The Role of Probabilistic Cues in L2 Processing: Verb bias in Spanish and English

Open Access
Dietrich, Amelia Jane
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
March 05, 2014
Committee Members:
  • Paola Eulalia Dussias, Dissertation Advisor
  • Paola Eulalia Dussias, Committee Chair
  • Carrie Neal Jackson, Committee Member
  • Rena Torres Cacoullos, Committee Member
  • John Lipski, Committee Member
  • Karen Lynn Miller, Committee Member
  • sentence processing
  • bilingualism
  • corpus linguistics
  • verb bias
  • psycholinguistics
  • Spanish
The majority of sentence processing research has focused on monolingual populations, but in today’s global society bilinguals are ubiquitous. In light of this, it is important to investigate sentence processing in bilinguals, and especially what factors might lead to high proficiency in a bilingual’s second language. One measure of this is how a bilingual achieves fully native-like performance in a bilingual’s second language. Verb subcategorization bias is one probabilistic cue shown to be used in monolingual sentence processing. Previous research with Spanish-English bilinguals has pointed towards the importance of immersion in the L2 for developing target performance with experience-based cues such as verb bias, but in populations immersed in their L1, transfer of L1 verb bias information has not been observed, even when highly activated cognate verbs are present in the stimuli. This raises the question: Is it the case that probabilistic cues like verb bias are not subject to L1 transfer during L2 processing, in contrast with what has been observed for lexical and grammatical properties? Or is there information missing from the assumptions about verb bias in the two languages in previous studies? To answer these questions, the studies in this dissertation had two principal goals: 1) To conduct a corpus study of complement-taking verbs in Spanish and English to determine the subcategorization biases of those verbs, as well as the linguistic contexts which co-occur with them; and 2) To use the information derived from the corpus study to inform and provide materials for an eye-tracking study which investigates how bilinguals use L1 and L2 verb bias information during online sentence parsing. Given that verb bias is one of many experience-based sources of information found to be useful during processing, bilingual groups immersed in their L1 and their L2 were included to examine how L2 immersion modulates the use of L1 and L2 information during processing in the L2. This dissertation is significant for a number of reasons. First, to my knowledge, this is the first large-scale corpus study investigating verb bias in Spanish. Through its design, this dissertation contributes a body of verb bias information in Spanish comparable to the existing information in English, thus paving the way for additional future verb bias research in other contexts when one of the languages studied is Spanish. In addition, this dissertation further strengthens and expands what is known about verb bias in English and cross-linguistically by developing among the first known multivariate models of verb bias in English and Spanish. The comparison of verb bias across corpora and cross-linguistically finds that while verb bias can differ from one language to another, the level of concordance across languages is actually quite high, and the level of concordance across corpora even higher, supporting a view that the structural features of verb bias are, to some extent, semantically driven. This dissertation is also among the first efforts to employ naturally occurring sentences selected from a corpus as stimuli in a lab-based processing study, thereby helping to close the perceived gap between laboratory results and what speakers do when processing naturally-occurring language. Twice, eye-tracking data from monolingual English speakers in this dissertation, using both laboratory-created and corpus-extract stimuli, replicated processing patterns found in previous experimental studies of verb bias and demonstrated that native speakers are sensitive to frequency-based cues to structure during online processing. Moreover, evidence from bilingual groups finds that with enough proficiency and immersion experience, second language learners are able to make use of those same cues in their target language. However, when bilinguals are exposed repeatedly to cognates which share form and meaning across languages, even the most proficient among them fall prey to L1 transfer effects encouraged by the heightened parallel activation of shared lexical items. In contrast, bilinguals who are highly proficient but have not been immersed in their second language long-term, resort to simplicity heuristics. From a theoretical standpoint, the comparison of processing patterns for sentences containing cognates and non-cognates informs models of second language processing and test the limits of L1 transfer as an L2 parsing strategy. The results herein provide evidence that ‘shallow structures’ are a likely a developmental phase, but not a permanent state of L2 learning which can be supplemented or even overcome through extensive language experience. It furthermore provides evidence of a usage-based approach to grammar and processing.