Compounding words in the syntax can produce phrasal phonology: Evidence from Japanese Aoyagi morphemes

Abstract Various proposals have been made to account for mismatches between syntax and prosody in natural languages. Prosodic prespecification (i.e., prosodic subcategorization) attributes such mismatches to morpheme-specific prosodic requirements (Bennett et al. 2018; Tyler 2019). On the other hand, Hsu (2019) and Revithiadou and Markopoulos (2021) argue that some patterns previously analyzed through subcategorization can instead be captured in Gradient Harmonic Grammar (Smolensky and Goldrick 2016) without a syntax-prosody mismatch. This paper contributes to the ongoing discussion about the syntax-prosody mismatch by addressing ‘Aoyagi prefixes’ in Japanese (e.g., gen ‘current’ in gen - daijin ‘current Minister’). While the ‘word-internal phrase boundary’ associated with Aoyagi morphemes has been attributed to prosodic subcategorization (Poser 1990), I argue that such subcategorization is unnecessary. The key evidence lies in the fact that all Aoyagi morphemes are accented. Vance (2008) and Ito and Mester (2013) independently observe that a prosodic phrase boundary emerges between the first and second elements only when the first element is an accented prosodic word in Japanese. Building on this correlation between accent and prosodic phrasing, I put forward an alternative analysis: first, I propose that Aoyagi morphemes are not prefixes, but the size of syntactic word (X 0 ), such that the entire Aoyagi construction should be analyzed as a syntactic compound [x 0 X 0 X 0 ] (Booij 2010). Given this structure, their prosodic behavior follows from an XP to φ mapping system (Ito and Mester 2013), where constraints on accent placement play a crucial role in mapping syntactic heads to phonological phrases, overriding Match constraints.