In the clinical and economic evaluation of health care, the value of benefit gained should be determined from a public perspective. The objective of this study was to establish relative valuations attached to different health states to form the basis for a 'social tariff' for use in quantifying patient benefit from health care. Three thousand three hundred and ninety-five interviews were conducted with a representative sample of the adult British population. Using the EuroQol health state classification and a visual analogue scale (VAS), each respondent valued 15 health states producing, in total, direct valuations for 45 states. Two hundred and twenty-one re-interviews were conducted approximately 10 weeks later. A near complete, and logically consistent, VAS data set was generated with good test-retest reliability (mean ICC = 0.78). Both social class and education had a significant effect, where higher median valuations were given by respondents in social classes III-V and by those with intermediate or no educational qualifications. These effects were particularly noticeable for more severe states. The use of such valuations in a social tariff raises important issues regarding the use of the VAS method itself to elicit valuations for hypothetical health states, the production of separate tariffs according to social class and/or education and the appropriate measure of central tendency.