Abstract A considerable number of reviews on community responses to noise have been carried out to compare dose–response relationships obtained from different noise sources and to investigate the effects of various factors on noise annoyance by using the data from different surveys. In order to compare the findings from various surveys precisely, it is very important to know how the different subjective or objective scales are transformed to unified scale. The present paper discusses the effect of four kinds of category scale on the annoyance response by using the data obtained from a social survey on community response to railway noise and compares the dose–response relationships between railway and road traffic noise obtained with the same scale. The extent of annoyance, such as % very annoyed, is strongly affected by the descriptors just below the annoyance range. This means that the descriptors are very important in constructing questionnaires and comparing the findings of different surveys. No systematic difference is found in dose–response relationships between railway and road traffic noises, using data obtained with the same method in the same area. This finding is quite different from those of European studies.