I argue that English modal auxiliaries have an array of logical types: they function as sentence operators, as predicates, and as variable binding operators. Chapter One introduces the root/epistemic distinction and earlier treatments of it in the literature of generative grammar and model theoretic semantics. An extensive descriptive section in Chapter Two establishes precisely the empirical characteristics of the distinction, which are confined to the semantics. I then account for these characteristics by positing two semantic types for modal auxiliaries: the first type is that of a sentence operator and is at work in the interpretation of modal sentences with an epistemic sense and for those with a deontic 'ought-to-be' sense; the second type is that of a relation between a property and an individual and it is at work in the interpretation of modal sentences expressing ability, disposition, rights and obligations (traditionally, the root interpretations.) The link between the interpretive class of the modal sentence and the type of the modal expression in Intensional Logic is not drawn in the lexicon, but is accounted for by native speakers' intuitions about what sort of notions make sense as modal properties and what as modal propositions. In Chapter Three, I turn to another class of interpretations of modal sentences, exemplified by the sentence 'A basketball player can be short'. After demonstrating that the modal auxiliary functions as a variable binder in such sentences, I go on to expand the inventory of types associated with modal auxiliaries to include the type associated with variable binding operators.

This dissertation investigates the ways in which natural language restricts the domains of quantifiers. Adverbs of quantification are analyzed as quantifying over situations (instead of being unselective quantifiers). The domain of quantifiers is pragmatically constrained: apparent processes of "semantic partition" are treated as pragmatic epiphenomena. The introductory Chapter 1 sketches some of the background of work on natural language quantification and begins the analysis of adverbial quantification over situations. Chapter 2 develops the central picture of "semantic partition" as a side-effect of pragmatic processes of anaphora resolution. I argue that the apparent effects of topic/focus articulation and presuppositional information on the interpretation of quantifiers are not the result of a direct and local mechanism of sentence grammar. Instead, I develop an analysis where the link is established via the anaphoric dependence of quantifier domains on the discourse context. Chapter 3 discusses the analysis of conditional clauses as quantifier restrictors, concentrating on the question whether conditional clauses restrict quantifiers directly (like common noun phrases restrict determiner-quantifiers) or indirectly (like topic/focus restrict quantifiers). A treatment is explored which has if-clauses constrain the value of the hidden domain variable of the restricted quantifier. Chapter 4, on unless-clauses, and Chapter 5, on only if- and even if-clauses, present some issues in the compositional analysis of complex conditional clauses. These chapters significantly expand the data coverage of the theory of A-quantification. Building on previous work of mine on exceptives, I analyze unless-clauses as exceptive operators on A-quantifiers. The analysis of only if-clauses, treated as conditional clauses that combine if with the focus adverb only, unearthes some interesting new properties. Chapter 6, finally, examines the phenomenon of donkey-anaphora in the light of the results of the previous chapters. I show that a solution to the proportion problem may become possible once we combine the situation-semantic approach to adverbial quantification with the pragmatic theory developed in Chapter 2 and further elaborated in the analysis of donkey anaphora in complex conditionals.

According to an Aristotelian view maintained in modern linguistics, a sentence consists of a predicate, a subject and some indication of tense. This work attempts a principled and coherent account of the Aristotelian sentence, in the framework of modular generative grammar (Chomsky 1981). From the starting point that predication is a licensing relation--the $\Pi$-relation--a category-neutral theory of predication becomes possible. This in turn has consequences for two domains of grammar: 'inside' the predicate, and 'outside' of it. Part I is concerned with the inside of the predicate. A category-neutral theory of predication has consequences for category features, for the representation of language-particular categorial inventories and for the role of selection and adjunction in phrase structure. A detailed case study compares the phrase structure of English and Standard Yoruba. The category-neutral framework employs bivalent Lexical projections in the analysis of English adjunct predicates, coordinate structures and triadic predicates, as well as serial verb constructions (Chinese, Igbo, Yoruba, Haitian) and consecutive constructions (Korean and Japanese), and develops a set of principles of event composition therein. Part II is concerned with the outside of the predicate, in particular tense-marking. The notion that copulas are inserted to turn non-verbs into predicates is critiqued. As an alternative, language-particular restrictions on tense-marking, default agreement and the distribution of copulas, aspect and negation are reduced to categorial selection. A detailed analysis is given of Standard American English tense-marking, followed by case studies of African American English, Modern Hebrew and Haitian. The device of homophonous grammatical morphemes is rejected as a matter of principle and on empirical grounds. The study closes by examining the role of default agreement in inflection and auxiliary placement in Igbo and Yoruba. An appendix gives a metrical account of Igbo tonal syntax. Other languages discussed include Modern Standard Arabic, Standard British English, Child English and Russian.

