Tinnitus is discussed, with particular reference to the problem of assessing severity. The authors argue that tinnitus severity can only usefully be determined by measuring the impact of tinnitus on an individual, and therefore propose a scale to estimate severity in these terms. Data presented on 112 members of a tinnitus self-help group, demonstrated the reliability of a Subjective Tinnitus Severity Scale (S.T.S.S.), with a coefficient alpha of 0.84. This indicates a high degree of internal consistency, i.e.: statistically this scale is measuring aspects of a single dimension. The validity was established in a separate sample of 30 clinic attenders, where mean S.T.S.S. scores were found to correlate highly with two independent clinical ratings of severity (r2 = 0.76, p<0.001, and r2 = 0.73, p<0.001). Additionally, in these patients S.T.S.S. scores were significantly associated with several audiometric variables, although the correlations were of low magnitude.