It is a common belief that saccades can be initiated by two different sources of information: (1) Stimulus driven ("bottom-up processing") and (2) Concept driven ("top-down processing"). Most research involves only one of these two sources of information, although -- since only one saccade can be initiated at one moment of time -- the saccadic system must co-ordinate the two sources of information for every single saccade.

We have developed an experimental technique for comparing the influence of the respective sources of information. The stimuli are presented within a circular window (4 deg diameter). In the condition of stimulus driven processing (1) eight competing stimuli are exposed for 1500 ms in a circular arrangement around the fixation cross. This is followed by a briefly presented target stimulus in the centre of the display. Observers have to decide whether the target stimulus was identical with one of the immediately preceding stimuli.

In the concept driven condition (2), the target stimulus is presented in each trial prior to the display of the test stimulus. The underlying assumption of this technique is that the priming with a target produces a top-down processing of subsequent test stimuli. By balancing the stimuli over the 8 spatial positions and recording eye movements at each trial, it is possible to make independent comparisons of the relative salience of the stimuli as well as of the preferred spatial position.

We report three experiments where the latencies and landing sites of the first saccades were recorded by means of a Dual-Purkinje-Image Eye Tracker and compared between the condition of top-down and bottom-up processing. Using a screen with a background of medium luminance, the stimuli varied in spatial frequency content (low-pass filtering and high-pass filtering), two levels of luminance (positive and negative) and of contrast, resulting in 8 stimuli per display. It is demonstrated that some of the stimulus characteristics (contrast and spatial frequency) play a prominent role in bottom up processing, but no effect of luminance was observed. Surprisingly, no effect of top-down processing could be demonstrated in the first experiment. However, by introducing stimulus content as an independent variable, resulted in additional top down processing. This effect could be increased by presenting the same target over a whole block of consecutive trials.