The Prediction of Therapeutic Change

  • Sidney J. Blatt
  • Richard Q. Ford
Part of the Applied Clinical Psychology book series (NSSB)


One of the more fascinating and important questions in psychotherapy research is the prediction of who is likely to gain from the treatment experience. Still applicable today is Frank’s (1979) comment:

After decades of research, the amount of well-established, clinically relevant knowledge about psychotherapeutic outcome still remains disappointingly meager. Although some relationships between determinants and outcome have attained statistical significance, few are powerful enough to be clinically relevant, and most of [these] ... are intuitively obvious. An example is the finding that patients who begin therapy at a high level of functioning terminate at higher levels than those who begin at lower levels. In other words, the healthier one is to start with, the healthier one is at the end (Garfield, 1978). ... [R]esearch ... to date suggest that the major determinants of therapeutic success appear to lie in aspects of patients’ personality and style of life (pp. 310, 312).


Path Analysis Path Coefficient Developmental Level Interpersonal Communication Indirect Path 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Autocorrelations are the correlations between two measurements of the same variables at different points in time (e.g., the correlation between Rorschach scores obtained 15 months apart, as well as the correlations between variables derived from clinical evaluations also conducted 15 months apart). Synchronous correlations are the conventional correlations between measurements of different variables at the same point in time (e.g., the correlation between test variables and clinical case record ratings at admission, as well as the test-case record correlations 15 months later).Google Scholar
  2. 2.
    The adequacy of a model can be tested by comparing the path coefficients with the simple correlations, but failing that test, there is no mechanical procedure that guides the construction of a better model. Trial and error may yield a better model, or one may be forced to conclude that relationships among measured variables contain large components of spuriousness from unmeasured variables.Google Scholar
  3. 3.
    The variables from the TAT were not included in these data analyses because they are sex linked. Including TAT variables would require considering males and females separately, thereby reducing the sample size below an acceptable level. Variables from the analyses of the HFDs were also not included in these analyses because this material was available on only a limited number of patients.Google Scholar
  4. 4.
    An alternate measure of object representation is the developmental index—the weighted sum of accurately and inaccurately perceived human responses. Results with these two summary measures of object representation essentially replicate the findings with the mean developmental level of accurately perceived and inaccurately perceived human responses on the Rorschach. The developmental mean, rather than the developmental index, was used in these analyses because the developmental index of inaccurately perceived responses (OR—) correlated significantly with the measure of thought disorder (r =. 67), whereas the developmental mean of inaccurately perceived responses (OR—) was relatively statistically independent of the level of thought disorder (r =. 21). Generally the developmental mean scores seem to provide a more accurate estimate of the quality of object representations and to be somewhat more effective predictors of outcome than the developmental index scores.Google Scholar
  5. 5.
    The seven criteria of clinical change were relatively independent at the first evaluation. The intercorrelation of the four Strauss-Harder symptom scales ranged from. 25 to. 42. The Menninger Factor I score was minimally correlated with the Menninger Factor II score but highly correlated with the Fairweather Scale (r = −.64). The Menninger Factor I score and the Fairweather were only moderately correlated with the four Strauss-Harder symptom scales (r =. 17 to. 45). The Menninger Factor II score was independent of the four Strauss-Harder scales (r =. 03 to. 18). The five test predictor variables were also relatively independent at initial assessment. The developmental mean for the concept of the object for OR+ and for OR—were minimally intercorrelated with each other (r =. 28) and with the three other test variables (correlations ranged from. 06 to. 33). Thought disorder correlated significantly with degree of reality testing (F+%) (r = −.47) and the MOA mean (r =. 49). Degree of reality testing also correlated significantly with MOA mean (r = −.41). At the second evaluation, the Strauss-Harder symptom scales were moderately intercorrelated (r =. 19 to. 47). Menninger Factor I (Interpersonal Relations) correlated significantly (r = −.62) with Menninger Factor II (Impulsivity), with the Fairweather Scale of Interpersonal Communication (r = −.50), and with the Strauss-Harder Neurotic and Psychotic scales (r = −.42 and −.47, respectively), but only moderately with Labile and Flattened Affect (−.23 and −.29, respectively). The Fairweather Scale correlated significantly with the Strauss-Harder Psychotic Scale (r =. 61) but only moderately with the other three Strauss-Harder scales (r =. 26 to. 37). Menninger Factor II had only moderate correlations with the four Strauss-Harder symptom scales (r =. 20 to. 37). At the second evaluation, the five test variables were minimally intercorrelated. The two concept of the object developmental means for OR+ and OR—were minimally intercorrelated with each other (r =. 10) and with the three other test variables (r =. 02 to. 18). Thought disorder had moderate correlations with degree of reality testing (r = −. 28) and MOA mean (r =. 48), and the degree of reality testing had minimal correlation (r = −.16) with the MOA mean.Google Scholar
  6. 6.
    The MOA is a 7-point scale, and the scores ranged from 1.00 to 5.67 at Time 1 and from 1.00 to 6.00 at Time 2. The change of mean MOA scores from Time 1 to Time 2 ranged from −4.00 to +2.50. The MOA is a reverse scale; thus, negative scores indicate positive change.Google Scholar
  7. 7.
    To specify more precisely the source of the relationship between the mean MOA score and interpersonal relations (the Menninger Factor I score), analyses were conducted between the mean MOA score at Time 1 and the four component scales that make up Menninger Factor I. The results indicate that the relationship between the mean MOA score at Time 1 and the Menninger Factor I score at Time 2 is determined equally by all four Menninger scales: Motivation for Treatment, Sublimatory Capacity, Superego Integration, and Object Relations (the primary T1 → C2 path coefficients were. 344,. 229,. 263, and. 287, respectively). These data indicate that the more an individual reports malevolent, unilateral interactions on the Rorschach at Time 1, the greater his or her constructive involvement with the treatment staff, including the therapist; the better regulated his or her behavior; and the more appropriate his or her interpersonal relationships at Time 2.Google Scholar

Copyright information

© Springer Science+Business Media New York 1994

Authors and Affiliations

  • Sidney J. Blatt
    • 1
  • Richard Q. Ford
    • 2
  1. 1.Yale UniversityNew HavenUSA
  2. 2.Austen Riggs Center, Inc.StockbridgeUSA

Personalised recommendations