Linguistics Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/2787
Browse
52 results
Search Results
Item How Grammars Grow: Argument Structure and the Acquisition of Non-Basic Syntax(2019) Perkins, Laurel; Lidz, Jeffrey; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation examines the acquisition of argument structure as a window into the role of development in grammar learning. The way that children represent the data for language acquisition depends on the grammatical knowledge they have at any given point in development. Children use their immature grammatical knowledge, together with other non-linguistic conceptual, pragmatic, and cognitive abilities, to parse and interpret their input. But until children have fully acquired the target grammar, these input representations will be incomplete and potentially inaccurate. Our learning theory must take into account how learning can operate over input representations that change over the course of development. What allows learners to acquire new knowledge from partial and noisy representations of their data, one step at a time, and still converge on the right grammar? The case study in this dissertation points towards one way to characterize the role of development in grammar acquisition by probing more deeply into the resources that learners bring to their learning task. I consider two types of resources. The first is representational: learners need resources for representing their input in useful ways, even early in development. In two behavioral studies, I ask what resources infants in their second year of life use to represent their input for argument structure acquisition. I show that English learners differentiate the grammatical and thematic relations of clause arguments, and that they recognize local argument relations before they recognize non-local predicate-argument dependencies. The second type of resource includes mechanisms for learning from input representations even when they are incomplete or inaccurate early in development. In two computational experiments, I investigate how learners could in principle use a combination of domain-specific linguistic knowledge and domain-general cognitive abilities in order to draw accurate inferences about verb argument structure from messy data, and to identify the forms that argument movement can take in their language. By investigating some of the earliest steps of syntax acquisition in infancy, this work aims to provide a fuller picture of what portion of the input is useful to an individual child at any single point in development, how the child perceives that portion of the input given her current grammatical knowledge, and what internal mechanisms enable the child to generalize beyond her input in inferring the grammar of her language. This work has implications not only for theories of language learning, but also for learning in general, by offering a new perspective on the use of data in the acquisition of knowledge.Item An Affiliative Model of Early Lexical Learning(2019) Tripp, Alayo; Feldman, Naomi; Idsardi, William; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In defining the language acquisition problem, traditional models abstract away effects of variability, defining the learner as acquiring a single language variety, which is spoken homogeneously by their speech community. However, infants are exposed to as many unique varieties of speech as they are speakers. Adult sociolinguistic competence is also characterized by the capacity to employ and interpret non-phonological linguistic distinctions which are associated with different social groups, including ‘code-switching’ or ‘style-shifting’ between languages and speech registers. This dissertation presents a model of infant lexical acquisition which assumes that learners monitor linguistic sources for variation in reliability. This model is adapted from Shafto, Eaves, Navarro, and Perfors (2012) which the authors used to describe the behavior of preschool children in selecting sources to learn labels from in K. Corriveau and Harris (2009) and M. Corriveau and Harris (2009). I show that this probabilistic model effectively simulates two experiments from the literature on preverbal infants’ perception of labeling, Rost and McMurray (2009) and Koenig and Echols (2003). Evidence suggests that the receptiveness of preverbal infants to novel lexical items is correlated with infant beliefs regarding the informant’s knowledgeability and social group membership. These simulations demonstrate that language learners may well be recruiting processes of epistemic trust to guide lexical acquisition much earlier than previously suggested. We should therefore expect even very young listeners to respond differently to dialects not solely as a function of exposure, but also as a function of attitudes towards the speech determined by the quality of that exposure. Developmental differences between populations in attention to non-linguistic affiliative cues are therefore expected to emerge early and have significant effects on language outcomes. Measures of online language proficiency may be vulnerable to significant bias owing to the activation of sociolinguistic biases in the presentation of test items. Differences in the breadth or specificity of listener preferences for speakers in turn predict differences in task complexity for learners of standard and non-standard dialects. A new research program in early sociophonetic perception, uniting accounts of selective trust with language learning has the potential to deepen understanding of both typical and disordered language