This dissertation investigates the use of null and overt subject pronouns in Cairene Arabic. The data for the study are: (1) a corpus of 1944 clauses from a movie; and (2) responses of 6 Egyptian consultants regarding their preference for the use/non-use of subject pronouns in those clauses. In indicative sentences, overt pronouns are used about one-third of the time with both first and second person subjects (31.7% and 29.8%, respectively). The percentage for third person subjects is only 6.4%. The claim, made frequently over the last 20 years, that null subjects can only occur in Arabic with predicates that have person marking is shown to be empirically incorrect. Numerous examples of null anaphora with predicates lacking person marking are presented and I argue that the information processing strategies needed to recover the intended subject with such predicates is not qualitatively different from strategies needed to recover null subjects with predicates that do have subject person marking. Referential ambiguity across predicate types was analyzed in order to determine whether or not this ambiguity was playing a role in motivating more overt pronoun use. I found no evidence that this ambiguity leads to increased overt pronoun use. The traditional explanation of overt pronoun use in terms of expressing "emphasis" and/or "contrast" is not illuminating. These terms are not semantic or pragmatic primitives and do not explain the underlying discourse circumstances that lead to overt pronoun use. I explain the use of overt pronouns as resulting from an extra "layer" of meaning implicit in the utterance with the overt pronoun and argue that the added meaning can frequently not be conveyed without the overt pronoun. Once speaker illocutionary intent is taken into consideration, the use of an overt pronoun frequently becomes pragmatically obligatory. The sources of that added, often indirect, meaning are identified through a careful analysis of examples in the corpus which consultants indicated were pragmatically obligatory. In addition, an information processing function of "predicate signaling" is proposed for overt pronouns occurring with definite NP predicates that could be interpreted as new subjects of predication, in the absence of an overt subject pronoun.

This dissertation examines the phonology and phonetics/phonology interface in Coatzospan Mixtec (CM). I focus on two major prosodies, glottalization and nasalization, in CM. First, I provide detailed phonological analyses of both within the context of Optimality Theory, OT (Prince and Smolensky 1993). This is important because often the treatment of a subset of data obscures more problematic aspects of a system. For example, the analysis of nasalization extends our understanding of how constraints can combine in a grammar. I motivate the conditional union of two Alignment (McCarthy and Prince 1993a) constraints to characterize attested patterns of root nasality, while ruling out impossible forms. The treatment of glottalization explores the implications of freedom of input in OT. I show that we cannot equate input with underlying; encoding the traditional sense of underlying representation requires viewing UR's as sets of optimal inputs lexical items. Regarding the phonetics/phonology interface, I pursue dual goals. Chapter 3 extends Grounding (Archangeli and Pulleyblank 1994a) to the opportunistically grounded relation between glottalization and stress. Although not inherently sympathetic to stress, glottalization is optimally realized under stress in the phonology of CM. Chapter 4 extends grounding by using sequential grounding (Smolensky 1993) to characterize the behavior of opaque consonants. Second, building on research in phonetic implementation (Pierrehumbert 1980, Keating 1990b), I show that a phonologically specified (+constricted glottis) must be implemented for only a part of the duration of the specified vowel. Similarly, orality targets in CM fricatives are also implemented at segment edges. The data support a view where targets are temporally located within segments (Huffman 1989). However, the location of targets may vary from edge to edge. Voiced fricatives implement orality upon release; voiceless fricatives do so at the onset of closure. The data also argue for a more complex notion of the relationship between phonetic data and phonological information than that of Cohn (1990). Partial implementation of a feature in a segment does not entail the phonetic rather than phonological presence of that feature. Phonetic data must be interpreted in the context of the phonological system from which they derive.

