Good question!
Apparently your question is on backward masking, which means that the masker follows the stimulus (probe) in time. Backward masking generally occurs at higher levels, typically the cortex. In case of visual stimuli this can be the primary visual cortex, or V1 (Mace et al. 2005). Ongoing processing of the probe is then thought to be interfered with by the subsequent presentation of the masker, as long as the masker-probe interval is not too long. The PNAS article by Covaks et al., 1995 shows this nicely at the cortical level: neurons working on the stimulus change their response patterns when a masker is presented afterwards. So although there may be connections folding back and forth onto themselves, a serial 'one-way' route where the processing of a stimulus changes upon arrival of the masker is the most parsimonious, and experimentally supported explanation. Cortical processes are often affected by consciousness (alertness/vigilance) so the presence of backward masking effects can be interpreted as a marker of cortical higher-level processing.
Forward masking, and simultaneous masking, on the other hand works, work at the neural level through adaptation processes. Although the effects of forward masking can be observed at the cortical level as shown by Brosch & Screiner 1997, forward masking is caused by processes occurring in the periphery (e.g., the cochlea or auditory nerve in hearing) and are unaffected by mental state. Hence, the presence of forward masking indicates that the neural process under investigation reflects a peripheral process that can also occur subconsciously, whereas the presence of backward masking generally point towards a higher-level neural process that is typically affected by mental state. Hope this helps.