Traffic noise and social surveys were carried out at eight London suburban sites. Dwellings at the selected sites were exposed to noise from freely flowing traffic at levels which ranged from 57 to 82 dB(A), measured as 18 hour Llo. The study was designed to obtain noise measurements and subjective responses from residents on four repeated occasions throughout the year. A total of 1363 interviews was conducted, 222 respondents each completing four interviews. Traffic noise was measured as Ll0, Ls0 and L,~ in both linear and A-weighted form. The eight noise indices were found to be very highly intercorrelated. The correlations between each of these and the subjective responses were all equally high. It was therefore not possible to select a "best" index of noise in terms of nuisance experienced by residents, even when the most highly reliable data, derived from averaging responses for the four repeated interviews, were employed. Subjective responses were measured by means of a 7-point scale of "dissatisfaction" and a 4-point verbal scale of "bother" and the former was found to be a significantly better correlate of noise exposure. The data from these eight sites, which underwent no change in noise levels during the survey, were compared with data obtained in "before and after" studies conducted over a roughly similar period. "Dissatisfaction" and "bother" scores for the "after" condition were found to differ from those which would have been predicted from stable and unchanging conditions. A number of possible reasons for this finding are discussed and it is suggested that prediction of the effects resulting from noise reduction procedures requires further study.