The MMPI–2 Restructured Clinical (RC) Scales and Redundancy: Response to Tellegen, Ben-Porath, and Sellbom

Abstract

In this article, we examine 5 criticisms of Tellegen, Ben-Porath, and Sellbom (2009/this issue) about our study demonstrating the redundant relationships of the Restructured Clinical (RC) scales with extant MMPI–2 scales. We discuss differences in univariate versus multivariate comparisons of the RC scales and our “proxy” scales using their data. We show that (a) both the RC and extant proxy scales identified in our analyses account for most of the variance in the Clinical scales; (b) the proxy scales are redundant with the RC scales; (c) the proxy scales matched the 6 RC scales in accounting for variance in the Clinical scales exactly in three cases, differed by ≤.02 in 2 cases, and reached a maximum of .11 in one case; (d) the item overlap between RC1 and HEA is not at issue but rather their correlation with Scale 1; and (e) the evidence for the construct validity of the RC scales is weak using findings on the incremental validity of RC4 as illustrative.

Notes

¹It should be noted that potential RC proxies are not in short supply. In his critique of the RC scales, Nichols (2006) called attention to the high levels of redundancy between these scales and 66 mostly content-based MMPI–2 measures including the Harris subscales, Content, PSY–5, and Supplementary scales. For each of the RC scales, no fewer than four redundant alternatives were provided. Within the Caldwell (1997) clinical data set (N > 50,000), the median correlation between a given RC scale and its alternative counterpart was .84 (M = .83). In their response, Tellegen et al. (2006) raised objections to the Caldwell sample on the grounds of its representativeness in terms of social desirability; its possible inclusion of child custody and/or employment screening data; and that, more generally, it is not representative of any specific clinical or other meaningfully defined population and is more appropriately described as a composite sample, an amalgam, than as a well-defined clinical sample (p. 152). However, Tellegen et al. (2006) did not challenge the values Rouse et al. (2008) provided nor, indeed, did they replicate any of Nichols' analyses on any of their own presumably more acceptable samples.