Abstract A video-tape of children's responses to an articulation test was played once in the ambient quiet of a television studio, and once while electronically mixed with five conditions of broadband noise. The responses were scored by trainees in speech pathology. Differences in scoring between the play and replay showed: errors that could be heard in both quiet and noise decreased, and more errors were obscured by noise as a function of greater noise levels. Responses scored as errors in noise, but not scored as errors in quiet, were maximized when the broadband noise was between 60 and 65 dB. It was concluded that the noise affected the listeners' judgments of the apparent defectiveness of the articulations they heard in complex and conflicting ways. The results illustrate the need for control of noise in experimental studies of articulation and in articulation testing.