This article is concerned with the epistemological and methodological problems related to the taxonomy of political parties whett based on noorganizational criteria. The study of parties represents a starting point for modern Political Science: i.e. the seminal researches of Bryce, Ostrogorski and Michels. However this important field of knowledge hasn't known that much progress since the classical Duverger's Political Parties. Why?Two kind of approaches are used in order to classify parties: individualistic versus holistic. «Individualistic classifications» often suffer from a lack of theoretical background. Same of them use a spurious criterium like party names: i.e. Radical means extreme-left in the USA and conservatism in Switzerland, secularist activism in Italy and evangelical left in the Netherlands etc. Some classifications are based on political platforms which is a meaningless criterium: «A general election campaign is about a choice between organizations, not ideas.» (Richard Rose) The third criterium is far better: policies really implemented by political parties. However when properly used it gives a typology of countries, not parties.«Holistic classifications» either functionnalist or marxist opposed each other: the first stressed on equilibrium, the latter on class warfare. One suggests Rokkan's four cleavages paradigm to classify parties: each side of a cleavage should correspond to a definite «Political Family». Which items to use in order to assess a party to a permanent cleavage line? (1) The historical function performed by the party at its creation. (2a) The sociological structure of the party's: electorate, membership and inner group; (2b) The linkage structure between the party and a given network of pressure groups, movements and associations.