The views that metaphor and metonymy are merely poetical or rhetorical embellishments and the opposite view that metaphor and metonymy are part and parcel of everyday language and thought, permeate and in fact constitute it, have been pitched against each other time and again during at least the last two centuries. This 19th-century tradition of reflection upon the linguistic, philosophical, and psychological aspects of metaphor and metonymy has so far been overlooked by those who, like Lakoff and Johnson (1980), make claim to a radically new approach to metaphor and metonymy overthrowing the view held by generative linguists and semanticists that metaphor is a deviant phenomenon of language. However, these modern theorists only pick up a long strand of thought that went underground after 1960. So as to fully understand the import of the concepts of metaphor and metonymy for contemporary semantic thought, it is necessary to reconstruct the contexts in which those concepts appeared for the first time in linguistic theory. Such a reconstruction is attempted here and may lead to a critical evaluation of contemporary concepts and theories. Figures of speech, such as metaphor and metonymy have been studied for 2000 years as part of rhetoric. In this article I want to study the role of metaphor and metonymy in the shaping of semantics as a linguistic discipline from the point when rhetoric thought merged with new insights into semiotics, semantics and the philosophy of language, at the end of the 18th century, especially in Germany, up to the first part of the 20th century when metaphor and metonymy were rediscovered inside structural linguistics, influenced by Gestalt psychology, the poetics of Roman Jakobson, and 'new' types of rhetoric developed both in the United States and in France. (Brigitte Nerlich)