This work is a sociolinguistic study employing quantitative methods to explore age, gender, and social class differences in the use of a range of discourse features. It is based on a gender-balanced sample of middle-class and working-class adolescents and adults, recorded under the same conditions in Glasgow, Scotland. Unlike studies of phonetic or morphological variation, the study of discourse variation requires samples of talk in action with speakers interacting with one another. The speakers, who knew each other, were recorded talking in the presence of the tape-recorder for approximately half an hour without the investigator being present. The recordings were transcribed in their totality and the transcripts searched for the occurrence of features such as the use of pronouns, adverbs, you know, I mean, as well as grammatical features such as questions and passive voice. The frequencies of use of the variables by the different social groups (e.g., middle-class women, adolescent boys) were calibrated and the results compared. Differences between adults and adolescents provided the greatest number of statistically significant results, followed by differences between males and females. The smallest number of statistically significant differences were related to social class. Qualitative analysis, however, revealed important social class differences in discourse styles. The study shows the danger of generalizing about social class or gender on the basis of a limited sample of a few discourse features.