Abstract: In this study, the check-all-that-apply (CATA) method, headspace solid phase microextraction (HS-SPME)-Arrow combined with gas chromatography-mass spectrometry (GC-MS) and gas chromatography-mass spectrometry-olfactometry (GC-MS-O) were used to analyze volatile flavor substances that affect the sensory differences among milk from four pastures. The results of the CATA questionnaire showed that milky aroma, creamy aroma, fragrant and sweet aroma, milky aroma, plastic odor, cooked odor and metallic odor were significantly different among the four milk samples (P < 0.05), and milk from pasture I were the most preferred by respondents, and the preference scores of milk from pastures III and IV were lower. Totally 14 aroma-active substances with an aroma intensity greater than or equal to 1 were obtained from the four milk samples through olfactometry. Using partial least squares regression analysis (PLSR), correlation analysis among sensory data, aroma-active substances and milk samples was conducted. The results showed that milky and creamy aroma from pasture I were prominent, which was strongly correlated with 2-heptanone, butanoic acid, decanoic acid, and hexanoic acid. Milk from pasture II showed a fragrant and sweet aroma, which may be related to the concentration of limonene. Milk from pasture III had prominent metallic, plastic, and milky odors, which may be related to 1-octene-3-alcohol. Milk from pasture IV had a cooked odor, which mainly came from hexanal and styrene.

Key words: pasture, check-all-that-apply, aroma-active compounds, gas chromatography-mass spectrometry-olfactometry