Which of the following is the most important determinant of the effectiveness of psychotherapy?

J Psychother Pract Res. 1998 Summer; 7(3): 236–248.

Abstract

Patient and therapist expectancies regarding the “typical session” were measured during a controlled trial of short-term, time-limited individual psychotherapy. Relationships between expectancy ratings and measures of the therapeutic alliance and treatment outcome were examined. Significant relationships were tested in the presence of a competing predictor variable, either pre-therapy disturbance (depression) or the patient's quality of object relations (QOR). Expectancies were associated strongly with the alliance but only moderately with treatment outcome. In most instances, expectancy and QOR combined in an additive fashion to account for variation in alliance or outcome. The patient's capacity for mature relationships and expectancies for therapy appear to be important determinants of treatment process and outcome. The clinical value of establishing accurate, moderate expectancies prior to therapy is considered.(The Journal of Psychotherapy Practice and Research 1998; 7:236–248)

Expectancies about psychotherapy include beliefs about the duration of treatment, the process of therapy, and the outcome of treatment. In 1959 Frank1 suggested that the beliefs or attitudes a patient brings to therapy have an important influence on the process and outcome of treatment. Expectancy variables have since occupied an awkward place in psychotherapy research: while continuing to hold promise as significant components of the change process, they have received only inconsistent empirical support.2

The most reliable finding in the literature is the direct relationship between the expected and actual duration of treatment.3 Confirming any significant effects of expectancy on therapy outcome has been difficult because of discrepant findings across studies. Methodological differences may help explain the inconsistency of results.2 For therapists' ratings of outcome, the effects of outcome expectancy appear negligible.4 Stronger findings have emerged when the patient's ratings are considered, with expectancy accounting for 8% to 12% of the variation in therapy outcome. Reviews of research on individual5 and group therapy6 conclude that expectancy variables do have some promise as predictor variables and should be considered more systematically by clinicians and researchers.

To deal with certain methodological difficulties, Perotti and Hopewell7 suggest the effects of expectancy should be differentiated according to the stage of therapy. Initial outcome expectancies are subject to revision as treatment progresses and thus may lose their predictive power. In contrast, initial expectancies regarding the therapy relationship may be more important because they represent the patient's preparedness for early engagement in, and presumably benefit from, the treatment process. We adopted this rationale for an examination of initial expectancy ratings collected during a controlled trial of short-term individual (STI) psychotherapy conducted in Edmonton.8 We predicted that our measures of expectancy would be strongly and directly associated with ratings of the therapeutic alliance, but only weakly if at all related to measures of therapy outcome.

We previously reported that the time-limited interpretive therapy evaluated in the controlled trial was effective on both statistical and clinical grounds.8 We also found direct relationships between patient and therapist ratings of the therapeutic alliance and treatment outcome.9 Similar direct relationships have been highlighted in reviews.10,11 Our present examination of the relationships of patient and therapist expectancies to alliance and outcome had four objectives:

1. To assess the simple relationships between initial patient and therapist expectancies regarding the “typical session” and measures of the therapeutic alliance.

2. To assess relationships between expectancy and therapy outcome.

3. To assess predictive relationships involving measures of the degree of confirmation or disconfirmation of initial expectancies by subsequent session evaluations collected during the course of treatment. Frank1 and his colleagues12 argued that the confirmation of expectancy should be directly related to therapy benefit.

4. To evaluate the simple relationships between expectancy and alliance or outcome against the prediction provided by two competing variables. One competing predictor variable was a quantitative measure of the patient's developmental level of interpersonal relations. Our clinical trial of STI therapy provided evidence that the patient personality variable quality of object relations (QOR) was directly related to the therapeutic alliance and treatment outcome.9 We used the patient's initial level of depressive symptoms, based on pre-therapy scores from the Beck Depression Inventory,13 as the second competing predictor variable.

Methods

The reader is directed to the original report of the controlled trial8 for methodological details.

Setting and Procedures

The setting for the clinical trial was the Psychiatric Walk-In Clinic, Department of Psychiatry, University of Alberta Hospitals Site in Edmonton. Patients were matched in pairs on QOR, age, and gender, and then randomly assigned to immediate or delayed therapy and to one of eight project therapists. During a 3-year period, 86 of 105 patients who began therapy completed the protocol. Sixty-four of these were chosen to form a sample that was balanced for QOR, treatment condition (immediate vs. delayed), and therapist.

