This dissertation presents a computational model of Optimality Theory (OT) (Prince and Smolensky 1993). The model provides an efficient solution to the problem of candidate generation and evaluation, and is demonstrated for the realm of phonological features. Explicit object-oriented implementations are proposed for autosegmental representations (Goldsmith 1976 and many others) and violable OT constraints and Gen operations on autosegmental representations. Previous computational models of OT (Ellison 1995, Tesar 1995, Eisner 1997, Hammond 1997, Karttunen 1998) have not dealt in depth with autosegmental representations. The proposed model provides a full treatment of autosegmental representations and constraints on autosegmental representations (Akinlabi 1996, Archangeli and Pulleyblank 1994, Ito, Mester, and Padgett 1995, Kirchner 1993, Padgett 1995, Pulleyblank 1993, 1996, 1998). Implementing Gen, the candidate generation component of OT, is a seemingly intractable problem. Gen in principle performs unlimited insertion; therefore, it may produce an infinite candidate set. For autosegmental representations, however, it is not necessary to think of Gen as infinite. The Obligatory Contour Principle (Leben 1973, McCarthy 1979, 1986) restricts the number of tokens of any one feature type in a single representation; hence, Gen for autosegmental features is finite. However, a finite Gen may produce a candidate set of exponential size. Consider an input representation with four anchors for each of five features: there are (2⁴ + 1)⁵, more than one million, candidates for such an input. The proposed model implements a method for significantly reducing the exponential size of the candidate set. Instead of first creating all candidates (Gen) and then evaluating them against the constraint hierarchy (Eval), candidate creation and evaluation are interleaved (cf. Eisner 1997, Hammond 1997) in a Gen-Eval loop. At each pass through the Gen-Eval loop, Gen operations apply to create the minimal number of candidates needed for constraint evaluation; this candidate set is evaluated and culled, and the set of Gen operations is reduced. The loop continues until the hierarchy is exhausted; the remaining candidate(s) are optimal. In providing explicit implementations of autosegmental representations, constraints, and Gen operations, the model provides a coherent view of autosegmental theory, Optimality Theory, and the interaction between the two.

This dissertation studies Bantu nominalizations drawing evidence primarily from Gikuyu and Bantu languages already in the literature. Gikuyu provides a very rich system of deverbal nouns which brings to the fore issues regarding word and phrase composition of deverbal elements, and the lexical integrity of words. Bantu nominalizations have received little attention in the literature in works such as Myers (1987), Kinyalolo (1991), Bresnan and Mchombo (1995). A very striking aspect of nominalized verbs in Gikuyu, (and Bantu) is that they bear both noun morphology (noun class marking), and verbal morphology (both inflectional and derivational). Deverbal nouns are many in Bantu languages and can not be taken to be idiosyncratic elements, without attempting to discover whether they are subject to principles that explain their large variety and numbers. In this study it is apparent that deverbal nouns do encapsulate the properties of nouns and verbs simultaneously. Upon nominalization Gikuyu shows that we get a set intersection of the properties of N and those of V. These properties are maintained from the sub-lexical level to the phrasal level. I propose that when the sub-lexical source of these nominalizations is established, it becomes apparent why deverbal nouns exhibit split category or mixed category status. This study also employs tests to check distributional and behavioral properties of all the items under scrutiny. I show that there are N/V ambiguous elements (infinitive/gerunds) and a family of mixed N/V elements whose category type can not be uniquely determined. Split and mixed category items challenge conventional premises of analysis which require every word to be associated to some unique category type. I have utilized the idea of Extended Heads in Grimshaw (1991) and Bresnan (1996) and the theory of Lexical-Functional Grammar (LFG) of Bresnan (1982), to account for the distinction between pure, split, and mixed category elements. Central to the analysis is the association of affixal information on words directly to grammatical function without first relating the affixes to category and projection.

