By repeatedly presenting an alien scent to territory-owning beavers, Castor canadensis, we tested two competing hypotheses about the function of scent marking: scent fence and scent matching. The scent-fence hypothesis predicts that territory owners should respond increasingly strongly over time towards a recurrent alien scent because of the ineffectiveness of previous responses. The scent-matching hypothesis predicts that the intensity of response should be the same or decrease because, without the presence of the intruding signaller coupled with the chemical signal, the presence of the scent itself does not advertise the ownership of a territory. The response level of resident beaver families was stable to strangers' anal gland secretions (AGSs) and decreased to strangers' castoreum during a period of 6 days. These results support the scent-matching hypothesis but not the scent-fence hypothesis. Copyright 1998 The Association for the Study of Animal Behaviour. Copyright 1998 The Association for the Study of Animal Behaviour.