Repetitive inspection with checklists: Design and performance

Tasks in a number of industries require long, complex procedural inspections. Inspectors are typically given checklists, but after gaining familiarity with the checklist they will often perform the task from memory, e.g. in civil aviation (Pearl and Drury, 1995). Our questions were whether such behaviour would occur under controlled conditions, and whether the design of the checklist contributed to the behaviour. This study used a 108-step procedure for an overnight check on a common airliner, taught to 24 students and then repeated on eight different days. After training to per-form the task in the order given on their checklist, with signoffs where specified, participants returned eight times to perform the task on simulated aircraft systems. Checklists were either arranged by function or by spatial location. There were either individual signoffs for each of the 108 items, or 37 signoffs for logical subsets of items. There was no difference in probability of defect detection between conditions, but the performance times and rates of both sequence and signoff errors changed significantly. All participants tended to follow a spatial sequence whatever their checklist, so that users of the functional checklist made more sequence errors (and were also slower) than those using the spatial checklist. Signoff errors did not differ between groups, although participants were quicker with fewer signoffs, and preferred that condition in post experimental ratings.