Where it's [at]: A phonological effect of phasal boundaries in the construct state of Modern Hebrew

The consonant of the feminine marker /at/ of Modern Hebrew is absent from certain configurations, but present in others (N-at+N constructs). In this paper, I propose to regard this phenomenon a case of allomorphy, conditioned both phonologically and morpho-syntactically. The consonant is analyzed as floating. In consequence, additional skeletal support is needed to explain its realization. One possible, independently motivated source for such support is Lowenstamm's (1996) “initial CV”: the floating /t/ attaches to the initial CV of the following word. Still, the question is raised why this happens mainly in that very specific configuration. N-at+N constructions are therefore compared to the minimally different Nat+Adj, and four differences are singled-out. After a prosodic solution is judged insufficient, syntactic structures are proposed for both constructions and the four differences are related to phase-structure, under the assumption that D is the first phasal head of the nominal architecture. Adopting Scheer's (2009) claim that Lowenstamm's initial CV marks phase boundaries, rather than word-boundaries, it is then shown that the /t/ remains afloat exactly when the phase structure motivated by the four differences renders the initial CV of the following phase inaccessible; but if phase structure allows it, the same /t/ can be linked to that following CV. The logic behind the allomorphy is thus both phonological (it relies on autosegmental analysis) and morphosyntactic - it follows from phase structure. On a general level, if so, evidence is adduced to the correctness of Scheer's proposal. On a more language-specific view, an elusive case of allomorphy, hitherto regarded as historical, receives synchronic motivation in the consideration of general principles of form and structure interrelations.