ebook img

Prediction, Error, and Adaptation During Online Sentence Comprehension by Alex Brabham Fine ... PDF

272 Pages·2013·14.74 MB·English
Save to my drive
Quick download

Preview Prediction, Error, and Adaptation During Online Sentence Comprehension by Alex Brabham Fine ...

Prediction, Error, and Adaptation During Online Sentence Comprehension by Alex Brabham Fine Submitted in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Supervised by Professor T. Florian Jaeger Professor Jeffrey T. Runner Department of Brain and Cognitive Sciences Department of Linguistics Arts, Sciences and Engineering School of Arts and Sciences University of Rochester Rochester, New York 2013 ii I dedicate this thesis to Carly, Gracie, Lawton, and Mylie. iii Biographical Sketch The author was born in Albemarle, NC, in June 1984. He attended the University of North Carolina at Chapel Hill, and graduated with a Bachelor of Arts in Linguistics (with Highest Honors) and a minor in German Literature (with no great distinction other than being really into Rilke) in 2006. He then worked as a research assistant at the Boston University School of Medicine from June 2006-July 2008. In the fall of 2008, he began doctoral studies in the Department of Brain and Cognitive Sciences and the Department of Linguistics at the University of Rochester. He received a Master of Arts degree in Brain and Cognitive Sciences in the Spring of 2011. In the Fall of 2010, he was awarded a National Science Foundation Graduate Research Fellowship. He pursued his research at the University of Rochester under the direction of Professors T. Florian Jaeger (Brain and Cognitive Sciences) and Jeffrey T. Runner (Linguistics). The following publications are a result of research the author conducted while pursuing doctoral studies at the University of Rochester: Farmer, T. A., Fine, A. B., & Jaeger, T. F. (2011). Implicit Context-Specific Learning Leads to Rapid Shifts in Syntactic Expectations. Proceedings of the 33rd Annual Meeting of the Cognitive Science Society. Farmer, T. A., Misyak, J. B., Fine, A. B., & Christiansen, M. H. (under revision). The “Ex-Factor”: Individual differences in linguistic experience help explain variability in measures of on-line syntactic processing skill. iv Fine, A. B., & Jaeger, T. F. (2011). Language comprehension is sensitive to changes in the reliability of lexical cues. Proceedings of the 33rd Annual Meeting of the Cognitive Science Society (pp. 925–930). Boston, MA. Fine, A. B., & Jaeger, T. F. (2013). Syntactic priming in language comprehension allows linguisic expectations to converge on the statistics of the input. Proceedings of the 35th annual meeting of the Cognitive Science Society. Berlin, Germany. Fine, A. B., & Jaeger, T. F. (2013). Evidence for implicit learning in syntactic comprehension. Cognitive Science. Fine, A. B., Jaeger, T. F., Farmer, T. A., & Qian, T. (under review). Rapid Expectation Adaptation during Syntactic Comprehension. Fine, A. B., Kleinschmidt, D., & Jaeger, T. F. (2012). A Bayesian belief updating model of syntactic expectation adaptation. 25th Meeting of the CUNY Conference on Human Sentence Processing. Fine, A. B., Qian, T., Jaeger, T. F., & Jacobs, R. A. (2010). Syntactic Adaptation in Language Comprehension. Proceedings of the 2010 Workshop on Cognitive Modeling and Computational Linguistics (pp. 18–26). Uppsala, Sweden: Association for Computational Linguistics. Kleinschmidt, D., Fine, A. B., & Jaeger, T. F. (2012). A belief-updating model of adaptation and cue combination in syntactic comprehension. Proceedings of the 34th Annual Meeting of the Cognitive Science Society. v Acknowledgments I’d first like to express my deepest gratitude to Florian Jaeger, who supervised and collaborated on every study reported in this thesis. Throughout graduate school, Florian has been an ardent supporter, a tough critic, and a patient teacher. I hope to repay some of my karmic debt to him by working as hard for my future students and collaborators as he has for me. I also thank Jeff Runner, with whom I had the pleasure of collaborating on a set of studies not included in this thesis. Jeff’s intelligence, calmness, and humanity helped me endure and find the humor in even the hardest periods of graduate school. I thank Mike Tanenhaus for his advice and encouragement over the years, and I thank him most of all for doing so much to create an environment that is such a stimulating and fun place to do psycholinguistics. Mike and Florian also deserve joint recognition for the seminars and lab meetings very early in my graduate career where the seeds of the main ideas in this thesis were planted. I thank the remaining members of my committee, Harry Reis and Greg Carlson, for their wisdom, kindness, and encouragement, and for teaching me a great deal about how to be a scholar and a teacher. I would like to thank Kathy Corser for all of her help and kindness throughout my time in Rochester, as well as Jen Gillis, Chris Dambra, and Chris Freemesser, for generally helping to make this department such an easy place to do research. vi I thank Bob Joseph, Daniela Plesa Skwerer, and Helen Tager-Flusberg at Boston University for teaching me how to work in an experimental psychology lab, and for taking a chance in hiring and mentoring someone with so little demonstrated ability to handle expensive equipment. I don’t think I would’ve had the opportunity to study in a program like the one at Rochester without their guidance. I thank my teachers at UNC—especially Misha Becker, Randy Hendrick, Laura Janda, and Craig Melchert—for encouraging me to pursue the study of language even when I was a shiftless undergraduate. I would like to single out Randy Hendrick, who tolerated my almost-daily visits to his office for four years without ever so much as hinting at my failure to comply with the unwritten dictates of appointment and office hour etiquette. I outline his contributions in more detail below, but I would like to take this additional opportunity to thank Andrew Watts for all of his help and patience over the years. The work reported here would not have happened—or