Computer speech-based training of literacy skills in neurologically impaired children: a controlled evaluation

Twenty-two reading-disabled children were randomly assigned to one of four training conditions to evaluate the effectiveness of a computer speech-based system for training literacy skills. The sample included 17 children with significant neurological impairment of various etiologies (including spina bifida and hydrocephalus, seizure disorder, brain tumors, cerebral palsy, and head injury) and five developmental dyslexics. The training employed a "talking" computer system that provides synthesized speech feedback during the course of learning. The training conditions included three word recognition and spelling-training programs and a math-training control program. Three different literacy-training procedures were compared, with the size of the trained print-to-sound unit varying as letter-sound (LSD: train-->t/r/ai/n); onset-rhyme (OR: train-->tr/ain) and whole word units (WW: train-->train). All literacy-training groups made significant gains in word recognition and spelling, with the LSD- and OR-trained subjects making the greatest word recognition gains on the words that could be trained with segmented speech feedback (i.e., words with regular spelling-to-sound patterns). All literacy-training groups demonstrated significant transfer on uninstructed rhymes of instructed regular words, with the greatest degree of transfer achieved by the LSD-trained subjects. These findings suggest that the neurologically impaired children were able to profit from instructional procedures that segment the printed word into units corresponding to onsets, rhymes, and phonemes and that this segmentation training may facilitate transfer-of-training for them.