The current study sought to examine the relative contributions of encoding and retrieval processes in accessing contextual information in the absence of item memory using an extralist cuing procedure in which the retrieval cues used to query memory for contextual information were "related" to the target item but never actually studied. In Experiments 1 and 2, participants studied 1 category member (e.g., "onion") from a variety of different categories and at test were presented with an unstudied category label (e.g., "vegetable") to probe memory for item and source information. In Experiments 3 and 4, 1 member of unidirectional (e.g., "credit" or "card") or bidirectional (e.g., "salt" or "pepper") associates was studied, whereas the other unstudied member served as a test probe. When recall failed, source information was accessible only when items were processed deeply during encoding (Experiments 1 and 2) and when there was strong forward associative strength between the retrieval cue and target (Experiments 3 and 4). These findings suggest that a retrieval probe diagnostic of semantically related item information reinstantiates information bound in memory during encoding that results in reactivation of associated contextual information, contingent upon sufficient learning of the item itself and the association between the item and its context information.