This dissertation proposes a new model of Lexical Phonology: the Two-Level Theory (TLT). The TLT consists of dividing phonological rules into two different sets. One set is active at the Lexical Level, and the other set is active at the Postlexical Level. Lexical rules are active at the Lexical Level and, possibly, at the Postlexical Level, too (the choice being language specific). This new model is a simplification of previous models that hold that rules found at the Lexical Level can be further subdivided into other strata, producing multilevel representations. This new model of Lexical Phonology is tested with regard to a number of phonological problems in Pipil (Chapter 2), in Spanish (Chapter 3), and in Malayalam (Chapter 4). These analyses within the TLT demonstrate that the multiple levels of previous analyses of comparable phenomena within these languages are unnecessary.

The ultimate objective of this project was to link psycholinguistic theories of sentence production to the study of L2 acquisition. To achieve this goal, two steps were taken: first, some of the specifics of the grammatical encoding processes that occur during L1 Japanese sentence production were investigated; second, L2 Japanese learners' sentence production processes were examined. Because any L2 data that we can observe is necessarily the product of both speakers' underlying knowledge and their capacity to process a language, it is important to investigate language processes in order to fully understand L2 development. I studied two processes of both L1 and L2 speakers: (1) lexical retrieval processes for adjectives and adjectival nouns (both of which roughly correspond to English adjectives, but differ from each other in their morphosyntax), by using tip-of-the-tongue (TOT) experiments; and (2) case particle selection, by examining speech errors (e.g., L1 and L2 errors that occurred during picture description experiments, in addition to naturally occurring L1 errors). L2 Japanese speakers also participated in several other tasks, including tasks that assessed their knowledge of adjectives/adjectival nouns and case particles. It was found that both L1 and L2 speakers of Japanese had access to the word category information of a target adjective or adjectival noun during TOT states. Also, L2 learners' error patterns in a picture description task roughly matched their knowledge of the morphosyntax of adjectives and adjectival nouns. Thus, most of the morphosyntactic errors of adjectives and adjectival nouns were attributed to L2 learners' knowledge, rather than their lexical retrieval processes or processing strategies. However, L2 speakers' self-repairs and verbal reports indicated that some morphosyntactic errors occurred due to processing factors. The examination of L1 case particle errors indicated that L1 speakers may use certain sentence processing strategies, two of which involve use of the nominative marker ga (as default), and the selection of a prototypical case particle for a semantic role of an NP. L2 speech errors revealed that L2 Japanese speakers employ word order based processing strategies during sentence production, and that they may also use prototypical particles for NPs having given semantic roles.

A central question in the study of language learning is how humans acquire syntactic categorical distinctions among words (e.g. noun, verb, etc.). Past research using miniature artificial grammars suggests that semantic information is not needed for this learning; distributional information alone can provide adequate input for learning. The current experiments extended this finding to a natural language. Adults who had never studied Russian listened to lists of Russian words for seven minutes. The words consisted of a content morpheme and a grammatical ending. The participants were not told the meanings of the words. Next they were tested on a series of legal and non-legal morpheme sequences, including sequences that were not in the training. Results showed that participants were able to distinguish between new legal and non-legal morpheme sequences, provided there were at least two category-markers in the input. This suggests that they were generalizing the words into categories. A corpus study showed that Russian probably contains the kind and quantity of markings required for category learning to take place. Reaction times were also analyzed.

This dissertation investigates how interpretations are differentiated between and within Japanese focalizers (toritate-joshi) and also their English counterparts. There are four main proposals in this thesis. First, I propose that focalizer interactions must be analyzed independently from general focus interactions. The ordinary assumptions made for 'focus' (e.g. 'new information', binary distinction against 'topic', etc.) do not capture focalizer interactions accurately, and raise serious questions such as that of 'second occurrence' (Partee 1991). This thesis illustrates how focalizer interactions are differentiated from 'focus' interactions. Second, I suggest that the interpretation of focalizers consists of two major parts: 'declaration' and 'indication'. The two-way representation is largely inspired by Horn's (1969) 'presuppositional' analysis, but with a different distinction between interpretations. This thesis separates context-free and 100% warranted propositions ('declaration') from propositions that may or may not appear depending on the context ('indication'). Third, I assume that the interpretation of focalizers is differentiated by the strength of the indications. Adopting Leech's (1974) argument that the probability of a potential (lexical) meaning is relative to context, I further develop the assumption that the strength of the potential meaning of a given indication is relative to context. A variety of interpretations of focalizers results from this relative strength of indications. Last, it is thoroughly claimed throughout the thesis that the interactions between the target of a focalizer ('self') and its comparative items ('others') must be deeply considered to capture the three claims above. One of the important adjustments to the previously proposed analyses is that the 'scalar value' (e.g. Fauconnier 1975a, 1975b) relates to how the interpreter considers 'others', and is not fully determined by the lexical entries of a focalizer. A variety of factors are actually involved in the interpretation of focalizers. Therefore, potential discussion topics and observations of focalizer interactions are also displayed in detail, without analyzing them, for future study.

