Abstract: A method and apparatus for continuous presence conferencing is disclosed having voice-activated quadrant selection. A control processor in a multipoint control unit selects and assigns anchored sites to anchored regions of a composite video image. The control processor further selects a switched site according to a selection process and assigns the switched site to a switched region of the composite video image. The selection process for selecting the switched site includes evaluating audio signals received from the conference sites to determine a dominant audio signal and selecting the site associated with the dominant audio signal. A rotating site is selected and is assignable to any of the regions. A video processor processes the video signals according to the composite video image assignments to provide composite video signals for transmission to sites in the conference.