The purpose of this paper is to advocate the use of discovery procedures in the study of conventional discourses. In order to distinguish discourses that are the result of socially ratified knowledge, and those that are not, it is necessary to establish some criteria. The investigative work documented in this paper represents an important contribution to the understanding of meaning in language: discourse analysis in its various guises seldom spells out the methodological basis for the interpretations it puts forward, and therefore little substantiation is available to establish the cognitive reality of discourses at all: a sad state of affairs given the prominent role of discourse in shaping individual and collective experience. Our innovative approach is achieved through looking at stereotypes as an epitome of socially shared knowledge. We conduct three preliminary experiments inspired by corpus linguistics, social psychology and pragmatics in order to assess the degree to which bringing together different methods can illuminate the existence of socially ratified discourses. The focus is therefore less on the depth and breath of these experiments than on their joint potential to reveal what is a conventional discourse, and what is not.