The paper examines one of the possible approaches to exploring the conceptual space represented by language signs and texts. The notion of the cognitheme as a unit of knowledge in the form of a proposition, functional for modelling the conceptual space, is defined and some principles of the cognitheme analysis are discussed. The cognitheme is considered as a unit of modelling mental entities reflected in the language, for example, such as the concept or the conceptual space connected with a text, and at the same time as a unit of conceptualization significant in its own right, revealing elements of knowledge important for a language community and thus fixed in language signs and texts. A feasible classification of cognithemes is described, examples illustrating this classification are given.