In this paper, I will discuss ambiguity among languages not as a communicative property, but rather as it can represent an issue for linguists who approach languages scientifically. In particular, I will describe two possible kinds of ambiguity within the field of linguistic typology. The first kind of ambiguity consists in the difficulty of defining categories in crosslinguistic perspective (multiplicity vs. uniqueness). In the literature, language comparability has been widely discussed, and I will present the most recent and convincing proposed solution exemplifying it through the phenomenon of pluractionality. The second kind of ambiguity deals with the nature of data that typologists usually examine, that is, data that are almost completely out of context. I will show how this ambiguity can lead to inaccurate or even wrong generalizations, and, finally, I will propose a method (‘discourse-sensitive typology’) allowing typology to go beyond these kinds of ambiguity and, thus, to make generalizations as reliable as possible.