This chapter critically examines the complex relationship between the description of discourse markers and corpus usage. The most widely accepted theoretical accounts of discourse markers today have made use of “intuition” (in a metalinguistic sense) to describe their meaning and functions, be it in Andrés Bello’s grammatical reflections, in argumentation theory or in text linguistics. From this starting point, studies on specific discourse markers have resorted to samples coming from corpora to achieve a more precise description or to elicitate features 255which might remain hardly accesible from a purely intuitive approach. The more specific the work, the more frequent this move from deduction to induction, as happens, for instance, in Sociolinguistics or in Applied Linguistics, but also in confirmatory studies on discourse markers. Metatheoretic reflexions stress the role attributed to “intuition” in linguistic description but fail to capture the need of data to falsify highly specific hypotheses on discourse markers, turning thus the theory-to-data relationship into a multi-layered process. The conclusions of this chapter are aimed to produce a better foundation for the study of Spanish discourse markers.