¹ Uhlig, Carl, Erläuterungen zur Wirtschaftskarte von Deutsch-Ostafrika (Berlin 1904), 3Google Scholar.

² Foran excellent elementary view, see Rummel, Rudolf, “Understanding Factor Analysis,” Conflict Resolution, xi (December 1967), 444-80CrossRefGoogle Scholar.

³ I have eschewed rotational schemes for the component structures to criteria formalizing the intuitions of psychologists for other purposes. In this context, I can offer no plausible argument for the usual normalized varimax criterion, which appears simply as an inefficient taxonomic device for which a classification of variables is achieved under a severe and quite unrealistic orthogonal constraint.

⁴ In the context of historical studies see McClelland, David, The Achieving Society (New York 1961), 61CrossRefGoogle Scholar.

⁵ Since I have used huge data matrices up to 290x26, I report only tables of component loadings.

⁶ Moffett, J. P., ed., Handbook of Tanganyika (Dar es Salaam 1958), 111Google Scholar.

⁷ The question of lag and lead is a difficult one to disentangle, since there are frequently strong interdependencies between transport and other phenomena. See Gould, Peter, The Development of the Transportation Pattern in Ghana (Evanston 1960), 106-14Google Scholar; also Edward Taaffe, Richard Morrill, and Peter Gould, “Transport Expansion in Underdeveloped Countries: A Comparative Analysis,” Geographical Review, LIII, 4, 528.

⁸ Cattell, Raymond, “The Meaning and Strategic Use of Factor Analysis,” Handbook of Multivariate Experimental Psychology (Chicago 1966), 206Google Scholar.

⁹ “While I have not seen such an argument in the literature, the ratios of successive eigenvalues quickly approach unity after the first component, with λ₁/λ₂ = 1.3 in contrast to λ₁/λ₂ = 6.2. If we diink of the eigenvalues as “stretchability coefficients,” a cut through the hyper-figure along the major axis indicates a strong directional thrust and ellipsoidal configuration. But successive cuts ordiogonal to the first indicate only rough circularity, so diat die locations of the odier axes are virtually arbitrary. Such arbitrariness of location may indicate that we are absorbing the irrelevant variance of uniqueness and error. Peter Gould, “On the Geographical Interpretation of Eigenvalues,” Transactions of the Institute of British Geographers, No. 42 (1967) 58.

¹⁰ Note that the 100 isoline has been included as a dotted line to identify the relatively weaker nodes of this period.

¹¹ With a λ₁/λ₂ ratio of 2.5, stronger evidence is now available for a bipolar component contrasting government functions (2, 3, 12, 13, 29, and 30), with those more generally in the private sector (6, 16, 31, and 32) together with the communications variables serving it (7 and 8).

¹² Soja, Edward W., The Geography of Modernization in Kenya (Syracuse 1968), 78Google Scholar, 92-93.

¹³ The ratio of the second and third eigenvalues is now 2.60.

¹⁴ McCall, Daniel F., “The Dynamics of Urbanization in Africa,” Annals of the American Academy of Political and Social Science, 298 (March 1955), 160CrossRefGoogle Scholar.

¹⁵ Morrill, Richard L., Migration and the Spread of Urban Settlement, Lund Studies in Geography, Series B, No. 26 (Lund 1965)Google Scholar.

¹⁶ Telephone flows among modernized nodes in Tanzania, for example, are predictable to a high degree by relating them in simple gravity-model form to the size of the modernization scores of the generating and receiving centers and die distance between tliem (R²_1.234 = 0.58). This can be expressed in standardized regression coefficients after logarithmic transformation:

¹⁷ A series of canonical correlations between the set of variables in the late sixties and those of all other time-slices confirmed the obvious as far as the sixty-four major nodes are concerned: namely, that many linear combinations of die variables can be found that relate the sets to an extremely high degree even when the time-span is over forty years.