2026-09-19T23:53:05Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/132678 2026-03-24 col_2142_5131 col_2142_11614 com_2142_5130 com_2142_11613 com_2142_8903

application/pdf en text Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2027-12-01 The student, Suyeon Hwang, accepted the attached license on 2025-12-01 at 15:55. The student, Suyeon Hwang, submitted this Dissertation for approval on 2025-12-01 at 16:10. This Dissertation was approved for publication on 2025-12-03 at 13:26. DSpace SAF Submission Ingestion Package generated from Vireo submission #23011 on 2026-02-19 at 18:46:28 When life gives you lemons, make a left: Investigation of contextualization of learning in speech production using speech errors Hwang, Suyeon 2025-12-03 Fisher, Cynthia L Fisher, Cynthia L Dell, Gary S Federmeier, Kara D Montag, Jessica L Willits, Jon A speech errors phonotactics implicit learning eng Language production is a flexible yet complex system and is constantly updated by new experience. Within this system, implicit phonotactic learning provides a window into how speakers acquire and generalize rules that govern sound patterns. This thesis investigates how such rules are represented, focusing on the role of mini-grammars, which are context-sensitive representational systems that allow learning of different patterns and protect the learning from interference. One hypothesis is that only self-interfering and English-inconsistent rules like 2nd-order vowel-contingent rules require segregation into a mini-grammar to be learned, whereas simple non-self-interfering 1st-order constraints integrate directly into the main phonotactic system. Experiments 1 and 2 confirmed this prediction: 1st-order constraints were learned rapidly and robustly but remained vulnerable to interference from English, suggesting no mini-grammar formation. Experiments 3 and 4 investigated whether 2nd-order rules, once learned inside a mini-grammar, generalize to structurally similar rules. Learners only learned the first rule and did not show evidence of learning the second rule, and later learning influenced the retention of the first rule. Computational modeling in Experiment 4 replicated these interference patterns but retained multiple constraints, unlike behavioral results. Thus, there was no clear evidence for generalization, suggesting that learned representations were specific to restricted consonants and could be reactivated to influence prior learning in similar contexts. 2025-12 Thesis https://hdl.handle.net/2142/132678 Copyright 2025 Suyeon Hwang Psychology Psychology University of Illinois Urbana-Champaign Ph.D. Dissertation