Many studies have reported variation in bird diet composition related to spatial variation in habitat quality and landscape heterogeneity. Most diet analyses involve collection of food items at a sub-sample of available sites. Because individuals at different sites may vary in their diet composition, the description of the diet of a population may depend on which sites are sampled. This potential bias associated with environmental heterogeneity has never been explicitly tested. We used dietary data from a long-term study of an owl population to simulate various scenarios of diet assessment in which the investigator samples food items from various sites without taking into account their variation in quality. The simulations showed that assessment of diet composition was highly dependent on the number and type of sites sampled for food items and that such bias caused by environmental heterogeneity disappeared when food samples were randomly collected at more than 60% of the available sites. Biases in diet description may affect a large number of studies, especially of territorial species in heterogeneous landscapes. Similar biases may apply to the assessment of demographic parameters.