Towards Language Independent Detection of Dyslexia with a Web-based Game
Maria Rauschenberger, Luz Rello, Ricardo Baeza-Yates, Jeffrey P. Bigham · 2018 · Proceedings of the 15th International Web for All Conference (W4A 2018) · doi:10.1145/3192714.3192816
Summary
This paper presents MusVis, a web-based game designed to detect dyslexia in children using musical and visual elements that are language independent, potentially enabling screening before children learn to read. Dyslexia affects 5-15% of the world’s population and early detection is critical because intervention is most effective before negative consequences like school failure accumulate — yet current detection methods require expensive psychologists or specialised hardware (eye trackers, MRI), are language-dependent (requiring adaptation for each new language), and can only be used once children are already learning to read. MusVis exploits two well-established findings from dyslexia research: children with dyslexia show differences in auditory processing (particularly for acoustic parameters like rise time, sound duration, frequency, and rhythm, which are related to the prosodic structure of language) and in visual-spatial attention (particularly for features involving visual symmetry, mirror letters, and symbol search). The musical part adapts the card game Memory: participants click sound cards to find matching pairs across four stages that each vary one acoustic parameter (frequency, length, rise time, rhythm). The visual part adapts Whac-A-Mole: participants see a target visual element for three seconds, then must click it as many times as possible within 15 seconds among distractors, across four stages with different visual types (symbol, z-shape, rectangle, face) featuring mirror, rotational, and symmetry variations known to be difficult for people with dyslexia. The game takes less than 10 minutes per part and uses gamification elements (points, rewards, progress bars, story) to maintain engagement.
Key findings
A study with 178 participants (67 diagnosed with dyslexia, 111 controls) aged 7-12 speaking Spanish (n=108), German (n=57), English (n=6), and Catalan (n=7) revealed eight game measures with significant differences between Spanish children with and without dyslexia, and four measures that showed language-independent significance. For the musical part, the 4th click time interval (p=7e-12, r=0.66) and 6th click time interval (p=2e-3, r=0.30), duration (p=1e-5, r=0.42), and average click time (p=2e-16, r=0.78) were all significant for Spanish, with children with dyslexia taking longer between clicks — consistent with slower auditory processing. However, musical measures were not language-independent, as German participants with dyslexia showed opposite trends on some measures, possibly due to cultural differences. For the visual part, four measures were language-independent across all languages: total clicks (p=3e-7), time to first click (p=3e-4), hits (p=4e-6), and efficiency (p=2e-4). Children with dyslexia consistently made fewer total clicks, took longer before their first click, had fewer hits, and were less efficient across both Spanish and German groups. Notably, accuracy and misses did not differ significantly between groups — children with dyslexia were not less accurate but were slower, consistent with the known slower reading rate in dyslexia. Children and parents provided overwhelmingly positive feedback (44 positive vs 7 negative comments), with children describing it as "the best day at school ever" and "fun and not boring."
Relevance
This paper addresses a significant global equity challenge: dyslexia screening is expensive, language-dependent, and typically delayed until children are already failing to learn to read. A free, web-based, language-independent screening game accessible from any browser could dramatically expand early detection, particularly in low-resource settings and for languages where formal dyslexia assessment tools do not yet exist. The finding that visual game measures (total clicks, time to first click, hits, efficiency) show language-independent differences between children with and without dyslexia is the paper’s most significant contribution, as it suggests the visual-spatial processing differences associated with dyslexia can be measured without any reading or language component. For accessibility practitioners, the game design principles are transferable: using acoustic parameters grounded in auditory processing research, visual elements designed around known perceptual differences, gamification to maintain engagement in what is essentially an assessment, and web-based delivery for universal access. The study’s limitations — small English and Catalan samples, cultural confounds in the musical part, potential for random correct answers in the memory game — point to needed future work, including a planned longitudinal study with pre-readers and a machine learning prediction model based on the eight Spanish indicators.
Tags: dyslexia · screening · serious games · language independent · auditory processing · visual perception · children · learning disabilities · early detection · phonological awareness