Sibilants are a known problem in speech recording. Even though they can often be decreased by a different placement of the microphones, there still is the necessity for methods that reduce these artifacts. Therefore new investigations on sibilants in German, English, Spanish, and French have been made to describe their properties in time and frequency domain. To find an adaptive algorithm which considers these properties a reliable method is needed for detection and classification of the different kinds of sibilants. It is shown that the psychoacoustic unit ‘‘sharpness’’ is well correlated with the appearance of these sounds and that specific loudness, an intermediate step in calculating sharpness, can be utilized to get information about the spectral properties of each sibilant. An algorithm is presented which employs sharpness to detect sibilants and reduces them using variable filters.