31 staff members within one VA medical center were briefly trained in using the Global Assessment of Functioning scale. At both pre- and posttraining participants provided Global Assessment of Functioning scores and identified their GAF-derivation strategies for two vignettes describing hypothetical patients. Analysis showed that initially there was inter- and intrarater inconsistency in strategy for deriving scores and that training increased consistency. Participants persisted in giving ratings that exceeded the criterion score. However, training significantly reduced Global Assessment of Functioning scores for one vignette, bringing scores closer to the criterion, and a significant number of people gave lower ratings after training. The study highlights common errors and points to the need for formal training in use of the scale.