Search

Second language learners constitute a significant and expanding portion of the global population and there is a growing demand for tools that facilitate language learning and instruction across various levels and in different countries. The development of large language models (LLMs) has brought about a significant impact on the domains of natural language processing and these advancements hold considerable potential for the realm of educational technology. While the potential use of large language models in education shows promise, it is crucial to conduct further exploration into their impact and limitations. This thesis explores the task of delivering explanation and feedback to learners through different formats, including essay feedback and a question-answering framework, and studies how LLMs can be leveraged for these tasks. The thesis is structured into three main parts: In the first part, we investigate the capabilities of large language models in generating feedback on students' essays. Our findings indicate that current state-of-the-art models are unable to provide learners with specific and actionable feedback. To address this issue, we propose a new corpus tailored for this task and demonstrate that utilizing this corpus in an in-context learning setup allows us to deliver more effective feedback to learners. In the second part, we study a different type of feedback (answering questions about the language, such as grammar or vocabulary questions) and introduce a new challenge dataset comprising metalinguistic questions and answers about the English language. These questions are posed by both English second language learners and native speakers. Our goal with this dataset is to use it as an evaluation benchmark and stimulate greater emphasis on addressing the complexities of metalinguistic question answering within the field. Using this benchmark, we investigate the extent to which language models can articulate their generalizations about language. In the third part, our focus shifts to multilingualism within the question answering framework. We propose using data from grammatical error correction corpora to establish benchmarks for assessing the multilingual capabilities of LLMs in addressing learners' language grammar questions. We ask whether LLMs can answer questions posed in a language other than the one the learner is asking about. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

The employment of genre-based pedagogy and Task-Based Language Teaching in second language education is representative of a paradigmatic shift towards a focus on meaning-making. Despite this shift, second language acquisition (SLA) research continues to predominantly rely on complexity, accuracy, and fluency metrics to assess learner production without considering if and how learners achieve their communicative goals and how they appropriate linguistic resources to do so. Though research has confirmed the benefit of textual borrowing from a model within genre-based pedagogy, scarce attention has been paid to the practice at the novice foreign language level. To fill these gaps, this mixed-methods study focused on 12 first-year foreign language learners of German who wrote personal emails after receiving a model text. Through textual analysis of their texts using Systemic Functional Linguistics (SFL) frameworks and semi-structured interviews, I investigated their linguistic choices and the resources borrowed from the model in support of their communicative goals. Because building the relationship with the reader is one of these goals, I employed the SFL MOOD and ATTITUDE frameworks to assess dialogic interaction and how they appealed emotionally to their readers. The participants adhered closely to the model's generic structure, though 25% of the participants omitted up to two stages, and borrowed lexicogrammatical features characteristic of the genre, including words, word groups, and clauses. The interviews revealed that the students most appreciated the model for the structure, though they also confirmed appropriating vocabulary and grammatical structures. Two interviewees conveyed apprehension about borrowing too closely from the model, indicating that textual borrowing may require additional contextualization. The MOOD analysis showed that participants primarily produced declaratives to give information and only sparingly integrated interrogatives and imperatives to elicit a response from their readers. To connect with their readers emotionally, all expressed evaluations of judgment, affect, and appreciation. In addition to demonstrating assessment methods that focus on the linguistic choices that contribute to achieving a communicative goal, the findings of this study show a need to assist learners in making form-meaning connections to foster strategic borrowing and language development. