Modification of Measures: Threats to Construct Validity and Recommendations for Improving Items

Poorly written items will result in poor psychometric properties of measures. We review the literature on item generation, develop a typology of threats to construct validity, and present the results of a comprehensive review of published measures. Our results demonstrate that many existing measures have significant problems that may represent threats to construct validity. Next, we provide an empirical demonstration of how a commonly-employed measure of affective commitment is improved by rewriting items to be relatively free from the threats to construct validity identified in the presented typology. Recommendations for improved item generation practice are offered.