The paper focuses on the relation between the analytical philosophy of science and modular and holistic approaches to cognitive linguistics, respectively. The authors show that Chomsky's as well as Bierwisch & Lang's and Lakoff & Johnson's approaches make substantial use of non-demonstrative inferences which the standard view of the analytical philosophy of science evaluates as fallacies. By outlining a metatheoretical framework focusing on plausible inferences, the authors argue that the inferences the theories mentioned make use of are plausible rather than fallacious. This finding illuminates basic aspects of theory formation in linguistics and motivates the revaluation of the methodological foundations of linguistic theories.