CoChat
Sign Up
Cited by 495Psychotherapy Techniques and ApplicationsEmpathy and Medical EducationCounseling Practices and Supervision

Therapist empathy and client outcome: An updated meta-analysis.

Open access
Elliott, Robert, Bohart, Arthur C., Watson, Jeanne C., Murphy, David·2018·Psychotherapy
DOI 10.1037/pst0000175
Open this paper in CoChat — free accountOpen PDF

Abstract

Put simply, empathy refers to understanding what another person is experiencing or trying to express. Therapist empathy has a long history as a hypothesized key change process in psychotherapy. We begin by discussing definitional issues and presenting an integrative definition. We then review measures of therapist empathy, including the conceptual problem of separating empathy from other relationship variables. We follow this with clinical examples illustrating different forms of therapist empathy and empathic response modes. The core of our review is a meta-analysis of research on the relation between therapist empathy and client outcome. Results indicated that empathy is a moderately strong predictor of therapy outcome: mean weighted r = .28 (p < .001; 95% confidence interval [.23, .33]; equivalent of d = .58) for 82 independent samples and 6,138 clients. In general, the empathy-outcome relation held for different theoretical orientations and client presenting problems; however, there was considerable heterogeneity in the effects. Client, observer, and therapist perception measures predicted client outcome better than empathic accuracy measures. We then consider the limitations of the current data. We conclude with diversity considerations and practice recommendations, including endorsing the different forms that empathy may take in therapy. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Integrity signals

Registry

Crossref · 10.1037/pst0000175

Cites

117 of 117 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 7 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The correlation between standard error of r and effect size was .06, indicating no bias from smaller studies with less precise effects producing larger effects.

    “The correlation between standard error of r and effect size was .06,indicating an absence of bias deriving from smaller studies with less precise effects producing larger effects.”
    Quote in the paperNot a substantial claim
  • Findingown data

    Having a smaller client sample predicted larger effect sizes (beta = -.24).

    “having a smaller sample of clients (-.24)”
    Quote in the paperNot a substantial claim
  • Findingown data

    The overall between-groups effect for client population categories was statistically significant (Q = 70.15, df = 4, 76; p < .01).

    “Although the overall between-groups effect was statistically significant (Q = 70.15,df = 4, 76; p < .01)”
    Quote in the paperNot a substantial claim
  • Findingown data

    The empathy-outcome association for mild, normal, or physical problems was r = .17 (k = 6).

    “Smaller effects were found for mild/normal/physical problems (.17; k = 6)”
    Quote in the paperNot a substantial claim
  • Findingown data

    The empathy-outcome association shows significant heterogeneity.

    “We found significant heterogeneity in the empathy-outcome association, pointing to the value of personalization and clinical judgment.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 12
  • May combine two claims. 18
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The I-squared statistic for the effect-level analysis was 72.2%.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The fail-safe number for the study-level analysis was 33.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Each of the client, observer, and therapist mean empathy effects was significantly greater than zero (p < .001).

    “each of these mean effects was significantly greater than zero (p < .001)but did not differ significantly from each other.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The empathy-outcome association held across a wide range of substantive and methodological variables, including client presenting problem or severity, therapy format, theoretical orientation, year of publication, and level or size of the unit at which empathy was measured.

    “We found that the empathy-outcome association held up across a wide range of variables, both substantive (client presenting problem/severity, therapy format, theoretical orientation) and methodological (year of publication, level/size of unit at which empathy was measured).”
    • Also stated in: § Abstract · hedged
      “In general, the empathy-outcome relation held for different theoretical orientations and client presenting problems; however, there was considerable heterogeneity in the effects.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    A 'more is less' effect was found: studies with more clients and a wider range of measures produced smaller effects.

    “We found a”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Across 290 nonindependent separate effects, the average empathy-outcome correlation was .21 (95% CI .18-.24), equivalent to d = .43.

