The reliability and validity of normative and ipsative approaches in the measurement of personality

Abstract

Ipsative questionnaires of personality have been attacked as necessarily inferior to normative approaches. Some even go so far as to accuse the authors of ipsative scales of reporting spurious statistics and of ‘cheating at patience'. Despite previous papers which claim to demonstrate that ipsative scales overestimate reliabilities, cannot be factored soundly and yield uninterpretable validity coefficients, this investigation shows with synthetic data that these generalizations are ill‐advised. The results demonstrate with simulated data that ipsative scores can be factored soundly, that reliability data are not overestimated, and that under moderate conditions of central tendency bias in normative items, ipsative scores actually correlate better with hypothetical ‘true’ scores than the normative form. When replicated on real data from a sample of 243 subjects, a high correlation was found between ipsative and normative scale scores, ipsative scaling did not produce spuriously high reliabilities, and both normative and ipsative data showed sensible and significant correlations with external rating criteria.