outcome. Three of these are the processes of empathy as a relationship condition, as a corrective emotional experience, and as a cognitive-affective processing condition. The fourth factor has to do with the role of the client as an active self-healer. Empathy as relationship condition. Empathy serves a positive relationship function. Feeling understood increases client satisfaction with therapy and thereby increases compliance (e.g., in cognitive-behavioral therapy). Feeling understood also increases feelings of safety in the relationship and makes it easier to self-disclose. Also, clients feel safer to approach difficult personal areas. In addition, there is evidence that empathy is correlated with staying in therapy as opposed to premature termination. For instance, Chafetz and associates (1962) found that after a single empathic counseling session alcoholic patients were over 10 times more likely to seek treatment, and 40 to 50 times as likely to stay in treatment. Empathy as corrective emotional experience. Empathy provides a kind of direct learning or "corrective relational experience." Theoretically, an empathic relationship may help strengthen the self and break isolation (Bohart & Greenberg, 1997). Experientially, it can help clients learn that they are worthy of respect, of being heard,

382

Empathy
and that their feelings and behaviors make sense, so that they can express their feelings and needs in relationships. In addition, empathic attunement may biologically affect both therapist and client (Levenson & Ruef, 1997). Empathy and cognitive-affective processing. Empathy has been found to promote exploration and meaning creation. It helps clients think more productively (Sachse, 1990), raises levels of productive experiencing (Klein & Mathieu-Coughlan, 1986), and facilitates emotional reprocessing (Greenberg & Paivio, 1997). Empathy and the client as active self-healer. Several recent writers have suggested that therapy works primarily by supporting and promoting clients' active self-healing efforts (e.g., Bohart & Tallman, 1999). Empathy contributes to this by promoting client involvement and openness to the process. In addition, it provides an "empathic workspace" in which clients can use their capacities for self-healing. Finally, it helps the therapist choose interventions compatible with the client's frame of reference (Hubble, Duncan, & Miller, 1999), which also promotes clients' active participation. Therapeutic Practices The most consistent evidence is that clients' perceptions of feeling understood by their therapists relate to outcome. This suggests it is important that therapists make efforts to understand their clients, and that this understanding be demonstrated through responses that address the needs of the client, as the client perceives them, on an ongoing basis. In addition, the findings that observer ratings of accurate empathy predict outcome suggest that therapist responses that add to or carry forward the meaning in the client's communication are useful. Several different specific types of empathic responses that do this have been identified. These are empathic understanding, empathic exploration, empathic evocations, and empathic conjectures (Greenberg, Rice, & Elliott, 1993; Greenberg & Elliott, 1997; Watson, 2001). Going beyond specific responses, the empathic therapist's primary task is to understand experiences rather than words. Truly empathic therapists do not parrot clients' words back or reflect only the content of those words; instead, they understand overall goals as well as moment-to-moment experiences, both explicit and implicit. Empathy in part entails capturing the nuances and implications of what people say, and reflecting this back to them for their consideration. People express themselves on multiple levels. In being empathic, therapists can focus on clients' feelings, perceptions, construals, values, and assumptions and on their views of other people and situations. Certain fragile clients may find expressions of empathy too intrusive, while highly resistant clients may find empathy too directive; still other clients may find an empathic focus on feelings too foreign. Therapists therefore need to know whenand when notto respond empathetically. Therapists need to continually engage in process diagnoses to determine when and how to communicate empathic understanding and at what level to focus their empathic responses from one moment to the next. Empathic therapists assist clients to symbolize their experience in words, and track their emotional responses, so that clients can deepen their experience and reflexively examine their feelings, values, and goals. Therapists need to help clients access as much internal information as possible. To this end they need to attend to what is not said, or what is at the periphery of awareness as well as what is said and is in focal awareness. References
ANDERSON, T. (1999). Specifying non-'specifics' in therapists: The effect of facilitative interpersonal skills in outcome and alliance formation. Paper presented at the 30th annual meeting of the International Society for Psychotherapy Research, Braga, Portugal. BARRETT-LENNARD, G. (1981). The empathy cycle: Refinement of a nuclear concept. Journal of Counseling Psychology, 28, 91-100. BARRETT-LENNARD, G. (1962). Dimensions of therapist response as causal factors in therapeutic change. Psychological Monographs, 76, 43. BEUTLER, L. E., JOHNSON, D. T., NEVILLE, C. W., & WORKMAN, S. N. (1972). "Accurate empathy" and the AB dichotomy. Journal of Consulting and Clinical Psychology, 38, 372-375. BOHART, A. C., ELLIOTT, R., GREENBERG, L. S., & WATSON, J. C. (in press). Empathy. In J. Norcross (Eds.), Psychotherapy relationships that work. New York: Oxford University Press BOHART, A. C., & GREENBERG, L. S. (1997). Empathy: Where are we and where do we go from here? In A. C. Bohart & L. S. Greenberg (Eds.), Empathy reconsidered: New directions in psychotherapy (pp. 419-450). Washington, DC: American Psychological Association. BOHART, A. C., & TALLMAN, K. (1999). How clients make therapy work: The process of active self-healing. Washington, DC: American Psychological Association. BRODLEY, B. T., & BRODY, A. F. (1990, August). Understanding client-centered therapy through interviews conducted by Carl Rogers. Paper presented at the annual con-

