—Error robust video communication for hand-held devices is a delicate task because of limited computational resources and hostile channel conditions in mobile environments. The loss of coded video data on the channel can result in spatio-temporal error propagation in the video. In addition, stringent end-to-end delays for conversational applications make this challenge even more difficult. In this work, we investigate several techniques which exploit feedback from the receiver to enhance the performance of conversational video in realistic mobile communication environments. Specifically, we show how a low-complexity interactive error tracking technique can be combined with multiple reference picture selection (RPS) based on the existing syntax of H.264/AVC. This technique outperforms other interactive error protection strategies by a margin of more than 2 dB for moderate channel loss rates, with minimal impact on end-to-end delays.