Proficient readers skip more words: examining to what extent EZ-Reader accounts for reader skill modulation of skipping rates

Abstract

Readers skip about 20-30% of the words they read (Rayner, 2009; Schotter et al., 2012). Due to having better lexical representations of words, better readers are thought to skip words more frequently than worse readers (Eskenazi & Folk, 2015). In this paper, a literature review is made of this hypothesis, looking at what influences skipping rates, and how reader skill could modulate this. Then EZ-Reader is explained, a mathematical model that models human reading behavior (Reichle et al., 2003). Subsequently, this model is used to run simulations on the GECO corpus (Cop et al., 2016), where simulation results are compared with empirical results for monolinguals (high-skill readers) and bilinguals (low-skill readers). Measures were taken of the total skipping rate and skipping rates for both low- and high-frequency words. There are conclusive findings that EZ-Reader can not account for reader skill modulation of skipping rates in the current implementation. Low-frequency skipping rates are predicted to be much higher than the empirical results for low-skill readers. The implications of these results are discussed, as well as their relevance towards AI.