Statistics and Human Geography

Abstract

Starting dates are often difficult to pinpoint exactly, but the “mid-Fifties” were certainly the time when statistical methodology began to burgeon in human geography. Although there was a lot of cross-fertilization and independent experimentation, the main strands can be picked out quite easily: the universities of Washington (Garrison, 1955; Berry and Garrison, 1958) and Lund (Hägerstand, 1952; Godlund, 1956a, b), and through the latter the remarkable and lonely efforts of one man in Finland (Ajo, 1953); some parallel efforts at Iowa (McCarty, 1956) and Northwestern (Taaffe, 1958; Thomas, 1960); and then one or two individuals who saw in genuine stochastic modeling a different and fruitful way of looking at the world (Curry, 1960). Years later a pioneering paper in Japan was recovered Matui (1932), but by then the particular statistical question had been asked again, and the answers greatly extended (Clark and Evans, 1954; Dacey, 1960). Most of the early statistical experiments were based upon regression analysis, representing attempts to estimate empirical relationships (usually linear), while maps of residuals were used to highlight the exceptions to such generalizations. No one had heard of such things as spatial autocorrelation in those days — including the statisticians, who were still helping the econometricians worry about the independence of their observations in time series. Spatial statistics were very much an open question, and more extended multivariate approaches, such as factor analytic and numerical taxonomic methods, were only just becoming practical possibilities with the growth of large computing facilities (Berry, 1960, 1961).

But philosophical knowledge can become genuinely relevant and fertile for ... [a] positive science only when ... [the geographer] comes across the basic traditional concepts and, furthermore, questions their suitability for that which is made the theme of his science.

Martin Heidegger, The Piety of Thinking, p. 21.