Cognitive metaphor research is characterized by the diversity of rival theories. Starting from this observation, the paper focuses on the problem of how the unity and diversity of cognitive theories of metaphor can be accounted for. The first part of the paper outlines a suitable metascientific approach which emerges as a modification of B. von Eckardt's notion of "research framework." In the second part, by the help of this approach, some aspects of the sophisticated relationship between Lakoff and Johnson's, Glucksberg's, and Gentner's theories are discussed. The main finding is that the data, the problems, the heuristics and the hypotheses which have been partly shaped by the rivals contribute to the development of the particular theories to a considerable extent. (Contains 1 figure and 4 tables.)