as being more reliable than the others, but it also failed to eliminate a scale from further consideration as a research tool because of its psychometric properties. These results suggest that researchers cannot base their choice of an alliance scale on its reliability indices; the scales all tend to receive strong support. The alliance ratings of patients, therapists, and observers all tended to have adequate reliability. Although the ratings of therapists seemed to be slightly less consistent than those of patients and observers, therapists' ratings of the alliance were still within the acceptable range. Across therapy sessions, patients tended to rate the alliance more consistently than did therapists or observers. On the basis of the present meta-analysis, it seems that patients tend to view the alliance as stable, whereas therapists and observers tend to indicate more change over time in their alliance ratings. The implications of this finding are clear: Because patients tend to view the alliance consistently throughout treatment, they are more likely to view the alliance as positive at termination if their initial assessment was positive. Thus, therapists must be effective at establishing positive alliances with their patients early in the therapy process. However, because of the small sample size of this comparison, the greater consistency of patient ratings across alliance sessions should be considered a tentative finding. Most of the alliance scales have been shown to be related to outcome. The Penn scales, the Vanderbilt scales, the WAI, and the CALPAS were moderately correlated with outcome, but the TARS failed to receive support. In addition, the Penn scales, the Vanderbilt scales, the WAI, and the CALPAS have received far more empirical scrutiny than any of the other alliance scales. Of these measures, the WAI is likely to be appropriate for most research projects. The scale was designed to measure alliance factors in all types of therapy and to measure the theoretical constructs underlying the alliance. The scale provides both an overall alliance score and also an assessment of Bordin's (1979) three aspects of the alliance: the bond, the agreement on goals, and the agreement on tasks. The WAI also provides an assessment of Horvath and Luborsky's (1993) two core aspects of the alliance measured by most scales: (a) therapist-patient affective attachments and (b) collaboration or willingness to invest in the therapy process. In addition, patient-, therapist-, and independent observerrated versions of the scale are available, as are shortened versions of these scales. The overall correlation of alliance and outcome did not seem to be influenced by publication status. Although the unpublished studies included in the meta-analysis had a slightly lower average correlation than did the published studies, the difference was not significant. Similarly, it is highly unlikely that enough unlocated studies with null results exist in file drawers to reduce the overall allianceoutcome correlation to a level of nonsignificance. Indeed, it would take 331 studies averaging null results to reduce the correlation of the alliance and outcome to .05. G$#.1/.\!
Daniel J. Martin, John P. Garske, and M. Katherine Davis, Ohio University
4'20&&;443&&32D.2E&&F&&19.3;&&2;.1.40&&GHIJ&&F& CQV
Made with FlippingBook Ebook Creator