The reliability, stability, and signal-to-noise ratio (SNR) of event-related potentials (ERPs) were investigated in children, adolescents, younger adults, and older adults in performance monitoring tasks. P2, N2, P3, and P2-N2 peak-to-peak amplitude showed high odd-even split reliabilities in all age groups, ranging from.70 to.90. Multigroup analyses showed that test-retest stabilities (across 2 weeks) of ERP amplitudes did not differ among the four age groups. In contrast, relative to adolescents and younger adults, SNRs were lower in children and older adults, with higher noise levels in children and lower signal power in older adults. We conclude that age differences in the SNR of stimulus-locked ERPs can be successfully compensated by the averaging procedure with about 40 trials in the average. However, age differences in baseline noise and split-half reliability should be considered when comparing age groups in single trial measures or time-varying processes with ERPs.