This dissertation investigates the conditions under which the verb moves to the highest inflectional category. The description and the morpho-syntactic theory of V to I raising are initially based on those Germanic languages that have a left-headed verb phrase (English, Yiddish and the Scandinavian languages), although data from Romance languages (Italian, French, and European and Brazilian Portuguese) are later shown to support the same conclusions. V to I raising occurs in those languages where, in the singular or plural of at least one tense, subject-verb agreement uniquely marks the person features (1st) and (2nd) by overtly distinguishing the forms for first and second person from each other, the form for third person and the infinitive. This correlation can be explained if it is assumed that the lexicon contains all and only referential elements and the referential features of inflection are those for person (and number). The paradigms of inflectional affixes which distinctively mark (1st), (2nd) and (singular) are referential and hence listed in the lexicon. Inserted under the highest inflectional head, these affixes trigger V to I raising in order to be bound. The paradigms of inflectional affixes which fail to distinctively mark one of the referential inflectional features are non-referential and hence not listed in the lexicon. Affixes of these paradigms are postsyntactic spell-outs of the abstract agreement features on the verb. The highest inflectional head remains empty throughout syntax and V to I raising is not triggered. This theory is corroborated by the fact that in the history of English and Mainland Scandinavian, V to I raising was lost after one of the referential inflectional features ceased to be distinctively marked. The distinction between lexical and post-syntactic inflectional affixes also plays a crucial role in the theory of empty referential subjects. Projections must be identified via distinct phonological or semantic matrixes in their specifiers or their heads (Speas (1993)). Accordingly, the s-structural subject position (i.e. the specifier of the highest inflectional projection) can remain empty in languages with referential, lexical agreement affixes but it must be overtly filled in languages with non-referential, post-syntactic agreement affixes.

This dissertation deals with the possibility of having syllable final consonants, including geminate consonants, and the interaction of such consonants with syllable weight and vowel length, by means of the moraic tier. The goal of this work is to have the range of patterns of consonant occurrence and vowel length follow from simple patterns in simple interactions. A chapter is devoted to making predictions about the phonological patterns possible in the system. These are made through the interaction of simple phonological constraints, with the adoption of Optimality Theory. Since the constraints interact in different ways in different languages, they need not be complicated themselves. Still, they produce the full range of expected patterns. Of special importance is the fact that intersyllabic consonant clusters and geminates are predicted to cross-classify. Two chapters are devoted to more detailed analyses of the languages Finnish, Fula, and Koya. In particular, these languages differ on which types of over-filled or hypercharacterized syllables they permit. Because hypercharacterized syllables are more than full, they are situations where we can gage the limits of syllabic constituency. An adequate approach to syllable theory must predict not only the possibility of the syllable types found in these languages, but the variation in types found between languages. A chapter is devoted to appendix consonants, which are non-moraic non-onset consonants. Here it is proposed that only appendix consonants display the coronal asymmetries that have been described for codas in a variety of languages. Furthermore, discussion of the structural analysis of the appendix consonants is undertaken. The aforementioned phenomena are all underlying. Another matter taken up is the nature of gemination of consonants conditioned by prosodic factors and geminates which arise as a result of concatenation. The language Wiyot, which has no distinctive length, is analyzed using the same basic schema as the languages with distinctive consonant or vowel length. Gemination by concatenation is analyzed by means of Alignment constraints. Finally, this dissertation includes discussion of alternative representations for consonant sequences and geminates.

