A canonical discriminant function analysis (CDA) was applied to essential oil compounds profiles to determine the growing location of marjoram samples. 12/21 essential oil compounds were integrated into two discriminant functions as the best predictors for geographical differences. Stepwise integration allowed the determination of the essential oil compounds mainly responsible for the geographical variation, viz. myreene, terpinene-4-ol, gamma-terpinene and p-cymene. The two discriminant functions themselves showed a strong association between the locations and the predictors. The two functions allowed a correct classification of 95% of samples (cross-validated). The high positive classification rate demonstrates the CDA's ability as a quality assurance tool enabling authenticity control. Copyright (C) 2004 John Wiley Sons, Ltd.