This paper estimates consumers' willingness to pay (WTP) for food-safety attributes in Korea. A choice experiment is designed to elicit WTP differences among apple attributes (with or without chemical fertilizer, genetically modified or non-genetically modified, level of pesticide, grade, price) according to food safety level. The results show that the estimated differential WTPs for apple attributes are meaningful. The results of this paper could be important information for farmers who want to know how to grow safe agri-product. Further, those results will contribute to government's food-safety policies. Like contingent valuation method, consumers also overestimate WTPs for apple attributes and this makes a hypothetical bias in choice experiment. Therefore, we must perform a test about hypothetical and real situation that consumers have to pay with real money.