Sensory characterisation of virgin olive oil and relationship with headspace composition

Thirty samples of virgin olive oil were analysed by sensory and instrumental methods. Free-choice profiling and quantitative descriptive analysis were used for odour analysis, and GC-MS for the analysis of headspace volatiles. Principal component analysis and partial least squares regression analysis were used for relating sensory and instrumental data. Similar clusters of samples were obtained from principal components analyses of both data sets. Partial least squares regression made good predictions from headspace data of some of the descriptors used in quantitative descriptive analysis. The prediction of cut grass odour was obtained in particular from five unidentified compounds. The five compounds had similar mass spectra, but GC-sniffport analysis showed that only one of them (possibly isomers or homologues) gave an aroma that could be related to the cut grass descriptor.