Search

In a program of inquiry that seeks to derive at least a subset of the design properties of the language faculty from its situation in the array of other cognitive systems of the mind/brain, research has to have a Janus-like character, simultaneously looking without at the (hypothesized) properties of the external systems, particularly the articulatory-perceptual (A-P) and conceptual-intensional (C-I) systems, as well as within the language faculty to examine how these external systems determine its generative processes. This dual perspective has, in recent times, been primarily interpreted as an imperative for the development of adequate "minimalist" theories of the external systems and the "imperfections" they induce in the language faculty, but given the depth of the grammar, issues about the reach of these imperfection-inducing properties into various components of UG also arise. That is, do all the conditions imposed by the performance systems have the identical status of bare output conditions (BOCs), and are all such BOCs necessarily relevant to the N →λ computation? Minimalist inquiry has tacitly assumed the answer to both these questions to be in the affirmative as far as the legibility conditions imposed by the C-I system is concerned, but this paper speculates that an interpretation like "discourse topic" may be arrived at without reference to LF and PF outputs and an interpretation like "topic for the following discourse" has no distinct status in the N →λcomputation. [ABSTRACT FROM AUTHOR]

The analysis of Hindi relative clauses presented in Srivastav (1991), if various gaps in it are filled in a straightforward way, has several implications not noted by Srivastav: it implies that the relative expression in the correlative construction should allow N's of arbitrary syntactic and semantic complexity while the resumptive expression should allow only simple N's and only a limited selection of them; that adnominal and extraposed relative clauses can be stacked while adsentential relative clauses cannot; that non-restrictive relatives do not conform to Srivastav's generalizations about the internal structure of adnominal relatives; and that iterated extraposition and long-distance extraposition of relatives should be possible. These implications are correct, but in less than the full generality implied by the most pristine version of the relevant assumptions. [ABSTRACT FROM AUTHOR]

In Sinhala (Indo-Aryan, Sri Lanka), focus and question constructions are formed with particles that attach to the focused phrase or question word. Although on the surface these particles are usually clause-internal, this paper presents evidence that these particles move before interpretation to the periphery of the clause. Along with strictly Sinhala-internal evidence, we will also consider evidence from Japanese, which (I will argue) shows the hypothesized particle movement in the surface syntax. That is, Sinhala and Japanese form a minimal pair with respect to whether the movement shown in Figure 1 is "overt" (Japanese) or "covert" (Sinhala). To forestall any disappointment, let me say up front that I will not have any concrete proposal as to exactly where the particle moves to, although I take up the issue in some detail in sections 10 and 13. The syntactic structure of focus and question constructions in Sinhala has been discussed in the previous literature, as has the potential similarity to Japanese. This paper builds on these works to varying degrees, primarily on Gair (1983), Gair and Sumangala (1991), Sumangala (1992), Kishimoto (1991, 1992, 1997, 1998), and Whitman (1997), Yanagida (1995). [ABSTRACT FROM AUTHOR]

Transitivity is an ambiguous term. It can refer to the argument structure of a predicate, distinguishing monovalent verbs with a single argument, from polyvalent verbs which obligatorily or optionally assign theta roles to more than one argument It can also refer to morphological and syntactic properties, such as the choice of case and the selection of auxiliary verbs. Ideally, these surface formal markers of transitivity would align with semantic argument structure, and surface properties of the sentence would fall out automatically from the semantic/lexical properties of the predicate. From the statement of the simple X ' notion of a VP projection containing objects, in Chomsky (1981), theoretical ideas have evolved about how verbs project a syntactic phrase in which the arguments are assigned theta-roles, have their case licensed and are in the right configuration to account for binding and other relations. The particular issue of syntactic projection of verbs which I will investigate in this paper is whether all (semantically) transitive verbs in Hindi/Urdu have the same kind of syntactic projection of the verb. In this language, the morphological criteria for transitivity do not all define the same set of sentences, if all the various types of polyvalent verbs are considered. For example, some direct objects and most subjects have nominative case, and can (depending on other conditions in the sentence) trigger verb agreement. Some verb types very clearly define a transitive subject, because it may have ergative case and consistently binds reflexives, but other subjects which bind reflexives must have dative case, and can also behave syntactically as non-subjects. I will propose two kinds of verbal projection, one which is a standard VP shell projection separating the light verb projection including the subject argument from the V and object, and another which is a simple VP, more like an ‘unaccusative' projection with an additional argument. The discussion of transitive verbs is wider here than in Mahajan (1990), which first gave an account of syntax and morphology in Hindi. within a Chomskyan grammar based on feature checking. [ABSTRACT FROM AUTHOR]