In this dissertation I show that the representations in Optimality Theory must be extended to allow multiple planes. The variable-position affixes in Afar occur as either prefixes or suffixes depending on the initial segment of the verb root. If the root begins with (i), (e(e)), (o(o)) or (u), the affix is a prefix (e.g., t-okm-e# (2-eat-perf) 'You (sg.) ate'); if it begins with (a) or a consonant, the affix is a suffix (e.g., rab-t-e# (die-2-perf) 'You died'). Additionally, plural not only appears as a prefix or a suffix, but when a suffix it can either precede or follow aspect (rab-n-e# (die-pl-perf) 'I died' vs. rab-e-n# (die-perf-pl) 'They died'). A parallel model in Optimality Theory is unable to account for the different order of affixes in forms such as rab-n-e# vs. rab-e-n#. The Multiplanar Model, which posits that output representations consist not only of a word plane but also an affix plane is able to account for this data. The representations for the two forms are: (y-e-n), (rab-e-n) and (n-e), (rab-n-e). In the first case, plural is specified as the rightmost morpheme by morphological constraints. Even though /y/, the third person marker, cannot surface on the word plane, it satisfies scONSET on the affix plane. This contrasts with the second case, where scONSET, being higher ranked than scPLURAL (scR), requires that plural occurs to the left of aspect to fill the onset position on the affix plane. I then show a serial monoplanar model can also account for this data. Finally, I compare the serial and multiplanar models, arguing that phonological evidence supports the Multiplanar Model.

This dissertation investigates the geminate consonant phenomena known as integrity and inalterability with an eye toward providing a general characterization of geminate behavior as well as a deeper understanding of geminates in a principled and systematic way under the Optimality Theoretic framework. The fundamental proposal made in this dissertation is to have the range of surface geminate patterns follow from varying the ranking of key constraints. Depending on the ranking of the key constraints, languages select different output forms from the same input form. Thus, the key constraints not only conspire to produce anti-integrity/anti-inalterability effects, but they also determine what a language does "do" with its input geminates (i.e. integrity/inalterability), giving rise to different resolutions to the geminate puzzle. A chapter is devoted to an indepth discussion of integrity effects in geminates. For this purpose, seven key constraints are proposed: MAX-IO, DEP-IO, ONS, PROSHIER, ALIGN(WD-R, M-R), PLONS, NOBREAKING. By varying the key constraints, we can make several predictions about the possible geminate patterns according to the positions in which they occur. Several patterns are exemplified in this dissertation. We have also provided a discussion of the gaps between what is predicted to exist and what cases are attested. Another chapter is devoted to a more detailed analysis of inalterability effects in geminates. In particular, it is claimed that geminate inalterability matters only when we deal with weakening processes (e.g. spirantization, sonorantization, etc.). It is also proposed that the constraints IDENT-IO(μSF) and NOBREAKING play a pivotal role in explaining typological differences between weakening and assimilation, and other types of inalterability/anti-inalterability concerning geminates. The most interesting part of this dissertation is that we can explain both integrity/inalterability and anti-integrity/anti-inalterability cases uniformly depending on the ranking of the key constraints, without assuming any ad hoc conditions or procedures. Thus, those anti-integrity and anti-inalterability effects are produced as a natural consequence of the interaction of the constraints, just as in the cases of integrity and inalterability. Finally, unlike previous rule-based approaches, our theory allows for a unified account of integrity and inalterability through the interaction of a set of key constraints, making predictions available about both phenomena.

The Goodman Reading Model differs from the word recognition model on the issues of (a) whether reading depends on perception of every single word; (b) whether prediction is used in the reading process; and (c) whether reading comprehension depends on individual words. The study tested the validity of the two models by investigating the reading process in Chinese through error detection. Two hundred subjects with equal numbers of college and middle school students participated in the experiment. The subjects at each educational level were randomly divided into error and meaning focus groups. The error focus groups were instructed to search for errors embedded in a Chinese text and the meaning focus groups to read for meaning of the same text within limited time. Then they were asked to recall the errors detected and the contents of the story in writing and to answer a questionnaire. After that they were given unlimited time to search for as many errors as they could. The main results showed that (a) all subjects failed to detect half of the errors under limited exposure and all errors under unlimited exposure; (b) the error focus subjects detected significantly more errors than the meaning focus subjects under limited exposure, but the meaning focus subjects scored significantly higher than the error focus subjects in recall of the story; (c) there was no significant difference between reading times in the number of errors detected but in the scores for the recall of the story; (d) the college subjects performed significantly better than the middle school subjects in error detection and reading comprehension; (e) more errors were detected in the contents word category than in the function word category; and (f) some extralinguistic factors had effects on the task performance. The results suggest (a) that characters and/or words are not recognized in a linear process in reading; (b) prediction is used under the influence of knowledge of various kinds; and (c) reading comprehension employs words but does not depend on individual words. The Goodman Reading Model is validated and proved applicable to reading in Chinese.

