ABSTRACT
Headspace solid-phase microextraction (SPME) and GC-MS were used to analyze 17 commercial French Cognac brandies (9 young and 8 well-aged, ranging in age from 3 to 55 years). Sixty-four volatiles were chosen on the basis of chromatographic separation and/or known odor importance. Chromatographic peaks were manually integrated and the peak area data analyzed using partial least-squares (PLS) regression to study relationships between volatile composition (X variables) and age (Y variable). When only those compounds with the highest significance were included and from these selected the variables (a total of 33) with the highest correlation loadings on the first two principal components, principal component 1 explained 82% of the variance of the measured compounds and 85% of the variance in age. These were considered the most important volatiles to distinguish products of different ages because young and old samples were separated along principal component 1. Norisoprenoids, terpenes, and acetate esters had weaker positive and negative loadings and were therefore left out. The PLS model could predict sample age accurately with the optimum 33 volatiles as well as with a smaller subset consisting of ethyl esters and methyl ketones.
