A working definition is proposed for "scientific text" in terms of text function, based on pragmatic criteria derived from various approaches to the sociology of science. The term "Language for Academic-Scientific Purposes" (LASP) is preferred because it allows for the inclusion of texts in the human sciences, at the same time implicitly excluding many "technical" texts that would be covered by the frequently applied concept of English for Science and Technology (EST). The definition and general assumptions of this paper are based on the concept of the scientific community as a self-contained and self-perpetuating group that needs to recruit new members from outside but that restricts access to its discourse to members of the group. A tentative discourse-oriented model is proposed for the analysis of LASP texts, which is then applied diachronically to two early scientific texts, one from 1665 and one from 1840. Contains 21 references. (LB)