Patients and Therapists

Diagnoses were made by the assessing therapist according to DSM-III14 after an initial assessment and consultation with a staff psychiatrist. For the sample of 64 patients, 72% received Axis I diagnoses, the most frequent being affective (27%), impulse control (7.8%), or anxiety (6.3%) disorder. An Axis II diagnosis was assigned for 27% of the sample, the most frequent being dependent (14%) or avoidant (5%) disorder. The average age of the patients was 32 years (SD = 8, range = 21–53 years), and 62% were female. Three psychiatrists, one psychologist, and four social workers served as therapists in the study. Their average age was 40 years, and they had practiced individual therapy for an average of 11.5 years.

Therapy

The time-limited therapy was dynamically oriented and followed a technical manual that drew on the approaches of Malan15 and Strupp and Binder.16 Interpretation and clarification were emphasized relative to support and direction. Twenty weekly sessions of 50 minutes' duration were planned; the average number of sessions attended was 18.8. The technical nature of the therapy was verified by a content analysis of therapist interventions for eight sessions (numbers 4, 7, 9, 11, 14, 16, 18, and 20), using the Therapist Intervention Rating System.17 On average, there were 44 interventions, 11 interpretations, and 5 transference interpretations per session, confirming that the therapists had been active, interpretive, and transference-oriented.

Predictor Variables

Expectancy Variables:

Patients completed a series of expectancy ratings as part of the initial outcome assessment. The first two sessions of STI therapy were commonly used for history-taking and development of rapport. Therapists completed expectancy ratings after the second therapy session. Expectancy ratings regarding the “typical session” were based on a modified version of Stiles's Session Evaluation Questionnaire (SEQ).18 As commonly used, the SEQ involves the rating of 12 semantic differential items (e.g., good–bad, easy–difficult) in response to the sentence stem, “This session was . . . .” Two scores, based on the underlying factor structure of the SEQ reported by Stiles,18 are obtained: Depth-Value represents the perceived usefulness of the session, and Smoothness-Ease represents the perceived comfort of the session. Scores range from a minimum of 1 to a maximum of 7. To represent expectancies at pre-therapy and early therapy, respectively, the patient and therapist rated the SEQ items in response to the sentence stem, “The typical therapy session will be . . . .” This approach allowed us to derive scores for expected session usefulness (Depth-Value) and expected session comfort (Smoothness-Ease).

Patients and therapists completed the usual form of the SEQ after each session. The two session evaluation scores were aggregated across all sessions for each participant. The difference (evaluation minus expectancy) was calculated for each measure for both patient and therapist, and represented the discrepancy from expected usefulness (Depth-Value) and comfort (Smoothness-Ease). Positive discrepancy scores indicated that the overall session evaluations exceeded initial expectancies (confirmation); negative scores indicated that the overall session evaluations failed to meet initial expectancies (disconfirmation).

Quality of Object Relations:

A personality variable, QOR is defined as a person's internal, enduring tendency to establish certain types of relationships with others.19 The dimension ranges across five levels of object relations (primitive, searching, controlling, triangular, and mature). In the clinical trial, the assessment of QOR comprised two 1-hour clinical interviews.

During the assessment, the lifelong pattern of relationships is examined. The interviewer considers the overall pattern of relationships in terms of behavioral manifestations, regulation of affect, regulation of self-esteem, and historical antecedents for each of the five levels. The interviewer then distributes 100 points among the five levels and derives a single global score ranging from 1 to 9.

At the primitive or low end of the 9-point scale, relations are characterized by inordinate dependence, extreme reactions to real or imagined loss, and destructiveness. At the mature end, relations are characterized by equity and the expression of love, tenderness, and concern. It is common for two overall scores of equal value to represent different patterns of object relatedness.

Since we conducted the STI therapy trial, the QOR assessment has been streamlined to a single hour of interview time, and reliability has been improved. In the clinical trial, the reliability between the interviewer and an independent rater using an audiotape was assessed for a sample of 50 cases. A stringent index of reliability, the intraclass correlation coefficient for the individual rater [ICC(1,1)], was used. A reliability coefficient of 0.50 was obtained.