The ba-construction is probably the best-known syntactic construction in Modern Standard Chinese, but little has been done on the acquisition of it by second language learners. My study fills this gap. The theoretical framework is Liu's (1997) aspectual analysis of ba. The study is experimental. The constraint on the ba-NP and the constraint on the ba-VP were investigated. The constraint on the ba-NP is that it has to be specific. The constraint on the ba-VP is that it has to be complex. Two of the structures that can satisfy the complexity constraint on the ba-VP, the resultative verb complement (RVC) and the perfective aspect marker le were investigated. The subjects were 65 students learning Chinese in the intensive Chinese program at the Defense Language Institute (DLI) in Monterey, California. They were divided into three groups according to the number of weeks that they had studied Chinese at the DLI. A group of 20 native speakers of Chinese also participated as a control group. The study involved two experiments: production and grammaticality judgments. In the production experiment, I designed pairs of actions on video tape and asked the subjects to describe the actions with two Chinese sentences. In the grammaticality judgment experiment, I played some of the video clips from the production experiment and asked the subjects to tell whether some written Chinese sentences accurately described the actions that they saw in the video scenes. I also asked the subjects to indicate their confidence in their judgments. Results were that even though the learners generally produced fewer ba-constructions than the native speakers in the production experiment, their judgments of most of the sentences were not significantly different from those of native speakers, indicating that they had some good knowledge of the construction. Moreover, learners with similar production patterns showed different patterns in their grammaticality judgments, suggesting that the two kinds of data complement each other in our understanding of language acquisition. Results also show that the confidence dimension captured subtle differences that would not have been captured by the grammaticality dimension alone in the grammaticality judgment experiment. It has been found that using the canonical SVO word order and the drop of the object are the two most common patterns used by learners to substitute for the ba-construction. Error patterns have also been identified. Some "unexpected" patterns of grammaticality judgments from native speakers suggest that the learners had not yet been exposed to some interesting phenomena in the language that interfered with the ba -construction. One such example is the topicalization of the object to the position between the subject and the verb. Some native speakers accepted such sentences while most learners rejected them probably because they had not been exposed to such structures yet. Finally, the variation among native speakers in their judgments of the ba-construction calls for a more systematic study of exactly how native speakers of Standard Chinese use the ba-construction, especially those who have been exposed to other Chinese dialects.

This dissertation seeks to test recent important theoretical ideas in the Principles and Parameters and Distributed Morphology frameworks against data from the relatively under-studied Uto-Aztecan languages. In this work I focus on the morphology of reduplication, noun incorporation and related derivational morphology, and the diachronic development of the polysynthetic morphological type in one sub-branch of the family (Corachol-Aztecan). With respect to prosodic morphology, I argue that the comparative Uto-Aztecan evidence suggests that reduplicants should be viewed as morphological pieces, and I analyze them as Vocabulary Items inserted into syntactic slots at Morphological Structure. I also argue that the evidence of cognate reduplication patterns across Uto-Aztecan supports a prosodic view of morphology, as well as the constraint-ranking approach to morphophonology. With respect to noun incorporation and derivational morphology, I argue that the comparative Uto-Aztecan evidence supports the view that denominal verbs are a sub-class of noun-incorporating verbs. I survey the noun incorporation types in Uto-Aztecan and classify NI in these languages into four types: N-V compounding, syntactic NI, classificatory NI, and "object polysynthesis". I offer a unified syntactic account of these types, maintaining that each is formed via head-movement in syntax. I provide a novel approach to hyponomous objects, suggesting that these are in argument positions, and that they are derived via the Late Insertion of material that is not cognate to the incorporated noun, but which is inserted into the lower copy of a movement chain. Non-theme "nominal" roots incorporated into verbs, such as instrumental prefixes, are analyzed as adverbial elements Merged directly into the verbal position. Finally, I argue that this theoretical analysis of NI leads naturally to a diachronic account of the development of polysynthesis in Nahuatl. I show that the crucial aspects of polysynthesis, subject and object pronominal marking on the verb as well as syntactic noun incorporation, have analogues elsewhere in Uto-Aztecan, and I offer a reconstruction of the likely stages of the development of polysynthesis in Nahuatl, each of which have attestation elsewhere in the family.

