Early components of visual evoked potentials (VEP) in EEG seem to be unaffected by target visibility in visual masking studies. Bridgeman's reanalysis of Jeffreys and Musselwhite's (1986) data suggests that a later visual component in the VEP, around 250 ms reflects the perceptual effect of masking. We challenge this view on the ground that temporal interactions between targets and masks unrelated to stimulus visibility could account for Bridgeman's observation of a U-shaped time course in VEP amplitudes for this later component. In an MEG experiment of metacontrast masking with variable stimulus onset asynchrony, we introduce a proper control, a pseudo mask. In contrast to an effective mask, the pseudomask should produce neither behavioral masking nor amplitude modulations of late VEPs. Our results show that effective masks produced a strong U-shaped perceptual effect of target visibility while performance remained virtually perfect when a pseudomask was used. The visual components around 250 ms after target onset did not show a distinction between mask and pseudomask conditions. The results indicate that these visual evoked potentials do not reveal neurophysiological correlates of stimulus visibility but rather reflect dynamic interactions between superimposed potentials elicited by stimuli in close temporal proximity. However, we observed a postperceptual component around 340 ms after target onset, located over temporal-parietal cortex, which shows a clear effect of visibility. Based on P300 ERP literature, this finding could indicate that working memory related processes contribute to metacontrast masking.