would have happened much more slowly and at a much greater spiritual cost—without Andrew. I read somewhere that if you ever discover that you are the smartest person in the room, you should find a new room. I thank my fellow students (both at Rochester and elsewhere) for never forcing me to find another room. For their consistent friendship, love, and intelligence I thank especially Esteban Buz, Thomas Farmer, Masha Fedzechkina, Maureen Gillespie, Peter Graff, Liz Hirshorn, Lizz Karuza, Chris Kim, Dave Kleinschmidt, Dan Pontillo, Ting Qian, Ben Van Durme, and Ilker Yildirim. Special thanks go to Judith Degen for her unwavering loyalty and friendship despite vii having such intimate knowledge of my shortcomings in the areas of toothpaste- purchasing, stove-cleaning, and dog-fur-sweeping. Of those people just mentioned, I am particularly indebted intellectually to Thomas Farmer, Dave Kleinschmidt, and Ting Qian, whose feedback and advice on the work reported here made it much better than it would otherwise be. I also thank my friends outside of my academic circle who were with me the whole time, and who helped me establish the emotional bulwark necessary to do behavioral research for a living, especially Greg Buzulencia, Joyce Chapman, Ben Ellis, John Francis, Abby Glogower, Dan Miner, Perry Sherouse, and Martin Zahra. I thank Inbal for the surprise ending. Finally, and most importantly, I thank my family for their unconditional love. I thank my mother and father for always supporting and never pressuring me. I thank my brother and sister for being my best friends. I probably forgot someone. Please don’t over-interpret it. To the extent that this thesis contains any durable insights about human language processing, the individuals mentioned above deserve credit. All deficiencies in the thesis are my responsibility alone. viii Abstract A fundamental challenge for human cognition is perceiving and acting in a world in which the statistics that characterize available sensory data are non-stationary. This thesis focuses on this problem specifically in the domain of sentence comprehension, where linguistic variability poses computational challenges to the processes underlying sentence comprehension. We begin by framing the problem of linguistic variation, especially in syntax, and then propose that humans respond to this variability by continuously adapting to and learning the statistical regularities of novel linguistic environments. Experiments 1-6 provide evidence that, in the face of a novel environment whose statistics violate subjects’ expectations, subjects adjust their linguistic expectations in order to allow their expectations to converge towards the statistics of the current environment. These experiments tacitly assume that adaptation is error-sensitive: subjects adapt when their expectations are violated. Experiment 7 explicitly addresses and tests this tacit assumption. In that experiment, we provide a quantitative operationalization of the syntactic “error signal” in language comprehension, and ask how the magnitude of this error signal correlates with immediate changes in linguistic expectations. This experiment also provides the context for a more in-depth discussion of what type of cognitive mechanism might give rise to the experimental results reported in this thesis. Experiment 8 attempts to more directly address the presumed environment- specificity of adaptation by asking whether distributional information indexed to a particular environment and to particular verbs can be retained over multiple days. The ix work reported here takes a step toward synthesizing three lines of research in psycholinguistics that have previously proceeded largely in parallel: (1) experience- or expectation-based processing, (2) syntactic priming, and (3) statistical learning in language. The work here also suggests an intriguing if somewhat tentative link between adaptation in syntactic comprehension on the one hand and adaptation in lower-level speech perception and non-linguistic perception, where adaptation phenomena have previously been more thoroughly examined. x Contributors and Funding Sources This thesis was supervised by a committee consisting of Professors T. Florian Jaeger, Jeffrey Runner, Michael Tanenhaus, Greg Carlson, and Harry Reis. All stages of experiment design, development, analysis and interpretation were conducted in collaboration with Florian Jaeger. The data reported in Experiment 1 was provided by Thomas Farmer, and earlier versions of the analyses reported there were previously published in a proceedings paper in 2011, as well as in a journal article currently under review, both listed in the Biographical Sketch. The data for Experiments 2 and 3 was collected with the assistance of Andrew Watts. The data from Experiment 2 was published in a proceedings paper from 2013, listed in the Biographical Sketch. The data for Experiment 4 was collected with the assistance of Andrew Watts and Jeremy Ferris. Earlier analyses of that data were published in a proceedings paper from 2010, listed in the Biographical Sketch. Experiment 5 was conducted with the assistance of Andrew Watts and has not yet been reported in a publication. Experiment 6 was conducted with the assistance of Narissa Gogos and Nicole Craycraft, and the results from that experiment were published in a proceedings paper from 2013, as well as in a journal article currently under review, both listed in the Biographical Sketch. The data reported in Experiment 7 was originally collected and generously shared with me by Malathi Thothathiri and Jesse Snedeker. Some of the analyses reported for that experiment were published in an article appearing in the journal Cognitive Science, listed in the Biographical Sketch. The data reported in Experiment 8 was collected with the

Alex Brabham Fine. Submitted in .. experimental conditions. Mean by-region length-corrected RTs for Experiment 2 for both conditions. competition, constraint satisfaction, and—in the most general terms—experience-based.
See more

The list of books you might like

Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.