AbstractAn alternative method of producing a doubly constrained trip distribution similar to that of a gravity model is described. This alternative model differs from conventional models in that it has an explicit, though naive, behavioural rationale. It also produces a doubly constrained distribution more rapidly than these conventional models, partly because it involves fewer iterations and partly because each successive iteration involves fewer zones than that preceding it. In the one full-scale test for which data are available, the new model efficiently simulates a survey derived distribution, while tests with hypothetical systems show that it produces a distribution which can be closely identified with that produced by a conventional model.