서브메뉴
검색
Evaluating Prediction-Based Theories of Bilingual Comprehension of Spanish/English Codeswitches
Evaluating Prediction-Based Theories of Bilingual Comprehension of Spanish/English Codeswitches
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153014
- ISBN
- 9798384045625
- DDC
- 150
- 서명/저자
- Evaluating Prediction-Based Theories of Bilingual Comprehension of Spanish/English Codeswitches
- 발행사항
- [Sl] : University of Michigan, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 156 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Boland, Julie E.;Lewis, Richard L.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2024.
- 초록/해제
- 요약This dissertation investigates how bilinguals use their two grammars to comprehend written intra-sentential codeswitches. I focus on adjective/noun constructions in Spanish and English where I manipulate the congruence of grammatical word order in the two languages across the codeswitch boundary. This allows me to test three codeswitching frameworks, the established Matrix Language Framework (Myers-Scotton, 2002) and two new frameworks that I propose, both of which integrate incremental predictions into their accounts of bilingual comprehension: the Current Word Hypothesis and the Surprisal Codeswitching Hypothesis. Each of the three frameworks propose that bilinguals use different types of information to predict upcoming language. The Matrix Language Framework proposes that bilinguals use the predominant language of the sentence to predict the upcoming word order of a sentence. The Current Word Hypothesis proposes that bilinguals use the language and grammatical category of the current word to predict the grammatical category of an upcoming word. The Surprisal Codeswitching Hypothesis proposes that bilinguals use the entire left context to predict upcoming words. Before testing the codeswitching frameworks, I identified which types of Spanish adjectives (pre-nominal; post-nominal; change: adjectives that change meaning based on their position; or no change: adjectives that do not change meaning based on their position) maximize the grammatical difference between Spanish and English. In an offline rating task, Spanish/English bilinguals preferred post-nominal and no change adjectives in the post-nominal position, and these were used in subsequent experimental stimuli. I then investigated bilingual processing of determiner/noun codeswitches where Spanish and English have the same word order and adjective/noun codeswitches where Spanish and English have different word orders in a stop-making-sense task. I established the task's viability for evaluating codeswitch comprehension and the predictions of the Matrix Language Framework and the Current Word Hypothesis. I then tested the two frameworks against each other in the same task and found overwhelming support for the Current Word Hypothesis. Finally, I compared surprisal as computed by GPT-3 to the human stop-making-sense data to evaluate if the Surprisal Codeswitching Hypothesis, Current Word Hypothesis, or Matrix Language Framework provide the best account for human codeswitch comprehension. Overall, I found support for codeswitching frameworks that include incremental predictions, though the Surprisal Codeswitching Hypothesis does not subsume the Current Word Hypothesis. Further, I evaluated the extent to which multilingual large language models (LLMs) such as GPT-3 can be used as a mental model for bilingual comprehension of codeswitches and found that LLMs can account for codeswitch effects but cannot fully account for the effects of other experimentally manipulated variables.In sum, this dissertation presents five main contributions: 1) I advance two new theoretical frameworks for understanding bilingual codeswitch comprehension, the Current Word Hypothesis and the Surprisal Codeswitching Hypothesis; 2) I validated the use of the stop-making-sense task on multilingual stimuli; 3) I found evidence that bilinguals flexibly switch between their mental grammars on a word-by-word basis; 4) I evaluated the viability of using multilingual LLMs as a mental model for bilingual comprehension of codeswitches; and 5) I found that while GPT-3 surprisal is a strong predictor of human responses to codeswitched sentences, the Surprisal Codeswitching Hypothesis provides an incomplete account of bilingual processing of codeswitches. Instead, bilinguals can flexibly switch grammars on a word-by-word basis.
- 일반주제명
- Psychology
- 일반주제명
- Linguistics
- 일반주제명
- Cognitive psychology
- 일반주제명
- Language
- 키워드
- Codeswitching
- 기타저자
- University of Michigan Psychology
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164537
■00520250211153014
■006m o d
■007cr#unu||||||||
■020 ▼a9798384045625
■035 ▼a(MiAaPQ)AAI31631499
■035 ▼a(MiAaPQ)umichrackham005844
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a150
■1001 ▼aVernooij, Natasha.
