Objective Human can focus attention to one sensory modality (i.e.auditory or visual) and reach a biased attentional state in a top-down manner, which will facilitate information processing in that sensory modality.Alpha band activity in human EEG was related to this intersensory selective attention in previous reports.Here we study the time course of intersensory selective attention by analyzing alpha-band ERD (event-related desynchronization) during the procedure.Methods Ten subjects (4 females) with a mean age of (21.3±2.7) years participated in the experiments.Cue-target paradigm was used, in which there were 3 kinds of cues: auditory (A), visual (Ⅴ) and audiovisual (AV).In each trial, the cue was firstly presented on a display and the subject was required to direct his/ her attention to a sensory modality as cued.After a variable cue-target interval (CTI), a compound audiovisual stimulus appeared as the target.Subjects performed a GO/NoGo task to detect the cued sensory part of the target in A or Ⅴ condition, and detect both sensory parts in AV condition.The EEG data were recorded from 30 scalp electrodes mounted in an elastic cap (EasyCapTM, BrainAmp amplifier, Brain Products GmbH, Munich, Germany).EEG was segmented to epochs according to different cues.Alpha-band ERD was calculated based on the artifact-free EEG epochs and then smoothed by averaging over time samples with a moving averaging window.To calculate the statistical significance of alpha-band ERD, 30 electrodes were divided to 13 cortical regions: PF(Fp1, Fp2), LF(F7, F3), MF(Fz), RF(F4, F8), LT(T7), RT(T8), LC(FC5, FC1, C3, CP5, CP1), MC(Cz), RC(FC6, FC2, C4, CP6, CP2), LP(P7, P3, TP9), MP(Pz), RP(P4, P8, TP10), and O(O1, O2, Oz).In addition to the pre-cue baseline, post-cue CTI was divided to three periods: begin, middle and end.Three-way repeated measures ANOVA was performed using SPSS, and the within-subject factors were defined as Sequence (baseline, begin, middle, end), Cue (A, V, AV) and Area (13 cortical regions listed above).Results Three-way repeated measures ANOVA of Sequence, Cue and Area showed a significant main effect of Cue on alpha-band ERD (F =5.062, P =0.018), and the factor of Area nearly reached the significant level (F =1.821, P =0.053).The interactions of Sequence×Area (F =1.766, P =0.006) and Sequence×Cue were also significant.The following analyses showed: (1) in the early post-cue period, there was no main effect for Cue (F =1.576, P =0.234); in both middle and late post-cue period, Cue had a main effect (middle: F =4.873, P =0.02; end: F =5.880, P =0.011).(2) for Ⅴ cues, Area had a main effect (F =2.504, P =0.004), while for A and AV cues, no main effect of Area was observed (A∶ F =0.965, P =0.487; V∶ F =1.486, P =0.141).Conclusion (1) Directing attention to visual modality caused a large scale alpha-band ERD over the fronto-central and parieto-occipital areas during the whole CTI.However, when attention was directed to auditory modality, parieto-occipital alpha-band ERD only occurred at the early CTI period while disappeared in the middle and late period of CTI.These results showed that maintaining attention on auditory or visual modality had different effects on alpha-band modulation.(2) All cues were presented thuough visual words, so the early CTI for A cues included the course of attention shift from visual to auditory modality, which could also cause alpha-band ERD.However, the mechanisms of this alpha-band ERD may be different from that caused by maintaining visual attention.