Recent models on cognitive aging consider the ability to maintain and update context information to be a key source of age-related impairments in various cognitive tasks (Braver & Barch in Neuroscience & Biobehavioral Reviews, 26: 809-817, 2002). Context updating has been investigated with a modified AX-continuous-performance task by comparing performance and brain activity between context-dependent trials (i.e., correct responses require updating of the preceding cue information) and context-independent trials (i.e., correct responses are independent of cue information). We used an event-related potential (ERP) approach to identify sources of age differences in context processing in the early and late processing of cue information. Our behavioral data showed longer latencies and higher error rates on context-dependent than on context-independent trials for older than for younger adults, suggesting age-related impairments in context updating. The ERP data revealed larger P3b amplitudes for context-dependent than for context-independent trials only in younger adults. In contrast, in older adults, P3b amplitudes were more evenly distributed across the scalp and did not differ between context conditions. Interestingly, older but not younger adults were sensitive to changes of cue identity, as indicated by larger P3b amplitudes on cue-change than on cue-repeat trials, irrespective of the actual context condition. We also found a larger CNV on context-dependent than on context-independent trials, reflecting active maintenance of context information and response preparation. The age-differential effects in the P3b suggest that both younger and older adults were engaged in updating task-relevant information, but relied on different information: Whereas younger participants indeed relied on context cues to update and reconfigure the task settings, older adults relied on changes in cue identity, irrespective of context information.