In order to evaluate the adequacy of several noise descriptors purported to account for the effects of time-varying noise on people, the annoyance of 24 three-minute samples of time-varying traffic noise as heard outdoors, indoors with open windows, and indoors with closed windows were rated by 28 subjects under laboratory conditions simulating a home environment. Included among the noise descriptors were statistical descriptors L10, average sound level Leq, and descriptors that take into account either the variability in the range of levels (NPL) or the rate of change of level with time Leq′, LB. Although correlation coefficients indicated that all descriptors predicted annoyance ratings satisfactorily, more sophisticated statistical analyses revealed that both L10 and Leq predicted annoyance best. An additional finding of this study was that all stimuli below an Leq, of 50 dB were generally acceptable; whereas, those exceeding this level were generally unacceptable to at least 50% of the subjects. All stimuli found acceptable were presented in the closed window condition. [Work supported by DOT/FHWA.]