    “average effects were somewhat smaller, at .21 (95% CI: .18-.24; equivalent to d = .43).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The fail-safe number for the effect-level analysis was 13.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingnone given

    The empathy-outcome finding has been repeated in more than 80 studies and in multiple meta-analyses.

    “This repeated finding, in more than 80 studies and now in multiple meta-analyses, leads to the following clinical recommendations:”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    There is no evidence that therapists' accurate prediction of clients' own views of their problems, experiences, or self-perceptions is effective.

    “There is no evidence that therapists accurately predicting clients' own views of their problems or experiences or self-perceptions is effective.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The fail-safe number needed to reduce the study-level weighted effect to r = .2 was 33 studies; for effect-level effects it was 12.

    “This value was 33 studies for the study level effect of .28; the comparable number for effect level effects was only 12.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The I-squared statistic for the study-level analysis was 76.8%.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Cochrane's Q for the study-level analysis was 348.68, statistically significant at p < .001.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Cochrane's Q for the effect-level analysis was 1039.2, statistically significant at p < .001.

    “Cochrane's Q 1039.2*348.68*I 2 72.2% 76.8% Fail safe number 13 33 * p < .001”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    Using global outcome variables such as client satisfaction produced larger empathy-outcome associations, possibly because client perceptions of empathy were confounded with client ratings of post-therapy satisfaction.

    “Using global outcome variables like client satisfaction resulted in larger associations between therapist empathy and outcome, possibly because of confounding between client perceptions of empathy and client ratings of post-therapy satisfaction.”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Interpretationown data

    Therapist empathy strongly and moderately predicts client outcomes in psychotherapy, with the link appearing across different therapy orientations, delivery formats, and client problem types.

    “As we have shown, empathy is a robust medium-sized predictor of client outcome in psychotherapy that holds across theoretical orientations, treatment formats, and client problems.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    Any one of the relational conditions without the others would create a distinctly different interpersonal climate and relationship.

    “Any one of these conditions without the others would provide a distinctly different interpersonal climate and relationship.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The problems found in empathy research are not unique to it but are common to all process-outcome research.

    “In fact, these and other problems are not restricted to empathy research but are common to all process-outcome research (Elliott, 2010).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Elliott, 2010

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Empathy usually accounts for more outcome variance than specific treatment methods, which have an estimated d = .2.

    “Overall, empathy typically accounts for more outcome variance than do specific treatment methods (compare Wampold's, 2015, estimate of d = .2for intervention effects).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Wampold's, 2015

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    Empathy will probably not be effective unless it is grounded in authentic caring for the client.

    “Empathy will probably not prove effective unless it is grounded in authentic caring for the client.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Abstract screening was done by four co-authors with a reliability sample of 200 studies (kappa = .61), and full-text evaluation was done by two co-authors (kappa = .45), with disagreements resolved by consensus.

    “The abstracts of the potential sources were screened by the four co-authors, with a sample of 200 studies to assess reliability (kappa = .61).”
    • Also stated in: § Screening and Analyses · plain
      “which were each evaluated by two of the co-authors (kappa = .45),with all disagreements resolved by discussion to consensus”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    The updated meta-analysis added 24 new studies to 58 carried over from the prior meta-analysis, yielding 82 client samples from 80 studies, 290 tests of the empathy-outcome association, and 6,138 clients seen for an average of 25 sessions.

    “These 24 studies were added to 58 studies that were carried over from our prior metaanalysis (Elliott et al., 2011), resulting in a total of 82 samples of clients (from 80 separate studies), aggregated from 290 separate tests of the empathy-outcome association and encompassing a total of 6138 clients, who were seen for an average of 25 sessions.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Elliott et al., 2011

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductcitation

    Study-level analyses averaged individual effects within client samples before further analysis, avoiding nonindependence and bias from variable numbers of effects across studies.