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Across three separate investigations, this dissertation documents work examining the role of higher-order processes and cognitive effort during the top-down compensatory repair of degraded speech, specifically perceptual restoration. Perceptual restoration is a phenomenon where a listener can perceptually restore or repair speech "missing" from an auditory signal. Because perceptual restoration improves speech intelligibility, this compensatory ability may allow some listeners to function better in challenging listening environments, such as a noisy restaurant. Prior investigations into perceptual restoration have found that there is considerable variability across individuals with regard to their ability to restore missing speech information. This work proposes that higher-order domain-general cognitive and domain-specific linguistic abilities may characterize unexplained variability during perceptual restoration tasks. This dissertation also characterizes cognitive resources devoted to task-related demands during perceptual restoration tasks, both during and after sentence listening. Understanding the role that cognitive effort plays in speech intelligibility and perceptual repair is important because two listeners can exert very different levels of effort at different times, even after listening has ended, while obtaining the same intelligibility score, highlighting aspects of speech recognition that are sometimes overlooked. This work employed speech interrupted by a gap of silence or a burst of noise to force listeners into perceptual restoration and repair. Because listeners with moderate or poorer hearing loss struggle with perceptual restoration tasks, these investigations recruited participants with normal hearing to mild sensorineural hearing loss. By asking listeners with a (near) normal auditory system to process a signal with only 50% of the original speech remaining reduces the impact of the peripheral auditory system and stresses higher-order reprocessing of speech information. The first investigation assessed a range of cognitive abilities as well as a participant's lexical knowledge and lexical access speed. Next, these abilities were related to a participant's ability to restore missing speech information across conditions that varied with regard to sentence predictability, interruption type, and linguistic complexity. This investigation found that a listener's complex working memory and lexical knowledge acted as the best predictor for perceptual restoration ability, followed next by inhibitory control and processing speed. These data suggest that domain-general cognitive processes such as working memory provide an avenue through which domain-specific linguistic mechanisms can repair and restore missing speech. Consistent with existing literature, listeners performed better when sentences were periodically interrupted by bursts of noise rather than gaps of silence, and performance was better for predictable/high context sentences than those where the last word was semantically unpredictable or not related to sentence context. The second investigation used well-validated neurocognitive measures that captured a participant's linguistic knowledge in greater detail. This study also investigated the role of other domain-general cognitive abilities, such as inhibitory control, processing speed, executive functioning, and problem-solving ability in addition to lexical knowledge and lexical access speed. Consistent with the results of the first investigation, complex working memory supported perceptual restoration regardless of condition, suggesting an important role of complex working memory regardless of interruption type, promoting the opportunity for speech recovery using linguistic processes. Notably, lexical knowledge and lexical access speed were important when speech was periodically interrupted by bursts of noise. This is consistent with data indicating that participants perceive the missing speech information as being continuous behind the noise burst, which may permit additional lexical activation and processing. While other domain general-cognitive abilities were not directly related to perceptual restoration, they were significantly correlated with complex working memory, suggesting a hierarchical relationship between these abilities and the role of more advanced cognitive processes. Finally, a third investigation characterized the level of cognitive effort required during perceptual restoration tasks using pupillometry, both during and after sentence listening. Cognitive effort during sentence listening has often been referred to as "listening effort" and reflects the time-course of effort devotion during the listening window. After the sentence has ended, ongoing cognitive effort may be devoted during a relatively short period of time known as the retention interval, which is thought to be when ongoing reprocessing and repair may occur. This investigation found that during the listening window, both interruption conditions resulted in additional pupil dilation, reflecting devotion of task-related cognitive resources compared to uninterrupted sentences. For a brief time near the end of the listening window however, the pupillary response was greatest for the noise burst interrupted sentences, which suggests additional processing occurring while the sentence is still playing. On the other hand, pupil dilation was greatest for the silent gap sentences near the end of the retention interval, which suggests sustained cognitive effort in attempting to resolve lexical or syntactic ambiguity. Taken together, these investigations offer valuable insight into the perceptual restoration process. These data provide clinicians and scientists with information regarding the cognitive and linguistic abilities that are key during perceptual restoration. This may allow for the development of auditory rehabilitation programs that emphasize perceptual restoration and cognitive training or the development of prediction algorithms for hearing aids or cochlear implants aimed at reducing cognitive load under certain listening conditions. