The recurrence of certain basic metaphors across diverse times and cultures is an interesting issue in the debate over the form and extent of innate human knowledge. Are the common themes of metaphor and narrative evidence for strong innate human knowledge, or the result of weaker constraints on learning and development? Machine learning has demonstrated that strong conceptual knowledge may be acquired from experience and weak initial learning biases. Drawing on these results and a computational model of metaphor and analogy formation, this paper examines the ability of emotions to serve as effective biases on the acquisition of useful, domain specific patterns of metaphor. In addition, it considers the mechanisms through which metaphor projects semantic structure onto emotional experience.