EEG event-related potentials (ERPs) and fMRI blood oxygenation level dependent (BOLD) new item contrasts were employed in six experiments to investigate the neural correlates of retrieval orientations - cognitive states or 'sets' that align cue processing with the demands of a particular retrieval goal. It is proposed that the neural correlates of different retrieval orientations can be investigated by comparing the activity elicited by the same class of retrieval cues when these were used to probe memory for different kinds of information (Rugg & Wilding, 2000), preferably with new items in memory tests, so as to minimize the confound between cue processing and processes associated with successful retrieval. The reported results on new unstudied items replicate and extend recent findings (e.g. Robb & Rugg, 2002) by showing that retrieval orientation effects are robust ERP effects across different study and test materials and semantic encoding tasks. However, the occurrence of the same effect is sensitive to manipulations of the retrieval tasks employed, and further a one-to-one mapping of the reported scalp ERP effect onto the observed fMRI activations is difficult to make. In terms of a functional interpretation, the retrieval orientation effect seems to reflect differences in the need to constrain cue processing. Specifically, it is proposed that the more a cue representation must be constrained to the semantic/conceptual level, the more negative-going will be the associated ERPs.