The study of case1 has been an integral part of the study of clausal structure for centuries. In the ancient Greek tradition, for example, the actants of a clause were defined primarily in terms of semantic case notions like accusative ‘accused/affected', dative ‘giving', or vocative ‘called/named'. In the Pāninian tradition (ca. 6th century BCE), the relationship between the arguments of a clause and their case realization was mediated via a level of semantic roles such as actor and undergoer. While the ancient notions have remained with us as part of the modern view of syntax, languages also came to be classified according to the kind of subject vs. object marking they displayed. In particular, a distinction is drawn between ergative and accusative types of languages.2 This distinction goes back to Fillmore (1968) and has been taken up in various forms in the literature (see Manning 1996 for an overview). [ABSTRACT FROM AUTHOR]

English has a fairly large number of zero derived denominal verbs like shelve, saddle and calve. Hale and Keyser (1983) have analysed pairs such as saddleN-saddlev as cases of incorporation, where the verb is derived by incorporation of the noun into an empty P and a light v. They take this to be successive cyclic head movement. In this paper, we try to relate the complete absence of denominal verbs like saddle in Malayalam to the fact that Malayalam has no empty preposition. The main proposal of this paper is that languages have to make a parametric choice: whether or not to allow an empty head of [-N] category in syntax. We show that Malayalam chooses the negative value while English chooses the positive value. Then we show that there are no zero verbs either and generalise the proposal to [-N] categories. Supporting evidence for this assumption comes from a range of apparently unrelated facts in the language. We show that each is linked to a zero verb/preposition in some way and that together they provide strong empirical support for the claim. The paper has nine sections. In section 2 we give an exposition of the incorporation analysis of denominal verbs. This is followed by a detailed account of Malayalam unaccusative and transitive verbs, in section 3. In section 4 we discuss the unergatives highlighting the differences with English, which paves the way for the formulation of the hypothesis. Then the dative shift is discussed, in the light of Baker's arguments to show that it is not like the diathesis alternations, for which we present further proof from Malayalam. Section 6 looks at complementisers and the Exceptional Case Marking. Syntax in morphology is addressed in the next section. Section 8 revisits some technical problems left open earlier and propose tentative solutions, and the last section deals with the learnability aspect of the proposed parameter. [ABSTRACT FROM AUTHOR]

Standard conceptions of the lexicon-syntax interface assume that morphologically complex words are constructed in the lexicon and then serve as the atomic objects for syntactic computation. On this view, morphologically complex words are the terminal nodes in a syntactic phrase-marker, their internal structure invisible to syntactic operations. The argument-taking properties of words can be altered by rules which apply inside the lexicon, often with a concomitant morphophonological change, but these properties cannot be affected by syntactic operations. In this paper, I explore an alternative grammatical architecture in which morphology applies to the output of the syntactic component (cf. Halle and Marantz 1993, Marantz 1997). Morphologically complex words, on this view, reflect properties of syntactic structure, which includes argument-structure information. The argument proceeds from an examination of Kannada "valencychanging" morphology, revealing that lexical properties alone cannot explain the distribution of the reflexive and causative morphemes. The analysis builds upon the conclusion of Lidz (1996, 2001c) that the Kannada "verbal reflexive" is best understood with respect to its relationship to causativity and not as a marker of semantic reflexivity. Moreover, given certain independently motivated assumptions about the representation of anaphora, Kannada reflexive morphology provides an argument that the morphological component takes syntactic representations as input and hence that morphological structure is an interpretation of syntactic structure, not the input to it. The resultant theory is one in which the morphological component determines, after some amount of syntactic computation, how a given syntactic representation should be pronounced. Simply put, lexical insertion applies late in the derivation, possibly at LF. The so-called reflexive morpheme argues strongly for a morphological component which is postsyntactic because it is only after certain syntactic operations have applied that the environment for the insertion of this morpheme is met. A theory in which morphologically complex words are constructed prior to syntax cannot explain the distribution of this morpheme. Thus, we need a grammatical architecture in which morphology takes articulated clausal structures as input. [ABSTRACT FROM AUTHOR]