development.Item How to use context for phonetic learning and perception from naturalistic speech(2019) Hitczenko, Katarzyna; Feldman, Naomi H.; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Infants learn about the sounds of their language and adults process the sounds they hear, even though sound categories often overlap in their acoustics. This dissertation is about how contextual information (e.g. who spoke the sound and what the neighboring sounds were) can help in phonetic learning and speech perception. The role of contextual information in these tasks is well-studied, but almost exclusively using simplified, controlled lab speech data. In this dissertation, we study naturalistic speech of the type that listeners primarily hear. The dissertation centers around two main theories about how context could be used: top-down information accounts, which argue that listeners use context to predict which sound will be produced, and normalization accounts, which argue that listeners compensate for the fact that the same sound is produced differently in different contexts by factoring out this systematic context-dependent variability from the acoustics. These ideas have been somewhat conflated in past research, and have rarely been tested on naturalistic speech. We start by implementing top-down and normalization accounts separately and evaluating their relative efficacy on spontaneous speech, using the test case of Japanese vowel length. We find that top-down information strategies are effective even on spontaneous speech. Surprisingly, we find that normalization is ineffective on spontaneous speech, in contrast to what has been found on lab speech. We, then, provide analyses showing that when there are systematic regularities in which contexts different sounds occur in - which are common in naturalistic speech, but generally controlled for in lab speech - normalization can actually increase category overlap rather than decrease it. Finally, we present a new proposal for how infants might learn which dimensions of their language are contrastive that takes advantage of these systematic regularities in which contexts different sounds occur in. We propose that infants might learn that a particular dimension of their language is contrastive, by tracking the acoustic distribution of speech sounds across contexts, and learning that a dimension is contrastive when the shape changes substantially across contexts. We show that this learning account makes critical predictions that hold true in naturalistic speech, and is one of the first accounts that can qualitatively explain why infants learn what they do. The results in this dissertation teach us about how listeners might use context to overcome variability in their input. More generally, they reveal that results from lab speech do not necessarily generalize to spontaneous speech, and that using realistic data matters. Turning to spontaneous speech not only gives us a more realistic view of language learning and processing, but can actually help us decide between different theories that all have support from lab speech and, therefore, can complement work on lab data well.Item Variation and learnability in constraints on A-bar movement(2019) Huang, Zhipeng (Nick); Phillips, Colin; Lasnik, Howard; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)A classic problem in linguistics is explaining how learners come to know so much about their native languages, despite receiving limited and noisy input. This learning problem becomes especially acute when the linguistic properties in question are obscure and show subtle variation across languages. Cross-linguistic variation means that learners must identify the appropriate points of variation for their language, even though the direct evidence that they need is often hard to detect or even non-existent. This dissertation presents two case studies on constraints in A-bar movement. Because constraints are by nature abstract and difficult to observe directly, a classic solution to the learning problem posed by constraints claims that knowledge of these abstract or negative linguistic properties is innate. However, a number of these constraints show cross-linguistic variation, raising questions about how they are represented and how linguistic experience might (or might not) shape linguistic knowledge. The first case study, discussed in chapters 2 and 3, involves cross-linguistic variation in the constraint that governs A-bar movement from relative clauses: some, but not all, languages allow A-bar movement from relative clauses under exceptional circumstances. I argue that these “porous” relative clauses that permit A-bar movement can be distinguished by a property that I call “tense dependence,” and discuss how this tense property might be formally related to A-bar movement. I show that this particular kind of variation presents a learning problem: in languages like English and Mandarin Chinese, learners have little direct positive evidence that such A-bar movement is possible. Using tense dependence, I propose that learners might circumvent this absence of direct evidence by relying on A-bar movement from a superficially unrelated structure: non-finite purposive clauses. The second case study, discussed in chapter 4, involves bridge verbs: within a given language, some verbs allow A-bar movement and others do not; in addition, the set of verbs that allow A-bar movement varies across languages. I present an acceptability judgment experiment that is aimed at clarifying existing generalizations about bridge verbs in English. With more secure generalizations in hand, I