Generative analyses traditionally acknowledge that comparative formation (Millhouse is better prepared today than Ralf was yesterday) potentially involves two types of ellipsis unattested in other contexts: Comparative Deletion (CD; Bresnan 1973), which elides the gradable property d-good , and Comparative Ellipsis (CE; Pinkham 1982), a deletion operation optionally applying to categories (prepared) in the comparative complement. Based on an investigation of structurally complex NP-comparatives, I present a syntactic and semantic account of comparatives which does not require reference to construction specific processes. According to the AP-Raising Hypothesis of CD, CD consists in feature-driven movement of the gradable property from the CD-site to its surface position. The analysis rests on two assumptions: First, CD is a syntactic process, and not a manifestation of semantic ellipsis (e.g. Kennedy 1997). Empirical support for this view derives from the specific scopal properties of the CD-site and categories inside the CD-site, which are tested on the basis of reconstruction and CSC-effects, WCO in attributive comparatives, and the distribution of de dicto readings for the CD-site. Second, I assume that the CD-site is associated with its antecedent by overt movement, not by ellipsis, and that the resulting chain is embedded in a right-branching phrase marker. This conception contributes to an understanding of a number of puzzles for NP-comparatives related to binding and locality as well as to the influence of word-order on the size of the CD-site and AP interpretation (Siegel 1976). The second objective of this dissertation consists in defending the Conjunction Reduction (CR)-Hypothesis of phrasal comparatives, which posits that phrasal comparatives derive from CD plus the application of Gapping, RNR and/or ATB-movement. The CR-Hypothesis subsumes the effects of CE under ATB-movement and CR, and rests on the premiss that the syntactic structure of reduced comparatives overlaps at one point of the derivation with the convergence of properties which defines coordination. Among others, the analysis captures the complex interdependencies between CE and serialization, and provides an explanation for the impact of the structure and size of ellipsis on the syntactic scope of the remnant. This latter observation presents a strong argument against competing direct approaches.

This dissertation analyzes the rich agreement system in Bafut DPs. I argue that we need two structural configurations: Spec-Head and Head-Head to account for DP agreement. The Head-Head relation is needed to check agreement on lexical categories via covert raising of features of adjectives and genitives, in a head-to-head fashion, to the noun in Num0. Features of functional categories (determiners and quantifiers) are checked via movement of NumP to Spec-DP through Spec-QP in the familiar way of Spec-Head agreement. This approach, which extends to Romance and Bantu DPs, extends Chomsky's (1995) proposals for checking Φ-features on arguments in clauses to the checking of non-argument agreement relations in the DP and suggests a way of dealing with a really rich agreement system without resorting to the projection of agreement phrases. The account of agreement follows naturally from the internal syntax of Bafut DPs. I argue that variations in distribution and interpretation of constituents of the DP results from movement to positions that correlate with different interpretations. In the structure, NP is embedded inside four functional projections: FocP, DP, QP and NumP. N-raising to Num0 and subsequent movement of NumP to Spec-QP and Spec-DP yield the unmarked noun-initial word order commonly attested in Bantu. Further raising of functional heads to Foc 0 produce the contrastive focus interpretation when functional heads exceptionally precede the noun. To have a broader picture of DP structure in Bafut, the dissertation also examines the structure of derived nominals. I argue that deverbal nouns and type I negative nominals are formed in the lexicon and inserted into the structure as plain nouns while Poss-ing gerunds and type II negative nominals are syntactically derived from an embedded VP. This account of derived nominals supports both the lexicalist and syntactic views on nominalizations. Although the dissertation does not draw explicit comparisons between DPs and clauses, intuitive parallels between them cannot be ignored. For instance, N0-to-Num0 raising accounts for agreement in DPs just as V0-to-I0 raising accounts for subject-verb agreement in clauses. To the extent that such intuitive parallels can be further documented, we may not need to posit a syntax of DPs distinct from that of clauses.

