This thesis examines the behaviour of sentential complements in Mohawk within the framework of Government and Binding Theory. Past proposals concerning the syntactic structure of sentential complements in Romance languages (and English) are explored in Mohawk. It is claimed that Mohawk only has full CP complements and no distinct types of embedded clauses (such as a subjunctive or infinitival). This is due to a morphological requirement (specified by the Minimal Word Constraint) on Mohawk verbs which dictates the need for obligatory agreement morphology. Tense/aspect co-occurrence restrictions are given to show what type of CP complements a verb can take. The evidence indicates that selection of complements is due to semantic and not syntactic reasons.

No description available.

This thesis discusses Internally Headed Constructions in Japanese, specifically, the "Counter-Equi NP" (CENP) construction and the "Internally Headed Relative Clause" (MC) construction. In both of these constructions, an NP within the embedded clause is interpreted as an argument of the matrix clause. / There are two major goals in this thesis. The first goal is to provide a unified syntactic and semantic analysis of the CENP and IHRC constructions. I argue that the CENP construction is basically the same as the IHRC construction, even though some previous researchers have assumed the contrary (Kuroda 1992, 1999, Ohara 1996, Shimoyama 1999, among others). Furthermore, I argue that both the CENP NP and the MC NP are arguments of the matrix verbs. / The second goal is to closely examine the syntactic and semantic properties of the CENP construction and the IHRC construction. As discussed by Hoshi (1995) and Shimoyama (1999), the interpretation of the argument which is "modified by the IHRC" in these constructions is quite similar to that of E-type pronouns (Evans 1977a,b; 1980). In addition, there are some facts that show that the event of the embedded clause is linguistically connected to the event of the matrix clause. Regarding all these properties, the CENP construction behaves in the same manner as the IHRC construction. In this thesis, adapting the E-type pronoun analyses of the IHRC (Hoshi 1995 and Shimoyama 1999), I propose an alternative analysis of the CENP construction and the IHRC construction to capture those properties. / Moreover, even though the CENP construction and the IHRC construction share many properties with each other, there are some differences between them (Shimoyama 1999, among others). In particular, the CENP cannot appear as the subject of the matrix clause, whereas the IHRC can. However, we observe restrictions on possible internal heads with regard to the subject IHRC, in contrast to the CENP. This thesis tries to account for those differences under a unified analysis of the CENP construction and the IHRC construction, together with an analysis of the subject MC, which adapts Shimoyama's (1999) analysis of the IHRC.

This thesis outlines a syntactic analysis of Noun Incorporation in Cree. In this construction, certain morphemes, 'medials', that appear as the nominal root of an external NP can alternatively appear within a verb. This thesis extends previous analyses of Algonquian medials by utilizing the theory of Incorporation developed in Baker (1988b). Within this theory of grammar, medials are base-generated as nouns within an 'object' NP and then optionally adjoined to the verb stem as a result of head (X$ sp{ rm o}$) movement. Established restrictions on head movement can account for many properties of NI, including paraphrasing, doubling, bare modifiers, possible thematic relations, and differences between NI and compounds. The efficacy of the syntactic approach validates a modular account of polysynthetic word formation. In addition, the distribution of Cree NI validates several putatively universal principles of theta-role assignment.

