In three experiments, we manipulated the processing demands of a concurrent task to test the hypothesis that an event-based prospective memory task satisfies a criterion of automaticity proposed by Hasher and Zacks (1979). As in the previous studies, a prospeetivc memory task (pressing a key whenever a target word was presented) was embedded within a short-term memory task (remembering seven words). In addition, participants performed a concurrent memory task which varied in difficulty. Participants repeated either 0 to 6 randomly generated digits or a single word. In all three experiments, short-term memory performance was influenced by the concurrent memory load. Prospective performance, in contrast, was not affected by the memory load even though an attempt was made to increase the difficulty of the prospective task by manipulating the specificity of the target instructions (Experiment 2) and the number of target words (Experiments 2 and 3). The results are discussed within the framework of automatic processing.