In this paper my goal is to investigate the following acquisition questions: Broadly, what can be learned from ambiguous evidence? More specifically, what learning outcomes or parameter settings do we expect if (essentially) all of the learner's input is ambiguous? That is, what similarities do we expect to see across grammars for learners of the same language? And what points of variability do we expect to see, and under what conditions? I illustrate the puzzle of ambiguous evidence by showing that there is a high degree of ambiguity in canonically verb-final languages. In particular, I show that there is ambiguity for two kinds of parameters: parameters concerning head-complement order and parameters concerning verb raising (i.e. head movement). By focusing on Korean, I claim that this ambiguity can be found in essentially all of the learner's input. I address the puzzle of ambiguous evidence with a probabilistic learning model. This model will provide a proof-of-concept illustration of how the acquisition questions above can be answered. We will see that the model will learn the grammar(s) of best fit to the input. Further, learning is systematically affected by parameter interaction. Results show that the model is highly successful at learning a consistently head-final grammar, but that there is variability as to whether there is verb movement. I conclude by drawing a connection between the modeling results here and the experimental results of Han et al. (2007).
Proceedings of the 33rd West Coast Conference on Formal Linguistics
edited by Kyeong-min Kim, Pocholo Umbal, Trevor Block, Queenie Chan, Tanie Cheng, Kelli Finney, Mara Katz, Sophie Nickel-Thompson, and Lisa Shorten Table of contents
ISBN 978-1-57473-469-0 library binding
viii + 426 pages
publication date: 2016
published by Cascadilla Proceedings Project, Somerville, MA, USA