Stability of noise annoyance

Noise has been shown to have negative health consequences and studies have linked these consequences in part to the subjective annoyance caused by the noise. Subjective noise annoyance is often measured using a 11-point numeric scale (e.g. annoyance scale in ISO 15666:2021 previously ISO/TS 15666:2003). Thereby the mean annoyance rating of a group of participants is commonly used as the default measure of the annoyance induced by a specific sound. However, the quality of these annoyance measures (which were developed only for the use in field studies) in terms of stability in a controlled laboratory setting has not been evaluated yet. The aim of the current study was therefore to evaluate the between- and within participant stability of noise annoyance, as well as the stability of the mean annoyance within a group of participants and between two groups of participants as well as between different experimental settings. The results revealed a low between- and a low within participant stability, but a high stability of the mean annoyance. The general sound context (other sounds included in the experiment) had an influence on the noise annoyance ratings. The results were used to develop a mathematical model of noise annoyance. According to this model, the mean annoyance is not the measure of central tendency of the tested group but the cleaned annoyance. The cleaned annoyance is thereby not an estimate of the actual annoyance a person is expected to experience in relation to a certain sound. It is rather a measure of the part of noise annoyance that is explained by the acoustical properties and the sound context. Our model can explain the results from the current study as well as results from previous studies. Recommendations are made concerning the evaluation of noise annoyance in laboratory studies and future development of the mathematical model.