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Classifier systems constitute one of the most prominent features of East and South-East Asian languages (Li, 2013). Classifiers categorize referent nouns based on salient semantic features such as humanness, animacy, shape, or others (Aikhenvald and Mihas, 2019). The work described in this dissertation constitutes a thorough quantitative investigation of the distribution and function of classifiers in Mandarin Chinese and beyond. The results of the studies settle longstanding theoretical debates about the categorial nature of classifiers and contribute the first in-depth investigation of their communicative functions. Our main theoretical contribution is the development of a reliable methodology for objectively distinguishing between syntactic categories based on measurable differences in word distribution and functions. With this method, we settle the longstanding theoretical debate about the existence of multiple classifier types in Mandarin (Cheng and Sybesma, 1999; Li and Thompson, 1989; Li, 2013): we show that mensural classifiers should be distinguished from other types of classifiers based on their measurably distinctive distributions and functions. The thesis further investigates the cognitive and communicative functions of classifiers. We show that classifier-noun combinations are optimized to accommodate general communicative pressures (Blevins et al., 2017) of learning and processing, by balancing predictable and discriminative classifier-noun combinations. Previous work on noun classification implies that gender systems are inherently optimized to accommodate communicative pressures on human language learning and processing (Dye et al., 2018). The authors state that languages can make use of either grammatical (e.g., gender) or probabilistic (pre-nominal modifier sequences) noun classification to smooth the uncertainty (entropy (Shannon, 1948)) of nouns in context. We show that a genderless language like Mandarin can rely on nominal classifiers in a comparable way: classifiers appear to play the same functional role as gender markers. Noun-classifier combinations are sensitive to same frequency, similarity, and co-occurrence interactions that structure gender systems. We also present the first study of the interaction between grammatical and probabilistic noun classification. Finally, this work explores the role of classifiers in post-nominal positions in Mandarin. Previous work suggests that linguistic elements in different positions will contribute differently to human learning and processing. Based on an artificial noun learning experiment, Ramscar (2013) shows how humans will tend to rely on prefixed elements to categorize nouns while suffixed elements will help them specify noun meanings. For nouns, meaning specification is especially useful if they carry more general meanings, i.e., for nouns that occur more frequently and in a broader range of contexts. An investigation of the differences between nouns occurring with pre- vs. post-nominal classifiers confirms these findings on natural language data. We find significant differences in the frequencies of the two types of nouns: pre-nominal classifiers tend to pair with lower frequency nouns to aid prediction, while post-nominal classifiers are associated with higher frequency nouns to refine their meanings. Additionally, sortal classifiers, which emphasize the intrinsic features of nouns, are predominantly used in pre-nominal positions for categorization. In contrast, mensural classifiers, known for their flexibility in quantifying nouns, are more commonly found in post-nominal positions to aid in meaning refinement. The implications of this work extend beyond language specific findings. By elucidating the complex interplay between classifiers and communicative efficiency, our study contributes to a deeper understanding of the cognitive processes underlying noun processing and learning. It also provides a framework to further investigate the role of classifiers in other classifier languages and structures. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