383

Leslie S. Greenberg et al.

vention of the American Psychological Association, Boston, MA. BURNS, D. D., & NOLEN-HOEKSEMA, S. (1992). Therapeutic empathy and recovery from depression in cognitivebehavioral therapy: A structural equation model. Journal of Consulting and Clinical Psychology, 60, 441-449. CHAFETZ, M. E., BLANE, H. T., ABRAM, H. S., GOLNER, J., LACY, E., McCouRT, W. F., CLARK, E., & MYERS, W. (1962). Establishing treatment relations with alcoholics. Journal of Nervous and Mental Disease, 134, 395-409. ELLIOTT , R. (2001). Research on the effectiveness of humanistic therapies: A meta-analysis. In D. Cain & J. Seeman (Eds.), Handbook of research and practice in humanistic psychotherapy (pp. 57-82). Washington, DC: American Psychological Association. ELLIOTT, R., FILIPOVICH, H., HARRIOAN, L., GAYNOR, J., REIMSCHUESSEL, C., & ZAPADKA, J. K. (1982). Measuring response empathy: The development of a multi-component rating scale. Journal of Counseling Psychology, 29, 379387. GREENBERO, L. S., & ELLIOTT, R. (1997). Varieties of empathic responding. In A. C. Bohart & L. S. Greenberg (Eds.), Empathy reconsidered: New directions in psychotherapy (pp. 167-186). Washington, DC: American Psychological Association. GREENBERO, L. S., & PAIVIO, S. C. (1997). Working with emotions in psychotherapy. New York: Guilford. GREENBERO, L. A., RICE, C., & ELLIOTT, R. (1993). Faciliating emotional change: The moment-by-moment process. New York: Guilford. GURMAN, A. S. (1977). The patient's perception of te therapeutic relationship. In A. S. Gurman & A. M. Razin (Eds.), Effective psychotherapy: A handbook of research (pp. 503-543). New York: Pergamon. HORVATH, A. O., & SYMONDS, B. D. (1991). Relation between working alliance and outcome in psychotherapy: A meta-analysis. Journal ofCounseling Psychology, 38,139149. HUBBLE, M. A., DUNCAN, B. L., & MILLER, S. D. (1999). Directing attention to what works. In M. A. Hubble, B. L. Duncan, & S. D. Miller (Eds.), The heart and soul of change: What works in therapy (pp. 407-448). Washington, DC: American Psychological Association. KLEIN, M. H., & MATHIEU-COUOHLAN, P. (1986). Measures of client and therapist depth of experience. In L. S. Greenberg & W. M. Pinsof (Eds.), The psychotherapeutic process: A research handbook (pp. 21-72). New York: Guilford. LEVENSON, R. W., & RUEF, A. M. (1997). Physiological aspects of emotional knowledge and rapport. In W. Ickes (Ed.), Empathic accuracy (pp. 44-72). New York: Guilford. MARTIN, D. J., GARSKE, J. P., & DAVIS, M. D. (2000). Relation of the therapeutic alliance with outcome and other variables: A meta-analytic review. Journal of Consulting and Clinical Psychology, 68, 438-450. MILLER, W., TAYLOR, C., & WEST, J. (1980). Focused versus broad spectrum behavior therapy for problem drinkers. Journal of Consulting and Clinical Psychology, 48, 590601. MITCHELL, K. M., TRUAX, C. B., BOZARTH, J. D., & KRAUFT, C. C. (1973, March). Antecedents to psychotherapeutic outcome. NIMH grant Report (12306). Arkansas Rehabilitation Research and Training Center, Arkansas Rehabilitation Services, Hot Springs, Arkansas. ROGERS, C. R. (1957). The necessary and sufficient conditions of therapeutic personality change. Journal of Consulting Psychology, 21, 95-103. ROGERS, C. R. (1980). A way of being. Boston: Houghton Mifflin. SACHSE, R. (1990). The influence of therapist processing proposals on the explication process of the client. PersonCentered Review, 5, 321-344. TRUAX, C. B., & CARKHUFF, R. R. (1967). Toward effective counseling and psychotherapy: Training and practice. Chicago: Aldine. WAMPOLD, B. E. (2001). The great psychotherapy debate: Models, methods, and findings. Mahwah, NJ: Lawrence Erlbaum. WATSON, J. C. (2001). Revisioning empathy. Theory, research, and practice. In D. Cain & J. Seeman (Eds.), Handbook of research and practice in humanistic psychotherapy (pp. 445-472). Washington, DC: American Psychological Association.

384