Filter Dissertations and Theses By:
Omitted Arguments and Complexity of Predication
Year of Dissertation:
Abstract OMITTED ARGUMENTS AND COMPLEXITY OF PREDICATION by Martin Port Adviser: Professor William McClure This work focuses on the licensing conditions and logical structure of understood-argument constructions, or complement-drop constructions, in English. There are two main types of such arguments: Indefinite Understood Arguments (IUA) and Definite Understood Arguments (DUA). IUA readings occur in such cases in He ate, He cooked. In such cases, the reference of the dropped element need not be known for the sentence to be satisfactorily interpretable. DUA readings are given in such examples as She followed, She won. Here the reference of the missing element must be known to the speaker/hearer; it must appear in discourse. Our central claim is that it is not necessary to resort to explanatory factors outside the lexical-semantic structure of verbs in order to account for the alternations. We propose that for both IUA and DUA, the structure of the understood argument is a complex structure involving existential quantification: "x [P(x)]". It is never a simplex, atomic element that could be represented by an individual constant. For IUA cases, we justify this complex structure by showing that it is mirrored in the structure of the alternating verbs. We note in particular that IUA verbs often undergo other alternations such as the material/product alternation, which we consider to be an indication of a complex lexical structure. For Definite Understood Arguments, we give two licensing factors that correspond to the variable and the predicate in the logical form. We motivate existential quantification in DUA constructions, by considering such contrasts as I know/I know about that vs. I believe/*I believe about that. Selecting for a PP-complement correlates with the possibility of DUA. We claim that in this example, the about-phrase provides a slot for a variable, since the phrase makes allusion to some object without stating explicitly what it is. We conclude by analyzing subject-drop alternations--the causative-inchoative alternation--in the light of our findings regarding complement drop. We note that this alternation occurs with verbs of a simplex structure; and we offer a description of the general system of argument drop in English.
Asking Questions in Learner English: First and Second Language Acquisition of Main and Embedded Interrogative Structures
Year of Dissertation:
This dissertation examines how adults and children learning English produce and judge English questions. The ultimate goal of this study is to contribute to an understanding of the extent, nature and causes of learners' persistent difficulties with some syntactic properties of the language they are acquiring. To examine whether word order errors in the production of English interrogatives by L2 learners stem from lack of knowledge or from difficulties with automatic implementation of L2 procedures under real time constraints, L2 learners' performance in tasks that tap into different abilities is compared. To examine whether errors in the production of English interrogatives by L2 learners can be imputed to L1 transfer, L1 Chinese and L1 Spanish production patterns are compared. Finally, to examine whether errors in the production of L1 learners can be attributed to properties of the adult input, the results from an elicited production study with 3-5 year olds are examined in light of the frequency of different word combinations in the adult input. Taken together, the present results indicate that difficulties with English interrogative structures (a) are a consistent phenomenon in L1 and L2 learners, (b) might be better accounted for in terms of non-target-like representations, rather than difficulties with implementation of L2 procedures, and (c) do not follow in a direct way from L1 transfer or properties of the input. Learners' errors depend on specific syntactic configurations (wh- vs. yes/no) and wh-words (why, when), suggesting that child and adult learners entertain similar grammatical hypotheses and make use of similar mechanisms for language acquisition.
