New York University
Friday, February 12th
Locality Domains for Contextual Allosemy in Words
At least since work within Lexical Morphology and Phonology, the issue of the connection between word structure and allomorphy has been heavily investigated by morphophonologists. Recent advances within Distributed Morphology (see in particular Embick 2010) have shown that the general cyclic architecture of a phase-based Minimalist Program syntax provides the proper locality domains for the interaction of information determining contextual allomorphy, although phonology-specific notions like adjacency also play a role, restricting possible interactions even more than what might be allowed within a cyclic domain. Less well understood are the parallel issues at the syntax/semantics interface, namely the computation of possible meanings of morphemes in context. Against some recent work disputing claims in Marantz (1997, 2000) linking the domain of special meanings to phases and against recent proposals that the locality domains for phonology and semantics might differ, this paper clarifies the issues in contextual meaning determination and supports the idea that the locality domains for contextual allosemy are just those for contextual allomorphy. As a specific notion of phonological adjacency further constrains allomorphic interactions, so too does a semantic specific notion of “adjacency” constrain allosemic interactions and may restrict possible interactions among morphemes even more strongly than the general cyclic architecture of phases.
The University of Massachusetts, Amherst
Saturday, February 13th
Distinguishing contrastive, new and given information
The question of how linguistic theory should break down the dimension of “information structure” that includes contrastiveness, newness and givenness continues to be a subject of debate. This talk reports on joint work with Lisa Selkirk and will defend the three-way distinction between given, new, and focus of contrast originally proposed in Chafe 1976. We argue that the status of a constituent as new is unmarked in the grammar, while constituents which are given or are a focus of contrast are marked as such in the syntactic representation which mediates between sound and meaning.
We show that a system which gives morphosyntactic representation to focus of contrast (FoC-marking) and to givenness (G-marking) but which leaves newness morphosyntactically unmarked has the right consequences for theories of the interfaces of syntax with sentence prosody on the one hand and semantics on the other. On the semantics side, renditions of the Rooth 1992 theory of alternatives focus and the Schwarzschild 1999 theory of givenness are combined with a set of syntax/semantics interface constraints to provide the interpretation and distribution of sentences whose constituents are FoC-marked, G-marked, and/or unmarked for either. On the phonology side, it is shown that all-new sentences receive a phonological interpretation that is based on general phonological principles, without any appeal to the morphosyntactic feature make-up of the sentence.
We also explore some of the typological predictions of our proposal: whether FoC-marking or G-marking are expressed in sentence prosody varies (independently) from one language to the next. Some languages show no prosodic reflexes of these morphosyntactic contrasts at all, instead defaulting to the types of unmarked sentence prosody found in all-new sentences.
The University of Texas at Austin
Sunday, February 14th
Manner and/or Result (in Syntax and/or Semantics)
A leading issue in work on lexical semantics is the question of possible and impossible word meanings: are there constraints on how word meanings are constructed that rule out certain otherwise plausible meanings a word can have? Rappaport Hovav and Levin (2009) (RHL) have recently proposed that while verbs may encode result meanings ("break, shatter, cut, enter") or manner meanings (e.g. types of actions; "run, swim, wipe, sweep"), languages do not seem to have verbs with meanings that combine manner and result together, e.g. no single verb lexicalizes a meaning such as "enter by running". This fact, RHL suggest, follows from basic assumptions about how verb meanings are constructed: verbs encode one and only one idiosyncratic root meaning, and since roots can either encode manner or result, but not both, there should therefore be a complementarity between these two meanings in any single verb. If true, this claim has important ramifications not just for what constitutes a word meaning, but also gives us insights into cross-linguistic lexicalization patterns such as those proposed by Talmy (1975, 2000) and Slobin (1996) regarding manner vs. change-of-location meanings in typologies of motion encoding (see Beavers, Levin, and Tham 2010).
In this talk I present joint work with Andrew Koontz-Garboden (University of Manchester), wherein we address the issue of whether this complementarity exists and if so, why it would. We suggest first and foremost that many previously proposed diagnostics for whether a verb encodes manner and/or result are insufficient to diagnose these two semantic components, and are also often interdependent in ways that make them inappropriate for supporting a claim of complementarity. Thus we develop a new set of truth-conditionally grounded diagnostics, and show that there do exist verbs encoding both components at once, including in particular verbs of manner of killing ("crucify, drown, behead, guillotine"). However, we also examine several morphosyntactic diagnostics for decompositional event structure in verb meaning --- including scopal operators (Dowty 1979, von Stechow 1996) --- and demonstrate that there is evidence supporting the single-rootedness claim of RHL that underlies their proposal. This suggests that while verbs may encode both meaning components at the same time on a truth conditional level, verbs do not literally have both manner and result roots at the same time. Thus we conclude that manner/result complementarity is a morphosyntactic fact about how roots are composed into larger meanings, but not a semantic fact and denotations, since actual roots may encode manner, result, or manner+result meanings. However, we do suggest that languages may show a tendency towards separating result from manner meanings for functional reasons, though other functional pressures may result in some crossed meanings in certain specialized semantic domains.