To build an Artificial Intelligence system that can assist us in daily lives, the ability to understand the world around us through visual input is essential. Prior studies train visual perception models by defining concept vocabularies and annotate data against the fixed vocabulary. It is hard to define a comprehensive set of everything, and thus they are hard to generalize to novel concepts and domains. In this thesis, I turn to language as a scalable and effective tool to build visually grounded models. Intuitively, natural languages are the most effective medium of learning and communication for humans. I will introduce two lines of work to train models to understand the visual world with language as supervision. The first line of work is inspired by masked language modeling such as BERT, and extends that to build contextualized representation models for vision and language. These models can be fine-tuned to perform vision-language tasks such as answering questions about an image. The second line of work uses language to supervise object detection models and enables object detection with prompts, where the users could specify custom needs and domain knowledge in a text prompt, and the model situates its predictions based on the text on the fly. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

This dissertation study explores a young Japanese heritage language speaker's learning process of formal writing in Japanese. With a rapid increase in immigrants into the U.S. and growing importance of diversity, many researchers have shown interest in immigrant children, i.e., heritage speakers, from educational and research perspectives and have found that many heritage speakers, even though they have a native-like fluency in informal conversations, cannot read and write in their heritage languages as native speakers at their age do. Regarding Japanese heritage speakers, Japanese supplementary schools play an important role in fostering their Japanese proficiency and cultural identities, but once children leave supplementary schools around age nine because of difficulty in curricula, it is hard for them to acquire advanced Japanese afterward. This dissertation provides a four-week intensive academic Japanese reading and writing program based on a usage-based language acquisition for the Japanese heritage speaker, Sakura, who has never attended Japanese schools. Her argument essays drastically improve, which empirically supports the efficacy of usage-based instruction for formal writing to some degree. The qualitative analysis of this study and its findings suggest multiple implications and possibilities for future research: (1) it is generally held that kango 'Sino-Japanese words' are more challenging for learners than "wago" 'Japanese native words,' but the most difficult words for Sakura are some extremely topic- and genre-specific "wago"; (2) new words are registered in her receptive lexicon through their frequent occurrences, but advancing them to productive lexicon necessitates varying contexts, collocations, and functions; (3) L1 transfer is not limited to beginners, but Sakura's writing also reveals an influence from English to satisfy her L1 linguistic sophistication; and (4) the most crucial factor in Sakura's improvement is her strong "mission-like" motivation with a sense of responsibility, which I call heritage motivation and which is not extractable from a conventional quantitative approach. This motivation has grown in her mind through a positive attitude, or somatic value, toward Japanese and their culture. It suggests a focal shift from "how teachers can teach learners" to "how teachers and others can nurture their positive somatic value leading to motivation." [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

In clinical and forensic evaluations, clinical neuropsychologists measure performance validity using standalone and embedded measures. Given the need to capture neuropsychological functioning through a variety of cognitive tests, newer, more sensitive measures should be evaluated for their utility in comparison to already established measures. Prior iterations of the Weschler Adult Intelligence Scale (WAIS) utilized Digit Span scaled score subtracted by Vocabulary scaled score as an embedded measure of validity, with success (Iverson & Tulsky, 2001; Mittenberg et al., 1995). This study retrospectively investigates the utility of WAIS-IV Vocabulary minus Digit Span (VC-DS) in clinical and forensic contexts compared to established criterion measures. Methods: A sample of 95 patients over the age of 18 were examined. Patients underwent a neuropsychological evaluation for varying medicolegal or clinical reasons. Chi-square test of independence was used to examine the relationship between VC-DS and criterion measures in medicolegal and clinical evaluations and across Racial and Ethnicity diverse groups. Results: A significant relationship was found between WAIS-IV Reliable Digit Span and VC-DS using liberal cut points (P=0.014). The relationship between RDS and VC-DS was found in the Caucasian cohort (p=0.34). A significant relationship was revealed between VC-DS and Test of Memory Malingering (TOMM) trial two (p=0.041)and the California Verbal Learning Test Second Edition forced-choice trial using a liberal cut point in forensic populations (p=0.005). Frequency analysis found that among the clinical and forensic evaluations, 74% passed three or more PVTs across the evaluation. Conclusion: The utility of VC-DS is limited, given the current findings and the statistical limitations. Future directions should include further investigation of how culturally and linguistically diverse groups and functionally impaired groups perform on VC-DS in relation to other established measures of validity. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