    “study-level analyses used averaged individual effects within client samples before further analysis, thus avoiding problems of nonindependence and eliminating bias due to variable numbers of effects reported in different studies (Lipsey & Wilson, 2001).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Lipsey & Wilson, 2001

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The funnel plot was quite symmetrical, making it unlikely that the overall effect would shift negatively if more powerful studies were published.

    “the funnel plot is quite symmetrical, making it unlikely that the overall effect would be shifted negatively if a larger number of more powerful studies were to be published.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductreasoning

    Empathy research has been limited by restricted range of predictor and criterion variables, confounding among study characteristics, and incomplete reporting of methods and results.

    “These include: (a) the questionable validity of some outcome measures (e.g., 15% of the studies included here used client-rated benefit or satisfaction with value of helpfulness of therapy as their main outcome measure); (b) restricted range of predictor and criterion variables; (c) confounding among many study characteristics (e.g., type of empathy and outcome variable, timing of empathy and outc”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The 2011 meta-analysis included 57 studies (224 effects) and 3,599 clients, and found an average weighted empathy-outcome correlation of .31.

    “the 2011 meta-analysis was conducted on 57 studies (224 effects) and encompassed a total of 3,599 clients.The average weighted correlation between empathy and outcome was an r of .31.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Elliott, Bohart, Watson & Greenberg, 2011

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    For the precedence condition of causal inference, the meta-analysis found mixed results: not all studies satisfied it, but when it was satisfied, overall effects were statistically significant though smaller.

    “(1) Precedence: The hypothesized causal variable must reliably precede the effect variable. Mixed: Not satisfied in all studies, but overall statistically significant (although smaller) effects are found when the precedence condition was satisfied.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    A content validity criterion was applied: only studies whose empathy scale had at least 50% of items clearly identifiable as empathy were included.

    “We decided in this meta-analysis to apply a content validity criterion by including only studies in which the”
    Quote in the paperSubstantial claim
27 of 27

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Empathic engagement predicts better client outcomes.

    “This empathic engagement predicts better client outcomes [22]; however, leaves MHWs at increased risk of EBS [19, 23].”

    From Interventions to address empathy-based stress in mental health workers: A scoping review and research agenda (2024) · cites this paper as [22]

  • People accessing mental health services have better outcomes if they experience compassionate and empathic relationships with staff caring for them.

    “People accessing mental health services have been found to have better outcomes if they experience compassionate [1] and empathic [2] relationships with staff caring for them.”

    From Interventions to address empathy-based stress in mental health workers: A scoping review and research agenda (2024) · cites this paper as [2]

  • Therapist empathy and alliance are robust predictors of therapy success across multiple studies of different therapies and clinical populations.

    “In fact, therapist empathy and alliance have been found to be robust predictors of therapy success across multiple studies of different therapies and clinical populations (Elliott et al., 2018;Flückiger et al., 2018;Martin et al., 2000).”

    From A mixed methods study of autistic adults' mental health therapy experiences (2023) · cites this paper as Elliott et al., 2018

  • Common factors shown to correlate with outcomes include therapeutic alliance, goal consensus and collaboration, and therapist empathy.

    “On the common factors side, widely studied factors shown to correlate with outcomes include therapeutic alliance52, goal consensus and collaboration between therapist and patient66, and therapist empathy67.”

    From A common factors perspective on mindfulness-based interventions (2022) · cites this paper as [67]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Effect-level fail-safe number differs between text and table

    Minor

    The text gives the effect-level fail-safe number as 12, but Table 2 gives it as 13.

  • Table 3 does not contain the cited measure comparison

    Minor

    The text directs readers to Table 3 for differences among empathy-measurement perspectives. The table labeled Table 3 instead presents causal-inference conditions.

  • The count of studies is conflated with the count of samples

    Minor

    The methods distinguish 82 client samples from 80 separate studies, but Table 1 labels its total of 82 as studies included.

