Abstract: The nature of spatial information used for memorizing and recalling places is largely unclear. Earlier studies tested integration of geometric and feature information mostly during reorientation in artificial environments without including time as a memory-critical component. Here, we tested a harbour seal in a delayed matching-to-sample task (DMTS) in a familiar environment under two spatial multiple-choice conditions. The feature condition consisted of a DMTS task with four comparison stimuli presented on fixed positions in a classic matching apparatus and was designed to make stimulus features the most prominent information. The landscape condition consisted of a DMTS task in a familiar environment with four places marked by comparison stimuli and allowed the use of all available spatial information including geometrical and feature information. The seal's performance was impaired by delays of 3, 6, 9 or 12 s only in the feature condition; a delay of 12 s resulted in chance level performance. Replacing the comparison stimuli at the apparatus with identical spheres resulted in impaired performance. Performance in the landscape condition was impaired neither by delays nor by replacing comparison stimuli with spheres. Landscape information obviously was encoded redundantly and could be recalled more reliably and longer than feature information, which reveals feature information to be a less valuable type of spatial information for memorizing and recalling places.