Does the Rorschach Performance Assessment System (R-PAS) Differ from the Comprehensive System (CS) on Variables Relevant to Interpretation?

Abstract

We examined the impact of the changes in administration and coding introduced by the Rorschach Performance Assessment System (R-PAS) relative to the Comprehensive System (CS) on the Rorschach response process, as manifested in variables relevant to interpretation. We also examined the efficiency of each system to obtain protocols in an optimal range of responses (R) for interpretation. As hypothesized, when comparing 50 CS and 50 R-PAS nonpatient protocols, R-PAS produced many more protocols in the optimal R range (18-27) than the CS (78% vs. 24%) and it eliminated the need for re-administration, which was required for five CS protocols. As expected, R was less variable with R-PAS, as were two variables derived from it, R8910% and Complexity. In addition, as expected because of different Form Quality tables, R-PAS showed notably fewer and less variable perceptual distortions than the CS, and an increase in more conventional perceptions. The other 58 variables showed no reliable differences in means or standard deviations, though modest power precluded definitive inferences about equivalence. Overall, our results support previous findings about the benefit of R-PAS to obtain protocols in an optimal range for interpretation, while keeping the core manifestations of the response process unchanged.

Notes

1 The R8910% variable is equivalent to the Affective Ratio from the CS.

2 The Complexity and Ego Impairment Index variables were developed for the CS, but never formally adopted by Exner (2003).

3 These expectations are for FQ raw scores. R-PAS provides norms for protocols collected using CS administration and CS FQ coding. We would not expect differences in the normed standard scores for any of the FQ variables; e.g., FQ-% should have a M of 100 for CS protocols using CS norms and a M of 100 for R-PAS protocols using R-PAS norms. Although other variables make use of FQ scores (e.g., PHR Proportion, SC-Comp), they have not shown differences in previous research (see Table 13.4 in Meyer et al., 2011). To maintain consistency with all the previous studies comparing R-PAS to the CS, we use raw scores rather than standard scores in our primary analyses.

4 According to the syntax file available at www.r-pas.org: Alternative Protocol Level Reliability Using Difference Scores Instead of Proportions - For Specification.

5 For instance, for a pooled SD of 10, a difference of two tenths of a SD would require one SD to be 11 and the other to be 9, leading to proportions of 9 / 11 = 0.818 and 11 / 9 = 1.222.

6 For instance, with a pooled SD of 10, a difference of 0.50 SDs corresponds to one SD of 7.5 and one of 12.5. These SDs form proportions of 7.5 / 12.5 = 0.600 and 12.5 / 7.5 = 1.667.

7 More recently, and quite independently, Norman, Sloan, and Wyrwich (2003) demonstrated not only how this effect size is the average minimally important perceivable difference across a range of quality of life instruments, but it also is tied to the limits of human perception. Specifically, it is the same limit Miller discovered in his famous work on sensory perception, which is that human perceivers are limited to roughly seven categories or levels of discrimination (e.g., degrees of saltiness, loudness of sounds). Statistically, one unit on a seven-point uniform distribution is about half a standard deviation (0.46 exactly).

Additional information

Funding

This study was supported by Grant #2014/04940-0, São Paulo Research Foundation (FAPESP).