Aims: Differences in wine flavour proceed primarily from grape quality. Environmental factors (climate, soil), cultivars and training systems modify many grape and wine quality traits. Metabolic profiling based on proton nuclear magnetic resonance (H-1-NMR) spectra has been proved to be useful to study multifactorial effects of the vine environment on intricate grape quality traits. The capacity of this method to discriminate the environmental effects on wine has to be demonstrated. Methods and results: : H-1-NMR spectra were mule from wines produced with grapes harvested at maturity of three cultivars (Cabernet-Sauvignon, Cabernet franc and Merlot) and three soil types (gravely, sandy and clayey) during two vintages (2002 and 2003). Data were analysed by multivariate statistical methods. Principal component analysis applied on the NMR spectra data were not always able to separate satisfactorily wines from the 3 soil types. Conversely, partial least square analysis separated clearly the 3 soil types independently of the vintage and cultivar. Conclusion: By comparing the NMR signals that contribute to the two first axes of the PCA and PLS analyses, a significant soil effect on NMR signals in wines is reported. However, the effect of the vintage on wine composition was greater then the effect of the soil type. Significance and impact of study: After validation on a larger number of wine samples this chemical profiling will be a useful new method to the qualify wines in relation to climate, soil, and cultivar effects which contribute to the terroir.