One of the most important outcomes in a child's education is the ability to read proficiently, which can improve future academic progress and occupational success. Due to the impairments in social and communication areas of development for individuals with autism spectrum disorder (ASD), literacy provides the possibility of improved communication as well as the building of relationships with others and their environment. Unfortunately, children with ASD, particularly those with moderate to severe symptomology (level 3) are often excluded from literacy instruction due to the substantial support needed for independence. Research has demonstrated that word comprehension is a critical component to overall reading comprehension, and that reading comprehension is an area where children with ASD are negatively impacted the most in comparison to their neurotypical peers. The gap in research exists with children who have level 3 ASD, because they often display maladaptive or unsafe behavior within the classroom setting and are dismissed from studies. To address this critical need of this underserved population and research gap, the DIP aimed to develop and explore the effectiveness of a vocabulary intervention program for elementary students with level 3 ASD, entitled "ABA Picture to Text Word Comprehension Intervention." Drawing on methods from Innovation/Intervention Research, the DIP employed a changing criterion design (Klein et al., 2015; Manolov et al., 2020; Tanious, 2022) to develop the intervention program. In addition, and Intervention Rating Profile was used to explore the social validity of the program by gathering perspectives of teachers, behavior analysts, and teacher assistants. The intervention took place at a Local Education Agency (LEA) with three male students diagnosed with ASD, who ranged in age from 6 years 2 months to 10 years 3 months. Findings indicate that participants increased word comprehension, decreased maladaptive behaviors, and increased socially appropriate behavior with socially valid perspectives from the immediate stakeholders. Use of Applied Behavior Analysis (ABA) teaching strategies, along with building of rapport with the participants were likely key factors in the results. The intervention provided a feasible example for improvement in elements of literacy acquisition as well as that it demonstrated the importance of well-designed academic interventions to the development of appropriate social behavior and reduction in maladaptive behavior in the classroom for children with level 3 ASD. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Research on language and cognitive processes in bilinguals over the past few decades has underscored the activation of words in the first language (L1) during comprehension and production of the second language (L2) (e.g., Dijkstra, Van Jaarsveld, & Ten Brinke, 1998; Van Heuven, Dijkstra, & Grainger, 1998; Hermans, Bongaerts, De Bot, & Schreuder, 1998; Lemhofer, et al., 2008). This activation extends bidirectionally, with words from the L2 also influencing the processing of the L1 (e.g., Van Hell & De Groot, 2008; Linck, Kroll, & Sunderman, 2009; Schmid & Jarvis, 2014; Schmid & Keijzer, 2009), leading to cross-linguistic interference (CLI) in bilinguals' language processing. However, the nature and extent of this interference may vary depending on the bilingual's dominant language and immersion context. L1-dominant bilinguals may struggle to control interference from their L1s, while the opposite may hold true for L2-dominant bilinguals, even when their L2 is not their native language. Consequently, the formerly dominant L1 of some bilinguals may exhibit signs of attrition. This dissertation explores the relationship between attrition and L2 acquisition, addressing two key questions: (1) to what extent does context (place of residence) and nature of bilingualism influences the amount and direction of cross-language interference? and (2) to what extent (if any) is the activation of form and meaning modulated by the nature of the lexical relationships and individual differences measures? And, if effects are found, how does sentence context modulate those effects? The contribution of this study stems from the novel examination of two types of semantic extensions: a) "new meanings" or words that share form but do not share any semantic relationship across languages, and b) "extended meanings" or words that share form and at least one semantic sense across languages. Additionally, synonyms, unrelated words, cognates, interlingual homographs and translation pairs are also included to explore form-meaning interactions. The comprehension of words with varying degrees of form-meaning relatedness is evaluated in two tasks: (1) a semantic-relatedness decision task (SRDT) within and across languages and (2) a self-paced reading task (SPR). Sentence cues in unilingual language contexts are manipulated in congruent and incongruent conditions to assess interference effects. Participants also complete a flanker task as a measure of cognitive control and a battery of proficiency measures. These tasks are implemented across five participant groups: (1) monolingual speakers of English in the U.S, (2) monolingual Spanish