The thesis examines the phonological properties of three reduplicative morphemes in Lushootseed, a Central Coast Salish language. It is argued that both the shape and segmental content of these morphemes follow from their classification as either root or affix. Expanding on proposals by McCarthy and Prince (1994a), a model of reduplication is developed in which reduplicative morphemes are specified only for morphological category: Generalized Template Theory (GT). The central hypothesis is that root reduplicants are more marked phonologically than affix reduplicants. This is manifest in differences in size, shape, segmental content, and segmental copying. The segmental content of reduplicative morphemes (RED) is attained via a correspondence relation between reduplicant and base. Set within Optimality Theory (Prince and Smolensky 1993), constraints which evaluate the identity of base and reduplicant are ranked and violable. The structural conditions on RED are the result of base-reduplicant identity constraints interacting with phono-constraints. There are no templates. Chapter 2 proposes that the fixed shape of reduplicative morphemes is best explained as a set of structural conditions, formally represented as constraint interaction. The same constraint interaction will derive a CV-prefix and a -VC suffix. Both are affixes and are less marked than the CVC root reduplicant. Moreover, -VC shape is not template spellable, but its structural conditions are typical of other suffixes in Lushootseed. In Chapter 4, a pattern of fixed segmentism is analyzed as an instance of unmarked structure emerging in reduplication. A default segment occurs with both affixal reduplicants, whenever the base would supply marked structure. In contrast, the root reduplicant permits the marked structure, as predicted by GT. Some of the conditioning environments for the default segment (stress and syllable structure) are examined in Chapter 3. An examination of doubly reduplicated stems in Chapter 5 shows that the base must be an adjacent string, present in the output only. The different patterns of copying are explained because the root allows the marked option of copying both root and affix material, while the affixes only copy root material.

This dissertation provides evidence that the notion of maximality plays a central role in the semantics of wh-constructions, in particular wh-questions, comparatives, and free relatives. It is argued that the semantics of each of these constructions involves reference to the maximal element of a certain set. The way maximality manifests itself depends on the algebraic structure of that set. Degrees are ordered linearly and the maximal element of a set of degrees is therefore the highest degree in that set. In a set in which the elements are only ordered in a join semi-lattice the maximal element is the sum of all the elements in the set. In chapter 2, I show how the maximality account of comparatives can account for their semantic properties, in particular with respect to the distribution of negative polarity items and the interpretation of disjunction. Chapter 3 extends the maximality analysis to questions and free relatives. In free relatives maximality has a 'universalizing' effect, whereas in questions it gives rise to the exhaustiveness. In chapter 4 discusses a particular kind of questions, namely how many-questions, focussing on the scope interactions between the how many-phrase and other elements in the sentence. A common thread running through the dissertation is the interaction between wh-movement and negation. 'Negative' elements can block wh-movement in certain cases, a phenomenon which is known in the literature as the negative island effect. In the rest of chapter 1 the negative island effect is introduced and the accounts of it found in the literature are briefly discussed. In chapters 2, 3, and 4 it is shown how maximality can account for the negative island effect. In chapter 5 certain remaining issues are discussed including the question whether the maximality account of negative island effects can be extended to other kinds of 'weak' or 'selective' islands, in particular wh-islands. A central issue in this matter is the proper division of labor between syntax and semantics. The chapter also discusses the role played by pragmatics in the explanation of the negative island effect and a maximality effect induced by focus.

This dissertation is an attempt to discern the nature of liquid consonants. The evidence amassed here principally addresses questions of the featural content of liquids along with the hierarchical relations among those features. One of the major findings of this study is the phonological validity of the class of liquids, defined as rhotics and sonorant laterals. Further, liquids are actually differentiated from each other by place features, not by manner features. I demonstrate that there is no need to add manner features to the theory to distinguish among liquids. The topic of Chapter 2 is laterals. The objective of this chapter is to see what phonologically unites the various lateral types and also what divides them. We see that laterals are defined by a Corono-Dorsal place of articulation complex. Both coronal and dorsal activity is motivated phonologically and phonetically for all types of laterals. I also show in this chapter that the feature (lateral) is redundant and phonologically invalid. Chapter 3 takes on the nebulous group of r-sounds, or "rhotics". Rhotics have long been grouped together, but have rarely been examined for phonological classhood. The evidence presented in this chapter supports a phonological class of rhotics, despite the lack of a single defining phonetic characteristic. I further claim that all rhotics have a non-primary Laminal node. Chapter 4 brings together laterals and rhotics and addresses the question of whether there is a class which subsumes both groups. I claim that liquids are indeed a natural class defined by the major class feature (liquid). Rhotics and sonorant laterals, but not obstruent laterals, comprise the class of liquids. Further, segments within the class of liquids are distinguished from each other by place of articulation specifications. All liquids are required to have a complex, or branching, Place node. In this chapter, there is also a discussion of the mechanisms involved in liquid dissimilation. Chapter 5 is a summary of the findings in the dissertation along with the phonological representations of the various liquid phonemes.

