Whole pink salmon were iced and stored under refrigerated conditions to allow decomposition to occur. After sensory grading, static headspace volatiles were analyzed and 22 of them were retained for multivariate analysis. Principal component analysis indicated that most of the variation (86.5%) was carried in the first four dimensions. Six compounds received high loadings for the second principal component that was associated with quality grade variation. when used in the computation of nonparametric Epanachnikov kernel functions, ethanol and 3-methyl-1 -butan01 played an effective role at segregating grades A, B, and reject with a high rate of correct classification (95.6%).