Introduction The critique of positivism is old news. Marxian and institutionalist economists have long believed that economics is an interpretive, value laden and historically conditioned discipline, and have argued against the simplistic empiricism and rationalism characteristic of the neoclassical majority (e.g., Dobb, 1973, ch 1). In recent years, however, the writings of Donald McCloskey (especially 1985) have brought this critique into much wider currency. This may be due to his orthodox credentials, his stature as an economic historian, the grace and wit of his writing style, or the sheer quantity of his publications. No matter; his point is well taken. Economics is another form of communication, an ongoing conversation conducted in terms of metaphorical allusions, established appeals to authority and other rhetorical devices. If economics is not physics, perhaps it is a form of literature, albeit one which would elicit little praise for the typical quality of its prose.(1) McCloskey's description of economics as rhetoric is seductive, but it is not without its dangers. The communicative nature of economics leaves open the question of what it is that economists communicate about. Though all economists engage in rhetoric, the differences in their questions and answers are nonetheless real. Sassower (1988, p. 554) suggests that McCloskey glosses over these differences by ignoring non-neoclassical paradigms. In a similar vein, Waller and Robertson (1990) accuse McCloskey of being anti-rhetorical in the sense that he is so tied to the premises of neoclassicism that he cannot acknowledge the validity of alternative perspectives. Dyer goes so far as to chide McCloskey for his intolerance for theoretical variety (1988, p. 163). Sebberson agrees that McCloskey's belief that rhetoric will not affect the substance of economics demonstrates his blindness to the relative character of his own point of view, and Sebberson wonders, happens if there is more at stake than just winning a given argument? What happens if the truth of the matter has to be discerned so that the best course of action may be followed (1990, p. 1021)? This paper explores the relation between argumentative rehetoric and analytical substance in the economics of discrimination. I shall try to follow the advice of Robert Solow to analyze the between particular lines of economic analysis and particular rhetorical conventions, though I would have put the word connection in the plural (1988, p. 35). A metaphor not only expresses an idea by combining disparate images; it also relates that idea to a larger system of social beliefs. Metaphors, in other words, convey an ideology as well as an analysis. For example, one popular model of racial inequality in the 1960s was based on a concept called internal colonialism. The metaphor of colonialism explained inequality as a consequence of unequal exchange between the center -- the white metropolis -- and the periphery -- the black ghetto. But it also provided a condemnation of inequality insofar as colonialism is an undemocratic exercise of domination and exploitation. In this way, metaphors establish a criteria for judgment as well as analysis. They reflect beliefs as well as ideas. The metaphor chosen will have moral as well as theoretical and empirical interest. Metaphors matter not just because they are metaphors, but because of the choice to employ one rather than another. McCloskey just has trouble taking the second step.(2) To illustrate this point, I will compare two classic texts on discrimination: Gunnar Myrdal's An American Dilemma (1972; first edition 1944) and Gary Becker's The Economics of Discrimination (1971; first edition 1957). These are undoubtedly the two most influential books ever written by economists on the subject of race, but two more different books could hardly be imagined. Myrdal headed one of the largest and most complex social research projects of his time (Southern, 1987, ch 2); his findings span two volumes and over a thousand pages of text with nearly five hundred additional pages of appendices, notes, bibliography and index. …