This dissertation examines the influence of morphological factors on lexical stress and pitch accent. Two basic types are recognized. In root-controlled accent, inherent accent in a root overrides inherent affix accent; as a result, affixal accent is only realized in words with unaccented roots. In affix-controlled accent, the presence of a particular affix triggers one of several accentual mutations in the base: deletion of accent, insertion of an accent (often known as pre- or post-accentuation), and accent shift or “flop”. I argue that these two types of accentual behavior, despite important differences, are united under the rubric of faithfulness constraints in Optimality Theory. Root-controlled accent is a consequence of the privileged faithfulness status of roots over affixes, as has been shown in other empirical domains such as vowel harmony. Affix-controlled accent is due to a novel type of constraint, anti-faithfulness, which evaluates a pair of related words and requires an alternation in the base of affixation. The principal case of root-controlled accent studied in this dissertation is the Uto-Aztecan language Cupeño. In addition, I show how the accentual systems of Japanese and Russian fall within the scope of root faithfulness constraints. The study of these cases leads to a substantive restriction on the range of edge effects in accent systems, and clarifies a role for root accentedness in blocking morpho-accentual processes. A number of properties of affix-controlled accentual processes are identified and shown to follow from the anti-faithfulness thesis. Affix-controlled accent is (I) morphologically triggered, (II) stem-mutating, and (III) grammar dependent. (I–II) follow from the assumption that anti-faithfulness operates on related words: forcing an alternation in a pair of words ensures that affix-controlled accent is morphological because it contrasts two word classes. Furthermore, anti-faithfulness only affects the interval of a word which occurs throughout a paradigm, namely the stem (II). Finally, anti-faithfulness does not fully specify how a set of words should differ accentually; the specific effects of anti-faithfulness therefore depend on the larger grammar in which it is embedded (III). Affix-controlled accent in Russian, Japanese, Cupeño, Limburg Dutch, and Aguaruna (Jivaroan) is investigated in a series of case studies. I argue that anti-faithfulness constitutes an integrated theory of the diverse morpho-accentual phenomena found in these languages, explains the important differences between the accentual properties of affixes and roots, and establishes parallels with non-accentual affix-controlled phenomena.

This dissertation studies the second language acquisition of SER and ESTAR in native English speakers and also the acquisition of TO BE in native Spanish speakers. Through two experiments a sample of students is tested on their performance in translation tasks from their native language onto their second language. The analysis of their performance is focused on the syntactic and semantic aspects of the responses. A working framework of SER and ESTAR is developed based on the data that was obtained in the two experiments and the subsequent statistical analysis. It is observed that there is a distinct difference in performance depending upon the direction of the translation task which is proportional to the degree of competence and the native tongue of a particular speaker. Finally both SER and ESTAR are compared to TO BE at a syntactic and semantic level and a final model is presented that explains the performance of the speakers in the sample as well as the underlying linguistic mechanisms.

The term ellipsis, broadly construed, applies to syntactic structures that seem to host obligatory positions not filled with overt lexical material. More narrowly construed. the term refers to a hypothetical grammatical process that deprives lexical material of its phonetic content. This dissertation explores ellipsis in both the broader and the narrower sense of the term. The three questions it addresses, with respect to different constructions., are the following. First, does the syntactic structure host a relevant silent position to begin with, and hence is it a case of ellipsis even in the broad sense of the term? Second, assuming the presence of silent structure, how does this structure come to be silent? Specifically, does it involve a grammatical process of ellipsis in the narrow sense? Third, assuming a process of ellipsis is at work, what are the constraints on this process? Chapter 1 lays out these issues in greater detail and gives an overview of the dissertation. Chapter 2 addresses the third question formulated above. Specifically, it examines the effects of ellipsis in what I call reduced conditionals in German. a construction not previously analyzed in the literature. It is argued that in these cases, the ellipsis process has the effect of forcing an index on its target to be anaphoric. Chapter 3 and chapter 4 deal with the first question by arguing that certain coordinations previously analyzed in terms of movement to the left of a coordinating particle are instead to be analyzed in terms of silent positions to its right. Chapter 3 makes this point with respect to disjunctions of the form… either…or, while chapter 4 analyzes certain asymmetric coordinations in German. Chapter 5 addresses the second question stated above. Specifically. it presents new evidence for the view that silent verb phrases in cases of so-called VP-ellipsis are not in fact the result of a process of ellipsis, but rather are silent proforms with the semantics of bound variables.

