The measurement of sociolinguistic behavior requires the formulation of a unit which can take into account continuous, ordered variation within and across discrete linguistic types--the linguistic variable. The linguistic variable, itself an abstraction, is realized in actual speech behavior by variants. The formulation of the linguistic variable may be viewed as a function of its correlation with extra-linguistic variables (socio-economic class, sex, age, contextual style, and racial isolation) or independent linguistic variables (linear environment and syntactic construction). The author tries to show that in order to account for systematic variation between the variants of a variable a consideration of extra-linguistic and independent constraints is imperative. Only a consideration of these two facets will reveal the fully systematic nature of variation and the various constraints on the relative stigmatization of certain variants. (DO)