Recently, it has been shown by Kujala and colleagues that an audio-visual training requiring the matching of visual and sound patterns has a remediating effect on dyslexia in children. In parallel, the training also altered the neural substrate of sound discrimination, as reflected in enhanced Mismatch Negativity (MMN) component amplitudes of the event-related brain potential (ERP). To shed light on the brain processes involved in the audio-visual training we developed an ERP-paradigm where predictive visual score-like symbols had to be matched to corresponding sounds. If the visual symbol was incongruent to the predicted sound, a brain response was elicited in healthy adults as early as 100 ms resembling the MMN component. It is argued that the auditory system can establish a representation of an expected stimulus on the basis of visual symbolic information which can be compared to the current auditory input at sensory levels of processing. In a subsequent study we examined dyslexic and control first grade Finnish school children with this Symbol-to-Sound matching paradigm. The brain response indicating the detection of an incongruency between the visual symbol and the predicted sound was markedly smaller in dyslexic children compared to control children over left hemisphere areas but not over right hemisphere areas. This is consistent with previous findings that also left hemisphere MMN amplitude is smaller in dyslexics and might indicate a left hemispheric deficit in dyslexia. It may thus be proposed that the audio-visual training helps remediating dyslexia by facilitating especially left-hemispheric functions.