This paper provides a corpus-based analysis of so-called “hedged performatives,” which, although frequently referred to in the literature, have never been the subject of an in-depth functional study. Using data from the Corpus of Contemporary American English, the focus is on I must/ have to say, I must/ have to admit, and I must/ have to confess, as the hedged performatives which are among the most frequent and score highest in terms of collocational strength. The qualitative analysis identifies two main functions, viz., downtoner and emphasizer. They are shown to derive from the interplay of three co(n)textual parameters: (i) “semantic valency” of the host clause (i.e., positive, negative, or neutral semantic content), (ii) “thematic orientation” of the host clause (i.e., toward the speaker, the addressee, or a third person/the situation), and (iii) conversational “alignment” of the speaker with the interlocutor (i.e., agreement or disagreement). It is further shown that hedged performatives play an important role in rapport management, serving (mainly positive face) politeness strategies, which are captured in terms of face-preservation, face-damage, and face-boost.