EMPIRICAL EVALUATION OF FORMAL THEORY.

The article focuses on empirical evaluation of formal theory. Researchers D.K. Hildebrand, J.D. Laing, and H.L. Rosenthal, henceforth H-L-R, propose an index to measure the adequacy of a theory that asserts or predicts emptiness of some cells in a two-way cross classification. There are two separate motivating arguments used by H-L-R to justify the independence-like products of marginals in the denominator of their index. The first argument establishes a kind of null or benchmark hypothesis for saying something about y in ignorance of both "x" and "y" to compare with statements of the theory. But that benchmark procedure seems to us arbitrary and inappropriate. The second argument for benchmark independence turns on arbitrary, tendentious definitions of the terms scope and precision for the theory. There have been many attempts to frame useful quantitative notions of degree of confirmation of a theory. H-L-R and also much of the literature mentioned treat highly simplified notions of theory. The very concept of a scientific theory is not simple, and in any case it seems to us confusing to deal with scientific theories as if they were reducible to cross-classifications with some cells asserted empty.