This study seeks to improve our understanding of the effects of microphone variations on speech recognition systems. The TIMIT corpus provides data recorded on close talking and far field microphones and over telephone lines. The SUMMIT system is configured for phonetic classification and recognition. At the last ICSLP, we presented an analysis of the data and experiments in phonetic classification using a baseline system and various preprocessing techniques. In this paper, we present experiments in phonetic recognition using an improved baseline system and compensation techniques that require varying amounts of microphone specific data.