Modeling the length effect for words in lexical decision: The role of visual attention

Abstract

The word length effect in Lexical Decision (LD) has been studied in many behavioral experiments but no computational models has yet simulated this effect. We use a new Bayesian model of visual word recognition, the BRAID model, that simulates expert readers performance. BRAID integrates an attentional component modeled by a Gaussian probability distribution, a mechanism of lateral interference between adjacent letters and an acuity gradient, but no phonological component. We explored the role of visual attention on the word length effect using 1,200 French words from 4 to 11 letters. A series of five simulations was carried out to assess (a) the impact of a single attentional focus versus multiple shifts of attention on the word length effect and (b) how this effect is modulated by variations in the distribution of attention. Results show that the model successfully simulates the word length effect reported for humans in the French Lexicon Project when allowing multiple shifts of attention for longer words. The magnitude and direction of the effect can be modulated depending on the use of a uniform or narrow distribution of attention. The present study provides evidence that visual attention is critical for the recognition of single words and that a narrowing of the attention distribution might account for the exaggerated length effect reported in some reading disorders.

Varieties of Cognitive Profiles in Poor Readers: Evidence for a VAS-Impaired Subtype

“…On the other hand, higher VAS abilities would facilitate parallel processing of the whole-word letter string, thus contributing to irregular word reading. VAS is known to reflect the amount of visual attention allocated to processing and how visual attention distributes over the letter string during reading (Ginestet et al, 2019; Lobier et al, 2013). Neuroimaging data support this interpretation, showing that VAS abilities relate to brain regions that are part of the dorsal attentional network (Lobier et al, 2012, 2014; Peyrin et al, 2011, 2012; Reilhac et al, 2013; Valdois et al, 2019b) but do not belong to the phonological brain network.…”

Orthographic learning of novel words in adults: effects of exposure and visual attention on eye movements

“…Current models (Pritchard et al, 2018;Ziegler et al, 2014) emphasise the impact of phonological decoding on orthographic learning while minimising the role of visual processing in this learning process. As acknowledged by Pritchard et al (2018), these models do not incorporate the sophisticated mechanisms of visual processing postulated by word recognition models, as the gradient of acuity (Whitney, 2001), lateral interferences (Davis, 2010;Gomez et al, 2008) or a visual attention component (Ans et al, 1998;Ginestet et al, 2019;Mozer & Behrmann, 1990). Evidence that visual attention contributes to orthographic learning would require the development of new models including both well-defined processes of phonological decoding, and fully specified mechanisms of visual processing, including a visual attention mechanism.…”

“…Performance on VA span tasks reflects the amount of visual attention capacity available for processing (Dubois et al, 2010;Lobier et al, 2013), which relates to the dorsal attention network (Lobier et al, 2014(Lobier et al, , 2012Reilhac et al, 2013;Valdois et al, 2019aValdois et al, , 2014. Individuals with higher VA span show more efficient word recognition skills, thus faster reading (Antzaka et al, 2017;Bosse & Valdois, 2009;Lobier et al, 2013;Valdois et al, 2019b), more accurate irregular word reading (Bosse & Valdois, 2009) and smaller length effects (van den Boer et al, 2013); for computational modelling, see Ginestet et al (2019). Exploration of their eye movements during text reading revealed lesser fixations and larger saccades, suggesting that more letters were simultaneously processed per fixation (Prado et al, 2007).…”

COSMO-Onset: A Neurally-Inspired Computational Model of Spoken Word Recognition, Combining Top-Down Prediction and Bottom-Up Detection of Syllabic Onsets

“…The decoding module of the COSMO-Onset model is inspired both by the BRAID model (Bayesian model of Word Recognition with Attention, Interference and Dynamics) of visual word recognition (Phénix, 2018;Ginestet et al, 2019) and by the classical, three-layer architecture of models of spoken word recognition, such as the TRACE model (McClelland and Elman, 1986). It can also be construed as a hierarchical (multi-layered) dynamic Bayesian network (Murphy, 2002), with an external component to control information propagation.…”