Metaphor, as treated here, is not a rhetorical device or strategy but the primary means by which dream and artistic images arise and give rise in turn to further narrative developments. The paper is chiefly concerned with dreams and the possibility that dream coherence--i.e., meaning, making sense--as we recognize it in the waking state may be a largely irrelevant to the organization and function of dreams that obviously have no artistic mission to be shared by others (readers, auditors) as forms of communication. I examine an experiment in dream-splicing conducted by Allan Hobson and his Harvard colleagues with a view to illustrating that bizarreness and discontinuity in dreams are not necessarily, as Hobson claims, signs of incoherence but may be the natural consequence of dreams exercising the power of association while the body is "off-line." If dream images arise from a virtually infinite experience in world-association, it is possible that dreams couldn't perform their function--whatever it may be--by offering coherent narratives of the sort that interest waking readers. In other words, the content of dreams--coherent or otherwise from the waking standpoint--may be less important than the act of dreaming itself. Additionally, I briefly discuss two leading theories of dream function--the recorrelation of memory and the vigilance theories and suggest, briefly, how they might apply not only to dreams but to artistic works as well.