Cloze probability norms for a dual-sentence prediction paradigm

A central component of sentence comprehension is prediction in which listeners and readers anticipate impending words. This prediction takes place as the sentence unfolds, indicating an incremental view of sentence processing based heavily upon prediction (Tyler & Marslen-Wilson, 1977). Additionally, the brain organizes semantic information such that words whose referents share semantic features are partially activated (McRae et al., 1997). Therefore, if we predict a word incorrectly, our language processing system is better equipped to deal with the unpredicted word if it shares features with the predicted word (Federmeier & Kutas, 1999).

This is often tested by using sentences with a lot of context to prime for a certain word. These sentences must have a high cloze probability, which is the odds that a person will predict the word based on the context of the sentence. However, a single sentence design, which has commonly been used, does not allow for expansion to test related words, as there are very few plausible endings. A two-sentence design with a heavily contextual first sentence and a carrier phrase opens the door for researchers to add many different plausible sentence endings, thereby adding more adaptability into the experimental design while maintaining a high cloze probability.

The aim of the present study was to extend this contribution by expanding the existing database of high-constraint, high cloze probability sentences to designs with two sentences, and is especially useful for experiments that examine featural similarity in greater depth. In total, 120 high-constraint sentences were validated and normed.