A proof by pointing user interface component allows a user to direct the course of a proof assistant by selecting terms with a mouse. Such a gesture is interpreted as a high-level tactical which triggers a sequence of low-level basic commands for the proof engine. The algorithm inherently relies on a structure-conscious environment; as a novelty we show how proof-by-pointing may easily be integrated into an interface without a structure editor. We discuss in detail the use of nested selectable text regions for user interaction, the modifications necessary to the proof-engine output, and the algorithm for interpreting selections as proof commands, with particular reference to a concrete implementation using XEmacs and LEGO.