The study examined whether concurrent and retrospective verbal protoco
ls possess the potential to provide a non-reactive and valid account o
f the cognitive processes involved in fault diagnosis. With this goal
in mind, a group of subjects performed a fault diagnosis task under co
ncurrent and retrospective verbalization and in a silent control condi
tion as well. In the task, concurrent verbalization led to a considera
ble increase in time to completion, but exerted no effect on overall a
ccuracy and the adopted strategy. Retrospective verbalization did not
have any effect on performance. The strategy-related data obtained und
er concurrent verbalization proved to be more valid than those obtaine
d under retrospective verbalization. On the basis of the results it is
suggested that, where possible, concurrent verbal reports should be c
ollected when trying to get a better understanding of the nature of th
e fault diagnostic process.