we found in the current meta-analysis is a little larger but similar to the values reported in previous research (Horvath & Symonds, 1991 r _ .26, k _ 26; Horvath & Bedi, 2002, r _ .21, k _ 100). The median effect size of ESs of the current data set was .28 suggesting that the group of effect sizes we collected was not strongly skewed. The overall effect size of .275 is statistically significant at p _ .0001 level indicating a moderate but highly reliable relation between alliance and psychotherapy outcome. In Addition, the impact of six categorical variables were investigated that have the potential of moderating the relation between alliance and outcome: alliance measure (CALPAS, VPPS, HAq, WAI, and Other); alliance rater (client, therapist, observer); time of alliance assessment (Early, Mid, Late, Averaged); outcome measure (BDI, SCL, Dropout); type of treatment (CBT, IPT, Psychodynamic, Substance Abuse); and publication source (journal, books/chapters, unpublished/thesis). There are several noteworthy features that apply to all of these results: All of the aggregate alliance-outcome correlations in each category are statistically significant beyond p _ .001. This result strongly supports the claim the impact of the alliance on therapy outcome is ubiquitous irrespective of how the alliance is measured, from whose perspective it is evaluated, when it is assessed, the way the outcome is evaluated, and the type of therapy involved. The quality of the alliance matters. The next most common feature is the finding that, with very few exceptions, within each of these subsets of data, the ES are very diverse in magnitude. It was noted earlier that heterogeneity of the ESs in a large-scale meta-analysis is not unusual. However, these results indicate that the high degree of variability remains practically unchanged within each level of these potential moderators. @)15-1.*$# The positive relation between the quality of the alliance and diverse outcomes for many different types of psychological therapies is confirmed in this meta-analysis. While the overall ES of r _ .275 accounts for a relatively modest proportion of the total variance in treatment outcome, the magnitude of this correlation, along with therapist effects, is one of the strongest and most robust predictors of treatment success empirical research has been able to. By including all research in which the authors refer to the process variable as alliance , the study might have collected and summarized a number of different kinds of things. A practical response to this conceptual problem is to conclude that this meta- analysis reports the results of alliance-outcome relation as it is researched at this time . In general, Studies on the alliance construct are an important contribution to psychotherapy research by creating diverse implications for therapeutic practice.
G$#.1/.!!
Prof. Adam O. Horvath. Simon Fraser University, Faculty of Education 8888 University Drive, Burnaby, BC, Canada V5A 1S6 Email: horvath@sfu.ca, Website: www.sfu.ca
4'20&&;443&&32D.2E&&F&&19.3;&&2;.1.40&&GHIJ&&F& CLV
Made with FlippingBook HTML5