On the Stability of Variability: Retest Reliability of Metatraits

Abstract

The variance in a subject's responses to the different -items on a trait scale has been interpreted as an indication of behavioral consistency, truthfulness, and/or traitedness. The present study examined the retest stability of intritem variances on three trait scales. Interitem variance scores showed fairly high retest reliability (in the. 7 range). Shorter, more homogeneous subscales had lower retest reliability scores. Interitem variances did not moderate the stability of trait scores. These results suggest that (a) people are consistent in their degree of variability on a given dimension, (b) highly variable responding to different items on a scale may be a consistent, meaningful pattern rather than an indication of incompetent or random responding, and (c) the simple deletion of high-variance (or "untraited") subjects from a sample will not automatically improve results. Additional findings suggested a positive but weak tendency for people who responded consistently (vs. variably) to the items on one scale to respond the same way to the items on other scales.