ERPs elicited by correctly classified unstudied items in tests of yes/no recognition memory were used to investigate the neural correlates of retrieval cue processing. Items in Experiment 1 consisted of pictures and their corresponding names, allowing study and test material to be factorially crossed in four separate study-test cycles. The ERPs elicited by unstudied pictures and words were, in each case, more negative-going when the study material belonged to the alternative rather than the same class of items. These findings demonstrate that previously reported ERP "retrieval orientation effects" depend on differences in similarity between study and test items, and not on the form of the sought for material. In Experiments 2a and 2b, study materials were auditory words and pictures, and the test items were visual words. In both experiments, ERPs elicited by unstudied test words were more negative-going when pictures rather than auditory words were the study material. Thus, ERP retrieval orientation effects do not depend on the employment of a copy cue condition. It is proposed that the effects reflect differences in the processing necessary to maximize overlap between cue and memory representations.