Recently, sociologists have expressed a renewed interest in the theoretical and empirical study of inequality, its determinants, and its effects. Recent studies include Gartrell (1977), Rubinson and Quinlan (1977), Blau (1977), Jencks and others (1972), and Chase-Dunn (1975). 1 In such studies the analysts usually choose a single index to measure inequality, such as the coefficient of variation or the Gini coefficient, and then use it to analyze their data. With the exception of Blau, few have made an explicit attempt to define the concept of inequality or to justify the chosen index as an appropriate measure of inequality. However, choosing a single index from the available ones implies that inequality is a unidimensional concept and that the chosen index is a valid measure of it. But it is not necessarily the case that different measures of inequality will correlate highly with the concept and with each other and that they will therefore rank distributions in the same order. Different measures may yield different results, and the differences may be considerable. We demonstrate this by analyzing the Kuznets data (1963) on the distribution of individual income for 12 countries in about 1950. Table 1 presents rank-order correlations (Kendall's tau) among four commonly used measures of inequality applied to data (Tables 2 to 4): the coefficient of variation (CV), the Gini coefficient (GC), the standard deviation of the logarithm (SDL), and the mean relative deviation (MRD). Formulas for these measures are given in the appendix. The first three measures are commonly used to measure income inequality; the mean relative deviation is used for this purpose and for measuring degree of segregation. 2 The correlations