Addressing Issues in the Development and Use of the Composite International Reference Values as Rorschach Norms for Adults

Abstract

This article describes 3 studies evaluating normative reference data for the Rorschach Comprehensive System (CS; Exner, 2003, 2007), with a particular focus on the viability of the Composite International Reference Values (CIRVs) that were compiled from 21 adult studies by Meyer, Erdberg, and Shaffer (2007). Study 1 documented how the CIRV norms are virtually identical when organized into 3 groups differentiated by the quality of their data collection effort, including an optimal group of 4 samples that relied on multiple experienced examiners and provided ongoing quality control over administration and coding. Analyses also showed that relative to the group of more optimal samples, the group of less optimal samples did not produce more variability in summary scores within or across samples or lower interrater reliability for coding. Study 2 used the existing CS reference norms to generate T scores for the CIRV means and documented how the CS norms make other samples of healthy nonpatients look psychologically impaired in multiple domains. Study 3 documented with examples from 4 different countries how 2 sets of within-country local norms produced notably different results on some variables, which compromises the ability of local norms to be used instead of the CIRVs. Taken together, the 3 studies provide support for the use of CIRVs in clinical practice as norms that are generalizable across samples, settings, languages, and cultures and that account for the natural variability that is present when clinicians and researchers contend with the ambiguity contained in the standard CS reference materials concerning the proper ways to administer and code. We conclude by urging CS users to rely on the CIRVs when making clinical inferences and to adopt alternative methods of ensuring they are following cohesively standardized administration and coding guidelines.

Notes

¹ Authors used different programs to generate results output, which meant that in some instances a study did not report data for a variable that had been reported in most other samples. The French data were not published in the 2007 supplement and that article did not report data for 30 of the 142 variables reported by other studies (i.e., the 14 Determinant codes differentiated by form dominance, PureF%, (H)+Hd+(Hd), An+Xy, total scores for the six Constellations, WD+, WDo, WDu, WD–, WDnone, EII–2, and HRV). For the last seven variables just listed, three or four other samples also did not report those results. However, using other scores, we could reasonably estimate sample-specific mean values for PureF%, (H)+Hd+(Hd), An+Xy, and HRV.

² As an example, the pooled within-sample standard deviation for R was 7.90. The standard deviations of the mean R values in the less optimal, midrange, and more optimal groups were 2.00, 1.88, and 3.37, respectively, which produces corresponding f values of .25, .24, and .43. Detailed results for these analyses and other results summarized in the text are available on request.

³ Depending on the variable under consideration, impairment can be indicated by low scores or high scores. For X+%, which indicates conventionality of perception, a low score is indicative of impairment, much like a low IQ score indicates cognitive impairment. Translating the observed results to the IQ score metric and interpreting the resulting value like an IQ score, the finding that the typical nonpatient is 2.7 SD below the official CS norms for X+% indicates that the typical nonpatient falls far into the range of what would be diagnosable intellectual disability or mental retardation, with an IQ equivalent score of 59.5.