Differential rehearsal account and the inhibition account were proposed to explain the directed forgetting effect. The differential-rehearsal account claims that to-be-remembered items receive more rehearsal than to-be-forgotten items. However, the inhibition account argues that the inhibition system prevents to-be-forgotten items from receiving further processing. Both accounts were supported by some electrophysiological evidences. But few study directly investigated the mechanism when the remembering/forget cue presented, and the brain potential features of the load effect on forgetting remains unclear. Therefore, we hypothesized that if forgetting is merely a passive process, then there would be no active effort required to process the to-be-forgotten information. Accordingly, there would be no load effect on the ERPs elicited by forgetting cues. On the contrary, if forgetting is an active process involving cognitive inhibition, then forgetting different amounts of information would require different levels of cognitive resource. Accordingly, the high load of forgetting was expected to be different from the low load of forgetting, reflected by the ERP amplitudes. Specifically, we predicted that the N2-P3 complexes might be sensitive to memory loads, because the N2 component has been related to forgetting attempts, while P3 has been associated with inhibition process and was sensitive to memory load. A novel short-term memory test was used in this study with Arabic digits (1-9) and letters (A-Z) as stimuli. Within each trial, two memory items were presented sequentially, followed by a cue indicating either the presented items should be remembered or not. The amount of memory information was much (high load condition) or less (low load condition). After an interval (a calculation operation), subjects were asked to make a response to the probe stimuli. The event related potentials (ERP) locked to the cues. Fourteen undergraduate students (8 male and 6 female, with a mean age of 21.4 years, SD = 1.6) were recruited for the present study. Repeated measures ANOVA with cue (remember vs. forget), load (high vs. low), caudality and electrode site as the within subject factors were performed on the mean amplitudes of the ERPs evoked by the cues. The ERP results revealed that: (1) remembering cues evoked more positive ERPs than forgetting cues during the 280-430 ms time window, with a maximum difference at frontal scalp sites; (2) the load effects were observed for both the remembering cues and the forgetting cues. Cues in the high load condition evoke more positive ERPs than that of the low load condition. The ERP results suggested that different cognitive resource was relocated to different cues, as well as to different amount of memory. The to-be-remembered items receive more cognitive resources than the to-be-forgotten items which reflected by the more positive ERPs. The marked load effect on forgetting might correlate to the active manipulation of working memory such as removing or inhibiting the abundant to-be-forgotten information. Different cognitive resource was recruited in the process of inhibiting different amount of irrelevant information. More cognitive resources were recruited in inhibiting more irrelevant information. These results indicated that the inhibition plays an important role in directed forgetting.