Students in grades 5 to 9 (ages 10 to 14; 6 girls, 27 boys) who had persisting specific learning disabilities in transcription (handwriting and spelling) completed three kinds of composition tasks requiring translation (thought to written language) on iPads using alternating transcription modes (stylus or keyboard) across every three lessons: personal narratives (6 lessons) and written summaries about read source material (integrated reading-writing) and heard source material (integrated listening-writing) (12 lessons). Before composing summaries, students clicked sequentially one at a time onto translation strategies, which they read and heard through earphones, and could click on again as needed during summary writing: (a) Level I composing of the very next sentence, and (b) Level II composing of a higher-level discourse structure. ANOVAs showed that Level I strategies were used significantly more often than Level II strategies; but the main effect for transcription mode was not significant. Written summaries of read source material had more errors in main ideas and factual details than heard source materials, but not more irrelevant statements. Applications of results are discussed for using computers for writing instruction, not just accommodations, for students with persisting transcription disabilities.