Animals often face great uncertainty as to the quality of foraging patches. There have been a number of theoretical studies investigating how non-omniscient predators, i.e. predators that are unable to assess foraging patch quality prior to patch exploitation, should forage in a heterogeneous environment, but empirical studies, especially in the field, are scarce. This paper describes the way in which white-fronted geese Anser albifrons forage on harvest remains of rice, focusing on the processes of patch selection and departure. Not only in autumn, but also in spring when rice depletion has progressed, patch rice density showed no positive effect on patch selection by geese, indicating the incapability of geese to select the most profitable patch. Instead, the geese tended to select patches with a large proportion of rice fields that were near the roost and a previously visited patch and bordered by a small number of windbreaks only. The rice consumption volume by geese increased with increasing initial rice density, while giving-up density was independent of initial rice density and was positively correlated to the mean rice density of the habitat. This suggests that the geese could compensate their lack of information on patch quality at the moment of patch selection by leaving less profitable patches earlier. We discuss the necessity of predictive models based on random patch selection and an appropriate departure rule to explain the distribution of individuals of species with limited information on patch quality.