For the current investigation, the overall QOR score (a continuous measure) was used as a predictor variable.

Initial Disturbance:

The pre-therapy score on the Beck Depression Inventory13 was used to represent initial disturbance, measured as severity of depressive symptoms prior to therapy. The BDI is a commonly used outcome measure with established psychometric properties.

Dependent Measures

Therapeutic Alliance:

The alliance was defined as the nature of the working relationship between patient and therapist. The two participants independently rated six 7-point items. Four “immediate” items were rated after each therapy session, and two “reflective” items were rated after each one-third of the therapy (at sessions 7, 14, and 20). Three immediate items addressed whether the patient had talked about private, important material, had felt understood by the therapist, and was able to understand and work with the therapist's interventions. The remaining immediate item concerned the overall usefulness of the session. The two reflective items addressed Luborsky's concept of the helping alliance (collaboration and helpfulness).20 Each set of six item ratings was aggregated across sessions or thirds; aggregate ratings were then subjected to a principal components analysis. One patient-rated alliance factor and two therapist-rated alliance factors (immediate, reflective) were derived.

Therapy Outcome:

The STI therapy outcome battery included several well-established self-report and interview measures of the patient's psychiatric symptomatology, interpersonal functioning, and personality functioning. The patient's individual target objectives were developed with the assistance of an independent assessor. Patient, therapist, and assessor ratings of target objective distress were included in the outcome battery. A total of 23 outcome variables were available; 19 were measured both before and after therapy (residual gain scores), and the remaining 4 were measured at post-therapy only (rated benefit scores). Seven variables were eliminated because of redundancy or a low response rate.

The results of a principal components analysis of 16 post-therapy outcome variables are presented in Table 1. The analysis identified four factors. The first three factors were retained to represent change due to treatment. Measures of improvement at post-therapy were the following: I, General Symptoms and Dysfunction (patient self-report); II, Individualized Objectives (patient, therapist, and independent assessor); and III, Social-Sexual Adjustment (assessor).

TABLE 1.

Approach to Analysis

The relationships among the predictor variables (expectancy, discrepancy, QOR, BDI), and between the predictor and dependent variables (alliance, outcome), were examined by using Pearson product-moment correlation coefficients. Expectancy and discrepancy variables having significant simple relationships with alliance or outcome were then considered in a series of hierarchical multiple regression analyses. The regression analyses assessed the strength of the relationship against the prediction provided by competing variables.

The regression analysis for each simple relationship (expectancy or discrepancy with alliance or outcome) followed the same sequence. On the first step, a competing predictor (QOR or BDI) entered the equation. The expectancy or discrepancy variable was entered on the second step. The interaction variable (product of the two predictors) was entered on a third step. The regression was then repeated with the order of entry of the two (main effect) predictors reversed. All predictor variables were centered (the sample mean subtracted from each patient's score) to control for a form of error variance, nonessential ill-conditioning,27 which is defined as shared variance that is not due to a real association in the population. Specifically, predictor and dependent variables with similar measurement scales would contribute to nonessential ill-conditioning and raise the likelihood of type I error.

Results

The noncentered means and standard deviations for the therapeutic alliance, expectancy, QOR, and initial disturbance are presented in Table 2. Overall, patients expected that sessions would be significantly more useful (t = 3.91, df = 61, P<0.0001) but significantly less comfortable (t = –2.21, df = 62, P<0.03) than their therapists did. The mean discrepancy between session evaluations and expectancies was significantly larger (indicating greater confirmation) for patients than for therapists, both for usefulness (t = 2.97, df = 61, P<0.005) and for comfort (t = 3.17, df = 62, P<0.002). In general, most patients reported that the experience of therapy sessions met or exceeded their initial expectations.

TABLE 2.

Correlations Between Predictor Variables

Table 3 presents the intercorrelations among the 10 predictor variables. Except for the two therapist expectancy ratings (THDV, THSE), which were independent, each remaining pair of variables (e.g., the two patient expectancy, two patient discrepancy, and two therapist discrepancy variables) were significantly correlated. Overall, expectancy ratings were significantly and inversely related to the respective discrepancy scores. These relationships indicated that the higher the initial expectancy, the greater the likelihood of disconfirmation; that is, of a failure of session evaluations to meet expectations.