This dissertation is concerned with split-ergative and related systems in which multiple case marking patterns are found within a single language. The central goal of the dissertation is to investigate the conditions under which different cases are assigned, and the relationship between case-assignment and the structural positions of arguments. This investigation is carried out within a version of Principles and Parameters Theory. Central to my argument is the idea that NP arguments possess features which determine their discourse prominence. These features include lexico-semantic features such as definiteness, animacy, and person, and morphosyntactic features related to aspectual interpretation. Ideally, the mapping of arguments onto argument positions should ensure that the structural prominence of NP arguments accurately reflects nonstructural notions of prominence. This, however, is not always possible since difference notions of prominence will not always coincide. I argue that the fundamental difference between structural and lexical (inherent) case lies in the types of NPs which they are capable of licensing. Lexical case is capable of licensing a wider range of NPs than structural case. Structural case will always be capable of licensing NPs whose prominence is accurately reflected in structure. However, in many languages structural case will be insufficient to license arguments whose prominence conflicts with their structural position. In such instances, lexical case must be employed. Split case marking systems arise in precisely those cases where a language is sensitive to prominence conflicts of a particular kind. I also argue that structural prominence must be construed as a relational notion rather than an absolute one. NPs achieve structural prominence or nonprominence by virtue of their relation to another NP. A consequence of this is that the subject of an intransitive verb will be treated as neither prominent nor nonprominent. As a result, split case marking will often cause transitive clauses to behave differently from intransitive ones, resulting in ergative marking. In languages where ergative case can be assigned to intransitive subjects, features of the verb may enter into prominence relations with NPs, providing an account for active case marking systems.

The present study has investigated the role of typological markedness in the processes of language change and language acquisition. In particular, the study focuses on changes within generations of speakers in the Puerto Rican community of Western Massachusetts regarding subjunctive mood, and the acquisition of this particular feature by speakers of Spanish as a second language. According to the typological school, markedness refers to universal categories that hold cross-linguistically (Greenberg 1966) in which certain functions are more frequent (less marked) and others less frequent (more marked). Studies in typology have favored the claim that the hierarchy of markedness of a particular feature relates to the hierarchy in which that feature is acquired as a first language (L1) as well as a second language (L2). Studies on bilingual communities in the United States have concluded that there is a change in progress among generations of speakers and that changes in a particular feature are related to the process of acquisition of that feature. The present study has questioned the extent to which these conclusions apply to modal selection in Spanish and the constraints that markedness might impose on the processes of language change and second language acquisition. Subjects (N = 87) were divided into two groups: (1) speakers of Spanish as L1, and (2) speakers of Spanish as L2. The first group is divided into generations of speakers. Knowledge of subjunctive use was measured by a grammaticality judgment test. The results of the test were statistically analyzed through logistic regression. The results of the statistical analyses for the grammaticality judgment test showed some evidence that in the process of recognition of the aspects that govern modal selection, L1 formal features of modal selection that appear later in the acquisition process are the first to weaken if the speaker is immersed in a L2 environment, and that this process is constrained by the marked character of some uses with respect to others. No conclusive evidence has been found to support the claim that L2 acquisition is patterned by markedness. However, the results indicate that in the second language acquisition process, markedness acts as a subconscious cognitive strategy.

In this dissertation, I develop a model of word segmentation in which systematic grammatical knowledge guides division of the speech stream into words. When the speaker's intended syllabification is unambiguously signaled by allophonic variation and phonotactic constraints, this information is used to segment the input. However, in the absence of phonotactic and allophonic cues to word boundaries, listeners still assign structure to the incoming acoustic signal. Language-specific rankings of a small set of universal constraints on syllable well-formedness are used to determine privileged alignment points for lexical search. As soon as a syllable onset is identified, the cohort of words consistent with that syllable onset is activated. This is a more efficient segmentation strategy than initiating lexical access at each phoneme, since a syllabic strategy results in comparatively fewer wasted access attempts. Supporting evidence for the grammatical model of word segmentation is presented in a series of wordspotting experiments. English listeners are shown to resolve allophonic and phonotactic ambiguity by using stress to determine syllabification. A stressed syllable can attract one or more consonants into its coda if followed by a stressless syllable, otherwise onsets are maximized. The Metrical Segmentation Strategy (Cutler & Norris, 1988) fails to account for these results since it ignores the effect of stress on syllabification. An important difference between the grammatical model and other current models of word segmentation, such as TRACE and Shortlist, is the claim that listeners use the grammar to parse the input into syllables, even in the absence of statistical and acoustic cues. TRACE does not recognize any level of structure between the phoneme and the word. Although Shortlist recognizes explicit cues to word boundaries, such as phonotactics, allophonics, and vowel quality, when such cues are absent lexical access is attempted at each phoneme.

