Accurate differentiation of mental fatigue levels by analysis of electroencephalogram features is still a challenge. This deficiency originates in the inability of conventional electroencephalogram features to reveal significant changes in fatigue levels. Alternatively, evaluation of the extent of alertness by characterization of the P300 component is widely done by performing a recognition task. The goal of this study is to classify pre- and post-task fatigue levels by tracking the spatial activation of their P300 sources in comparison with differentiation of P300 waveform features. To track these sources, the P300 wave was extracted from the background electroencephalogram from all channels using conventional time-locked synchronous grand averaging over all time frames and subjects. Next, standardized low resolution electromagnetic tomography and shrinking standardized low resolution electromagnetic tomography were both applied to the extracted P300 wave of all channels to estimate the activity of the P300 sources. Thirty healthy subjects were recruited and their electroencephalogram signals were recorded from thirty channels during pre-task (30 minutes), task (60-90 minutes), and post-task (30 minutes) states. During each recording period, an equal number of audio and visual stimuli were applied to the subjects who were performing both audio and visual recognition tasks. Empirical results showed a significant decrease in P300 source activation over tempo-parieto-occipital areas (secondary association area) in the post-task mental fatigue level when compared with the pre-task. It is of interest that in most channels, no significant change in the amplitude/latency of P300 was observed between the two fatigue levels.