speakers in Argentina, (3) L1 Spanish-L2 English bilinguals in the USA, (4), L1 Spanish-L2 English bilinguals in Argentina and (5) L1 English- L2 Spanish learners in the USA. Results from this study indicate that immersion context influences the direction of language interference, with more L2 to L1 effects observed for immersed bilinguals compared to non-immersed counterparts, and with both groups being vulnerable to CLI effects compared to monolinguals. Despite extended exposure in an L2 dominant context, bilinguals can maintain their L1 proficiency, contradicting L1 attrition views. The activation of form-meaning relationships is directly related to the nature of the lexical relationships, with orthographically and semantically related words (e.g., cognates, extended meanings) showing processing advantages over orthographically related but semantically unrelated words across languages (e.g., false friends, new meanings). Within-language semantic competitors cause more interference during processing compared to unrelated words. Inhibitory control mechanisms or language dominance do not significantly attenuate CLI effects from form-meaning competitors in the non-target language. Results from the SPR revealed that sentence context can modulate interference effects, facilitating the resolution of lexical ambiguities and form-meaning relationships. However, language cues from a sentence alone are not sufficient to constrain CL competition, highlighting the importance of semantic information conveyed by the sentence context for language non-selectivity to occur. Overall, bilingual characteristics, including the type of bilingual, linguistic exposure/experience and levels of L1-L2 proficiency all seem to influence language processing. This dissertation contributes to the field of psycholinguistics of bilingualism, language attrition and second language acquisition. It highlights the importance of understanding language-attrition as a bridge between first- and second-language lexical processing. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

More than nine million students enroll in community colleges annually (Fink et al., 2023) making them accessible entry points for individuals of diverse backgrounds (Fink et al., 2023). Upward transfer from community colleges to four-year institutions creates critical pathways to baccalaureate degree attainment, but only 33% of community college students succeed in transferring to four-year institutions (Velasco et al., 2024). Transfer success from community college to four-year institutions is further confused by the disconnect between the measures of time to completion rates and common terminology that exists among community colleges versus four-year institutions (Burridge et al., 2023). This study investigated the knowledge and confidence first-semester transfer-aspiring Northeast Community College students had about navigating the transfer process at the start of their transfer journey. This study also sought to increase first-semester transfer-aspiring Northeast Community College student confidence and knowledge about the transfer process through a communication plan. Utilizing a mixed methods design, 48 new aspiring community college transfer students were introduced via a communication plan to terminology, institutional agents, types of agreements, benefits, and academic planning needed to successfully transfer to four-year institutions. Analysis of anonymous, matched pre- and post-surveys revealed that aspiring transfer students were overconfident in their general transfer knowledge, while most students did not know specific transfer knowledge, especially related to institutional knowledge needed to successfully transfer. Results from a focus group revealed that the communication plan helped students identify gaps in transfer knowledge to help build early momentum to better navigate the transfer process. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

In this thesis, I use a combination of simulations and empirical data to demonstrate that a small set of structural and functional principles - the basic tenets of predictive coding theory - succinctly accounts for a very wide range of properties in the language processing system. Predictive coding approximates hierarchical Bayesian inference via a biologically plausible mechanism of competition whose unique properties (i) allow it to capture important asymmetries in bottom-up and top-down processing and (ii) offer a critical insight into the functional relationship between behavioral and neural measures of word processing. In addition to accounting for well-known properties of the language processing system, this mechanism of competition allows us to formulate and test a novel empirical prediction - that N400 attenuation can be observed in the presence of behavioral interference - which challenges a widely held assumption that N400 attenuation is essentially synonymous with facilitated processing. Importantly, the computational model that we used to simulate the above psycholinguistic phenomena was tightly constrained by predictive coding principles and the particular class of algorithm we employed (PC/BC-DIM; Spratling, De Meyer, & Kompass 2009), emphasizing how a fundamental, biologically plausible perceptual mechanism that is shared across the cortex can also give rise to psycholinguistic phenomena. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

