The importance of honey has been recently increased because of its nutrient and therapeutic effects, but the adulteration of honey in terms of botanical origin has increased, too. The floral origin of honeys is usually determined using melisso-palynological analysis and organoleptic characteristics, but the application of these techniques requires some expertise. A number of papers have confirmed the possibility of characterizing honey samples by selected chemical parameters. In this study high-resolution nuclear magnetic resonance (HR-NMR) and multivariate statistical analysis methods were used to identify and classify honeys of five different floral sources. The 71 honey samples (robinia, chestnut, citrus, eucalyptus, polyfloral) were analyzed by HR-NMR using both 1H NMR and heteronuclear multiple bond correlation spectroscopy (HMBC). Spectral data were analyzed by application of unsupervised and supervised pattern recognition and multivariate statistical techniques such as principal component analysis (PCA) and general discriminant analysis (GDA). The use of 1H-(13)C HMBC coupled with appropriate statistical analysis seems to be an efficient technique for the classification of honeys.