A SOURCE memory procedure was used to investigate whether event-related potentials (ERPs) reflect the retrieval of episodic information (recollection) on tests of word-stem cued recall. ‘Cued recall ERP effects’ for recalled items were found to differ only in their magnitudes, and not their scalp distributions, according to the accuracy of source memory. This finding strongly supports the hypothesis that cued recall ERP effects reflect quantitative variation in the amount or quality of information which can be retrieved about recently experienced events. Topographic analyses also revealed that the distribution of the cued recall ERP effects changed over time, reflecting the activity of at least two temporally and neuroanatomically dissociable neural populations contributing to recollection on this task.