The tradition of large-scale survey methodology in the study of variation has left a gap between the linguistic data and the social practice that yields these data. Since sociolinguistic surveys bring away little information about the communities that produce their linguistic data, correlations of linguistic variants with survey categories have been interpreted on the basis of general knowledge of the social dynamics associated with those categories. The success of this approach has depended on the quality of this general knowledge. The examination of variation and socioeconomic class has benefited from sociolinguists’ attention to a vast literature on class and to critical analyses of the indices by which class membership is commonly determined. The study of gender and variation, on the other hand, has suffered from the fact that the amount of scientific attention given to gender over the years cannot begin to be compared with that given to class. Many current beliefs about the role of gender in variation, therefore, are a result of substituting popular (and unpopular) belief for social theory in the interpretation of patterns of sex correlations with variation.