... statistic indicates the extent to which each surprisal estimate accounts for RT, and can thus serve as a measure of the psychological accuracy of each model However, this kind of analysis assumes ... probabilistic language model Journal of Machine Learning Research, 3:1137–1155 Marisa Ferrara Boston, John Hale, Reinhold Kliegl, Umesh Patil, and Shravan Vasishth 2008 Parsing costs as predictors of reading ... the ability of lexicalized surprisal to explain a significant amount of variance in RT data for naturalistic texts, over and above that accounted for by other low-level lexical factors, such as...