返回Google圖書搜尋

The Reliability of Standard Reading Time Analyses and Understanding the Nature of Maintained Information in Speech Processing

Zachary James BurchillT. Florian JaegerLinda Liu

出版	University of Rochester, 2023
URL	http://books.google.com.hk/books?id=zpZd0AEACAAJ&hl=&source=gbs_api

註釋"This thesis comprises two topics. Chapter 2 focuses on how properties of a commonly analyzed type of data?reading times?violate assumptions used in common analysis approaches and the impacts of this mismatch. Using large-scale power simulations, we show that these violations do have practical impact on power and Type I errors, and that researchers cannot ignore the theory behind the analysis choices they make when analyzing reading times. Chapters 3 and 4 focus on information maintenance in speech recognition. In speech understanding, listeners must infer increasingly abstract representations from the speech signal. While this process requires discarding information about less abstract representations as higher level ones are inferred, listeners do not immediately discard this information after the next step in the inference process, though this is not well understood. One particularly critical question is what type of information listeners can maintain, and at which levels of representation. In this thesis we show that listeners can maintain phonetic information (or lower) beyond word offset and provide multiple paradigms for investigating this question further."--Page xvii.