The hereditary basis of dyslexia makes it possible to identify children at risk early on. Pre-reading children genetically at risk received during 14 weeks a home- and computer-based training in phonemic awareness and letter-sound relationships in the context of reading instruction. At posttest training effects were found for both phonemic awareness and letter knowledge. Trained at-risk children (n=31) made more progress than untrained at-risk controls (n=26) and kept up with untrained not-at-risk controls (n=16). However, the headstart of the trained group did not affect beneficially first and second grade reading and spelling proficiency. Following the start of phonics-based instruction at school, the trained at-risk children could not be discriminated from the untrained at-risk controls, and they were delayed relative to the not-at-risk controls. In order to promote long-term benefits and prevent undoing the advantage of early intervention, delivery should be both home- and school-based and more effort needs to be put in throughout first grade.