Five empirical questions about predictability effects in reading, and what their answers might mean (1/2)

Adrian Staub (UMass Amherst)
Five empirical questions about predictability effects in reading, and what their answers might mean

In these two lectures I take up five specific empirical questions. Integrating my own work and others’, I address the following: (a) Is cloze probability the best, or the only, useful measure of predictability?  (b) What is the form of the relationship between predictability and reading time? (c) What are the effects of predictability manipulations on distributions of reading times? (d) Does the effect of predictability interact with the effect of word frequency? (e) How is the effect of predictability modulated by manipulations of parafoveal preview? I will argue that taken together, the answers to these questions suggest two important theoretical conclusions.  First, while language comprehension does involve forward-looking, anticipatory processes, these are best understood as reflecting diffuse pre-activation of multiple lexical representations, rather than discrete prediction of specific words; and second, most if not all of the effect of predictability operates at the very earliest stages of lexical processing, or even at pre-lexical stages.