discuss the extent to which bridge effects have a pragmatic origin, bringing in data from an informal survey of English and Dutch native speakers that looks at the effect of context on long-distance A-bar movement. Echoing existing work, the survey shows what appears to be a case of cross-linguistic variation between English and some Dutch varieties for cognitive factive verbs. To account for this instance of cross-linguistic variation, I suggest that English learners might have limited access to direct evidence, and discuss what learning mechanisms a learner might need to draw the language-appropriate conclusions based on sparse evidence. Chapter 5 discusses the consequences these case studies have for our formal accounts of these constraints. I evaluate existing proposals and argue that the range of variation observed requires more flexibility than what many existing proposals can offer. Chapter 6 concludes.Item THE ROLE OF STRUCTURAL INFORMATION IN THE RESOLUTION OF LONG-DISTANCE DEPENDENCIES(2018) Malko, Anton; Phillips, Colin; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)The main question that this thesis addresses is: in what way does structural information enter into the processing of long-distance dependencies? Does it constrain the computations, and if so, to what degree? Available experimental evidence suggests that sometimes structurally illicit but otherwise suitable constituents are accessed during dependency resolution. Subject-verb agreement is a prime example (Wagers et al., 2009; Dillon et al., 2013), and similar effects were reported for negative polarity items (NPIs) licensing (Vasishth et al., 2008) and reflexive pronouns resolution (Parker and Phillips, 2017; Sloggett, 2017). Prima facie this evidence suggests that structural information fails to perfectly constrain real-time language processing to be in line with grammatical constraints. This conclusion would fall neatly in line with an assumption that human sentence processing relies on cue-based memory (e.g McElree et al., 2003; Lewis and Vasishth, 2005; Van Dyke and Johns, 2012; Wagers et al., 2009, a.m.o.), the key property of which is the fragility of memory search, which can return irrelevant results if they look similar enough to the relevant ones. The attractiveness of such an approach lies in its parsimony: there is independent evidence that general purpose working memory is cue-based (Jonides et al., 2008), so we do not need to postulate any language specific mechanisms. Additionally, the processing of multiple linguistic dependencies can be analyzed within the same theoretical framework. Cue-based approach has also been argued to be the best one in terms of its empirical coverage: some of the experimental evidence was assumed to only be explainable within it (the absence of ungrammaticality illusions in subject-verb agreement is the main example, to which we will return in more detail later). However, recently several other approaches have been suggested which would be able to ac- count for these cases (Eberhard et al., 2005; Xiang et al., 2013; Sloggett, 2017; Hammerly et al., draft.april.2018). These approaches usually assume separate processing mechanisms for different linguistic dependencies, and thus lose the parsimonious attractiveness of cue-based memory models. They also take a different stance on the role of structural information in real-time language processing, assuming that structural cues do accurately guide the dependency resolution. A priori there is no reason why they could not turn out to be true. But given the theoretical attractiveness of cue-based models in which structural information does not categorically restrain processing, it is important to critically evaluate these recent claims. In this thesis, we focus on reflexive pronouns and on the novel pattern reported in Parker and Phillips (2017) and Sloggett (2017): the finding that reflexive pronouns are sensitive to the properties of structurally inaccessible antecedents in some specific conditions (interference effect). The two works report consistent findings, but the accounts they give take opposite perspectives on the role of structural information in reflexive resolution. Our aim in this thesis is to assess the reliability of these findings and to experimentally investigate cases which would hopefully provide clearer evidence on how the structure guides reflexives processing. To this aim, we conduct two direct replications of Parker and Phillips (2017) and four novel experiments further investigating the properties of the interference effect. None of the six experiments provided strong statistical support for the previous findings. After ruling out several possible confounds and analyzing numerical patterns (which go in the expected direction and are consistent with previous results), we conclude that interference effect is likely real, but may be less strong than the previous studies would lead to believe. These results can be used for setting more realistic expectations for future studies regarding the size of the effect and statistical power necessary to detect it. With respect to our main goal of distinguishing between cue-based and alternative accounts of the interference effect, we tentatively conclude that cue-based approaches are preferred; however, one has to assume that some structural features are able to categorically rule out illicit antecedents. Further highly powered studies are necessary to verify and confirm these conclusions.Item Adjunct Control: Syntax and processing(2018) Green, Jeffrey Jack; Williams, Alexander; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation analyzes the syntax and processing of adjunct control. Adjunct control is the referential relation between the implicit (PRO) subject of a non-finite adjunct clause and its understood antecedent, as in the temporal adjunct in ‘Holly1 went to bed [after PRO1 drinking milk]’, or the rationale clause in ‘August1 sat on the couch [in order PRO1 to read library books]’. Adjunct control is often assumed to involve a syntactic ‘Obligatory Control’ (OC) dependency, but I show that some adjuncts also permit what is referred to as ‘Non-Obligatory Control’ (NOC), as in the sentences ‘The food tasted better [after PRO drinking milk]’ and ‘The book was checked out from the library [in order PRO to read it]’, where PRO refers to some unnamed entity. I argue that for some adjuncts, OC and NOC are not in complementary distribution, contrary to assumptions of much prior literature, but in agreement with Landau (2017). Contrary to implicit assumptions of Landau, however, I also show that this OC/NOC duality does not extend to all adjuncts. I outline assumptions that Landau’s theory would have to make in order to accommodate the wider distribution of OC and NOC in adjuncts, but argue that this is better accomplished within the Movement Theory of Control (Hornstein, 1999) by relaxing the assumption that all adjuncts are phases. Even in adjuncts where both OC and NOC are possible, OC is often strongly preferred. I argue that this is in large part due to interpretive biases in processing. As a foundational step in examining what these processing biases are, the second part of this dissertation uses visual-world eyetracking to compare the timecourse of interpretation of subject-controlled PRO and overt pronouns in temporal adjuncts. The results suggest that PRO can be interpreted just as quickly as overt pronouns once the relevant bottom-up input is received. These experiments also provide evidence that structural predictions can facilitate reference resolution independent of next-mention predictions.Item Argument Roles in Adult and Child Comprehension(2018) Ehrenhofer, Lara; Phillips, Colin; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Language comprehension requires comprehenders to commit rapidly to interpretations based on incremental and occasionally misleading input. This is especially difficult in the case of argument roles, which may be more or less useful depending on whether comprehenders also have access to verb information. In children, a combination of subject-as-agent parsing biases and difficulty with revising initial misinterpretations may be the source of persistent misunderstandings of passives, in which subjects are not agents. My experimental investigation contrasted German five-year-olds’ argument role assignment in passives in a task that combined act-out and eye-tracking measures. Manipulating the order of subject and voice (Exp. 4.1, 4.3) did not impact German learners’ success in comprehending passives, but providing the cue to voice after the main verb (Exp. 4.2) led to a steep drop in children’s comprehension outcomes, suggesting that the inclusion of verb information impacts how young comprehenders process argument role information. In adults, many studies have found that although argument role reversals create strong contrasts in offline cloze probability, they do not elicit N400 contrasts. This may be because in the absence of a main verb, the parser is unable to use argument role information. In an EEG experiment (Exp. 5.1), we used word order to manipulate the presence or absence of verb information, contrasting noun-noun-verb reversals (NNV; which cowboy the bull had ridden) with noun-verb-noun reversals (NVN; which horse had raced the jockey). We found an N400 contrast in NVN contexts, as predicted, but surprisingly, we also found an N400 contrast in NNV contexts. Unlike previous experimental materials, our stimuli were designed to elicit symmetrically strong and distinct verb predictions with both canonical and reversed argument role assignments. These data suggest that adult comprehenders are able to overcome the absence of a main verb when probability distributions over combined verb-argument role information can contribute to generating role-specific verb candidates. The overall investigation suggests that prediction and comprehension of argument role information is impacted by the presence or absence of verb information, which may allow comprehenders to bridge the divide between linguistic representations and world knowledge in real-time processing.Item Relating lexical and syntactic processes in language: Bridging research in humans and machines(2018) Ettinger, Allyson; Phillips, Colin; Resnik, Philip; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Potential to bridge research on language in humans and machines is substantial - as linguists and cognitive scientists apply scientific theory and methods to understand how language is processed and represented by humans, computer scientists apply computational methods to determine how to process and represent language in machines. The present work integrates approaches from each of these domains in order to tackle an issue of relevance for both: the nature of the relationship between low-level lexical processes and syntactically-driven interpretation processes. In the first part of the dissertation, this distinction between lexical and syntactic processes focuses on understanding asyntactic lexical effects in online sentence comprehension in humans, and the relationship of those effects to syntactically-driven interpretation processes. I draw on