This article combines theoretical and methodological considerations. It addresses the question how formulaic language manifests itself from a pragmatic point of view and how it can be analyzed with corpuslinguistic methods. First, it is shown that formulaic language – on the performance level – appears and is visible in the form of language patterns. Furthermore, an overview of the conceptual diversity prevailing in this context is provided. Subsequently, the question is how to make formulaic language empirically tangible, and the strengths and weaknesses of various corpuslinguistic approaches are discussed. In the third section, some methodological consequences, which result from a (largely) inductive corpuslinguistic analysis, are highlighted. Firstly, it is argued that even an inductive approach in practice necessarily includes deductive operations involving, on one hand, the corpus structure, and on the other, the analysis itself. Secondly, it is described how the study design affects the kind of formulaic language that can be identified. Thirdly, the possibilities and limitations of establishing formulaic language at the level of word forms and at the morphosyntactic level are discussed.