■24510▼aEvaluating Prediction-Based Theories of Bilingual Comprehension of Spanish/English Codeswitches
■260 ▼a[Sl]▼bUniversity of Michigan▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a156 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Boland, Julie E.;Lewis, Richard L.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2024.
■520 ▼aThis dissertation investigates how bilinguals use their two grammars to comprehend written intra-sentential codeswitches. I focus on adjective/noun constructions in Spanish and English where I manipulate the congruence of grammatical word order in the two languages across the codeswitch boundary. This allows me to test three codeswitching frameworks, the established Matrix Language Framework (Myers-Scotton, 2002) and two new frameworks that I propose, both of which integrate incremental predictions into their accounts of bilingual comprehension: the Current Word Hypothesis and the Surprisal Codeswitching Hypothesis. Each of the three frameworks propose that bilinguals use different types of information to predict upcoming language. The Matrix Language Framework proposes that bilinguals use the predominant language of the sentence to predict the upcoming word order of a sentence. The Current Word Hypothesis proposes that bilinguals use the language and grammatical category of the current word to predict the grammatical category of an upcoming word. The Surprisal Codeswitching Hypothesis proposes that bilinguals use the entire left context to predict upcoming words. Before testing the codeswitching frameworks, I identified which types of Spanish adjectives (pre-nominal; post-nominal; change: adjectives that change meaning based on their position; or no change: adjectives that do not change meaning based on their position) maximize the grammatical difference between Spanish and English. In an offline rating task, Spanish/English bilinguals preferred post-nominal and no change adjectives in the post-nominal position, and these were used in subsequent experimental stimuli. I then investigated bilingual processing of determiner/noun codeswitches where Spanish and English have the same word order and adjective/noun codeswitches where Spanish and English have different word orders in a stop-making-sense task. I established the task's viability for evaluating codeswitch comprehension and the predictions of the Matrix Language Framework and the Current Word Hypothesis. I then tested the two frameworks against each other in the same task and found overwhelming support for the Current Word Hypothesis. Finally, I compared surprisal as computed by GPT-3 to the human stop-making-sense data to evaluate if the Surprisal Codeswitching Hypothesis, Current Word Hypothesis, or Matrix Language Framework provide the best account for human codeswitch comprehension. Overall, I found support for codeswitching frameworks that include incremental predictions, though the Surprisal Codeswitching Hypothesis does not subsume the Current Word Hypothesis. Further, I evaluated the extent to which multilingual large language models (LLMs) such as GPT-3 can be used as a mental model for bilingual comprehension of codeswitches and found that LLMs can account for codeswitch effects but cannot fully account for the effects of other experimentally manipulated variables.In sum, this dissertation presents five main contributions: 1) I advance two new theoretical frameworks for understanding bilingual codeswitch comprehension, the Current Word Hypothesis and the Surprisal Codeswitching Hypothesis; 2) I validated the use of the stop-making-sense task on multilingual stimuli; 3) I found evidence that bilinguals flexibly switch between their mental grammars on a word-by-word basis; 4) I evaluated the viability of using multilingual LLMs as a mental model for bilingual comprehension of codeswitches; and 5) I found that while GPT-3 surprisal is a strong predictor of human responses to codeswitched sentences, the Surprisal Codeswitching Hypothesis provides an incomplete account of bilingual processing of codeswitches. Instead, bilinguals can flexibly switch grammars on a word-by-word basis.
■590 ▼aSchool code: 0127.
■650 4▼aPsychology
■650 4▼aLinguistics
■650 4▼aCognitive psychology
■650 4▼aLanguage
■653 ▼aCodeswitching
■653 ▼aIncremental predictions
■653 ▼aLarge language models
■653 ▼aSurprisal Codeswitching Hypothesis
■653 ▼aTheory comparison
■653 ▼aSpanish/English bilinguals
■690 ▼a0621
■690 ▼a0633
■690 ▼a0290
■690 ▼a0679
■71020▼aUniversity of Michigan▼bPsychology.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164537▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


