A Computer-Based Method to Improve the Spelling of Children with Dyslexia
Luz Rello, Clara Bayarri, Yolanda Otal, Martin Pielot · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS 2014) · doi:10.1145/2661334.2661373
Summary
This paper presents a novel method for improving the spelling skills of children with dyslexia through error-based exercises delivered via an iPad game called DysEggxia (Piruletras in Spanish). Unlike traditional approaches that present correct words as positive examples, DysEggxia shows children misspelled words and asks them to identify and correct the errors. The exercises were designed using linguistic analysis of errors found in two corpora of texts written by children with dyslexia — 2,654 errors in English and 1,171 in Spanish — revealing six error types: letter insertion, omission, substitution, transposition, word boundary errors (splits and run-ons), and morphology errors. The game contains 5,000 exercises (2,500 per language) across six exercise types matching these error categories, with five difficulty levels calibrated to dyslexia-specific challenges (word frequency, length, phonetic similarity, morphological complexity). The rationale is that since people with dyslexia cannot consciously detect errors while reading, training them to identify typical dyslexic errors could develop compensating strategies for better writing. The interface was carefully designed following dyslexia accessibility research: black text on cream background, Helvetica font at minimum 18pt, animations instead of written instructions to minimize reading demands, and gamification elements (a growing penguin character, achievements).
Key findings
An eight-week within-subject experiment with 48 Spanish-speaking children with diagnosed dyslexia (ages 6-11) compared DysEggxia against Word Search (a control word game) in a crossover design with 12 sessions of 20 minutes per phase. Results showed that DysEggxia significantly reduced writing errors compared to the control condition: the rate of errors per word decreased by 20% (p = 0.029), and the severity of errors within incorrect words decreased by 14.5% (p = 0.011). However, DysEggxia did not significantly improve reading skills (p = 0.41), which the authors attribute to the exercises being designed specifically for spelling rather than reading, and the game requiring minimal reading beyond exercise words. Subjective perceptions of reading and writing ability did not change significantly either — consistent with the known difficulty people with dyslexia have in self-assessing their literacy performance, which is why dyslexia is often called a "hidden disability." The iOS version had been installed over 17,000 times and adopted by three schools into their curricula at time of publication.
Relevance
This is the first scientifically validated computer-based method shown to improve spelling skills in children with dyslexia — previous technology interventions had focused on reading. The error-based exercise approach is counterintuitive (presenting incorrect words rather than correct ones) but effective, suggesting that training error detection builds compensating strategies. For accessibility practitioners, the paper demonstrates how corpus-based linguistic analysis of disability-specific error patterns can inform the design of targeted therapeutic tools. The careful interface design choices (font, size, color, no-reading-required instructions) exemplify how to create accessible educational software for users with cognitive disabilities. Limitations include that results are validated only for Spanish (English has a more opaque orthography, so effects may differ), and the study lacked a no-game control condition. The approach of using NLP and corpus linguistics to create personalized accessibility interventions is a model worth replicating for other cognitive disabilities.
Tags: dyslexia · spelling · serious games · literacy · education · cognitive accessibility · children · tablet accessibility