  • Study-level heterogeneity has two reported values

    Minor

    The text reports a study-level I2 of 72%, whereas Table 2 reports 76.8%. The 72.2% value in the table belongs to the effect-level analysis.

4 of 4

Cites

Flagged first, then by citations

1 of the 118 references couldn’t be matched to a known paper.

Research works printed but not linked: 25

  1. Barrington, B. L. (1967). The differential effectiveness of therapy as measured by the Thematic Apperception Test. In C. R. Rogers et al. (Eds.), The therapeutic relationship and its impact (pp. 337-352). Madison: University of Wisconsin Press. . [combined with Kiesler, Mathieu, Klein & Schoeninger, 1967, and Van der Veen, 1967]
  2. Beutler, L. E., Crago, M., & Arizmendi, T. G. (1986). Research on therapist variables in psychotherapy. In S. L. Garfield & A. E. Bergin (Eds.), Handbook of psychotherapy and behavior change (3 rd ed., pp. 257-310). New York: Wiley.
  3. 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.
  4. Gurman, A.S. (1977). The patient's perception of the therapeutic relationship. In A.S.Gurman & A.M. Razin (Eds.), Effective Psychotherapy: A handbook of research (pp. 503-543). New York: Pergamon.
  5. Hamilton, J. C. (2000). Construct validity of the core conditions and factor structure of the Client Evaluation of Counselor Scale. The Person-Centered Journal, 7, 40-51.
  6. Hansen, J. C., Moore, G. D., & Carkhuff, R. R. (1968). The differential relationships of objective and client perceptions of counseling. Journal of Clinical Psychology, 24, 244-246.
  7. Ickes, W. (Ed.). (1997). Empathic accuracy. New York: Guilford.
  8. Johnson, S.M., Bradley, B., Furrow, J., Lee, A., Palmer, G., Tilley, D.G., & Woolley, S.R. (2005). Emotionally Focused Marital Therapy Workbook. New York: Brunner Routledge.
  9. Kiesler, D. J., Klein, M. H., Mathieu, P. L., & Schoeninger, D. (1967). Constructive personality change for therapy and control patients. In C. R. Rogers et al. (Eds.), The therapeutic relationship and its impact (pp. 251-294), Madison: University of Wisconsin Press. (Combined with Barrington, 1967, and Van Der Veen, 1967) *
  10. Lerner, B. (1972). Therapy in the ghetto. Baltimore: Johns Hopkins University Press.
  11. Martin, D. G. (2000). Counseling and therapy skills (2 nd ed.). Prospect Heights, Illinois: Waveland Press.
  12. Melnick, B., & Pierce, R. M. (1971). Client evaluation of therapist strength and positive- negative evaluation as related to client dynamics, objective ratings of competence and outcome. Journal of Clinical Psychology, 27, 408-410.
  13. Muller, J., & Abeles, N. (1971). Relationship of liking, empathy and therapists' experience to outcome in psychotherapy. Journal of Counseling Psychology, 18, 39-43.
  14. Orlinsky, D. E., Grawe, K., & Parks, B. K. (1994). Process and outcome in psychotherapy-noch einmal. In A. E. Bergin & S. L. Garfield (Eds.), Handbook of psychotherapy and behavior change (4 th ed., pp. 270-378). New York: Wiley.
  15. Orlinsky, D.E., Rønnestad, M.H., & Willutzki, U. (2003). Process and Outcome in Psychotherapy. In M.J. Lambert (Ed.), Bergin and Garfield's Handbook of Psychotherapy and Behavior Change (5th ed) (pp. 307-389). New York: Wiley.
  16. Parloff, M. B., Waskow, I. E., & Wolfe, B. E. (1978). Research on therapist variables in relation to process and outcome. In Garfield, S. L., & Bergin, A. E. (Eds.), Handbook of psychotherapy and behavior change (2 nd ed., pp. 233-282). New York: Wiley.
  17. Peake, T. H. (1979). Therapist-patient agreement and outcome in group therapy. Journal of Clinical Psychology, 35U, 637-646.
  18. Sachse, R. (1990b). The influence of therapist processing proposals on the explication process of the client. Person-Centered Review, 5, 321-344.
  19. Shlien, J. (1997). Empathy in psychotherapy: A vital mechanism? Yes. Therapist's conceit? All too often. In A. C. Bohart & L. S. Greenberg (Eds.), Empathy reconsidered: New directions in psychotherapy (pp. 63-80). Washington, DC: American Psychological Association.
  20. Strupp, H. H., Fox, R. E., & Lessler, K. (1969). Patients view their psychotherapy. Baltimore: Johns Hopkins Press.
  21. Truax, C.B., Carkhuff, R. R., & Kodman, F., Jr. (1965). Relationships between therapist- offered conditions and patient change in group psychotherapy. Journal of Clinical Psychology, 21, 327-329.
  22. Truax, C. B., & Mitchell, K. M. (1971). Research on certain therapist interpersonal skills in relation to process and outcome. In A. E. Bergin & S. L. Garfield (Eds.), Handbook of psychotherapy and behavior change (1 st ed., pp. 299-344). New York: Wiley.
  23. Truax, C. B., & Wittmer, J. (1971). The effects of therapist focus on patient anxiety source and the interaction with therapist level of accurate empathy. Journal of Clinical Psychology, 27, 297-299.
  24. Truax, C. B., Wittmer, J., & Wargo, D. G. (1971). Effects of the therapeutic conditions of accurate empathy, nonpossessive warmth, and genuineness on hospitalized mental patients during group therapy. Journal of Clinical Psychology, 27, 137-142.
  25. Watson, J.C., & Prosser, M. (2002). Development of an observer rated measure of therapist empathy. In J. C. Watson, R. Goldman, & M Warner (Eds.), Client-centered and experiential psychotherapy in the 21st century: Advances in theory, research and practice (pp.303-314). Ross on Wye: PCCS Books.

