This article considers the scientific process whereby new and better clinical tests of executive function might be developed, and what form they might take. We argue that many of the traditional tests of executive function most commonly in use (e.g., the Wisconsin Card Sorting Test; Stroop) are adaptations of procedures that emerged almost coincidentally from conceptual and experimental frameworks far removed from those currently in favour, and that the prolongation of their use has been encouraged by a sustained period of concentration on “construct-driven” experimentation in neuropsychology. This resulted from the special theoretical demands made by the field of executive function, but was not a necessary consequence, and may not even have been a useful one. Whilst useful, these tests may not therefore be optimal for their purpose. We consider as an alternative approach a function-led development programme which in principle could yield tasks better suited to the concerns of the clinician because of the transparency afforded by increased “representativeness” and “generalisability.” We further argue that the requirement of such a programme to represent the interaction between the individual and situational context might also provide useful constraints for purely experimental investigations. We provide an example of such a programme with reference to the Multiple Errands and Six Element tests. (JINS, 2006, 12, 194–209.)