One of the central issues in the theory of speech community is whether speech community is a naturally-existing entity or a research construct without any restricted empirical basis. The issue is attacked here by way of a language survey. The survey was on language-choice behavior in public places, conducted in the city of Nanjing in 2002. The survey results show: (a) Nanjing residents’ verbal behavior exhibits a specific order of social convention and the urban population thus makes up an effective body of social communication. (b) The ordered behavior reflects a community-wide evaluative system that governs linguistic heterogeneity. (c) The thinning-out of the regularity among the individuals’ behaviors is typically found both in the spatial and the temporal marginals of the urban population. The study supports the hypothesis that a speech community is a naturally-existing entity. An attitudinal-behavioral and impactal unity is the core of such existence. With the approach taken here, a speech community can be discovered with certain well-defined empirical procedures. The wider significance invoked is that the organizational system of speakers is an important linguistic system alongside the other linguistic systems.