Features from stimuli presented at a high rate in a single spatial position (Rapid Serial Visual Presentation, RSVP) can migrate forming a wrong combination or illusory conjunction. Several serial and parallel models have been proposed to explain the generation of this type of errors. The behavioral results fit better the two-stage parallel model than other serial and parallel models. However, they have not been studied the psychophysiological correlates that distinguish successful bindings from Illusory Conjunctions. The goal here is to collect electrophysiological records during this task to determine the degree to which they converge with the evidence from behavioral results. One RSVP task required to identify the only uppercase word in a stream of lowercase words at a rate of 12 items/sec. As in previous experiments, more intrusions from post-target items than from pre-target items were observed. The results from eventrelated potentials are also more supportive for the two-stage parallel model than for the serial or other parallel models, as reflected in the differential waves associated to correct and wrong combinations.