This study focuses on the analysis of the Lombard effect in visual and audiovisual speech recognition. Previous studies have shown that the performance of an audio-only automatic speech recognizer decreases in noisy environments because of the Lombard reflex. A few studies have considered the visual changes due to the Lombard reflex, but the role of the Lombard reflex in automatic visual speech recognition has not been investigated so far. The authors show that the Lombard reflex plays an important role not only in audio, but also in automatic visual speech recognition, and this factor should be considered while designing a robust audiovisual speech recognizer.