P300 event-related brain potentials (ERPs) were elicited using a simple discrimination task in which participants discriminated two different equiprobable visual stimuli with button-press responses (n = 20). A total of ten trial blocks were presented at 10-min intervals. P300 amplitude declined significantly, but peak latency did not change reliably across trial blocks. P300 amplitude demonstrated a reliable cyclical fluctuation across trial blocks, although P300 latency did not. Intra-trial block ERP variability was assessed by computing the correlation coefficients between the target and standard stimuli for amplitude and latency measures across participants within each trial block. P300 amplitude correlations were weakest at the Fz electrode, more strongly associated at Cz, and were most strongly correlated at Pz across trial blocks. P300 latency correlations were somewhat weaker and similar in strength across electrodes sites. The correlational patterns for both P300 amplitude and latency demonstrated reliable cyclical variation. The N100 component produced strong and consistent correlations for both amplitude and latency, whereas the P200 and N200 component measures evinced cyclical correlational patterns similar to the P300 across trial blocks. These results suggest that the stability of P300 and other component measures can vary appreciably within and across trial blocks in a manner that reflects ultradian variation in arousal level.