We investigated the effect of randomizing different spatial and temporal parameters on saccadic reaction times (SRTs) in five human subjects, to explore the relative occurrence of express and regular saccades. Parameters randomized in various test sessions were: target direction (right/left), intertrial interval, fixation foreperiod and gap duration (two or three 400 msec) in gap and overlap trials. For the sake of comparison the parameters under consideration were kept constant in non-random control sessions. We found that express saccades were still present in the random test sessions but their relative frequency (and the number of regular saccades) obtained in a given test session depends on the type of randomized parameters: randomizing the intertrial interval or the fixation foreperiod in the gap task yielded modest but significant changes in the SRT distributions, express and fast regular saccades being present in both the control and the random conditions. Randomization of the fixation foreperiod in the overlap task, on the other hand, caused a quite drastic increase in the SRTs. Randomization of gap and overlap trials did not cause considerable effects on express and fast regular saccades in the gap trials, and the SRTs in the overlap trials were significantly increased only in two subjects. When two or three gap durations were randomly interleaved, we found effects that ranged from "negligible" (usually for the longest gap in a given test session) to highly significant differences as compared with the corresponding control condition. The results suggest that express saccades--as fast regular saccades--are visually guided saccades which occur when a certain state of saccade preparation has been reached before target onset. This state depends on the amount of activation in the brain structures involved in the control of attention and fixation, and the decision processes involved in saccade preparation.