Previous experiments had obtained theoretically interesting differences between the short-term retention of temporal, spatial, and item information. Two experiments are reported which compared the learning, relearning, and long-term retention of temporal, spatial, and item information. In Experiment 1, 20 common nouns were presented sequentially within a vertical array on a computer screen. Subjects reported the identity of the nouns, their temporal sequence, or their spatial location within the array. Retention was then tested after a one week delay after which subjects relearned the list. Spatial information was learned most slowly, followed by temporal, and then item information. In Experiment 2, the presentation array was changed to two 3 1 3 matrices. Faster learning was found for spatial compared to temporal information, the opposite of what was observed in Experiment 1. The results are explained in terms of the benefits of spatial distinctiveness. α§ 1997 Academic Press Previous studies of short-term retention 1964). Third, the serial position function is from our laboratory have revealed important bow-shaped and symmetrical, so that recall differences in the coding and durability of inexhibits both a primacy and a recency advanformation about temporal sequence, spatial artage (e.g., . All three of rangement, and item identity in a list (see, these fundamental findings are linked, howe.g., Healy, , 1982; Healy, ever, to the recall of items in temporal se-Cunningham, Gesi, Till, & Bourne, 1991). quence. When the same items are recalled in-There have been three fundamental findings stead according to their spatial arrangement, in studies of short-term memory: First, the the time course of forgetting is no longer time course of forgetting is extremely rapid, rapid, there is no evidence for phonemic codso that recall approaches chance after only ing, and the serial position function is no 20-30 s (e.g., Brown, 1958; Peterson & Pelonger symmetrically bow shaped (e.g., Healy, terson, 1959). Second, recall depends primar-1975. Instead, when items are recalled ily on phonemic encoding, so that recall errors by spatial, rather than temporal, positions, the are based on confusions among similar-soundtime course of forgetting is protracted, recall ing items (e.g., Baddeley, 1966a; Conrad, errors do not involve phonemic confusions but are based instead on temporal-spatial patterns (e.g., , and the serial position Grant Sinclair is now at Old Dominion University. This functions are either flat or take on a inverted research was supported in part by Army Research Institute Contracts MDA903-90-K-0066, MDA903-93-K-0010, bow shape (e.g., . General and DASW01-96-K-0010 to the University of Colorado. models of short-term memory, like the influ-The authors thank Douglas Nelson and three anonymous ential perturbation model (Estes, , 1991, , reviewers for helpful comments concerning an earlier verin press; Lee, 1992;, sion of this article. We thank Michael Kos for his prowere designed specifically to account for the gramming of the experiments and Antoinette Gesi for her help with the selection of the stimulus materials used three fundamental findings with temporal orin the experiments. Address correspondence and reprint der recall and thus are inconsistent with the requests to Grant P. Sinclair, Department of Psychology, findings of spatial order recall.