The influence of lexicon size and font characteristics on information optimization in reading: an orthographic prediction model investigation

The influence of lexicon size and font characteristics on information optimization in reading: an orthographic prediction model investigation

First Author: Benjamin Gagl -- Goethe University Frankfurt
Additional authors/chairs: 
Simon Schug; Chistian Fiebach
Keywords: visual word recognition, Font type, Vocabulary, Computational modelling, Abstract representations
Abstract / Summary: 

Neurophysiological evidence suggests energy-efficient visual processing. Taking this constraint into account, we developed and evaluated (on behavioral, fMRI and EEG data from multiple languages) an orthographic prediction model of reading according to which redundant perceptual input is "explained away" so that subsequent stages of word recognition are based on an optimized representation: the orthographic prediction error. Currently, the orthographic prediction error is determined by the size of the lexicon (i.e. words included for the redundancy estimation) and the visual occurrence (i.e. font). Here we investigate the characteristics of the influence of lexicon size and font type on the orthographic prediction error. Both investigations were performed using linear mixed models to fit reaction times from a lexical decision task (35 normal fluent readers of German; 800 words, 400 pseudowords and 400 consonant strings). Contrasting model fit was realized by estimating orthographic prediction error effects, on the basis of multiple implementations using parameterizations of lexicon size and font type. Lexicon size below 500 words resulted in an exponential increase of model fits leveling out for larger lexicon sizes. For the fonts comparison we found highest model fits for the presented font (Courier New) and a systematic decrease from fonts with comparable (i.e. other mono-spaced) to dissimilar (i.e. proportionally-spaced) characteristics. Thus, information optimization is influenced by font systematically depending on visual characteristics and lexicon size only when the numbers of words was assumed below 500.