AuthorDrake, Shiloh N.
artificial grammar learning
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractThis dissertation addresses the question of whether non-adjacent morphological dependencies are as difficult to learn as non-adjacent phonological dependencies. Non-adjacent dependencies have been investigated in the past, and have proven to be at best difficult to learn (Bonatti, Peña, Nespor, & Mehler, 2005; Gómez, 2002; LaCross, 2011, 2015; Newport & Aslin, 2004), and at worst, completely unlearnable (Newport & Aslin, 2004: experiment 1). LaCross (2011, 2015) showed that speakers of a language employing non-adjacent dependencies were able to learn an artificial grammar employing analogous non-adjacent dependencies easily, suggesting there may be a linguistic bias that makes speakers more aware or capable of unconsciously parsing non-adjacent dependencies so long as they speak a language that employs vowel harmony. The research in this dissertation studies three subject populations with two tasks and two grammars to discover whether speakers of a language utilizing root-and-pattern morphology also have the ability to unconsciously parse non-adjacent dependencies pred- icated on morphological structure. Chapter 2 uses a segmentation or statistical learning task similar to the experiments mentioned above, while Chapter 3 uses a word elicitation task to establish a more fine-grained representation of what experiment participants learn after a very short exposure. The experiments show that there may be a cognitive bias toward concatenative morphology even among Arabic and Maltese speakers, but also that Arabic and Maltese speakers are willing to adjust CV skeleta and syllabic structure when deriving plural forms from singular forms. The methods that they use when producing novel plural forms are similar to those found in their L1, showing that this type of bias is predicated on morphophonological structure in the participants’ L1. The results together support a root-based lexicon for Arabic and Maltese and aggressive morphological decomposition (Boudelaa & Marslen-Wilson, 2001, 2004a, 2004b, 2015; Deutsch, Frost, & Forster, 1998; Frost, Deutsch, & Forster, 2000; Frost, Forster, & Deutsch, 1997; Ussishkin, Dawson, Wedel, & Schluter, 2015) even in novel words. Additionally, this work supports the notion of morphological abstraction, abstract grammatical features (such as past or plural) may be expressed by multiple allomorphs, particularly in the context of learning a new language. I extend this work to suggest that a processing model of Distributed Morphology (Halle & Marantz, 1993; Harley & Noyer, 1999; inter alia) would be appropriate both to model the results here and to better explain morphological processing disorders. Although Distributed Morphology has not been extensively tested as a processing model, recent research shows compatibility with existing psycholinguistic models (Gwilliams & Marantz, 2015; Stockall & Marantz, 2006) and has better explanatory power for deficits in morphological processing (Tat, 2013).
Degree ProgramGraduate College