Is cue-based memory retrieval ‘good-enough’?: Agreement, comprehension, and implicit prosody in native and bilingual speakers of English
Year of Dissertation:
This dissertation focuses on structural and prosodic effects during reading, examining their influence on agreement processing and comprehension in native English (L1) and Spanish-English bilingual (L2) speakers. I consolidate research from three distinct areas of inquiry—cognitive processing models, development of reading fluency, and L1/L2 processing strategies—and outline a cohesive and comprehensive processing model that can be applied to speakers regardless of language profile. This model is characterized by three critical components: a cognitive model of memory retrieval, a processing paradigm that outlines how resources may be deployed online, and the role of factors such as prosody in parsing decisions. The general framework of this integrated ‘Good-enough Cue’ (GC) model assumes the ‘Good-Enough’ Hypothesis and cue-based memory retrieval as central aspects. The ‘Good-Enough’ Hypothesis states that all speakers have access to two processing routes: a complete syntactic route, and a ‘good enough’ heuristic route (Ferreira, Bailey, & Ferraro, 2002; Ferreira, 2003). In the interest of conserving resources, speakers tend to rely more on heuristics and templates whenever the task allows, and may be required to rely on this fallback route when task demand is high. In the proposed GC model, cue-based memory retrieval (CBMR) is the instantiation of the complete syntactic route for agreement and long-distance dependencies in particular (Lewis & Vasishth, 2005; Wagers, Lau, & Phillips, 2009; Wagers, 2008). When retrieval fails using CBMR (due to cue overlap, memory trace decay, or some other factor), comprehenders may compensate by applying a ‘good-enough’ processing heuristic, which prioritizes general comprehension over detailed syntactic computation. Prosody (or implicit prosody) may reduce processing load by either facilitating syntactic processing or otherwise assisting memory retrieval, thus reducing reliance on the good-enough fallback route. This investigation explores how text presentation format interacts with these algorithmic versus heuristic processing strategies. Most specifically, measuring whether the presentation format of text affects readers’ comprehension and ability to detect subject-verb agreement errors in simple and complex relative clause constructions. The experimental design manipulated text presentation to influence implicit prosody, using sentences designed to induce subject-verb agreement attraction errors. Materials included simple and embedded relative clauses with head nouns and verbs that were either matched or mismatched for number. Participants read items in one of three presentation formats: a) whole sentence, b) word-by-word, or b) phrase-by-phrase, and rated each item for grammaticality and responded to a comprehension probe. Results indicate that while overall comprehension is typically prioritized over grammatical processing (following the ‘Good-Enough’ Hypothesis), the effects of presentation format are differentially influential based on group differences and processing measure. For the L1 participants, facilitating the projection of phrasal prosody (phrase-by-phrase presentation) onto text enhances performance in syntactic and grammatical processing, while disrupting it via a word-by-word presentation decreases comprehension accuracy. For the L2 participants however, phrase-by-phrase presentation is not significantly beneficial for grammatical processing—even resulting in a decrease in comprehension accuracy. These differences provide insight into the interaction of cognitive taskload, processing strategy selection, and the role of implicit prosody in reading fluency, building toward a comprehensive processing model for speakers of varying language profiles and proficiencies.
Effects of L1 Spanish and L2 English sublexical and lexical processing on L2 English word reading speed and accuracy, and reading comprehension
Year of Dissertation:
This study explores the word identification processes used by adult bilingual Spanish/English readers of varying proficiencies in both L1 and L2, and how these relate to reading comprehension levels in L1 and L2. It has been proposed in word reading research that lexical processing (accessing the word directly from the lexicon) is more efficient than sublexical processing (assembling the phonology of a word letter by letter). This investigation is concerned with the type of bilingual processing used by readers of varying ability. Three groups of Spanish/English bilingual readers, who differed across both L1 and L2 in their reading comprehension levels, performed a reading and listening task in both Spanish (Experiment 1) and English (Experiment 2). A third monolingual control group performed the task in their L1, English (Experiment 3). The on-line procedure involved presenting words visually and aurally, for participants to make a yes/no choice as to whether the words match. This bimodal task, not previously used in second language research, presented equal difficulty for readers of all proficiency types. The written words appeared in either uniform (lower) case or alternating (aLtErNaTiNg) case, to test reliance on lexical processing during word identification. Overall, L1 Spanish readers of all proficiency levels used similar strategies for L1 word identification, but the less proficient readers showed this effect when orthographic form was disrupted, i.e., when the stimuli were presented in alternating case. In L2, the less proficient readers tended to rely more on a sublexical strategy, while the high proficiency group used a lexical strategy. Crosslinguistic correlations indicated that readers who showed more disruption (as evidenced by higher reaction times) to alternating case stimuli in L1 Spanish had lower proficiency levels in English, indicating that L1 reading skill influences L2 reading proficiency. These findings combined support dual route models of reading, and have important implications for educational decisions: both sublexical and lexical skills are emergent in L1 and L2, and should be focused on in instructional settings.