This work provides facts about word order in this language. In addition to listing the permissible word orders in matrix clauses and simple Wh-questions, I also document the semantic changes which result from the variation in word order. These facts of word order are analyzed within the Minimalist Program. Throughout this work I support the concept of no optionality and that movement has to be well-motivated. I propose that overt movement is not only morphologically motivated, as claimed in the Minimalist Program, but it is also semantically motivated. Such a proposal better accounts for the word order facts, both structural and semantic. In this proposal, I do not have to resort to optionality to explain the variation in word order. I look at the semantic difference between SV and VS word orders. In the SV order, an entity is chosen and a statement is given about it; in VS order, an event is displayed which has an entity that takes part in it. This variation in word order reflects the thetic/categorical distinction. In these word order, the verb raises overtly to the functional heads: T and AGR. This is an example of an overt movement which is morphologically driven. The subject in preverbal position in Gulf Arabic has to be information that is previously given in the discourse while if the subject is in postverbal position, it could either be new or given. Hence, I propose that the subject raises to be topicalized. Overt movement of the subject is not motivated by morphology but by semantics. Accordingly, Case and nominal agreement features are considered weak and they do not motivate the subject to move overtly. In Gulf Arabic, argumental Wh-phrases can be both fronted and left in-situ. I propose that the fronted argumental Wh-phrases raise to (Spec,CP) to be focused. Focus movement is an example of a semantically driven movement. Adjunct Wh-phrases in Gulf Arabic are base-generated in (Spec,CP).

Navajo is a Southern Athapaskan language spoken by approximately 160,000 people in Arizona and New Mexico. This dissertation examines the syllable structure alternations in the prefixes of the Navajo verb. Specifically, the distribution of open and closed syllables in the verbal prefixes are at issue. This distribution is seen to follow from the interaction of constraints on phonological well-formedness including Markedness, Faithfulness and Alignment constraints, under Optimality Theory. The dissertation makes the following empirical and theoretical points. Empirically, the analysis is based on a description of the surface forms of the Navajo verb, without recourse to diachronic or comparative data. In this respect, the analysis is in line with the kind of reasoning that would have to be undertaken by the language-learner, and which must form some part of the phonological knowledge of native speakers of this language. Furthermore, the analysis is undertaken without reference to the specialized terminology which permeates linguistic analyses of Navajo in particular, and of the language family in general. Thus the description and analysis of the data are presented in such a way that a non-specialist in the language family might understand the data and analysis. Theoretically the analysis shows the utility of Optimality Theory in dealing with the complex interactions between morphology and phonology that characterize this language. It is demonstrated that the interaction of cross-linguistically motivated constraints on well-formedness results in the attested surface patterns. It is further argued that this analysis fills out a typological prediction of Optimality Theory by attesting one of the possible rankings of Markedness, Faithfulness and Alignment constraints. The basics of Navajo syllable structure, and of Optimality Theory are presented first, followed by a discussion of the fundamental morphological and phonological properties of the system. A set of Navajo verbal paradigms is then analyzed and input forms of the morphemes which participate in syllable structure alternations are derived. The Optimality Theoretic analysis is presented, and the dissertation concludes with a discussion of the issues raised, and of a set of alternative analyses of the same data.

