Two instructional experiments used randomized, controlled designs to evaluate the effectiveness of writing instruction for students with carefully diagnosed dyslexia, which is both an oral reading and writing disorder, characterized by impaired "word" decoding, reading, and spelling. In Study 1 (4th to 6th grade sample and 7th to 9th grade sample), students were randomly assigned to orthographic or morphological spelling treatment but all students were taught strategies for planning, writing, and reviewing/revising narrative and expository texts. Both treatments resulted in improvement on three measures of spelling and one measure of composition. Morphological treatment resulted in better improvement in spelling pseudowords, whereas orthographic treatment resulted in better improvement in rate of reading pseudowords. In Study 2 (4th to 6th grade sample), students were randomly assigned to explicit language treatment (phonological working memory + phonological-orthographic spelling + science report writing training) or nonverbal problem solving treatment (virtual reality-based computer simulation, drawing maps, and building clay models). Both treatments used science content material and resulted in significant improvement in spelling and reading pseudowords (accuracy and rate). The surprising finding was that hands-on, engaging science problem solving led to more improvement in phonological working memory than did specialized phonological instruction. Only when spelling instruction emphasized orthography or morphology (Study 1) did real word spelling improve.