Moving one's eyes while reading is one of the most complex everyday tasks humans face. To perform efficiently, readers must make decisions about when and where to move their eyes every 200-300ms. Over the past decades, it has been demonstrated that these fine-grained decisions are influenced by a range of linguistic properties of the text, and measuring eye movements during reading has become one of the primary methods of studying online sentence comprehension. However, it is still largely unclear why linguistic variables affect the eye movement record in the ways they do. The present work begins to answer this question by presenting a rational framework for understanding eye movement control in reading, in which probabilistic language knowledge plays a crucial role. Specifically, the task of reading is taken to be one of sentence identification: readers move their eyes to efficiently obtain visual input, which they combine with probabilistic language knowledge through Bayesian inference to yield posterior beliefs about sentence form and structure. Simulations with implemented models within this framework demonstrate that it can provide a principled account of many aspects of reading behavior, including the influence of a number of linguistic variables. In addition, the framework suggests a novel explanation for one of the least understood aspects of eye movements in reading - regressive eye movements - and we present evidence from an eye tracking corpus to support this proposal

During reading, readers intentionally do not fixate a wordwhen highly confident in its identity. In a rational model ofreading, word skipping decisions should be complex functionsof the particular word, linguistic context, and visual informa-tion available. In contrast, simple heuristic of reading onlypredicts additive effects of word and context features. Here wetest these predictions by implementing a rational model withBayesian inference, and predicting human skipping with theentropy of this model’s posterior distribution. Results showeda significant effect of the entropy in predicting skipping abovea strong baseline model including word and context features.This pattern held for entropy measures from rational modelswith a frequency prior but not from ones with a 5-gram prior.These results suggest complex interactions between visual in-put and linguistic knowledge as predicted by the rational modelof reading, and a dominant role of frequency in making skip-ping decisions.

The standard model is that word identification in reading isa holistic process, most efficient when words are centered inthe visual field. In contrast, rational models of reading predictword identification to be a constructive process, where readersefficiently gather visual information about a word, and maycombine visual information about different parts of the wordobtained across multiple fixations to identify it. We tease apartthese accounts by arguing that rational models should predictthat the most efficient place in a word to make a second fixa-tion (refixation) depends on the visual information the readerhas already obtained. We present evidence supporting this pre-diction from an eye movement corpus. Computational modelsimulations confirm that a rational model predicts this find-ing, but a model implementing holistic identification does not.These results suggest that refixations can be well understoodas rationally gathering visual information, and that word iden-tification works constructively.