Centering theory (Grosz et al., 1983) establishes a structural relationship between the form of referential expressions and attentional focus. It predicts specifically that highly focused entities tend to be realized with less marked forms (e.g. pronouns) and that factors such as grammatical status or surface position affect prominence. The three psycholinguistic experiments reported here aimed to test a claim put forward in linguistic studies on French evidential adverbials (Schrepfer-André, 2005) concerning the functioning of the preposed position as a potential frame influencing the following pronominal resolutions: a preposed Prepositional Phrase (PP), but not an inserted PP, is considered to frame an informational bloc in which antecedents are easily accessed compared to antecedents located outside the bloc. The results of the first experiment are similar to those reported by Gordon et al. (1993) with the Repeated Name Penalty Paradigm which they interpreted as suggesting that "initial and non subject" or "non-initial and subject" entities are equally prominent. The second experiment, however, shows that the preceding results are too robust to be attributed only to surface order. The third experiment. suggests that the semantic affinity between the evidential PPs and the VP of subsequent utterances could enhance the preposed PP's effect: opinion VPs seem to be primed by preposed evidential PPs, inducing a possible narrative transformation from an objective story about the subject referent to a story about the speaker's inner thoughts. Altogether, these results tend to show that preposed PPs fulfil a specific framing function (Charolles, 1987) in certain conditions.