The ongoing tonogenetic sound change in Seoul Korean involves transphonologization in phrase initial position, where the fundamental frequency (F0) of the vowel following aspirated or lenis stops becomes associated with the aspirated-lenis stop contrast (phonologization), while the originally contrastive Voice Onset Time (VOT) values merge (dephonologization). One critical aspect that remains unexplored in this context is the impact of lexical competition, often evaluated through minimal pairs. This dissertation explores the influence of lexical competition on hyperarticulation in Korean stops within this ongoing tonogenetic sound change. It aims to investigate how hyperarticulation influences sound change and how speakers adapt to the change. The study conducted production experiments on both phrase-initial and phrase-medial stops, alongside analyzing a large dialogue corpus dataset. The statistical analyses of experimental data reveal that speakers hyperarticulate the VOT of aspirated stops, resulting in a greater VOT contrast, and tend to hyperarticulate the F0 of lenis stops, leading to a greater F0 contrast when they have a lenis minimal pair. Similarly, corpus analyses indicate hyperarticulation of VOT and F0 in aspirated stops, resulting in an expanded contrast for both of these cues as well. These findings suggest two main points: (1) the hyperarticulation effect observed in the current dissertation is 'context-free' contrastive hyperarticulation, suggesting that this process is related to speaker-internal processes, and (2) the presence of a minimal pair inhibits VOT merger due to its continuing high functional load in the speech community. In phrase medial position, the aspirated-lenis contrast is not reported to involve any F0 contrast. However, statistical results of experimental data on phrase-medial stops show that speakers who rely more on F0 for distinguishing phrase-initial aspirated and lenis stops also now use F0 for distinguishing phrase-medial aspirated and lenis stops. These observations suggest an emerging tonogenetic sound change extending the F0 contrast to the phrase-medial position, marking genuine tonogenesis. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

All individuals develop new skills over time and one important developmental stage is when individuals can learn in the absence of direct reinforcement. Behavior Analytic theories have stressed the importance of these emergent response repertoires and have consistently studied them across fields (i.e. Incidental Bidirectional Naming, Arbitrarily Applicable Relational Responding, Equivalence). Emergent responses are responses that are not explicitly taught during instruction. Recently, behavior analytic perspectives have studied their similarities across perspectives, one area of interest being Incidental Bidirectional Naming (Inc-BiN). Researchers have identified multiple types of Naming, including Pre-Unidirectional Naming (Pre-UniN), Incidental Unidirectional Naming (Inc-UniN), and Incidental Bidirectional Naming (Inc-BiN). Study 1 investigated the basic question about whether verbal development -- in this case degree of Inc-BiN -- is related to academic performance (mathematics and reading). The researcher correlated participants' listener and speaker responses for two Brief Inc-BiN Probes with academic performance (iReady® Diagnostic Mathematic and Reading scores) for 41 participants. Results demonstrated significant positive correlations between degree of Inc-BiN and reading and mathematics performance. Given this correlation, two experiments examined the relationship between verbal development and two specific instructional design tactics that focus on the emergence of novel behavior. Experiment 2 investigated the effectiveness and appropriateness of a relational training procedure (matrix instruction) on participants' emergent intraverbal responses. The researcher used learn units to teach intraverbal atomic units and measured generalized operant responding. The researcher used a matrix with five prefixes and five root words, which combined to make 25-word combinations. The researcher taught five combinations (1 prefix plus 1 root) and tested the other 20-word combinations. The researcher implemented a multiple probe design and measured emergent intraverbals and degree of Inc-BiN prior to and following intervention. Results demonstrated that emergent intraverbal responses varied in accordance with participant degree of Inc-BiN. Experiment 3 investigated the effects of another relational training procedure (Equivalence-Based Instruction; EBI), across participants with ranging degrees of Inc-BiN. During teaching, the researcher taught two of six relations in a class to mastery using learn unit instruction across 13 participants. Results in Experiment 3 demonstrated that, after EBI, participants emitted a greater number of emergent intraverbals when they had higher degrees of Inc-BiN and lower emergent intraverbals when they had lower degrees of Inc-BiN. Experiment 2 and 3 demonstrate similar results that participants with Inc-UniN and Inc-BiN emitted a high percentage of emergent relations following learn unit instruction across the three studies. Data highlight the importance of Inc-BiN's relation to academic performance in general (Experiment 1) and in predicting participants' success with behavior analytic instructional design tactics (Experiment 2 and 3). [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