A CORPUS-BASED SOCIOLINGUISTIC STUDY OF SUBJECT PRONOUN PLACEMENT IN SPANISH IN NEW YORK
Year of Dissertation:
This dissertation presents a variationist sociolinguistic study of the variable placement of subject personal pronouns before or after verbs in Spanish in New York City (e.g. ella canta; canta ella, both `she sings'). It pursues a line of inquiry that partially replicates recent work by Otheguy & Zentella (2012). The research addresses, on the basis of a different morphosyntactic phenomenon, the same question of whether bilingualism in NYC is giving rise to language contact phenomena under which patterns of English usage are affecting the use of Spanish. In addition, this project expands the original database extracted from the Otheguy - Zentella corpus to include not only finite verbs but also non-finites, and it expands as well the tools of the analysis by introducing additional linguistic predictor variables. This project will demonstrate that, with respect to the feature under study, there are indeed two distinct groups of newcomers entering the City, those from the Caribbean and those from the Latin American Mainland. It will also show that while continuity with Latin American ways of speaking remains the strongest force shaping Spanish in NYC, changes in the internal grammar of the speakers begin with slight force in the first generation and continue in the second. On the basis of extensive internal evidence, this project advances our knowledge of variables conditioning the placement of subject pronouns. It is shown that the influencing variables for Mainlanders coincide for the most part with those for Caribbeans; that the speech of Caribbeans is not as rigidly set on an SVO word order as it has been claimed in the literature; and that the speech of Mainlanders is more set on an SVO order as well, rather than on the required VSO word order claimed by several authors in the literature. Lastly, the study demonstrates that the linguistic variables constraining pronoun placement in the speech of newcomers begin to change in significance and hierarchy in the first generation with more exposure to the City, and become insignificant for the second generation of speakers.
Echolocation: Using Word-Burst Analysis to Rescore Keyword Search Candidates in Low-Resource Languages
Year of Dissertation:
State of the art technologies for speech recognition are very accurate for heavily studied languages like English. They perform poorly, though, for languages wherein the recorded archives of speech data available to researchers are relatively scant. In the context of these low-resource languages, the task of keyword search within recorded speech is formidable. We demonstrate a method that generates more accurate keyword search results on low-resource languages by studying a pattern not exploited by the speech recognizer. The word-burst, or burstiness, pattern is the tendency for word utterances to appear together in bursts as conversational topics fluctuate. We give evidence that the burstiness phenomenon exhibits itself across varied languages. Using burstiness features to train a machine-learning algorithm, we are able to assess the likelihood that a hypothesized keyword location is correct and adjust its confidence score accordingly, yielding improvements in the efficacy of keyword search in low-resource languages.
Obstruent Voicing and Tone in Siklis Gurung
Year of Dissertation:
This thesis examines proposals for the tone system of Gurung, a Sino-Tibetan language spoken in Nepal, in light of data collected from a speaker of Siklis Gurung. Although Gurung is widely acknowledged to be a tonal language, existing descriptions of Gurung disagree as to how these tone categories should be defined and whether word-initial obstruent voicing is phonemic or allophonic. The data presented in this paper suggests that, despite claims otherwise, voicing is phonemic in some dialects of Gurung. It also suggests that Siklis Gurung is best analyzed as having three tone categories: a low tone that occurs with breathy phonation; and high and mid tones that occur with modal phonation. Following models of tonogenesis, the emergence of these three tones is attributed to the split of one of two proto-tones due to the loss of a word-initial obstruent voicing contrast. This hypothesis is compared to theories outlined in the literature.
Attachment Ambiguities in Hebrew Complex Nominals: Prosody and Parsing
Year of Dissertation:
This dissertation investigates the prosody-syntax interface in two Hebrew complex nominals. Their meaning is identical but their syntax and phonology differ. The free state (FS) nominal is similar to English (e.g., the coach of the wrestler). The construct state (CS) nominal lacks the preposition, has stress on the second noun, and the construction constitutes one phonological word despite its internal syntactic structure. When followed by a relative clause (RC) the two constructions (1a,b) are syntactically/semantically ambiguous in the same way; but they contrast prosodically because one of the phrasing patterns permitted for FS (a boundary between the two nouns) is inhibited for CS. (1) a. FS: ha-me'amen shel ha-mit'agref she-parash le'axar ha-taxarut b. CS: me'amen ha-mit'agref she-parash le'axar ha-taxarut (the-)coach of the-wrestler who-retired after the-fight Research on the relative clause attachment ambiguity has shown that different languages and constructions vary in attachment preferences; it is important to investigate factors that may contribute to attachment decisions, because this variation threatens the universality and innateness of the human sentence processing routines. In recent years, much research has focused on a prosodic explanation: The Implicit Prosody Hypothesis (Fodor, 2002) proposes that attachment choices in silent reading are affected by implicit prosody mentally projected onto the sentence by readers, on the basis of language-specific phonological rules. Since implicit prosody cannot be directly observed, it is investigated by studying overt prosody in speaking and listening. This dissertation thus provides data on prosodic phrasing in Hebrew and its relation to the processing of the RC-attachment construction. Three experiments are reported: a production experiment studied speakers' preferred prosodic patterns for each construction; a listening study examined how the preferred interpretation differed for different prosodic contours; a silent reading experiment investigated the interpretation when no overt prosody was provided. For the FS sentences the results of the three experiments concurred, indicating that RC-attachment is sensitive to prosody, both explicit and implicit. For the CS, prosody was also found to play a role but some other factor(s) intrinsic to the construction resulted in consistently higher RC-attachment.
