A quantitative model for the generation of saccades is presented which is based on the accepted notion that saccade metrics are coded in the deeper layers of the superior colliculus by the population of recruited movement-related cells. The model has important features derived from electrophysiological data in the monkey: a nonhomogeneous and anisotropic afferent-mapping function relating the outside visual world to collicular coordinates and a dome-shaped activity profile in the deeper layers describing the spatial extent of the recruited movement-related cell population. It is proposed that each cell generates a small movement contribution which is determined both by the neuron''s location in the collicular map and its firing rate. The individual cell contributions are then added vectorially to yield the total saccade. The model can generate saccades to single visual stimuli in all directions and simulate decimal stimulation experiments. To account for some nonlinear properties of the saccadic system in the computation of saccade metrics, a nonlinear version of the linear summation model is discussed.