We examined the effect of excessive Internet use on the time-frequency characteristic of the electroencephalogram by wavelet trans-formed and non-negative matrix factorization (NMF). The event-related potentials (ERP) of normal subjects and excessive Internet users were acquired using the oddball paradigm experiment. We a.pplied the wavelet transformed and event-related spectral perturbation to ERP in order to extract the time-frequency values. The F-test statistics of time-frequency values were then decomposed into two com-ponents by NMF. Excessive Internet use resulted in a significant decrease in the P300 amplitudes (P < 0.05) and a significant increase in the P300 latency (P < 0.05) in all electrodes. The major effect of excessive Internet use on gamma oscillation occurred at ～300 ms after stimuli at 40-50 Hz on the parietal central region. Thus, these data suggest that excessive Internet use affects information coding and integration in the brain.