Objective: Embedded validity indicators (EVIs) are cost-effective psychometric tools to identify non-credible response sets during neuropsychological testing. As research on EVIs expands, assessors are faced with an emerging contradiction: the range of credible impairment disappears between the 'normal' and 'invalid' range of performance. We labeled this phenomenon as the invalid-before-impaired paradox. This study was designed to explore the origin of this psychometric anomaly, subject it to empirical investigation, and generate potential solutions.
Method: Archival data were analyzed from a mixed clinical sample of 312 (MAge = 45.2; MEducation = 13.6) patients medically referred for neuropsychological assessment. The distribution of scores on eight subtests of the third and fourth editions of Wechsler Adult Intelligence Scale (WAIS) were examined in relation to the standard normal curve and two performance validity tests (PVTs).
Results: Although WAIS subtests varied in their sensitivity to non-credible responding, they were all significant predictors of performance validity. While subtests previously identified as EVIs (Digit Span, Coding, and Symbol Search) were comparably effective at differentiating credible and non-credible response sets, their classification accuracy was driven by their base rate of low scores, requiring different cutoffs to achieve comparable specificity.
Conclusions: Invalid performance had a global effect on WAIS scores. Genuine impairment and non-credible performance can co-exist, are often intertwined, and may be psychometrically indistinguishable. A compromise between the alpha and beta bias on PVTs based on a balanced, objective evaluation of the evidence that requires concessions from both sides is needed to maintain/restore the credibility of performance validity assessment.
Keywords: Performance validity assessment; alpha and beta bias; embedded validity indicators; near-pass; soft fail.