1. By recording eye movements with the search coil technique and subjecting them to an accurate infinite Fourier transform algorithm, we describe the Fourier spectra of human and monkey saccades. In both species we find heretofore undescribed features consisting of a regular pattern of local minima in the power plot, which cannot be attributed to noise. The frequency of these minima is well correlated with saccade duration. 2. Computer simulation shows that if the pulse component of the saccade is considered to be rectangular, then the first of these minima (called M1) occurs at a frequency that is the reciprocal of the duration of the pulse. 3. Comparing the position of this component during individual monkey saccades with electrophysiological recordings of motoneurons during the same saccades leads to the conclusion that these minima are related to the burst components in ocular motoneuron discharges. Specifically, the reciprocals of the frequencies of these minima are correlated with the duration of the burst component in the motoneuron discharge. 4. In the Fourier spectra of human saccades, the relationship of the frequency of M1 to saccadic duration is a function similar to that in the monkey. This adds to the evidence that the human saccade also is driven by a pulse-step signal. 5. In both monkeys and humans, T1, the reciprocal of the frequency of M1, is shorter than both the saccade duration and the burst duration of individual motoneurons, even though neurophysiological studies in monkeys generally report the saccadic burst duration to be equal to the saccade duration. This probably arises because the saccadic pulse is not rectangular, with the extremes contributing very little energy to the Fourier spectrum. By further computer modeling we show these shape effects explicitly: as the rise- and falltime increase, making the pulse less rectangular, T1 becomes shorter; in addition, as the asymmetry of rise and fall increases, the depth of the minima is reduced. We conclude that T1 measures the "effective pulse" duration of the motoneuron. 6. There is a difference in the relationship of effective pulse duration to the saccade duration between short and long saccades. For saccades shorter than approximately 40 ms in the human and 50 ms in the monkey, the pulse width as measured by this technique varies little with saccade duration. For longer saccades, effective pulse width increases linearly with duration. We agree with others that for short saccades the pulse is both height- and width-modulated; but for longer saccades, height modulation saturates and only width modulation remains.(ABSTRACT TRUNCATED AT 400 WORDS)