A review of the literature showed that in vigilance or oddball tasks, changes over time in event-related potentials (ERPs) and performance measures often seem to be unrelated, but a number of studies had some shortcomings. In the present study a visual vigilance experiment was carried out, in which single-trial ERPs and performance data of 40 males were obtained. A relationship between time trends in behavioral and ERP measures was found: an early P3 amplitude and response latency (RT) showed an inversely varying relation over time. Analysis of covariance showed that the two linear trends tap the same aspect of processing: both trends became insignificant when adjusted for common variance. A negative correlation between mean values of P3 amplitude and RT rather than for change scores has been observed in previous studies, but has been ignored in the literature. However, correlations with RT have also been reported for other ERP deflections. Although there were several correlations between mean scores, most time-induced changes in ERP parameters appeared to be unrelated to worsening performance, for which several explanations are advanced. The ERP results do not support the hypothesis that a decrement in performance is caused by increasing difficulty discriminating targets from nontargets. A gradual decline in effort or resources allocated to the task might be an alternative explanation of performance deterioration. However, an effort-hypothesis cannot easily be tested. Effort is often invoked post hoc, and has previously been associated with many ERP components. The relationship between ERPs and the signal detection measures "sensitivity" d' and "response bias" beta is discussed.