A disproportionate number of preservice elementary teachers exhibit high levels of math anxiety (Reid et al., 2018). Not only does this inhibit their ability to understand mathematical concepts, but it also negatively impacts their pedagogical approach to math (Beilock & Maloney, 2015; Reid et al., 2018). Teachers who experience math anxiety encounter challenges when it comes to effectively explaining concepts, including the appropriate use of mathematical vocabulary. The use of mathematical language, which includes content specific vocabulary, holds significant importance in shaping students' comprehension of concepts (Riccomini et al., 2015). However, the relationship between math anxiety and correct usage of mathematical vocabulary is a relatively unexplored area. This research used a mixed-methods approach to explore the role of mathematical discourse in enhancing the content knowledge of community college preservice teachers, facilitating their acquisition of mathematical vocabulary, and mitigating math anxiety. Surveys were used to evaluate preservice teachers' level of math anxiety and their beliefs about the importance of using correct vocabulary when explaining concepts. Data were gathered both before and after students completed a mathematical content course for early childhood and elementary education. Interviews with the course instructors were used to ascertain the degree to which discourse was used as a way to strengthen preservice teachers' math vocabulary and conceptual understanding. The findings revealed a decrease in math anxiety as students' content knowledge increased and confidence in their ability to use mathematical vocabulary grew. Students who found both the course and mathematics challenging exhibited a higher degree of math anxiety by the conclusion of the course. Furthermore, students found the class discussions in which they practiced using mathematical vocabulary invaluable for increasing their content knowledge and preparing to explain concepts to their future students. Findings from this research can be used to inform teacher educator programs on the effectiveness of incorporating discourse to increase preservice teachers' mathematical vocabulary and content knowledge and indirectly affect their math anxiety. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

We study how to automatically generate cloze questions from given texts to assess reading comprehension, where a cloze question consists of a stem with a blank space holder for the answer key, and three distractors for generating confusions. We present a generative method called CQG (Cloze Question Generator) for constructing cloze questions from a given article, utilizing neural networks and WordNet with an emphasis on generating multi-gram distractors. WordNet is a hypernym-hyponym network of synsets, where each synset is a set of lemmas with the same gloss labeled by a lexname. CQG harnesses word-sense disambiguation, text-to-text transformation, and WordNet's synset taxonomies and lexical labels to select an answer key from a sentence, segment it, and generate instance-level distractor candidates (IDCs) using a transformer and sibling synsets. After ranking the IDCs based on contextual embedding similarities, synset, and lexical relatedness, CQG forms distractor candidates and checks if they align with people's writing conventions to determine whether they can be distractors. CQG significantly outperforms SOTA results, confirmed by the high quality of the generated distractors assessed by human judges. The effectiveness of CQG, however, is confined by WordNet's limited vocabulary. There are tens of thousands of new lemmas that are not yet included in WordNet. It is therefore desirable to construct an automated system that can add new lemmas to WordNet with their glosses being the only information available, which can be readily obtained from Wikitionary and other open sources. This task is challenging. We tackle this challenge by devising a system called WordNeter that predicts a lexname for the given gloss, determines if a new synset should be formed, predicts a hypernym for the new synset, and updates the existing hypernym-hyponym relations in WordNet. WordNeter excels with a 93.6% F1 score on predicting lexnames for given glosses and 64.8% exact matches of the predicted direct hypernyms with the true direct hypernyms, which significantly outperforms GPT-3.5-Turbo and other models. Even without exact matches of hypernym predictions, most predicted hypernyms are still suitable for generating high-quality distractors. Integrating WordNeter with CQG expands CQG's ability to generate satisfactory distractors for cloze questions with answer keys outside WordNet's current vocabulary, advancing the methodology of cloze question generation. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]