Deriving Word Order in Code-switching: Feature Inheritance and Light Verbs
Year of Dissertation:
Marcel den Dikken
This dissertation investigates code-switching (CS), the concurrent use of more than one language in conversation, commonly observed in bilingual speech. Assuming that code-switching is subject to universal principles, just like monolingual grammar, the dissertation provides a principled account of code-switching, with particular emphasis on OV~VO variation in two typologically similar language pairs, Korean-English and Japanese-English bilingual speech. Taking the view into consideration that linguistic variation is a result of variation in the domain of functional categories rather than lexical roots (e.g., Borer 1984; Chomsky 1995), the role of light verbs in word order in code-switching is further investigated and tested against Korean-English and Japanese-English bilingual speakers' introspective judgments of the code-switching patterns presented to them in the form of a questionnaire. The results provide strong evidence indicating that the distinction between lexical and functional or light verbs play a major role in deriving different word order, OV and VO in Korean-English and Japanese-English code-switching, respectively, supporting the hypothesis that parametric variation is attributed to differences in the features of a functional category in the lexicon. In particular, the explanation pursued in this dissertation is based on feature inheritance, proposed in recent developments the Minimalist Program. To account for OV~VO variation in Korean-English and Japanese-English code-switching, feature inheritance, primarily proposed for the C-T domain by Chomsky (2000, 2001, 2008), is extended to the v-ASP domain, thereby developing it into a full-fledged mechanism for the two phases, C and v, of the clause. Two principles of feature inheritance (feature selection and feature expiration) and three operational rules (earliness, economy, and multiple agree under antisymmetry) are proposed to show that feature inheritance is designed to make a derivation proceed economically and efficiently in the syntax. Based on this, the dissertation presents how head-initial structure in English (C-S-V-O) and head-final structure (S-O-V-C) in Korean and Japanese are derived, and argues that the OV-VO variation in Korean-English and Japanese-English code-switching is due to a result of object shift: if object shift occurs, OV is derived. On the other hand, if object shift fails, the underlying VO structure will surface.
The Semantics of Adjectives of Quantity
Year of Dissertation:
This work investigates the semantics of the adjectives of quantity (Q-adjectives) many, few, much and little, with the goal of providing a unified semantic analysis of this class, and in doing so exploring the semantics of quantity and degree more broadly. My central claim is that Q-adjectives must be analyzed as degree predicates - gradable predicates of scalar intervals - with much of the semantic content traditionally ascribed to these words instead contributed by a set of null functional elements and operations. This proposal allows a compositional analysis of Q-adjectives across the wide range of syntactic positions in which they occur, including quantificational (many dogs bark), predicative (John's friends are few), attributive (the little rice that remains), and especially differential (many fewer than 100 students), the latter of which is problematic for theories that take Q-adjectives to be either quantifying determiners or cardinality predicates. The same mechanism also accounts for the operator-like behavior of few and little and the availability of much as a dummy element (much-support), and allows quantification over individuals to be analyzed via simple Existential Closure, without giving rise to spurious `at least' readings for few/little. I further show that patterns in the interpretation and distribution of Q-adjectives can be related to properties of the scales of whose intervals they are predicated. The vagueness of Q-adjectives and their apparent cardinal/proportional ambiguity can be accommodated via the manipulation of two elements in the scalar representation, the structure of the scale (bounded vs. unbounded) and the location of the standard of comparison, with no need to posit multiple lexical entries. Aspects of scale structure are also responsible for contrasts in distribution among individual Q-adjectives (e.g. a few vs. *a many; the problems were many vs. *the difficulty was much). This thesis thus argues for the relevance of degrees and scales to the analysis of natural language meaning, while adding to recent work investigating subtle syntactic and semantic differences between superficially similar quantificational expressions, and developing compositional analyses of complex expressions of quantity. It further supports a view of nominal syntax in which functional elements contribute semantic content.