This paper presents experiments with the identification and automatic detection of parasitic sounds in speech signals for the purposes of unit-selection-based speech synthesis.