Effects of visual complexity on reading: Modelling traditional and simplified Chinese character reading development

Effects of visual complexity on reading: Modelling traditional and simplified Chinese character reading development

First Author: Li-Yun Wendy Chang -- National Taiwan Normal University
Additional authors/chairs: 
Ya-Ning Chang
Keywords: (Chinese) characters, Visual complexity, Reading development, Computational modelling, Reading fluency
Abstract / Summary: 

Purpose
Previous cross-orthography studies have shown that visual complexity is critical to reading development. Children who learn to read visually more complex orthographies tend to outperform children who learn to read relatively simple orthographies on visual-spatial tasks (McBride-Chang et al. 2011). However, whether there is a direct link between visual complexity and reading development is less clear. We investigated this issue using computational modelling.

Method
We trained and tested two connectionist models which learned to map orthography and phonology with thousands of simplified and traditional Chinese characters, respectively. We controlled the complexity of linguistic units but varied the complexity of orthography units. All the training procedures for both models were exactly the same. We used linear mixed effects models to analyse the models’ reading fluency on the shared characters between the two Chinese scripts.

Results
The simplified model learned faster and produced fewer errors than the traditional model. A significant three-way interaction among training time, character type and number of strokes showed that at early training there was no difference between the two models on the stroke effect. However, as training proceeded, a different pattern emerged, the traditional model was better at naming less complex characters (number of strokes less than 15) but worse at naming very complex characters.

Conclusions
We demonstrated a link between visual complexity of orthographic units and reading fluency with careful control for other linguistic units. Learning to read was more difficult for the model trained with complex orthographic units compared to that with simple orthographic units; however, with sufficient amount of exposure, the benefit of learning complex units might emerge. These findings provide evidence of perceptual processing that suggests orthographic complexity plays a role in reading.