This dissertation examines the phenomenon of consonant cluster reduction in young children's speech from both an experimental and a theoretical perspective. After first arguing that previous, articulatory accounts of children's cluster reductions are not satisfactory, I propose an alternative hypothesis based on Sonority Theory. Contrary to an articulatory approach which might predict that children reduce consonant clusters to whichever consonant is easier to produce, the Sonority Hypothesis predicts that children reduce clusters to whichever consonant produces the most optimal syllable. An optimal syllable is one that begins with a maximal rise in sonority from the initial consonant to the vowel and ends with a minimal (or no) sonority descent, where consonants are classified as more or less sonorous according to a Sonority Hierarchy. This hypothesis is then tested in two experiments where subjects were asked to repeat names for imaginary animals either of the form CCVC or CVCC. In this way, cluster reductions were elicited from children ranging in age from 29-36 months old. A post-test was also conducted on each child to ensure that both consonants of any given cluster were contained in the child's consonant inventory. Results of both experiments support the Sonority Hypothesis. Consequent to the experimental investigation, I examine several larger issues in language acquisition that are raised by this research, such as the importance of cross-linguistic and child language parallels in acquisition, and the question of variability in child data. This discussion raises the further question of how best to account for these types of disparate properties in child language. As a means of addressing these concerns, I present one possible approach by offering a complete phonological analysis of cluster reduction in an Optimality Theoretic framework. I then examine the success of this account with respect to the issues raised earlier. In concluding this dissertation, I suggest that by also considering the effects of performance factors on children's early productions we can arrive at a fully explanatory theory of phonological acquisition that addresses all of these significant issues.

This is a study of topicalization (OSV) and object preposing (SOV) in Mandarin Chinese. Whether topics in Mandarin are base-generated or derived by movement has been controversial. We provide a detailed discussion and show that they are derived by movement. Topicalization in Mandarin is generally assumed to be A'-movement. Based on Mahajan's (1990) criteria, we however find that it may be A or A'-movement for it exhibits reconstruction effects, lacks WCO effects and topics can bind anaphors. Yet we still argue that it is uniformly A'-movement based on the logophoric analysis of anaphor binding and the reconstruction effects. We explain the lack of WCO effects by adding the constraint "a trace v is a formal variable if the chain is triggered by a (+wh) / (+quant) feature" to Higginbotham' s (1983) Accessibility Condition. Topic traces are thus classified as A'-anaphors for topicalization is triggered by (+topic). With this feature-based definition, we admit a third type of position--A'-positions may be operator or non-operator positions, depending on the features they host. Topics in Mandarin are normally assumed to move to IP-adjunct position. We however suggest that they move to (Spec, CP) because: topicalization is triggered by the topic feature, the occurrence of an embedded topic depends on the verb that selects the embedded clause, and no topic is allowed after an adjunct complementizer. Object preposing in Mandarin is assumed to be movement to adjunct positions in some studies. We however find it movement to specifier positions because it is an A-movement triggered by the focus feature. We suggest that the preposed object lands on (Spec, ModalP) or (Spec, VP). We divide the preposed objects into (+C-focus) and (+focus) objects; (+C-focus) objects have emphatic markers or contrastive conjuncts but (+focus) objects do not. The focus features reside in Modal or V to trigger object preposing, with selectional restrictions; e.g., modals only select verbs with (+C-focus), which explains why objects in (Spec, VP) must have emphatic markers or contrastive conjuncts. Related phenomena such as clause-boundness of object preposing are also discussed.