Three experiments investigated the mechanisms by which we estimate Euclidean distances on the basis of kinaesthetic cues. In all experiments, blindfolded participants followed straight and curvilinear paths with a stylus. Then, with a straight response movement, they estimated the distance between the end-points of the previously explored path. Experiment 1 was designed to validate the hypothesis-made on the basis of results from a previous study-that errors in the kinaesthetic estimations of distances (detour effect) originate from the difficulty to decompose the displacement vector into relevant and irrelevant components, which would become more severe at points of inflection. Using elliptic paths (no inflections), we demonstrated that errors are indeed reduced considerably. The role of the orientation of the work plane was investigated in Experiment 2 in which the same paths used in our previous study were oriented in the frontal rather than the horizontal plane. The results indicate that the detour effect is independent of the orientation. Moreover, despite the asymmetry that gravity introduces between upward and downward movements, errors in the two directions are almost identical. Experiment 3 addressed two issues. First, we demonstrated that introducing a delay between the exploration of the path and the response did not alter significantly the pattern of errors. By contrast, we demonstrated that errors are severely reduced when the number of paths to be explored is reduced by half. The results of the three experiments are discussed within the context of current theories of sensori-motor coding.