This contribution investigates the French prepositional system from two different angles. On the one hand, it considers the stereotypical environment of certain prepositions, on the other hand, it explores the most typical prepositional collocates of some nouns. Both types of studies are conducted on the basis of statistical, probabilistic analyses of a literary corpus (novels from the second half of the 20th century). Our method makes it possible to compare the combinatory profiles of a great number of prepositions. This, in turn, allows to establish sub-groups (clusters) of prepositions according to distributional criteria. We then study semantic interactions between some nouns of emotion and the prepositions that they can be governed by. The result of this interaction determines the behaviour of prepositional phrases from a syntactic point of view.