Research on context-mediated facilitation of recognition memory distinguishes between the effects of reinstating the exact same context previously associated with a target and a context that is familiar but not directly associated with the target. As both effects are difficult to produce reliably in recognition experiments, attention has turned to measures that may explain inconsistencies, such as the extent to which instructions encourage association between targets and contexts. The aim of the current study was to examine the distinctive and interactive effects of three factors that may lead to variability in context effects (CEs), namely type of instructions given at learning, delay between learning and test, and exposure time for targets and contexts at learning. Using a comprehensive paradigm developed by Vakil and colleagues, with photographs of faces serving as target and context stimuli, both exposure time and delay were shown to be associated with the occurrence of CEs and appeared to interact with one another in determining the nature of these effects. Unlike several previous studies, false alarms did not increase when foils were presented with familiar contexts. Also unexpectedly, the instruction manipulation did not appear to strengthen target-context binding. It may instead have increased attention to contexts at the expense of targets, as suggested by the finding that direct memory for context improved under associative instruction conditions. Overall, the study demonstrates the importance of understanding and controlling various factors that may potentially influence the emergence of both reinstatement and familiarity-based CEs, among them exposure time and learning-to-test delay.