Although many of the variables used in sociological propositions can be conceptualized as having continuous underlying values, it is common practice to accept rank category measures of these variables as adequate surrogates for the underlying values. This widespread practice results in an unacceptable amount of distortion in the metric of underlying variables. Using computer simulations, the influence of four factors on the degree of distortion is examined: (1) the number of categories, (2) the marginal proportions for the categories, (3) the shape or form of the underlying distribution, and (4) how the categories are scored. Given the amount of distortion found, it is suggested that sociologists concentrate on the improvement of their measures. Although many of the variables used in sociological propositions can be conceptualized as continuous variables, in general there is little effort paid to creating measures of these variables that even approach the interval level of measurement. This is the case in spite of the fact that we often employ statistical procedures which require interval level measures, e.g., regression/correlation analyses. For example, even if we conceive of variables such as political involvement (Hensler and Stipak); social differentiation, demographic complexity, religious differentiation (Silver); strike identification, political position (Lyons); attitudes toward drug use (Akers et al.); political attitudes (Robinson and Kelley); physical discomfort, physical demands (Cullen and Novick); salience of a topic (Herberlein and Baumgartner); poverty (Hofferth and Moore) and so on, as continuous interval level variables, we (and the researchers cited in this paragraph) usually measure them quite crudely by using rank caterory or even dichotomous classifications based upon judges' opinions or respondents' answers. 1