Saccade target selection must be understood in relation to the obvious fact that vision naturally occurs in a continuous cycle of fixations interrupted by gaze shifts. The guidance of eye movements requires information about what is where in the visual field. The identities of objects are derived from their visible features. Single neurons in the visual system represent the presence of specific features by the level of activation; the reliability of the discriminating signal from single neurons varies over time. Each point in the visual field is represented by many populations of neurons activated by all types of features. Topographic representations are found throughout the visual and oculomotor systems; neighboring neurons tend to represent similar visual field locations or saccades. Selecting one out of many stimuli to which to direct gaze requires comparing stimulus attributes across the visual field. The existence of retinotopic maps of the visual field makes possible local interactions to implement such comparisons /41/. For example, a lateral inhibition network can extract the location of the most conspicuous stimulus in the visual field /30,40,81/. Coordinated with this parallel visual processing is activation in structures responsible for producing the movement such as FEF and the superior colliculus. A saccade is produced when the neurons at one location within the motor maps become sufficiently active. One job of visual processing, then, is to ensure that only one site within a movement map becomes activated. This is done when the neurons signalling the location of the desired target develop enhanced activation while the neurons responding to other locations are attenuated. Saccade target selection often converts an initially ambiguous pattern of neural activation into a pattern that reliably signals one target location. The ambiguity may be reduced through prior knowledge of the likely target location or identity, and extraretinal signals reflecting such expectations can modulate the responsiveness of afferent visual neurons. Specifying the metrics of a saccade and triggering the movement are coordinated but dissociable processes. Speed-accuracy trade-offs can thereby be produced allowing the visuomotor system to produce a saccade that is inaccurate because it is premature relative to the target selection process. While there are many gaps in our knowledge, the questions to ask seem reasonably clear. Because saccade target selection involves visual processing and eye movement programming combined with mnemonic influences, only continued experimental ingenuity will disentangle the various and variable contributions of individual neurons.(ABSTRACT TRUNCATED AT 400 WORDS)