TABLE 3.

Two additional patterns of intercorrelation were identified. First, confirmation of the patient's expectancy of session comfort was associated with confirmation of the therapist's expectancies of both session comfort and usefulness. Second, confirmation of the therapist's expectancy of session comfort was directly associated both with lower patient expectancies of usefulness and confirmation of patient-expected usefulness. These relationships indicated a degree of patient–therapist interdependence in the evaluation of whether initial expectancies were confirmed by the actual experience of therapy sessions.

QOR was independent of the expectancy and discrepancy variables and was inversely related to initial disturbance. Patient expectancies were inversely related to initial disturbance: the greater the patient's depressive symptoms at pre-therapy, the lower the expectancies of session usefulness and comfort.

Simple Predictions

Table 4 presents the simple relationships among the 10 predictor variables (QOR, initial disturbance, 4 expectancy, 4 discrepancy) and the 6 dependent variables (3 therapeutic alliance, 3 post-therapy outcome).

TABLE 4.

From previous work,9 we knew that QOR was directly associated with the patient-rated alliance, the therapist-rated reflective alliance, and improvement on two of the three outcome factors (I and II). The BDI score was inversely associated with the patient-rated alliance.

Expectancy and Alliance:

Three expectancy–alliance relationships were identified, each involving a distinct pair of expectancy and alliance variables associated with the same rating source. First, the patient's expectancy of usefulness was directly associated with the patient-rated alliance. Second, the therapist's expectancy of usefulness was directly associated with the therapist-rated immediate alliance. Third, the therapist's expectancy of session comfort was directly associated with the therapist-rated reflective alliance. These correlations indicated that expectancy accounted for 18% to 40% of the variation in alliance ratings.

Expectancy and Outcome:

Expectancies regarding session comfort were directly associated with improvement at post-therapy. Three significant relationships were identified, each involving one of the three outcome factors. The patient's expectancy of session comfort was directly associated with benefit on General Symptoms and Dysfunction (I) and Social-Sexual Adjustment (III). The therapist's expectancy of session comfort was directly associated with benefit on Individualized Objectives (II). These correlations indicated that expectancy accounted for 7% to 10% of the variation in outcome scores. This was considerably less than the variation of alliance accounted for by expectancy.

Discrepancy and Alliance/Outcome:

Confirmation of each of the patient's initial expectancies (usefulness, comfort) was directly associated with the therapist's rating of the reflective alliance. Discrepancy scores were not significantly associated with therapy outcome.

Multivariate Relationships

Expectancy–Alliance:

Three relationships were tested: patient-expected usefulness and patient alliance; therapist-expected usefulness and therapist immediate alliance; and therapist-expected comfort and therapist reflective alliance. For the first relationship, initial disturbance (depression) was considered as the first competing predictor variable. Initial disturbance was predictive of the patient-rated alliance, as described above. However, in the presence of the expectancy variable (patient usefulness), this contribution did not attain significance. In all of the remaining analyses, the pre-therapy BDI score was found not to account for significant proportions of criterion variance. As a competing predictor variable, initial disturbance will not be addressed further.

Remaining with the patient-expected usefulness–patient alliance relationship, our next step in the analysis was to consider QOR as a competing predictor variable. Table 5 presents the regression analysis. Both predictors (QOR, expectancy) were significant, but the interaction was not. Proportions of alliance variance accounted for were averaged across the pair of regression analyses conducted to test each expectancy–alliance relationship. QOR accounted for 7%, and the expectancy variable for an additional 26%, of the variation in the patient-rated therapeutic alliance.

TABLE 5.

Therapist-expected usefulness emerged as the only significant predictor of the therapist-rated immediate alliance. For the relationship between therapist-expected comfort and the therapist-rated reflective alliance, a similar pattern of findings was evident: both predictors (expectancy and QOR) were significant, and the interaction was not. QOR accounted for 7%, and the expectancy variable for an additional 16%, of the variation in the therapist-rated reflective alliance.

