On the interaction of portmanteaux and ellipsis

Abstract

This thesis is an investigation of what happens when an ellipsis boundary tries to split a portmanteau. It focuses on two patterns: elliptical indivisibility, and elliptical divisibility. Elliptical indivisibility is exemplified by the Hungarian portmanteau negative 3rd person indicative copula, which gets pronounced in its entirety even when the ellipsis site is thought to contain the copula. Elliptical divisibility is exemplified by the Bengali portmanteau negative perfect, which splits into a default sentential negation and a silent perfect when the complement of negation is elided. Investigations of both ellipsis sites shows evidence for complex unpronounced structure, suggesting that the variation in elliptical (in)divisibility arises not from having different kinds of ellipsis operations, but from different portmanteaux forming operations. I propose that indivisible portmanteaux are the result of either fusion or non-terminal insertion in a Late Insertion model of the postsyntax, while divisible portmanteaux are the result of two cases of contextual allomorphy.

From the study of Hungarian elliptical indivisibility, we learn that ellipsis sites must be post-syntactically accessible, at least to some morphological operations, since portmanteau formation in Hungarian is shown to be post-syntactic. From the study of Bengali elliptical divisibility, we learn that locality and directionality restrictions on contextual allomorphy must be loose enough to allow allomorphy to be triggered by non-local inward sensitivity to morphosyntactic features. From the synthesis of the two case studies, we learn that to successfully model both elliptical divisibility and indivisibility, a single ellipsis silencing mechanism is sufficient, as long as different portmanteau forming operations are used.