Learning variation and systematicity in language Christine Soh Yue

Format:

Book

Thesis/Dissertation

Author/Creator:

Yue, Christine Soh, author.

Contributor:

University of Pennsylvania. Linguistics., degree granting institution.

Language:

English

Subjects (All):

Linguistics.

Psychology.

Cognitive psychology.

0290.

0621.

0633.

Local Subjects:

Linguistics.

Psychology.

Cognitive psychology.

0290.

0621.

0633.

Genre:

Academic theses

Physical Description:

1 online resource (214 pages)

Contained In:

Dissertations Abstracts International 87-12B

Place of Publication:

Ann Arbor : ProQuest Dissertations and Theses, 2026

Language Note:

English

Summary:

When acquiring a language, the input rarely provides the learner with a clean set of labels and rules; instead, it offers a tangle of exceptions, ambiguity, and variability. To succeed, the learner must do more than simply record and repeat their input -- they must distinguish the meaningful signals of a systematic grammar. The present dissertation addresses this challenge by proposing and evaluating two mechanisms of learning that bring a learner from their first word to grammatical rules. First, learners learn by rote association using a memory-constrained hypothesis-testing mechanism; second, learners learn categorical and variable rules through the productive generalization of one or multiple forms, respectively. These two mechanisms both use the distributional information in the input but offer a sharp distinction: while rote-learning depends on token frequency (id est, the number of exposures), rule-abstraction relies on the type frequency (id est, the diverse distribution over items).I focus on two challenges in the acquisition process as case studies supporting this integrated theory: the mapping problem in early word learning and the search for productivity in variable rule learning (Chapter 2). I present the rote-learning model alongside simulations of existing experimental data (Chapter 3). Predictions made by the rote-learning mechanism are evaluated through four cross-situational word-learning experiments that capture challenges encountered in real-world learning scenarios, such as the absence of helpful referential context, the presence of conflicting evidence, and intervening word-learning exposures (Chapter 4). The rule-learning model is presented with a corpus analysis of variable Spanish Differential Object Marking (Chapter 5) and then evaluated through a pair of artificial language learning experiments, where the results support the unified account of acquiring rules driven by the search for productivity (Chapter 5). When the type distribution supports a single productive generalization, the learner generalizes categorically; and when it supports multiple generalizations over the same contexts, the learner acquires systematic variation. In sum, this dissertation argues that language learning is not a monolithic statistical process but rather a sequenced interaction of rote mapping and rule abstraction

Notes:

Source: Dissertations Abstracts International, Volume: 87-12, Section: B.

Advisors: Yang, Charles D.; Schuler, Kathryn D. Committee members: Trueswell, John C.

Ph.D. University of Pennsylvania 2026

Vendor supplied data

Local Notes:

School code: 0175

ISBN:

9798247979876

Access Restriction:

Restricted for use by site license