Hungarian 'focus position' is typically thought of as a central example of a 'discourse configurational' phenomenon, since it not only involves the expression of information-structural (or 'discourse semantic') meaning through the manipulation of word order but also interacts syntactically with other elements of the sentence. In this thesis, I argue that this kind of phenomenon highlights fundamental theoretical problems with conventional assumptions about the relationships between linguistic form and different kinds of meaning and demonstrate that these problems have led to empirical inadequacies in the syntactic analysis of Hungarian. I propose an alternative analysis that makes use of a dynamic, incremental parsing-based approach to grammar, which in turn allows for the influence of inferential pragmatic operations (investigated in terms of Relevance Theory) at all stages in the process of interpreting linguistic form. This opens up possibilities of structural and interpretive underspecification that allow for the interpretation of the 'focus position' to be unified with the information-structural interpretation of sentences that do not contain a syntactically focused expression. This analysis explains the interaction of syntactic foci with other pre-verbal items. The burden of explanation is thus shifted away from specialised, abstract syntactic representations and onto independently necessary aspects of cognitive organisation. The use of 'discourse semantic' primitives---whether in terms of focus or exhaustivity---to encode the effects of the 'focus position' is shown to be both theoretically problematic and empirically inadequate. The information-structural meanings associated with the position must be viewed not as the input to interpretive processes but instead as the result of inferential processes performed in context. Reanalysis of the syntactic evidence shows the relevant position to be not merely pre-verbal, but underlyingly pre-tense, showing that the unmarked position of the main verb is essentially the same as that of syntactically focused expressions. This leads to an analysis whereby both 'neutral', topic-comment readings and cases of narrow focus emerge from inferences over a common interpretive procedure. This procedure is identified as 'main predication': the point in the parsing of a sentence at which the application of a single predicate effects the conversion of a mere description of an event into a truth-conditional assertion. Main predication is represented using neo-Davidsonian, event-based semantics (the effect of the main predicate being equivalent to that of the application of an existential quantifier over an event variable in the neo-Davidsonian approach) and made dynamic by the use of the epsilon calculus. This analysis predicts the postposing of any (otherwise pre-tense) 'verbal modifier' (VM) in the presence of a syntactic focus and the apparent information-structural ambiguity of VMs when they are pre-tense. Certain constraints on the distribution of quantifiers are also predicted, one such constraint being adequately characterisable only within a semantically underspecified, procedural account. The behaviour of the negative particle "nem" is also given a maximally simple explanation. The apparently variable scope of the negative operator is explicable without ad hoc syntactic mechanisms: the apparent wide scope reading associated with 'sentential' negation follows inferentially from narrow scope negation of temporal information. The syntactic positions of negation are predictable on this basis. In addition, the assumption of consistent narrow scope negation correctly predicts that VMs must postpose or receive a narrow focus reading in the presence of "nem".

Graduate

State-of-the-art parsers suffer from incomplete lexicons, as evidenced by the fact that they all contain built-in methods for dealing with out-of-lexicon items at parse time. Since new labelled data is expensive to produce and no amount of it will conquer the long tail, we attempt to address this problem by leveraging the enormous amount of raw text available for free, and expanding the lexicon offline, with a semi-supervised word learner. We accomplish this with a method similar to self-training, where a fully trained parser is used to generate new parses with which the next generation of parser is trained. This thesis introduces Chart Inference (CI), a two-phase word-learning method with Combinatory Categorial Grammar (CCG), operating on the level of the partial parse as produced by a trained parser. CI uses the parsing model and lexicon to identify the CCG category type for one unknown word in a context of known words by inferring the type of the sentence using a model of end punctuation, then traversing the chart from the top down, filling in each empty cell as a function of its mother and its sister. We first specify the CI algorithm, and then compare it to two baseline wordlearning systems over a battery of learning tasks. CI is shown to outperform the baselines in every task, and to function in a number of applications, including grammar acquisition and domain adaptation. This method performs consistently better than self-training, and improves upon the standard POS-backoff strategy employed by the baseline StatCCG parser by adding new entries to the lexicon. The first learning task establishes lexical convergence over a toy corpus, showing that CI’s ability to accurately model a target lexicon is more robust to initial conditions than either of the baseline methods. We then introduce a novel natural language corpus based on children’s educational materials, which is fully annotated with CCG derivations. We use this corpus as a testbed to establish that CI is capable in principle of recovering the whole range of category types necessary for a wide-coverage lexicon. The complexity of the learning task is then increased, using the CCGbank corpus, a version of the Penn Treebank, and showing that CI improves as its initial seed corpus is increased. The next experiment uses CCGbank as the seed and attempts to recover missing question-type categories in the TREC question answering corpus. The final task extends the coverage of the CCGbank-trained parser by running CI over the raw text of the Gigaword corpus. Where appropriate, a fine-grained error analysis is also undertaken to supplement the quantitative evaluation of the parser performance with deeper reasoning as to the linguistic points of the lexicon and parsing model.

The purpose of this paper is to describe the phenomenon of multiple subjects in Korean. In SOV languages like Korean and Japanese, where double subjects (S + S + V), occur, the second subjects has a subject marker in the surface form, but functions as the object of that structure. This relationship is termed as ergativity: the hypothesis of this study is that ergativity accounts for double subject constructions. When the hypothesis is applied to double subjects, almost all of them are identified as ergative structures.Another typical characteristic of the Korean language is that when topicalization is applied to a possessive structure in the subject position, it produces two successive NPs. Subjectivalization can further be applied to both NPs resulting in two subjects. The same phenomenon occurs when it applied to a structure which has a series of NPs or Locatives. In a few rare cases, these double subjects precede the ergative structure which already has other double subjects. Then the structure might have as many as four subjects in the surface structure.

