In these experiments, eight cats had to choose between two paths leading to a baited feeding-bowl. Transparent or opaque screens were interposed between the starting-point and the goal; their arrangement was devised so as to manipulate the length and the angular deviation or detour amplitude of the paths. The results showed that cats exhibited a preference for one of the routes in most of the situations, although both were always rewarded. When the goal was visible (transparent screens), their choice of paths depended almost exclusively on the initial direction of the path with respect to a straight line to the goal: they chose the less divergent path even if this was the longer one. However, when the goal was hidden (opaque screens), cats took account both of the length and the angular deviation of each path; they chose the most appropriate route, with respect first to its length, then to its angular deviation. These results suggest that two kinds of mechanisms underlie the selection of paths in these spatial situations. Our conclusions fit with O'Keefe & Nadel's (1978) hypothesis concerning the existence of two orientation systems, the ‘taxon’ and the ‘locale’ systems, involved in the organization of travel.