Cited by

Sorted by citations

Related

Authors
Elliott, R., Bohart, A. C., Watson, J. C., & Murphy, D.
Affiliations
University of Nottingham (GB)
Corresponding author
Elliott, Robert
ORCID
Venue
Psychotherapy
Publisher
American Psychological Association (APA)
Published
2018-12
Volume
55
Issue
4
Pages
399-410
ISSN
  • 0033-3204 (ISSN-L)
  • 1939-1536
DOI
10.1037/pst0000175
OpenAlex
W2897362601
Semantic Scholar
f8cb722ce1760eb9a601849b1586d8f6fccac372
Cited by
495
References
118
Citations by year
2026: 97 · 2025: 86 · 2024: 58 · 2023: 85 · 2022: 65 · 2021: 53 · 2020: 31 · 2019: 17 · 2018: 2 · 2014: 1
FWCI
30.26
Type
journal article
Language
en
Open access
Locations
  • Psychotherapy (publishedVersion)
  • PubMed (publishedVersion)
  • Nottingham ePrints (University of Nottingham) (acceptedVersion)
  • Strathprints: The University of Strathclyde institutional repository (University of Strathclyde) (submittedVersion)
  • Repository@Nottingham (University of Nottingham) (submittedVersion, cc-by)
Topics
  • Social Sciences › Psychology › Clinical Psychology › Psychotherapy Techniques and Applications
  • Health Sciences › Medicine › Psychiatry and Mental health › Empathy and Medical Education
  • Social Sciences › Psychology › Social Psychology › Counseling Practices and Supervision
Keywords
therapist empathy, client outcomes, psychotherapy outcome, empathic accuracy, consumer perception, viewer perception, therapist perspectives, theoretical orientation
MeSH terms
Attitude of Health Personnel, Mental Disorders, Empathy, Humans, Professional-Patient Relations, Psychotherapy, Treatment Outcome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review