Expectancy–Outcome:

Three relationships were tested: patient-expected comfort and General Symptoms and Dysfunction (I); patient-expected comfort and Social-Sexual Adjustment (III); and therapist-expected comfort and Individualized Objectives (II). Analyses with QOR as the competing predictor again resulted in important findings.

Table 6 presents the results of the regression analysis for General Symptoms and Dysfunction (I). QOR and patient-expected comfort both emerged as significant predictors, but the interaction did not. Each predictor accounted for roughly 7% of the variance in symptomatic improvement. Patient-expected comfort emerged as the only significant predictor of Social-Sexual Adjustment (III). For Individualized Objectives (II), both predictors emerged as significant, and the interaction did not. QOR accounted for approximately 11% of the variance in improvement, and therapist-expected comfort accounted for an additional 8% of outcome variance.

TABLE 6.

Discrepancy–Alliance:

The two patient discrepancy variables having significant relationships with the therapist's reflective alliance were themselves highly correlated (r = 0.55, df = 60, P<0.0001). To maintain consistency with the other analyses, separate regression analyses were conducted to test the strength of each discrepancy–alliance relationship on its own.

For the regression analysis involving discrepancy scores for patient-expected usefulness, QOR and the discrepancy variable both emerged as significant predictors of the alliance, but the interaction did not. For the analysis involving the discrepancy scores for patient-expected comfort, there was evidence for significant independent contributions by each predictor and for the interactive effect. Table 7 presents the result of the regression analysis. QOR accounted for approximately 8%, patient comfort accounted for approximately 7%, and the interaction accounted for an additional 9% of the variance in the therapist-rated reflective alliance. The interaction indicated that the greater the confirmation of the patient's expectancy (the more positive the discrepancy between experienced and expected comfort), the stronger the direct effect of the patient's QOR on the therapist's general perception of the alliance.

TABLE 7.

Expectancy and Alliance as Joint Predictors of Outcome

We returned to the expectancy variables at this point in the analysis. We were interested in whether expectancies would still significantly account for outcome variance when the prediction afforded by the therapeutic alliance was considered first. Three hierarchical regression analyses were conducted. For outcome factor I (General Symptoms and Dysfunction), the predictors were QOR, each of the alliance variables in turn, and patient-expected comfort. QOR accounted for 7% of outcome variance, as above, but when alliance and expectancy were in the equation the direct effect of QOR was no longer significant. Alliance accounted for 7% to 13% of outcome variance; each alliance variable provided for significant prediction in the regression. The patient expectancy rating, when entered last, accounted for an additional 6% to 14% of outcome variance and was also a significant predictor in each analysis. For outcome factor II (Individualized Objectives), the predictors were QOR, the alliance variables, and the therapist's expected comfort. QOR accounted for 12% and the alliance for 19% to 22% of outcome variance, but therapist expectancy did not provide for a significant additional contribution. For outcome factor III (Social-Sexual Adjustment), the predictors were the alliance variables and patient-expected comfort. Only the expectancy variable accounted for significant outcome variance (9%–11%). These additional analyses indicated that patient expectancy, but not therapist expectancy, provided for a significant prediction of outcome over and above the prediction afforded by the alliance.

Discussion

We studied patient and therapist expectancy ratings as potential predictors of the therapeutic alliance and treatment outcome. The analyses demonstrated that expectancies regarding the experience of therapy sessions are strongly and directly related to the quality of the therapeutic alliance. Relationships between expectancy and outcome proved to be less strong but still substantial. In the multivariate analyses, expectancy variables frequently combined additively with quality of object relations in accounting for variation in alliance and outcome. In an analysis examining the joint prediction of outcome, QOR, the alliance, and patient expectancy were found to independently contribute to therapy benefit. We will consider the results and their clinical implications in the sequence that was followed in the preceding section.

The simple descriptive analyses (direct comparisons of patient and therapist ratings, correlations among the predictor variables) proved to be quite informative. High expectancies were clearly related to the experience of disconfirmation—that is, disappointment with actual therapy sessions. In direct comparisons of the expectancy ratings, patients expected significantly more session usefulness but significantly less session comfort than therapists. To put this another way, therapists had moderate expectancies about therapy sessions relative to patients. The two therapist expectancy variables were found to be independent of one another, which also suggested that the therapists had a more differentiated picture of the therapy process. In effect, it is likely that therapists “know what to expect” as therapy begins. This clinical understanding of the therapy process should be employed during the preparation phase to modify any patient expectations that appear to be overly optimistic or idealized.