The purpose of this study was to determine if the 'language listening' of children ages three years and three months (3.3) to three years and nine months (3.9) as measured by their behavioral response is influenced by the word-order of an interrogative sentence when presented in normal word-order, misplaced word-order, and scrambled word-order.'Language listening' for the purpose of this investigation was defined as the interpreted meaning of a young child as measured by the correct behavioral response to verbal stimuli ordered in specific syntactical variations. The verbal stimuli were written in three variations of word order. These patterns were normal word order which was the regular order of an interrogative sentence; misplaced word order was a pattern in which all parts of the verb and noun were interchanged in position; and scrambled sentence were positioned randomly without any set pattern of order.The subjects in this study were selected from the available population of children attending five nursery schools located in the metropolitan area of Muncie, Indiana. Selection of subjects and categorization into groups was based upon the index of the mean length of utterance of each individual member. This index was computed from a language sample containing one hundred utterances that were collected in the Screening Session.'Language listening' was measured by the relevant responses of subjects performance to specific requests to respond with an appropriate toy to the question asked and the directions given. Scripts used in the Data Collection Session combined a series of nine behavioral tasks with three of each of the types of word order--normal, misplaced, and scrambled.One major limitation of this study was the index used to measure the verbal maturity of the subjects in this study. This index is an average of the utterance used by the child and disguises the verbal expansion ability of the child and the sophistication of the child's verbal ability. Another limitation of this study was the selection of the behavioral tasks. It was observed that the subjects' performance in some cases was made from an anticipated response rather than responding from actual understanding of the tasks. It appeared that the behavioral tasks were oversimplified to be used the subjects included in this study.A research design employing the use of a Latin Square was constructed to combine the behavioral tasks with the types of word order. An analysis of variance was used for the analysis of data. The F-ratio, derived from an analysis of variance, was used to test statistical significance of the null hypotheses. The .05 level of significance was designated as the standard of significance.The findings of this investigation revealed that there is no statistical significance between 'language listening' and the type of word order. It was also determined in this study that the type of word order does not differentially affect children with varying verbal maturity. The findings did agree with research in the area in regard to children with a nonfluent level of verbal maturity. Agreement was found to support the research that with children having a nonfluent level of verbal maturity, the type of word order that is used as verbal stimuli does not affect the meaning that they glean from the stimuli.Among the recommendations offered was that further research be conducted with children in the early stages of language listening. It was also recommended that further research be conducted using an index of verbal maturity that is descriptive of the actual verbal ability of the child.

This thesis explores an applicative approach to two constructions in Chinese: the “oblique object” construction and the double object construction (DOC). The DP following the verb in an “oblique object” construction has generally been viewed as an object of the verb; however, its properties, especially object-associated properties, have not been tested systematically. This study tests the properties of the oblique object, and finds that the oblique object in the “oblique object” construction differs from a typical theme object in a range of syntactic properties. Pylkkänen (2002) hypothesizes that applicatives fall into two categories, high applicatives and low applicatives. Syntactically, the high applicative head is merged above the VP, with the VP as complement and the applied argument as specifier; semantically, high applicatives denote a relation between the applied argument and the event described by the VP. Low applicatives are base-generated within the VP, merging with the direct object as complement and the applied argument as specifier; semantically, a low applied argument bears a transfer-of-possession relation to the direct object. According to Huang (2007), Chinese has two types of DOCs, giving DOCs and stealing DOCs. Giving DOCs indicate situations in which the indirect object is given something or some advantage; stealing DOCs express situations in which the indirect object suffers some loss, or is adversely affected. In analyzing Chinese DOCs from an applicative approach on the basis of Pylkkänen’s high/low applicative hypothesis, Cheng and Wen (2008) suggest that Chinese DOCs are high applicatives, while Sun and Li (2010) suggest that they are low applicatives. This thesis suggests a finer classification of Chinese DOCs, and shows that Chinese DOCs include both high and low applicatives. Specifically, giving DOCs can be analyzed as low recipient applicatives. Stealing DOCs in which the indirect object bears a possessive relation to the theme can be analyzed as low source applicatives, and stealing DOCs in which the indirect object bears no direct semantic relation to the theme correspond to high malefactive applicatives. / Graduate