Testing non-cognitive attributes in selection centres: how to avoid being reliably wrong.

In this issue, Gafni et al.1 provide a strong and timely case for evaluating the reliability and validity of multi-station selection centres (SCs, sometimes referred to as assessment centres). In developing this important research topic, we highlight: (i) important differences between SCs and objective structured clinical examinations (OSCEs); (ii) the importance of demonstrating the construct validity of selection methods, and (iii) psychometric issues surrounding the reliability of observer ratings and the use of generalisability theory (G theory).