Probabilistic Selection of Input in Morphophonological Acquisition

• Author / Creator

  Chen, Tsung-Ying

• This dissertation sets out to explain the development during morphophonological acquisition and its possible learning outcomes by constructing a Probabilistic Selection of Input (PSI) rich lexicon learning model, in part based on psycholinguistic evidence that rich language details are lexically encoded. Contrary to traditional UR assumptions based on lexical economy, PSI stores and associates all surface allomorphs of a morpheme in a rich lexicon as possible inputs of the morpheme. Through the lexical associations between stored allomorphs, the leaerner assigns a probability between 0 and 1 to each allomorph and probabilistically select the phonological input of the morpheme. Output pattern variation along the acquisition course is thus analyzed as results of different input preferences and corresponding grammar shifts at sequential stages, and a successful morphophonological learning stands for an adult-like lexical generalization (i.e. input probabilities) and phonological grammar captured by learners. Diachronic morphophonemic changes can nevertheless occur with a shift in input preferences over learning generations, which gradually leads to permanent grammar shifts.
  PSI is tested with computer simulations using corpus data as a training corpus in various case studies, including the acquisition of Dutch stem-final voicing alternation (Chapter 3), the diachronic change of Mandarin Tone 3 (Chapter 4), and the emergence of Korean stem-final obstruent variations (Chapter 5). Learning outcomes similar to the performance by native speakers in elicitation tasks are demonstrated in the PSI simulations as a result of temporarily or permanently selecting different stored allomorphs as phonological inputs.

• Subjects / Keywords

  • Morphophonology
  • Computer modeling
  • Language acquisition
  • Korean
  • Mandarin
  • Hybrid model
  • Dutch

• Graduation date

  Fall 2014

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R3MD5G

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.