The electrophysiological correlates of recognition memory for new associations were investigated in two experiments. In both experiments subjects first studied unrelated word pairs. At test, they were presented with old words in the same pairing as at study (same pairs), old words in a different pairing from study (rearranged pairs), and pairs of new words. In Experiment 1 the test requirement was to discriminate between old and new pairs and, for any pair judged old, to then judge whether the pair was the same or rearranged. In Experiment 2 the requirement was merely to discriminate between old and new pairs. Event-related potentials (ERPs) were recorded for correctly classified same, rearranged and new pairs. The ERPs elicited by same pairs exhibited a similar pattern of effects in both experiments. Relative to the ERPs to new pairs, these effects took the form of sustained positive shifts with two distinct scalp maxima, over the left temporo-parietal and right frontal scalp respectively. ERPs to rearranged pairs showed effects which were similar in scalp topography, but markedly smaller in magnitude. This pattern of ERP effects closely resembles that found previously for test items defined as recollected on the basis of their attracting a successful source judgement. The findings therefore suggest that associative recognition memory shares some of the recollective processes that are engaged by the requirement to retrieve contextual information about a study episode. The findings from Experiment 2 indicate that the processes associated with the recollection of associated pairs are engaged regardless of whether the retrieval of associative information is an explicit task requirement.