The extent to which protein synthesis is involved in working memory was investigated with the protein synthesis inhibitor anisomycin (ANI). Rats were trained to perform accurately on a 12-arm radial maze when delays of 240 min were interposed between choice 6 and choice 7. Bilateral hippocampal cannulas were then implanted. Accuracy on choices 7-12 was studied when ANI or saline was injected either 30 min before choice 1 or 5-10 min after choice 6 (Experiment 1). Pretrial injection of ANI significantly impaired performance following the 240-min delay, whereas ANI injected during the delay had no such effect. In Experiments 2 and 3, the ANI-induced amnesia was replicated, and the temporal course of development of the amnesia was determined. Pretrial administration of ANI did not significantly affect retention after a 2-min delay but did produce amnesia after delays of 15 min or longer. These data suggest that protein synthesis is important for the formation of temporary memories, provided the retention interval is long enough. It is suggested that working memory includes both short-term and long-term components. Protein synthesis appears to be important for formation of the long-term component, hut not the short-term component, of working memory.The current literature shows that there is no generally accepted theory describing the neurochemieal bases of learning and memory. An example is the lack of agreement on exactly what different forms of memory exist. Several dichotomies of memory have been proposed, for example, working and reference memory (Olton, 1983) and episodic and semantic memory (Tulving, 1972), most of which have been formulated from behavioral observations of humans or animals. Whether these behavioral distinctions actually represent quali-