The question of whether globally amnesic subjects can learn new semantic (factual) information is controversial. Some students of amnesia believe that they can, others that they cannot. In this article we report an extensive experiment conducted with the amnesic patient K.C. in which we examined the role of repetition and associative interference in his learning of new semantic information. In the course of 8 study sessions distributed over 4 weeks, we taught K.C. novel, amusing definitions of 96 target words (e.g., “a talkative featherbrain—PARAKEET”). We varied systematically the degree of both pre-experimental and intraexperimental associative interference, as well as the amount of study. The results of the experiment showed that K.C. can learn new semantic knowledge, and retain it over a period as long as 30 months indistinguishably from control subjects. The results further showed that the efficacy of such learning depends critically on both repetition of the material and the absence, or minimization, of pre-experimental and intraexperimental associative interference. These findings suggest that the extent to which at least some amnesic patients can acquire and retain new semantic knowledge depends on the conditions under which learning occurs, and that unqualified statements regarding the deficiency or absence of such learning in amnesia are not justified.