Frequency and Predictability Effects for Line Final Words

Abstract

Computational models of eye movement control during reading have revolutionized the study of visual, perceptual, and linguistic processes underlying reading. However, these models can only simulate and test predictions about the reading of single lines of text. Here we report two studies that examined how placeholders for lexical processing (frequency and predictability) influence the processing of line-final words. The first study was a linear mixed-effects analysis of the Provo Corpus, which included data for 84 readers reading 55 multi-line texts. The second study was a pre-registered eye movement experiment, where 32 participants read 128 items where frequency, predictability, and position (intra-line vs line-final) were orthogonally manipulated. Both studies were consistent in showing that reading times were shorter on line-final words. While there was mixed evidence for frequency and predictability effects in the Provo Corpus, our experimental data confirmed additive effects of frequency and predictability for line-final words which did not differ from those for intra-line words. We conclude that while models that make additive assumptions about the role of frequency and predictability may be better suited to modelling the current findings, additional assumptions are required if models are to be capable of modelling shorter reading times on line-final words.