This paper aims to show how medical scientists may use metaphor in ways closely parallel to poets. Those who believe metaphor has any role at all in science may describe its use in various ways. Associationists think metaphors are based upon likenesses, and collapse the notions of model and metaphor together. But, as an example from the work of Louis Pasteur suggests, metaphor need not be based upon likenesses. Rather it may play a role in making possible a model's explanatory significance. Models may presuppose metaphors. The Pasteur example also suggests metaphor may play a part in creating likenesses through its role in classification and reclassification. It is in these ways that the use of metaphor in medical science most closely parallels that in poetry.