Rethinking embedded honorification: Syntactic insights from Korean si

This paper revisits the embedding asymmetry between argument honorification and addressee honorification, with a focus on Korean si and supni. Korean argument honorific marker si is embeddable while the addressee honorific marker supni is not. First, this paper presents novel Korean data demonstrating that embedded si, typically classified as a content-oriented argument honorific marker, exhibits sensitivity to utterance context, including the presence of a direct addressee and the relative ranking of the speaker and addressee. I dub this innovative use of si as non-canonical use of si. This paper discusses how this non-canonical use of si reacts to the association between utterance sensitivity and embeddability proposed in Portner et al. (2019). At first glance, the novel Korean data on si appears to challenge Portner et al. (2019)’s generalization that utterance-oriented markers are not embeddable. However, I show how this non-canonical si can be reduced to its canonical counterpart, preserving the broader theoretical distinction between content-oriented and utterance-oriented markers. Syntactically, I argue that the non-canonical si retains the downward agreement dependency characteristic of its canonical counterpart. I propose that the observed utterance sensitivity arises from long-distance binding between the matrix addressee and a goal beyond the subject position. While si typically agrees with the subject in its canonical use, I argue that in its innovative use, si can target a broader set of discourse-salient event participants, such as affectees/beneficiaries. This approach captures the indirect addressee honorification effect while maintaining si’s core status as an argument honorific marker.