computational methods for simulating these lexical effects and their relationship to interpretation processes. In the latter part of the dissertation, the lexical/syntactic distinction is focused on the application of semantic composition to complex lexical content, for derivation of sentence meaning. For this work I draw on methodology from cognitive neuroscience and linguistics to analyze the capacity of natural language processing systems to do vector-based sentence composition, in order to improve the capacities of models to compose and represent sentence meaning.Item A Structural Theory of Derivations(2018) Stone, Zachary; Lasnik, Howard; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Operations which take in tuples of syntactic objects and assign them output syntactic objects are used to formalize the generative component of most formal grammars in the minimalist tradition. However, these models do not usually include information which relates the structure of the input and output objects explicitly. We develop a very general formal model of grammars which includes this structural change data, and also allows for richer dependency structures such as feature geometry and feature-sharing. Importantly, syntactic operations involving phrasal attachment selection, agreement, licensing, head-adjunction, etc. can all be captured as special kinds of structural changes, and hence we can analyze them using a uniform technique. Using this data, we give a rich theory of isomorphisms, equivalences, and substructures of syntactic objects, structural changes, derivations, rules, grammars, and languages. We show that many of these notions, while useful, are technically difficult or impossible to state in prior models. It is immediately possible to define grammatical notions like projection, agreement, selection, etc. structurally in a manner preserved under equivalences of various sorts. We use the richer structure of syntactic objects to give a novel characterization of c-command naturally arising from this structure. We use the richer structure of rules to give a general theory of structural analyses and generating structural changes. Our theory of structural analyses makes it possible to extract from productions what structure is targeted by a rule and what conditions a rule can apply in, regardless of the underlying structure of syntactic objects or the kinds of phrasal and featural manipulations performed, where other formal models have difficulty incorporating such structure-sensitive rules. This knowledge of structural changes also makes it possible to extend rules to new objects straightforwardly. Our theory of structural changes allows us to deconstruct them into component parts and show relationships between operations which are missed by models lacking this data. Finally, we extend the model to a copying theory of movement. We implement a traditional model of copying ‘online’, where copies and chains are formed throughout the course of the derivation (while still admitting a feature calculus in the objects themselves). Part of what allows for this is having a robust theory of substructures of derived objects and how they are related throughout a derivation. We show consequences for checking features in chains and feature-sharing.Item Memory retrieval in parsing and interpretation(2017) Schlueter, Ananda Lila Zoe; Lau, Ellen; Linguistics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation explores the relationship between the parser and the grammar in error-driven retrieval by examining the mechanism underlying the illusory licensing of subject-verb agreement violations (‘agreement attraction’). Previous work motivates a two-stage model of agreement attraction in which the parser predicts the verb’s number and engages in retrieval of the agreement controller only when it detects a mismatch between the prediction and the bottom-up input (Wagers, Lau & Phillips, 2009; Lago, Shalom, Sigman, Lau & Phillips, 2015). It is the second stage of retrieval and feature-checking that is thought to be error-prone, resulting in agreement attraction. Here we investigate two central questions about the processing system that underlies this profile. First, to what extent does error-driven retrieval end up altering the structural representation of the sentence, as compared to an independent feature-checking process that can result in global inconsistencies? Using a novel dual-task paradigm combining self-paced reading and a speeded forced choice task, we show that comprehenders do not misinterpret the attractor as the subject in agreement attraction. This indicates that the illusory licensing reflects a low-level number rechecking process that does not lead to restructuring. Second, what is the relationship between the information guiding the retrieval process and the terms that define agreement in the grammar? In a series of speeded acceptability judgment and self-paced reading experiments, we demonstrate that the number cue in error-driven retrieval is as abstract as the terms in which agreement is stated in the grammar, and that semantic features not relevant to the dependency in the grammar are not used to guide retrieval of the agreement controller. However, data from advanced Chinese learners of English suggests that it is not the case that all features relevant to the grammatical dependency will necessarily be used as retrieval cues. Taken together, these results suggest that the feature-checking repair mechanism follows grammatical principles but can result in a final structural representation of the sentence that is inconsistent with the grammar.