In many studies travel behaviour (for example, commuting) is analysed on the basis of a utility function with the distance (d) travelled as one of the arguments. An example is U = U(d,Y-cd,T-td) where Y and T denote money and time constraints, and c and t money and time costs per unit distance. This standard approach is not without problems, however, since it ignores the fundamental fact that most transport has a derived character: travelling kilometres is not an activity that gives utility per se, but only because these kilometres bring people to certain places they want to visit. In this paper we develop a method that provides a justification for utility functions such as shown here by showing that these can be made consistent with theories that take into account the derived character of transport. Implications of our approach are discussed for commuting distances of different types of jobs. Our approach gives an explanation for the paradox that highly educated workers tend to have long commuting distances. Given their high value of time one would expect short commuting distances, but the low spatial density of their jobs appears to dominate the outcome.