Accurate information about gaze direction is required to direct the hand towards visual objects in the environment. In the present experiments, we tested whether retinal inputs affect the accuracy with which healthy subjects indicate their gaze direction with the unseen index finger after voluntary saccadic eye movements. In experiment 1, subjects produced a series of back and forth saccades (about eight) of self-selected magnitudes before positioning the eyes in a self-chosen direction to the right. The saccades were produced while facing one of four possible visual scenes: (1) complete darkness, (2) a scene composed of a single light-emitting diode (LED) located at 18 degrees to the right, (3) a visually enriched scene made up of three LEDs located at 0 degrees, 18 degrees and 36 degrees to the right, or (4) a normally illuminated scene where the lights in the experimental room were turned on. Subjects were then asked to indicate their gaze direction with their unseen index finger. In the conditions where the visual scenes were composed of LEDs, subjects were instructed to foveate or not foveate one of the LEDs with their last saccade. It was therefore possible to compare subjects' accuracy when pointing in the direction of their gaze in conditions with and without foveal stimulation. The results showed that the accuracy of the pointing movements decreased when subjects produced their saccades in a dark environment or in the presence of a single LED compared to when the saccades were generated in richer visual environments. Visual stimulation of the fovea did not increase subjects' accuracy when pointing in the direction of their gaze compared to conditions where there was only stimulation of the peripheral retina. Experiment 2 tested how the retinal signals could contribute to the coding of eye position after saccadic eye movements. More specifically, we tested whether the shift in the retinal image of the environment during the saccades provided information about the reached position of the eyes. Subjects produced their series of saccades while facing a visual environment made up of three LEDs. In some trials, the whole visual scene was displaced either 4.5 degrees to the left or 3 degrees to the right during the primary saccade. These displacements created mismatches between the shift of the retinal image of the environment and the extent of gaze deviation. The displacements of the visual scene were not perceived by the subjects because they occurred near the peak velocity of the saccade (saccadic suppression phenomenon). Pointing accuracy was not affected by the unperceived shifts of the visual scene. The results of these experiments suggest that the arm motor system receives more precise information about gaze direction when there is retinal stimulation than when there is none. They also suggest that the most relevant factor in defining gaze direction is not the retinal locus of the visual stimulation (that is peripheral or foveal) but rather the amount of visual information. Finally, the results suggest an enhanced egocentric encoding of gaze direction by the retinal inputs and do not support a retinotopic model for encoding gaze direction.