Correlations between patient and therapist discrepancy scores indicated that there is a clear dyadic interdependence when session experiences are evaluated against expectancies. Patient discrepancy scores were significantly more positive than were therapist discrepancy scores. For the patients, the actual experience of therapy was generally in line with or exceeded their expectations, suggesting that for most of them, therapy was a reasonably positive experience.

Overly optimistic or idealized expectations thus may not be a frequent occurrence, but they should definitely be addressed if they are identified early in the treatment process. Ensuring that the patient has reasonable expectancies about the treatment experience will militate against disappointment. Although this point was not addressed by our analyses, it is also possible that reasonable expectancies that are shared by the patient and therapist would be even more strongly associated with the quality of the therapeutic collaboration.

Substantial expectancy–alliance relationships were identified. For patients and therapists, expectancies of session usefulness were directly associated with the strength of the respective alliance ratings. Beginning therapy with the expectation that individual sessions will be productive may help ensure that the therapy relationship is also productive, or at least is perceived as productive. The therapist's expectancy of session comfort was directly associated with his or her rating of the reflective alliance. This relationship suggests that if the therapist believes he or she will be comfortable in the therapy, again perhaps as a result of a productive preparation, more general perceptions of the treatment relationship will also turn out to be positive.

Expectancy–outcome relationships were notably smaller in absolute value than expectancy–alliance relationships. This discrepancy supports the findings of Perotti and Hopewell,7 which suggest that expectancies may have more direct effects on the establishment of the therapeutic alliance than on the actual outcome of treatment. Expectancies regarding session comfort were nonetheless clearly associated with treatment benefit. For patients, who completed these ratings prior to meeting the therapist, expectancies of session comfort may have reflected “preparedness” and a positive intention to engage in meaningful self-examination. For therapists, who completed ratings after two sessions, expectancies regarding comfort may have reflected positive impressions of the patient and of the potential for collaboration.

Patient expectancies of comfort were directly associated with symptom improvement and overall adjustment in social activity and intimate relationships. Expecting sessions to be relatively comfortable may indicate openness to the relationship with the therapist and the process of therapy. A simple assessment of the patient's expectancy of session comfort could be used as an early indicator of potential change in symptomatic and interpersonal distress. Therapist expectancies of comfort were directly associated with positive change on individualized objectives for therapy. Therapist expectancies of comfort may reflect an estimation of the potential for collaboration on the patient's problems, involving judgments about appropriateness and capacity for therapy, the usefulness of any preparation, and the therapist's own experience with treatment for similar problems.

Relative to expectancy ratings, the discrepancy scores were less fruitful as predictor variables. Confirmation of the patient's expectancies was directly associated with the therapist's reflective alliance. If the patient finds that sessions meet or exceed expectations, the therapist's general perception of the therapeutic alliance is positive. A reasonable confirmation of the patient's expectancies may represent a therapist objective for the early stages of therapy.

Multivariate analyses aimed at testing the robustness of the simple relationships involving the expectancy and discrepancy variables. The competing predictors included an index of the patient's capacity for healthy interpersonal relationships (QOR) and an established measure of initial symptomatic distress (BDI). The first set of regression analyses considered the three expectancy–alliance relationships. Initial disturbance was eliminated as a significant predictor in one analysis, and it did not have a significant relationship with the criterion in any subsequent analyses. QOR was significant as a competing predictor in two of three analyses, in each case accounting for roughly 7% of the variation in the quality of the therapeutic alliance. In sharp contrast, expectancy was a significant predictor in all three analyses and accounted for a large proportion (16%–40%) of variation in the alliance. The prediction provided by QOR and expectancy was additive.

This finding has implications for the selection and preparation of patients for short-term interpretive therapy. A capacity to establish a good working relationship (selection) and the expectation that work will occur comfortably and productively during therapy sessions (preparation) are strongly associated with a positive therapeutic alliance.

