Doctoral Dissertation Research: Expectations and Noisy-Channel Processing of Relative Clauses in a verb-initial language

Expectations about the grammatical structure of a sentence and the meaning of that sentence strongly affect how humans process and interpret language. In some cases, sentence structures and meanings align with our expectations and sentences are easier to process: people can understand them more easily and are more likely to interpret them correctly. However, sometimes our expectations conflict with the actual interpretation of the sentence and make the sentence more difficult to process. This research investigates how the brain generates these expectations and how they are used in understanding sentences. The project studies this in an understudied language in experimental research, monitoring people’s eye movement as they read sentences, with a diverse set of native speakers in the U.S. and abroad. This research improves our understanding of the various factors that interact to affect how humans process language, which helps inform developments in areas such as technology and medicine. This doctoral dissertation project specifically answers the question, how do language comprehenders make use of statistical expectations about sentence structure in light of conflicting input (e.g., input using less common sentence structures)? The project investigates this question by further testing existing theories of language processing, such as memory- and expectation-based theories of processing difficulty, alongside other theories of language comprehension, in particular good-enough and noisy-channel processing theories. The project tests these theories by recording eye movements during reading, and identifying how these eye movements change while reading a difficult sentence as compared to an easy sentence. The researchers additionally test the readers’ understanding of each sentence with comprehension questions, evaluate how a reader’s expectations about the structure and the meaning of a sentence affect their final interpretation of the sentence, and evaluate findings through Bayesian mixed effects regression models. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.