This dissertation investigates the phenomenon of dissimilation from a theoretical perspective, with special attention to crosslinguistic patterns. After first arguing that the previous accounts based on the Obligatory Contour Principle (OCP) (Leben 1973, McCarthy 1979, 1986) are not satisfactory, I propose an alternative theory of identity avoidance, G sc ENERALIZED OCP (GOCP) which generalizes the applicability of the traditional OCP to a wider range of phenomena, not just autosegmental (i.e. featural) ones. My proposal asserts that identity avoidance between two elements in sequence is fundamental to linguistic theory, an idea that can be characterized by a universal constraint governing various types of dissimilatory phenomena. This concept is implemented within the framework of Optimality Theory (Prince and Smolensky 1993, McCarthy and Prince 1993a,b), which provides the flexibility for constraints to be both violable and rankable. Contrary to the traditional OCP based approach which is bound by various representational properties such as feature geometry and underspecification, the proposed approach abandons this representational dependency in favor of the richly articulated constraint-based system. Based on the data collected from 57 language cases, I then examine the various factors that play a role in dissimilation, including the elements involved, their adjacency relations, and the domain of dissimilation. I demonstrate that the GOCP constitutes a consistent formal apparatus on the one hand, and the versatility to accommodate the complexity of dissimilation patterns on the other. Moreover, it is shown that the present approach formally unifies the characterization of both the similarity effects and blocking effects by directly incorporating Local Conjunction as a uniform mechanism of accounting for the Similarity effect, OCP-subsidiary feature phenomena. As a result, these phenomena need not require novel theoretical devices for each case, but rather are construed as instances of the combination of multiple GOCP constraints.

This dissertation is an analysis of the consonantal phonology of Gosiute, a member of the Numic group of Uto-Aztecan languages. The Numic languages are characterized by consonant alternations and distributional patterns which are rooted in patterns of phonetic naturalness. In this dissertation I provide an analysis of these patterns of distribution and alternation within the framework of Optimality Theory (Prince and Smolensky 1993). This dissertation accomplishes three things. First, it provides the most detailed treatment to date of the consonant system of a Numic language. Second, it demonstrates the efficacy of Grounding Theory (Archangeli and Pulleyblank 1994) in the analysis of the consonantal patterns under investigation. Third, it shows that Optimality Theory is up to the task of providing a framework for the analysis of large portions of the phonological system of a single language.

Recent analyses in reduplication have questioned the viability of template constraints to account for reduplicant shape in Optimality Theory. Such template constraints require the mapping of a reduplicant to prosodic unit such as the foot, syllable, or mora. Such template constraints make incorrect predictions regarding the types of reduplicative patterns and incorrectly match morphological types to prosodic types. The alternative is to eliminate template constraints and allow the shape of reduplicants to be determined by more general structural constraints in language. In this dissertation, I make two major contributions to this body of work. One major contribution is the presentation of data regarding bare-consonant reduplication (Semai, Marshallese. Coushatta, Yokuts, Secwepemc). In this data, reduplicants surface as a copy of a single consonant (C) (eg. Marshallese yibbiqen 'chunky (distributive)') or a string of two consonants (CC) (eg. Yokuts giy'igyifta 'touch repeatedly'). The reduplicants in these data are not clearly delimited by a prosodic unit, and therefore, provide support for the position that template constraints are not only undesirable, but empirically inadequate. The second contribution to this body of work is an alternative method of analysis that accounts for reduplicant shapes by the interaction of constraints that are independently necessary to account for the ordering of morphemes in a morphologically-complex form. Under this proposal, reduplicants are "compressed" between morphemes and the edges of the morphological word. This compression model uses constraints of the Generalized Alignment schema of constraints (McCarthy & Prince 1993b). The model is more empirically adequate than alternative a-templatic analyses. The compression model is extended to cases of reduplication in which the reduplicant is not a consistent prosodic unit across a paradigm (Hopi). Also this model is shown to be adequate to account for cases of reduplication that are more transparently matched to a prosodic unit (Ilokano). Such extensions of the compression model make predictions about types of non-concatenative morphology that have empirical evidence.