The second set of regression analyses tested the three expectancy–outcome relationships. A similar pattern of findings emerged. QOR was predictive of improvement in two of three analyses. In all three analyses, the expectancy variable made a significant single or additional contribution to the prediction. Expectancy accounted for roughly 8% of outcome variance in each instance.

The third set of regression analyses considered the relationships between confirmation of the two patient expectancies and the therapist-rated reflective alliance. When the patient discrepancy score for expected usefulness was used as a predictor, the familiar pattern of findings emerged: both quality of object relations and the discrepancy variable accounted for significant proportions of alliance variance, but the interaction did not. The therapist's rating of the general quality of the therapeutic alliance was elevated when the patient presented with a good capacity for interpersonal relationships and a belief in the usefulness of the therapy process.

When the patient discrepancy score for expected comfort was used as a predictor, all three effects (QOR, patient discrepancy, and the interaction) emerged as significant. Thus, confirmation of the expectancy that sessions would be comfortable increased the likelihood that the patient's capacity for satisfying relationships would be put to use in the work of therapy. Ensuring that the patient is comfortable with the demands of the therapy process prior to and during sessions allows for the development of the best possible patient–therapist relationship. This multiplicative effect represented an important independent contribution to the prediction of the therapist-rated reflective alliance.

The final set of regression analyses was prompted by our interest in the joint prediction of outcome by three variables: the quality of object relations, the therapeutic alliance, and expectancy. If expectancy accounted for outcome variance over and above the contributions of QOR and the alliance, this would underscore the importance of the relationship. The results showed that symptomatic improvement was strongly predicted by the alliance and the patient's expectancy of session comfort; QOR was eliminated as a predictor when these variables were present in the regression equation. Change on individualized objectives was predicted by QOR and the alliance, but not by therapist expectancy. Change in broader overall adjustment was predicted solely by the patient's expectancy of session comfort.

Taken together, the results of these additional analyses suggest two conclusions. First, patient expectancies are strong predictors of therapy outcome, but therapist expectancies are not. Second, the patient's capacity for a good relationship, the patient's expec0tancy that the therapy sessions will be comfortable, and the actual experience of a strong therapeutic alliance all represent consistently strong determinants of therapy benefit.

The strength of our findings with measures of patient and therapist expectancy was somewhat of a surprise, particularly given the simplicity of the expectancy rating. The findings clearly argue for the preparation of patients for short-term, time-limited individual psychotherapy. Referring therapists, or the treating therapist at the time of a treatment contract, should seek to reinforce moderate patient expectancies. Overly high expectancies are likely to be painfully disconfirmed and perhaps increase the likelihood of a treatment dropout. Reasonable expectancies represent one goal for the patient's preparation for therapy. In terms of expectancies regarding session usefulness, the patient should understand that each session contributes to overall benefit and by itself is unlikely to have dramatic effects on the presenting problem.

In terms of expectancies regarding session comfort, the patient should be clear that some degree of session difficulty is associated with the hard work of a successful psychotherapy. After therapy has actually started, one aspect of the therapist's activity should be to engage the patient in a “good” working process28 and reinforce the patient when this is achieved.

Confirming an early expectancy that sessions can be productive and comfortable may make it more likely that the patient and therapist will be able to establish a good working relationship. This confirmation can also allow the patient's capacity for healthy relationships to come more fully to the fore in the therapy process. In turn, the patient's actual experience of a strong alliance can be the foundation for a successful treatment outcome.

References

1. Frank JD: The dynamics of the psychotherapeutic relationship. Psychiatry 1959; 22:17–39 [PubMed] [Google Scholar]

2. Beutler LE, Machado PPP, Allstetter Neufeldt S: Therapist variables, in Handbook of Psychotherapy and Behavior Change, 4th edition, edited by Garfield SL, Bergin AE. New York, Wiley, 1994, 229–269

3. Jenkins SJ, Fuqua DR, Blum CR: Factors related to duration of counseling in a university counseling center. Psychol Rep 1986; 58:467–472 [PubMed] [Google Scholar]

4. Bonner BL, Everett FL: Influence of client preparation and problem severity on attitudes and expectations in child psychotherapy. Professional Psychology: Research and Practice 1986; 17:223–229 [Google Scholar]

