Abstract: Marsh tits (Parus palustris) store single food items in scattered locations and recover them hours or days later. Some properties of the spatial memory involved were analyzed in two laboratory experiments. In the first, marsh tits were offered 97 sites for storing 12 seeds. They recovered a median of 65% of them 2-3 hr later, making only two errors per seed while doing so. Over trials, they used some sites more often than others, but during recovery they were more likely to visit a site of any preference value if they had stored a seed there that day than if they had not. Recovery performance was much worse if the experimenters moved the seeds between storage and recovery. A fixed search strategy that had some of the same average properties as the tits' search behavior also did worse than the real birds. In Experiment 2, any tendency to visit the same sites on successive daily tests in the aviary was placed in opposition to memory for storage sites by allowing the tits to store more seeds 2 hr after storing a first batch. They tended to avoid individual storage sites holding seeds from the first batch. When the tits searched for all the seeds 2 hr later, they tended to recover more seeds from the second batch than from the first, i.e., there was a recency effect.