The use of the term "qualitative" in science and, above all, in physics has a pejorative ring. It was a physicist who reminded me, not without vehemence, of Rutherford's dictum, "Qualitative is nothing but poor quantitative." But consider the following example. Let us suppose that the experimental study of phenomenon F gives an empirical graph g with equation y = g(x). To explain F the theorist has available two theories, T1 and T2; these theories give graphs y = g1(x) and y = g2(x), respectively. Neither of these graphs fits the graph y = g(x) well (see graph); the graph y = g1(x) fits better quantitatively in the sense that, over the interval considered, the sum of the differences between g and g1 is smaller than the sum of the differences between g and g2, but the graph g2 has the same shape and appearance as g. In this situation one would lay odds that the theorist would retain T2 rather than T1, even at the expense of a greater quantitative error, feeling that T2, which gives rise to a graph of the same appearance as the experimental result, must be a better clue to the underlying mechanisms of F than the quantitatively more exact T1. Of course this example is not a proof, but it illustrates the natural tendency of the mind to give to the shape of a graph some intrinsic value; it is this tendency that we shall develop here to its ultimate consequences.