This article presents research—an experiment and the resulting prototype—on a method for treating gestural input so that it can be used for multimodal applications, such as interacting with virtual environments. This method involves the capture and use of natural, empty-hand gestures that are made during conventional descriptive utterances. Users are allowed to gesture in a normal continuous manner, rather than being restricted to a small set of discrete gestural commands as in most other systems. The gestures are captured and analyzed into a higher-level description. This description can be used by an application-specific interpreter to understand the gestural input in its proper context. Having a gesture analyzer of this sort enables natural gesture input to any appropriate application.