ABSTRACT
The problem considered in this paper is the design of selective and diagnostic testing procedures to optimize final performance in complex, procedural tasks. We are specifically concerned with high-performance tasks that require extended training to develop proficiency. Examples of applicable task domains are air traffic control, electronic troubleshooting, typing, and piloting; inapplicable task domains are practicing law, writing technical materials, and teaching history. For applicable task domains, we argue for a closer relationship between training, prediction, and diagnostic assessment. The principles discussed have limited utility for tasks that rely primarily on a declarative-knowledge base rather than performance components, and have greater utility in training environments where there is a strong investment in developing procedure-based skill.
