The theoretical and empirical backgrounds for the utility of the P300 event-related potential (ERP) as a measure of cognitive aging are summarized. P300 latency data from 32 different normative aging studies are then reviewed and assessed with meta-analytic procedures. Evaluation of moderator variables indicates that sample characteristics, stimulus factors, and task conditions contribute significantly to the "normal" change in peak latency that occurs with aging. These findings are critiqued in the context of previous reports, and implications are outlined for future applications of ERPs to normative aging. It is concluded that P300 latency can provide useful information about cognitive aging but that specific variables must be considered to obtain more precise results.