Linguistic Variation Yearbook - Volume 7, Issue 1, 2007

We argue that the smallest shifts in grammar should give us the most precise insight into grammatical mechanisms. The often tiny differences captured by dialects become a natural focus for linguistic theory, precisely because of their variability and instability in the history of language and in the grammatical commitments of individual speakers. The notion of stability refers to where dialects differ, where registers add or drop features, where historical change moves, what proves difficult in the steps in acquisition, and where second language learners easily stumble. In a word, we argue that nodes are stable and features are unstable. In particular, node labels, which reflect head features in a feature bundle, are stable, while non-head features are unstable and subject to variation. We show that specific nodes in African American English, such as Negative Focus (CP) in negative inversion constructions, are stable (Don’t nobody play baseball.). Also, particle positions (he threw (up) the ball (up)) and various features, like those associated with tense inflection (if it is under a V node not a T node) or presuppositionality (if it is under a CP node), are inherently unstable. Many other predictions for dialect, acquisition, and diachronic patterns follow as well.

The term Focus Dependency describes an important phenomenon at the syntax-semantics interface: Elided material can exhibit bound-variable-like behavior when its antecedent is a focussed phrase in the same sentence. In the past, focus dependency has been analyzed as actual binding or by means of copying. This paper presents a new account of focus dependency that relies on the syntactic idea of structure-sharing. Structure-sharing allows sub-phrases to be syntactically linked to more than one position of a phrase marker. The proposal better explains focus dependency than existing accounts considering data from sentence-boundedness, insensitivity to c-command, extraction from the focus-dependent material, and the formal link to the antecedent. It also achieves a theoretical unification with other phenomena where structure sharing has been made use of, specifically movement.

A restrictive theory of syntax needs both a restrictive theory of structures and a restrictive theory of operations.Much recent effort has gone into narrowing the class of allowable structures and a lot has been learned. This paper proposes that operations are linearly ordered on an essentially constituent by constituent basis. A universal constraint on the ordering of operations in language is proposed whose function is to fix the order in which operations apply. This constraint is deployed using a generalized prohibition against improper movement. The proposal captures some but not all effects of what has traditionally been called the freezing principle. It is argued that empirically exactly the right cut is made. It is further argued that the proposal rules out an entire class of remnant movement derivations, including the analysis of cross-serial dependencies in Nilsen (2003) and the analysis of order-preservation in Koopman and Szabolcsi (2000).

I discuss the variation concerning agreement with coordinated subjects in Dutch dialects. I show that a verb or a complementizer in several variants of Dutch agrees with the first conjunct of a coordinated subject and in other variants with the coordinated subject as a whole. I argue that this variation can be accounted for by the interaction between the syntactic derivation and the post-syntactic morphological component.More specifically, I argue that syntax establishes an agreement relation with both the coordinated subject as a whole and the first conjunct of the coordinated subject. Subsequently, during the post-syntactic morphological derivation, one of these agreement relations will be overtly expressed on the Probe. The decision as to which one of the two relations is spelled out depends on the affix inventory of the language or dialect. More specifically, the subset principle is extended in such a way that, confronted with the situation in which a Probe is related to two Goals, an affix is inserted for the relation which results in the most specific agreement morphology. The analysis is extended to the typologically unrelated languages Irish and Arabic.

I argue that despite their traditional verb-first vs. verb second partition, Welsh and Breton both instantiate a ban on verb-first and I present an analysis of these two languages as fundamentally verb second. In this view, so-called verb first orders prototypically illustrated byWelsh result from inconspicuous strategies to fill in the preverbal position, whereas traditional verb second prototypically illustrated by Breton results from conspicuous strategies to fill in the preverbal position. I show that both conspicuous and inconspicuous verb second orders are present in bothWelsh and Breton. The difference in word order between Welsh and Breton is reduced to (i) a lexical parameter, that is availability of a free preverbal expletive particle inWelsh, and (ii) a syntactic parameter: Breton allows for the creation of expletives by short movement, a parameter shared with Icelandic and other languages instantiating stylistic fronting.

This paper argues that evidential clitics in St’át’imcets (a.k.a. Lillooet; Northern Interior Salish) must be analyzed as epistemic modals.We apply a range of tests which distinguish the modal analysis from the main alternative contender (an illocutionary operator analysis, as in Faller 2002), and show that the St’át’imcets evidentials obey the predictions of a modal analysis. Our results support the growing body of evidence that the functions of encoding information source and epistemic modality are not necessarily distinct. The St’át’imcets data further provide a novel argument against the claim that evidentiality and epistemic modality are separate categories. Many authors argue that evidentials differ from modals in that the former do not encode speaker certainty (see, e.g., de Haan 1999; Aikhenvald 2004).We argue that modals are also not required to encode speaker certainty; we provide evidence from St’át’imcets that marking quantificational strength is not an intrinsic property of modal elements.