Objectives: The primary aim was to investigate whether computer-assisted reading intervention somehow can affect event-related potentials (ERP) and mismatch negativity (MMN) in hearing impaired (HI) children with hearing aids (HAs) and normal hearing (NH) children. Methods: The study included 15 HI children with sensorineural hearing loss (SNHL) using bilateral HAs and 14 NH children as a reference group; all children between the ages of 5 and 8. A multi-feature MMN-paradigm, Optimum-1, with a standard stimulus alternating with 5 different deviants was used. ERPs were recorded pre and post intervention, i.e. one month of repeatedly computer-assisted training (GraphoGame). MMN was calculated from the average ERP of each deviant minus standard. Data were based on samples within a specific time interval, 80-224 ms, and repeated measures ANOVA was used to analyze possible interactions. Results: There was a significant difference between groups before training, though, the mean obligatory responses or MMN was not statistically significantly different before versus after training, neither among the NH nor the HI children. Further, the HI children did generally achieve lower levels in GraphoGame compared to the NH children. Altogether, our findings indicate differences between the groups and that training may affect the neurophysiological processing in the brain, gaining the HI children. Both MMN and positive mismatch response (pMMR) were seen among both the HA and NH children, irrespective to deviant type. Individually, changes of the MMN and pMMR after training seem unpredictable. Conclusion: There are statistically significant differences in both the obligatory responses in ERP and the MMNs between the NH and HI groups before the computer-assisted training. Though, these differences disappear after the intervention. This suggests possible training effects regarding the central auditory processing among the HI children.