Person:

Nevins, Andrew

This paper provides supporting evidence for a number of hypotheses made in recent models of derivational syntax. The phenomenon under study is shm-reduplication in English, a particular instance of the more general, cross-linguistic pattern of echo reduplication. It is argued that the two elements in a reduplicated structure form a chain of two left-peripheral positions that, due to a distinctness requirement within a Spell-Out unit for the Transfer to PF, cannot be mapped onto linear order. A number of seemingly unrelated facts are derived rather naturally: (i) English shmreduplication cannot appear in an argument position; (ii) the two copies involved in shmreduplication are strictly adjacent; (iii) the phonological phrasing of the two copies is not the intonation of a compound; (iv) the discourse context felicitating shm-reduplication is not out-ofthe-blue; (v) no echo reduplication process yields the reverse order (e.g., with the echo reduplicant preceding the base); and (vi) echo reduplication is never the exponent of a Case- or wh-feature.

Merchant (2001) proposes that preposition stranding under sluicing is allowed only in those languages that also allow P-stranding in regular wh- questions. Spanish and Brazilian Portuguese (BP) seem to falsify this generalization, as both are non-Pstranding languages that allow P-stranding under sluicing. Our claim is that, despite initial appearances, Spanish and BP do not constitute counterexamples to Merchant’s generalization. We propose that there are two sources of sluicing in Romance: wh-movement plus IP-deletion (Merchant 2001), and clefting plus IP deletion (Merchant 1998), the latter being the underlying source for P-stranding sluicing. The apparent P-stranding effect follows from the fact that, as opposed to regular interrogatives, clefts in BP and Spanish do not involve P-stranding at all. We reinforce this conclusion by showing that, in those cases where a cleft base is independently banned, P-stranding under sluicing becomes impossible too.

In this paper we review evidence from a variety of sources, including deneutralization studies, that indicate that the choice of underlying representations is governed by causal reasoning, statistical inference, orthographic knowledge, and hypercorrection, but rarely, if ever, by a principle of minimizing faithfulness violations.

The EPP cannot be reduced to the Inverse Case Filter (contra Boskovic 2002) and remains for the moment axiomatic in phrase-marker construction. Feature-splitting, Agree, and the EPP allow for a range of derivational options that have been constrained by the Activity Condition (Chomsky 2001 et preq). The phenomenon of non-nominative subjects (including A-movement of structurally-Case marked elements) runs afoul of the activity condition. This paper explores the hypothesis that all derivational options ruled out by the activity condition can be ruled out by independent principles of locality and a constraint against multiple Case valuation.