Prior research has suggested an ERP correlate of implicit memory for words consisting of a centro-parietal positivity around 400 ms. We attempted (1) to replicate this ERP modulation in a different task, involving only trials with correct responses, and (2) to compare the findings to the domain of faces. Two experiments were conducted with a modified Sternberg task, in which both targets and nontargets were presented repeatedly. In Experiment 1, positive ERP differences between repeated and new nontargets were observed, which were domain-specific in topography and, for words, replicated the previously reported findings. In Experiment 2, the amplitude of the modulation for words, but not for faces, was unaffected by a variation of the level of processing during encoding, supporting the implicitness of the processes underlying the ERP modulation to nontarget words.