Research on the syntax of negation has produced a lively debate in the literature. One of the central issues concerns whether designated Neg functional projections exist in the phrase structure of natural languages or whether negation elements should be more properly assimilated to adverbial (modificatory) phrases (cf. Holmberg and Platzack 1995 for Scandinavian). If Neg projections do exist, the question arises as to where they are situated within the clause, either as a universal fact, or a parametrizable fact about different languages (see Zanuttini 1991, Ouhalla 1990 for alternatives). Negative forms themselves have been analysed as forming the head of NegP, or sitting in its specifier (cf. Haegeman 1995 for West Flemish), again, possibly as a matter of parametric variation (Zanuttini 1991). However, one of the unstated assumptions of this debate has been that the various different syntactic options under hypothesis carry the same interpretative force. In other words, it is assumed that the logic of clause level negation is uniform and uncontroversial, and that the syntactic representation is what is in doubt, or maybe parametrizable in certain respects. On the other hand, in the literature, there is some evidence of different classes of clausal negative elements, even within the same language. In particular, many languages have different negative forms for non-finite clauses, or for imperatives (Zhang 1990). In this paper, I will present some evidence from Bengali to argue that clausal negation can actually take two semantically different forms. Specifically, I will propose that negation is not a sentential operator, as has been standardly assumed, but selectively binds particular variables introduced by particular syntactic heads. The evidence for this position will come from the scopal interactions between negation, the binding of the tense variable and the binding of the event variable. I will argue that these scopal interactions are not freely ordered, but rather betray the rigidity of clausal architecture and its mapping to the semantics, and which introduce unexpected semantic incompatibilities. [ABSTRACT FROM AUTHOR]

The Kannada clause raises interesting questions from a synchronic as well as diachronic perspective. In earlier work (Amritavalli 2000), beginning with the puzzle of how tense was interpreted in negative sentences instantiating the ‘matrix infinitive' and the ‘matrix gerund,' I proposed that the finiteness feature of the Kannada clause resides in a Mood Phrase that hosts one of the heads agr (Indicative mood), neg, or modal. The category tense, now divested of finiteness, i.e. its function of "anchoring the sentence in time," is consequently subsumable under that of temporal aspect in this language. Each of the heads of the Mood Phrase is seen to select an Aspect Phrase complement, some apparently finite in containing morphemes standardly taken to be ‘tense,' now reanalyzed as temporal aspect; and some obviously nonfinite, i.e. the ‘matrix infinitive/gerund'. The postulation of a Mood Phrase with a finiteness feature thus allows us to analyze a variety of superficial clause-types in Kannada in terms of a single underlying clause structure. This work is summarized in section 2. In section 3 of this paper, I argue that the Mood P is a relatively new functional projection in the language. An older functional projection hosting agr, which was the sole marker of finiteness, looks to have yielded ground to a new projection of Mood which allows finiteness to be marked by the neg and modal elements in addition to agr. Simultaneously, a new functional projection or projections have evolved of temporal Aspect; and this is perhaps the first step in the evolution of a category of Tense in the language. These changes in the functional architecture of the clause are invoked to explain a variety of changes in the distributional facts concerning the cooccurrence of agreement with negation, agreement with modality, and negation with an interpretation of tense, in the history of Kannada. At a prior stage of the language, which had a ‘negative conjugation' and a ‘contingent future tense', markers for negation, temporal aspect and modality occur in a single, undifferentiated position in the clause. Negation does not permit temporal specification, and negation cooccurs with agreement. In the current language, both these facts are reversed. Our account of the differentiation of the features of negation and temporal aspect, and their evolution into separate functional projections, allows us to strengthen Grimshaw's (1991) proposal of extended projections with an interesting detail. The original proposal that functional projections share categorial features with related lexical projections does not address the question why this should be so; nor does it give a rationale, other than an observational one, for the identification of particular lexical and functional categories as categorially related. We argue the functional elements in the extended projection of a lexical category to literally originate in the lexical category, at a functional level of zero. It is lexical features that "migrate" or reinstantiate themselves in functional projections in the XP. In particular, the Kannada facts suggest that the feature content of heads of individual functional projections in the clause for negation, temporality and modality has its counterpart not only in the lexical aspectual features of a verb as an integral part of its meaning, but also in isolable morphemes with corresponding meanings which nevertheless lack individual clausal, functionally differentiated positions of their own. Section 4 of the paper elaborates the clause structure developed in the preceding two sections by taking into account nonfinite negation in the matrix clause, and the occurrence of temporal aspect on "dummy verbs." This structure is also seen to accommodate the semi-lexical auxiliary and the serial verb constructions in the language. [ABSTRACT FROM AUTHOR]