S. AUROUX, A. BOES, avec la collaboration de Ch. PORSET: Court de Géhelin (1 725-1 784) et le comparatisme. Deux textes inédits. This paper deals with manuscripts from the French Protestant Antoine Court de Gébelin (1725-1 784), one of the first scholars to have developed a historical and comparative approach to world languages (Monde primitif, 1773-1784), before the great compilations of Hervas (1 784, 1800-1805) AdelungjVater (1806-1807) and Balbi (1826). With these two texts we have material that enables us to construct elements of a seriaI history of the comparative method. The first text (which deals with the Tahitian language and assumes its Malaysian origin) is particularly important for this purpose: it exists in three different versions and we can replace · it in the wider context of a whole series on this topic (Reeland, De Brosses, Pereire, Humboldt, Buschmann). In the introduction we give a probable date for the texts and an essay on Court's methodology. Our epistemological approach is influenced by Lakatos' scientific research programmes. After a description of Court'g programme, the characteristic feature of which is the assertion of a universal, organic and · primitive language, we attempt to answer the question of its failure. We assert that the rejection by later comparatism of such a hypothesis cannot be based on empirical falsification. Court's theory is an empirical one, and the scholar refers to facts which largely confirm it. Each of the three versions deals with new facts which support the claim to the primitive language and also new assertions on the relationship between particular languages. But one cannot at the same time collect facts for the universal connection between languages and empirical evidence for localized language grouping. This epistemological contradiction is the main reason which explains Court's failure and perhaps the new directions taken at the outset of the nineteenth century.