Mathematical analysis shows that if the pattern of rehearsal in free-recall experiments (of necessity, the pattern observed when participants rehearse aloud) be continued without any further interruption by stimuli (as happens during recall), it terminates with the retrieval of the same 1 word over and over again. Such a terminal state is commonly reached before some of the words in the list have been retrieved even once; those words are not recalled. The 1 minute frequently allowed for recall in free-recall experiments is ample time for retrieval to seize up in this way. The author proposes a model that represents the essential features of the pattern of rehearsal; validates that model by reference to the overt rehearsal data from B. B. Murdock, Jr., and J. Metcalfe (1978) and the recall data from B. B. Murdock, Jr., and R. Okada (1970); demonstrates the long-term properties of continued sequences of retrievals and, also, a fundamental relation linking recall to the total time of presentation; and, finally, compares failure to recall in free-recall experiments with forgetting in general.