MULTIVARIATE STATISTICAL-ANALYSIS OF GAS CHROMATOGRAMS TO DIFFERENTIATE COCOA MASSES BY GEOGRAPHICAL ORIGIN AND ROASTING CONDITIONS

Multivariate statistical methods were applied to the differentiation o f cocoa from 13 geographical origins, taken at four roasting steps and supplied by five different manufacturers. An analysis of variance app lied to 37 peak areas showed that in only nine instances was the varia bility of the response influenced by the origin, the degree of roastin g, and the supplier. These peaks, identified by gas chromatography-mas s spectrometry, were used as variables to perform principal components analysis, hierarchical clustering, and discriminant analysis. It was established that unroasted masses fall into five groups, which essenti ally differ in the amounts of two components, hexane and 2-methoxy-4-m ethylphenol. These two chemical components may therefore be good origi n markers. On the other hand, the analysis of cocoa at the end of the roasting process demonstrates the importance of the thermal treatment conditions. Indeed, the use of hot-air roasters favours the synthesis of aldehydes. Discriminant analysis also shows the influence of time a nd temperature on the production of 2,3-diethyl-5-methylpyrazine, whic h has a typical odour of roasted peanut. This study shows that gas chr omatography is an ideal technique for the objective discrimination of cocoa origin and roasting conditions.