This dissertation addresses whether synchronic phonology encodes phonetic naturalness. I argue that while phonological grammars allow phonetically unnatural patterns as a result of coincidental historical changes, they are nevertheless biased toward phonetic naturalness. The key observation revealed by experimentation is that phonetic naturalness emerges when speakers create novel phonological patterns. I refer to the observation as "the emergence of phonetic naturalness". To explain the emergence of phonetic naturalness, I propose within Optimality Theory (Prince and Smolensky 2004) that the universal set of constraints is phonetically natural, and constraint rankings are also phonetically natural by default. To accommodate unnaturalness, individual languages can create unnatural constraints and reverse default constraint rankings. Chapter 1 lays out the issues and Chapter 2 develops a grammatical model. Chapter 3 discusses devoicing of voiced geminates in Japanese loanword phonology. Only voiced geminates, but not voiced singletons, can devoice to dissimilate from another voiced obstruent. I argue that the difference in neutralizability is grounded in a perceptibility scale related to the voicing contrast in singletons and geminates, but that the cause of devoicing, OCP(voice), is phonetically unnatural. Chapter 4 discusses a novel pattern of Japanese mimetic gemination. When speakers create emphatic forms, they prefer geminate stops the most, geminate fricatives next, and geminate nasals least. The preference hierarchy harmonically aligns with the perceptibility scale for singleton-geminate contrasts for different manners of consonants. Chapter 5 examines the distribution of vowels that are correspond with Ø in novel Japanese rap rhyming. The likelihood of vowels being treated as Ø directly correlates with their perceptual similarity to Ø. Chapter 6 shows that in rap rhyming the more perceptually similar two consonants are, the more likely they are to rhyme. The patterns studied in Chapter 5 and 6 show that phonetically natural patterns emergence in novel verbal art. The case studies suggest that when speakers create novel phonological patterns, they deploy phonetic knowledge, and hence phonetically natural patterns assert themselves.

I explore first language acquisition of scalar implicatures as a class of meanings. I begin by proposing a question-under-discussion (QUD) account of scalar implicatures in which the role of Relevance in the generation of scalar implicatures is clearly articulated. Next, I propose an account of how the child constructs scales that SIs are based on, and suggest ways in which the child's semantic and pragmatic knowledge unrelated to scales, such as that of polarity, aids her in constructing scales. My experimental results did not support the theoretical distinction between Horn and pragmatic scales that is often made in the literature; I found that the time-line of the acquisition of SIs is governed not by this theoretical distinction between classes of scales but rather by the idiosyncratic challenges that individual scales pose. I provide experimental evidence and theoretical arguments in favor of the Context-based approach to SI computation on which SIs are computed by going through the steps of pragmatic reasoning, and against the Default approach on which SIs are computed by default. I found that children start out by computing SIs only in contexts where they are relevant, which is predicted on the Context-based but not on the Default approach. Another theme is the instantiation of Principle B in the child's grammar, which I consider in the context of how the child constructs scales of the form . SIs based on these scales rule out the coreference reading in Principle B environments (B-contexts). I argue that constructing these scales is a necessary step the child goes through in learning to interpret B-contexts and Exceptional Coreference Contexts and present an experiment on the two types of contexts. The overarching goal of this work is to show that picking up on oppositions between lexical items, and arriving at pragmatic and syntactic generalizations about these lexical items on the basis of these contrasts, is one of the major learning mechanisms involved in first language acquisition.