This thesis investigated some of the eye movement factors related to the development and use of eye pointing devices with three dimensional displays (stereoscopic and linear perspective). In order for eye pointing to be used as a successful device for input-control of a 3D display it is necessary to characterise the accuracy and speed with which the binocular point of foveation can locate a particular point in 3D space. Linear perspective was found to be insufficient to elicit a change in the depth of the binocular point of fixation except under optimal conditions (monocular viewing, accommodative loop open and constant display paradigm). Comparison of the oculomotor responses made between a stereoscopic 'virtual' and a 'real' display showed there were no differences with regards to target fixational accuracy. With one exception, subjects showed the same degree of fixational accuracy with respect to target direction and depth. However, close target proximity (in terms of direction) affected the accuracy of fixation with respect to depth (but not direction). No differences were found between fixational accuracy of large and small targets under either display conditions. The visual conditions eliciting fast changes in the location of the binocular point of foveation, i.e. saccade disconjugacy, were investigated. Target-directed saccade disconjugacy was confirmed, in some cases, between targets presented at different depths on a stereoscopic display. However, in general the direction of saccade disconjugacy was best predicted by the horizontal direction of the target. Leftward saccade disconjugacy was more divergent than rightward. This asymmetry was overlaid on a disconjugacy response, which when considered in relative terms, was appropriated for the level of vergence demand. Linear perspective depth cues did not elicit target-directed disconjugate saccades.

Nos entrées sensorielles, comme nos mouvements, sont entachés d’incertitudes. Pourtant, notre système nerveux central semble être aussi précis que possible compte tenu de ces incertitudes: il les gère de manière optimale, par exemple en pondérant des informations sensorielles redondantes en fonction de leur fiabilité, ou en prenant en compte ses incertitudes motrices lors de la réalisation de mouvements. Si les modalités des combinaisons d’informations redondantes sont bien connues lors de tâches statiques, elles le sont moins en conditions dynamiques, lors de mouvements. La partie expérimentale de cette thèse a permis de confirmer l’existence de mécanismes d’estimation et de contrôle optimaux des mouvements chez l’humain. En effet, nous avons mis en évidence l’intégration optimale d’information visuelle lors de la réalisation de saccades à la lumière: lors de séquences de saccades, le système visuomoteur est capable d’utiliser l’information visuelle pour mettre à jour ses estimations internes de la position de l’œil. Une étude complémentaire des sources de variabilité des saccades suggère un rôle similaire pour la proprioception extra-oculaire. Par une troisième expérience, novatrice, nous avons montré que le toucher est pris en compte en temps réel lors de mouvements de la main en contact avec une surface. Nous avons également inféré une mesure de la variance de l'information tactile. Enfin, à partir des connaissances sur la représentation des variables sensorimotrices dans le système nerveux, nous avons construit plusieurs réseaux de neurones qui implémentent de manière proche de l'optimum statistique la planification et le contrôle de mouvements / Our sensory inputs, as well as our movements, are uncertain. Nevertheless, our central nervous systems appears to be as accurate as possible: these uncertainties are handled in an optimal fashion. For example, redundant sensory signals are weighted according to their accuracy, and motor uncertainties are taken in account when movements are made. The characteristics of the combination of redundant sensory signals are well known for static tasks. However, they are less known in dynamic conditions. The experimental part of this thesis allowed to confirm the use of statistically optimal sensorimotor processes during movements. We showed that visual information can be integrated during sequences of saccades, the oculomotor system being able to use visual information to update its internal estimate of eye position. A complementary study on the sources of variability for saccadic eye movements suggests a similar role for extra-ocular proprioception. In a third original experiment, we showed that tactile input is optimally taken in account for the on-line control of arm movements during which fingertips are in contact with a textured surface. Last, we built several neuronal networks models simulating optimal movement planning. These networks were based on current knowledge about probabilistic representations in the nervous system