This paper presents a logical approach to nonmonotonic reasoning based on the concept of a nonmonotonic consequence relation. It introduces the notion of a conditional knowledge base, which consists of a set of conditional assertions of the form "if ... then ..." representing defeasible knowledge about how the world generally behaves. The paper aims to define a plausible set of all conditional assertions entailed by a conditional knowledge base. It defines and studies rational consequence relations, a restricted class of consequence relations that are more suitable for reasonable nonmonotonic inference procedures. The paper argues that any reasonable nonmonotonic inference procedure should define a rational relation. It shows that rational relations can be represented by ranked preferential models or non-standard probabilistic models. The rational closure of a conditional knowledge base is defined and proven to be cumulative and computationally tractable. The paper also discusses the global properties of the rational closure operation and provides an efficient algorithm for computing it. The main contribution of the paper is the rational closure construction, which provides a uniform and elegant way to answer the question of what assertions should be considered as entailed by a given knowledge base.This paper presents a logical approach to nonmonotonic reasoning based on the concept of a nonmonotonic consequence relation. It introduces the notion of a conditional knowledge base, which consists of a set of conditional assertions of the form "if ... then ..." representing defeasible knowledge about how the world generally behaves. The paper aims to define a plausible set of all conditional assertions entailed by a conditional knowledge base. It defines and studies rational consequence relations, a restricted class of consequence relations that are more suitable for reasonable nonmonotonic inference procedures. The paper argues that any reasonable nonmonotonic inference procedure should define a rational relation. It shows that rational relations can be represented by ranked preferential models or non-standard probabilistic models. The rational closure of a conditional knowledge base is defined and proven to be cumulative and computationally tractable. The paper also discusses the global properties of the rational closure operation and provides an efficient algorithm for computing it. The main contribution of the paper is the rational closure construction, which provides a uniform and elegant way to answer the question of what assertions should be considered as entailed by a given knowledge base.