This dissertation investigates the semantics and LF-representation of wh-phrases by attending to two phenomena: the effect of Focus in wh-phrases and reconstructed readings ofwh-phrases. First, I exploit the semantics of Focus to show how certain peculiarities of Sluicing follow without the need for special LF-operations, contrary to Chung-Ladusaw-McCloskey (1995). I claim that remnant wh-phrases in a sluiced interrogative clause usually bear focal stress and I define a set of alternative semantic values for a focused wh-Determiner. From this, two consequences follow: the remnant wh-phrase has to contrast with its correlate in the antecedent clause--which derives the restriction on possible correlate phrases--and the denotation of the Sluicing clause and the denotation of the antecedent clause have to be identical in certain respects--which derives the inheritance of content and islands cases. Second, I turn to the question of whether reconstructed readings of how many phrases and which phrases derive from Syntactic Reconstruction (SynR) or from Semantic Reconstruction (SemR). I present two challenges for the SemR account of reconstructed scope readings of how many phrases. First, numerous examples are provided that show that Principle C Connectivity correlates with reconstructed scope readings, a fact which is predicted under the SynR approach and unexpected under the SemR view. Second, I investigate reconstructed scope readings of how many phrases in VP Phonological Reduction and argue that they cannot be derived within the SemR line without generating unwelcome results. Two lines have been pursued in the literature to capture functional readings of which phrases: the choice function approach (Reinhart (1993, 1997)), which involves SynR of the restrictor of the which-phrase, and the skolem function approach (Engdahl (1986)), which amounts to SemR. Again, it is shown that Principle C Connectivity correlates with the embedding needed for the variable to get bound, which supports the SynR approach. However, I present two problems for the current implementation of the choice function SynR line: local presupposition accommodation readings are wrongly excluded, and intensional readings cannot be derived from transparent which phrases. I propose a new architecture of choice functions that derives these new data.

This dissertation presents Error-Selective Learning, an error-driven model of phonological acquisition in Optimality Theory which is both restrictive and gradual. Together these two properties provide a model that can derive many attested intermediate stages in phonological development, and yet also explains how learners eventually converge on the target grammar. Error-Selective Learning is restrictive because its ranking algorithm is a version of Biased Constraint Demotion (BCD: Prince and Tesar, 2004). BCD learners store their errors in a table called the Support, and use ranking biases to build the most restrictive ranking compatible with their Support. The version of BCD adopted here has three such biases: (i) one for high-ranking Markedness (Smolensky 1996) (ii) on for high-ranking OO-Faith constraints (McCarthy 1998); Hayes 2004); and (iii) one for ranking specific IO-Faith constraints above general ones (Smith 2000; Hayes 2004). Error-Selective Learning is gradual because it uses a novel mechanism for introducing errors into the Support. As errors are made they are not immediately used to learn new rankings, but rather stored temporarily in an Error Cache. Learning via BCD is only triggered once some constraint has caused too many errors to be ignored. Once learning is triggered, the learner chooses one best error in the Cache to add to the Support---an error that will cause minimal changes to the current grammar. The first main chapter synthesizes the existing arguments for this BCD algorithm, and emphasizes the necessity of the Support's stored errors. The subsequent chapter presents Error-Selective Learning, using cross-linguistic examples of attested intermediate stages that can be accounted for in this approach. The third chapter compares ESL to a well-known alternative, the Gradual Learning Algorithm (GLA: Boersma, 1997; Boersma and Hayes, 2001), and argues that the GLA is overall not well-suited to learning restrictively because it does not store its errors, and because it cannot reason from errors to rankings as does the BCD. The final chapter presents an artificial language learning experiment, designed to test for high-ranking OO-faith in children's grammar, whose results are consistent with the biases and stages of Error-Selective Learning. .