GENERALIZATION IN COMPARATIVE AND HISTORICAL SOCIAL SCIENCE 55positivism not as a counterpart but as a contrast to the “historical method.” Treating the former
as burdened by fallacies, Boas (1973 [1896]) endorsed the latter as a proper basis for the de-
tailed primary studies of a narrower scope that he and his students were to make a new ortho-
doxy for American anthropologists.
A different line of movement away from evolutionary positivism was, however, to have a
larger import for the long-term fate of cross-societal inquiry in America. The turn of the century
saw the beginning of a remaking of positivist social science. Although the belief that natural
science provides a model for all scientific inquiry would persist, the content of that model was
changing. Nineteenth-century positivists had presupposed the existence of natural laws of social
progress when they sought to weld comparative and historical findings together into synthetic
schemes. For positivists in the new century, however, explicit notions of “progress” would come
to appear laden with subjective meaning and values in a fashion inappropriate to the conceptual
structure of rigorous science. In need of a new basis on which to construct knowledge, positivists
would increasingly look to the observation and statistical analysis of correlations as a foundation
for properly “scientific” social science. An important role inaugurating this transition was played
by Franklin Giddings of Columbia University.^6 His Principles of Sociology presented the com-
parative and historical methods as applying a single logic, that of the “method of concomitant
variations”: both were said to center on the “systematic observation of coherences among phe-
nomena,” with “comparative method” focusing on coherences “distributed in space” and “his-
torical method” on coherences “through periods of time.” This portrayal was accompanied by the
suggestion that these methods “may become precise when they can become statistical” (1896,
64). In succeeding years, as Giddings learned of Galton and Pearson’s work pioneering the tools
of correlational analysis, his belief in the potential of statistical work deepened. He would come
to acclaim the “statistical method” as “an inestimably valuable form of the comparative and his-
torical methods” that promised to “bring our knowledge of society up to standards of thorough-
ness and precision comparable to the results attained by any natural science” (1904, 175–76).
Giddings’ work was a forerunner of a set of epistemological changes that gained momentum
over the next generation as social science advocates for the new statistical techniques grew in
number and promoted a new image of “scientific method.” We can label this shift as the rise of
modernism, where “modernism” denotes, as it does in the history of art or architecture, a novel,
distinctively twentieth-century perspective.^7 For the idea of a break from the nineteenth cen-
tury that this label implies, the appeal to the “method of concomitant variations” made by
Giddings is especially noteworthy. In this appeal Giddings drew on the account “Of the Four
Methods of Experimental Inquiry,” in John Stuart Mill’s System of Logic (1865 [1843]). But in
using one of these methods as a model for the logical structure of social scientific inquiries,
Giddings made a major departure, for Mill had held that none of the methods was a satisfactory
model for social science.^8 Novel in 1896, the belief that a logical reconstruction of experimen-
tal inquiry can, and should, provide a model for all scientific inquiry was to become commonly
accepted among interwar expositors of scientific method. As already evident with Giddings,
this belief developed intertwined with the argument that, when experimental inquiry itself can-
not be pursued, the best approximation to the “scientific” kind of knowledge that it models is
offered by statistical analysis.
The blend of belief and argument offered by social science proponents of this new view of
scientific method was, however, far from sweeping all before it in the interwar years. Much of the
empirical thrust that marked American social science up into the 1930s was worked out through
the burgeoning of largely qualitative case-study and field research, perhaps most notably associ-
ated with the “Chicago School.” Many practitioners in these traditions found the threat posed by