If you are interested, please attend the Q&A at Poster Session B (Friday 12:10 – 2PM). Here is the zoom link: https://mit.zoom.us/j/804326843 Email me your questions: ybian@mit.edu (Yuan Bian) Abstract: In the sentence processing literature, Gibson et al. 2013 found that people balance the prior probability of a sentence with the likelihood of the current sentence being corrupted by noise. Specifically, (1) people are more willing to adopt nonliteral inference if the plausible alternative involves positing fewer/minor changes to the literal sentence in terms of Levenshtein edit distance; (2) people are more likely to make a nonliteral inference if the literal sentence is one deletion away from the plausible alternative compared to one insertion (Bayesian size principle). These experiments often probe people’s understanding of a (declarative) prompt sentence by asking a following comprehension question. Hence, understanding how people process the questions themselves in context is crucial to the design of these experiments. In this poster, we replicated the effect of (1) & (2) in sentences of question format, as well as the declarative prompts.