Questionnaire studies were conducted in a residential area before and after the erection of a 2.25 m high noise barrier of conventional type along a heavily traveled road (19,600 vehicles/24 h). The interval between studies was two years. Houses closest to the barrier received a sound-level reduction from -70.0 to 62.5 dB Lden at the most exposed facade. The sound-level reduction decreased with distance to the road, and was negligible for houses at more than 100 m distance. Up to this distance, the noise barrier reduced residents' noise annoyance outdoors and indoors as well as improved speech communication outdoors. Indoors, speech communication and sleep disturbance were slightly but nonsignificantly improved. Predictions of the number of annoyed persons from published exposure-response curves (in Lden) agreed with the percentage of residents being annoyed when indoors, before and after the barrier. Conversely, the percentage of residents being annoyed when outdoors clearly exceeded the predictions. These results suggest that these exposure-response curves may be used in predicting indoor situations, but they should not be applied in situations where outdoor annoyance is at focus.