While the EuroQOL instrument (EQ-5D) is being considered and used in clinical trials in the United States and Canada, and in large international multicentre studies, population weights for the instrument have never been established in North America. The primary purpose of this investigation was to derive a set of US-based population weights for the standard set of health states described in the EQ-5D health questionnaire. Valuations for EQ-5D health states were obtained via a postal survey using the visual analogue scale (VAS) format. A sample of 3,500 adults from the continental US were surveyed. A response rate of 25.8% was obtained. Mean and median valuations for 45 distinct health states, including dead and unconscious, were calculated. An ordinary least squares (OLS) regression-based weighting scheme was constructed to impute scores for states not directly valued in the questionnaire formats. Valuations for the standard EQ-5D health states displayed similar characteristics to VAS valuations obtained in previous investigations in European countries. The OLS model fit the observed data relatively well, achieving an adjusted R2 of 0.42. However, the diagnostic testing indicated that the initial model was misspecified. Subsequent alternative models alleviated some, but not all, of the problems of misspecification. The EQ-5D valuations from adult Americans in this sample appeared to behave in much the same fashion as in previous valuation studies. However, the generalisability of results to the entire adult American population may be limited. Violations of assumptions of the OLS regression model indicate the need for further investigation into the modelling technique used in deriving a single index score.