Virgin olive oil samples were analysed by instrumental and sensory analysis. Solid phase microextraction (SPME) was applied for sampling volatile compounds that were analysed by GC-MS. Quantitative descriptive analysis was used for odor analysis of 36 oil samples. Principal Component Analysis (PCA) and Projection to Latent Structures by means of Partial Least Squares analysis (PLS) were used for a preliminary investigation of data matrix and for relating sensory and instrumental data, respectively. One hundred and twenty-one volatile compounds were sampled by SPME and identified by GS-MS. PCA applied to the sensory data clearly discriminated the oils into two clusters: the first by cut-grass, floral, green apple and fruity; the second by hay-like and fatty sensory descriptors. The PLS applied to the sensory and instrumental data defined a good correlation between the cut-grass odor and (E)-2-hexenal, hexanal and (Z)-3-hexenal, and between floral odor and hexyl acetate, respectively. The hay-like sensory descriptor, on the contrary, did not show any correlation with the head space data. GC-sniff-port analysis, moreover, showed for selected samples, odors included in the two clusters, two odor described as cooked meat and putrid, found only in the hay-like oils. The floral/ester sensory note, correlated to 2-pentenal and 3-hexenyl-acetate, were described only in the oils classified as cut-grass/floral. So far, however, no compounds were described as hay-like in any of the oil samples analysed. Furthermore, oils characterized by hay-like sensory notes showed a low concentration of the compounds that correlated with the cut-grass and floral sensory notes, such as hexanal, (E)-2hexenal, (Z)-3-hexenal, 3-hexenyl acetate, 2-pentenal and hexyl acetate.