This study uses the facts of Irish lenition, gemination processes and stress placement constraints to refute the theory of the syntax-phonology interface proposed by Truckenbrodt (1999) where it is claimed that the only structure visible to phonology at the interface is that of phrases. I use these same facts in support of Match Theory (Selkirk 2009; to appear) which allows a direct 1:1 mapping between syntactic and phonological structure at the word, phrase and clausal levels. Further, I go on to propose strength conditions on the boundaries of prosodic words dependant on whether those words are maximal, or non-maximal recursive word structures. I conclude that while *STRUC constraints eliminate redundant word bracketing structure, it does not target recursive word bracketing provided that that bracket contain at least some segmental information. This fact will account for Geminate Inalterability (Ní Chosáin 1991; Green 2008) found in Irish coronal clusters as well as secondary stress placement present only in recursive word structure. These facts can only be handled by a theory that allows a direct mapping of all types of syntactic structures to prosodic structure and not just syntactic phrases to phonological phrases.