5. Luborsky L, Crits-Christoph P, Mintz J, et al: Who Will Benefit From Psychotherapy? Predicting Therapeutic Outcomes. New York, Basic Books, 1988

6. Piper WE: Client variables, in Handbook of Group Psychotherapy, edited by Fuhriman A, Burlingame GM. New York, Wiley, 1994, 83–113

7. Perotti LP, Hopewell CA: Expectancy effects in psychotherapy and systematic desensitization: a review. JSAS: Catalog of Selected Do0cuments in Psychology 1980; 10: Ms No 2052

8. Piper WE, Azim HFA, McCallum M, et al: Patient suitability and outcome in short-term individual psychotherapy. J Consult Clin Psychol 1990; 58:475–481 [PubMed] [Google Scholar]

9. Piper WE, Azim HFA, Joyce AS, et al: Quality of object relations vs. interpersonal functioning as predictors of therapeutic alliance and psychotherapy outcome. J Nerv Ment Dis 1991; 179:432–438 [PubMed] [Google Scholar]

10. Henry WP, Strupp HH, Schacht TE, et al: Psychodynamic approaches, in Handbook of Psychotherapy and Behavior Change, 4th edition, edited by Bergin AE, Garfield SL. New York, Wiley, 1994, 467–508

11. Horvath AO, Symonds BD: Relation between working alliance and outcome in psychotherapy: a meta-analysis. Journal of Counseling Psychology 1991; 38: 139–149

12. Frank JD, Gliedman LH, Imber SD, et al: Patients' expectancies and relearning as factors determining improvement in psychotherapy. Am J Psychiatry 1959; 115:961–968 [PubMed] [Google Scholar]

13. Beck AT, Steer RA: Beck Depression Inventory Manual. New York, Harcourt Brace Jovanovich, 1987

14. American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, 3rd edition. Washington, DC, American Psychiatric Association, 1980

15. Malan DH: The Frontier of Brief Psychotherapy. New York, Plenum, 1976

16. Strupp HH, Binder JL: Psychotherapy in a New Key: A Guide to Time-Limited Dynamic Psychotherapy. New York, Basic Books, 1984

17. Piper WE, Debbane EG, de Carufel FL, et al: A system for differentiating therapist interpretations and other interventions. Bull Menninger Clin 1987; 51:532–550 [PubMed] [Google Scholar]

18. Stiles WB: Measurement of the impact of psychotherapy sessions. J Consult Clin Psychol 1980; 48:176–185 [PubMed] [Google Scholar]

19. Azim HFA, Piper WE, Segal PM, et al: The Quality of Object Relations Scale. Bull Menninger Clin 1991; 55:323–343 [PubMed] [Google Scholar]

20. Luborsky L: Principles of Psychoanalytic Psychotherapy: A Manual for Supportive-Expressive Treatment. New York, Basic Books, 1984

21. Hirschfeld RMA, Klerman GL, Gough JG, et al: A measure of interpersonal dependency. J Pers Assess 1977; 41:610–618 [PubMed] [Google Scholar]

22. Rosenberg M: Conceiving the Self. New York, Basic Books, 1979

23. Piper WE, Debbane EG, Garant J: An outcome study of group psychotherapy. Arch Gen Psychiatry 1977; 34:1027–1032 [PubMed] [Google Scholar]

24. Spielberger CD: Manual for the State-Trait Anxiety Inventory. Palo Alto, CA, Consulting Psychologists Press, 1983

25. Derogatis LR: SCL-90-R Manual I. Baltimore, MD, Johns Hopkins University School of Medicine, Clinical Psychometrics Unit

26. Weissman MM, Paykel ES, Siegel R, et al: The social role performance of depressed women: a comparison with a normal sample. Am J Orthopsychiatry 1971; 41:390–405 [PubMed] [Google Scholar]

27. Aiken LS, West SG: Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA, Sage, 1991

28. Hoyt M: Therapist and patient actions in “good” psychotherapy sessions. Arch Gen Psychiatry 1980; 37:159–161 [PubMed] [Google Scholar]

Articles from The Journal of Psychotherapy Practice and Research are provided here courtesy of American Psychiatric Publishing