Date created: | Last Updated:

Category: Project

Description: During reading or listening, people can generate predictions about lexical and morphosyntactic properties of the upcoming input based on available context. Psycholinguistic experiments that study predictability or control for it conventionally rely on human-based approach and estimate predictability via the cloze task. Despite its ubiquitous use, the cloze task is criticized for its lexical biases and the lack of information about very improbable continuations. The present study investigated the alternative corpus-based approach for estimating predictability via language predictability models. First, we compared 5-gram and LSTM predictability models, trained on three text corpora and found that the LSTM performed better. Then we obtained cloze and corpus-based probabilities for all words in 144 Russian sentences, correlated the two measures, and found a strong correlation between them. Finally, we estimated how much variance in eye movements registered while reading the same 144 sentences was explained by each of the two probabilities. In addition to the traditionally studied lexical predictability, i.e. the activation of a particular word form, we analysed morphosyntactic predictability and its effect on reading over and above lexical predictability. We found that for predicting reading times, cloze and corpus-based measures of both lexical and morphosyntactic predictability are highly comparable. Therefore, corpus-based probabilities can substitute for cloze probabilities in reading experiments, thus sparing resources needed to collect cloze probabilities. Our results also indicate that in languages with rich inflectional morphology, such as Russian, pre-activation of word’s morphosyntactic features is much more common than prediction of their full identity.