OBJECTIVES: Visual analogue scales (VAS) are widely used to value health. The relative merits of VAS include minimal cognitive burden, ease of administration and low cost. Additionally, psychometric performance can be quite robust. However, the legitimacy of VAS-generated values in economic evaluation is often challenged despite recent literature that advocates a more sympathetic assessment. Integral to this debate is the question of whether VAS-based methods generate cardinal scales necessary for economic evaluation. This study reports on that issue. METHODS: A UK postal survey was used to elicit values for all 243 health states defined by EQ-5D. A factorial block design was used to construct 21 questionnaire versions in which a sample of 14 EQ-5D health states were valued by each respondent. Data were obtained from 825 participants. Health states were ranked in terms of their VAS scores on a within-respondent basis and analysed using a conditional logit model to obtain estimated values on an interval scale. The estimated values were compared with directly observed VAS values rescaled so that the best state (11111) was 100 and the worst state (33333) was 0. RESULTS: The estimated values derived from this analysis were as much as 10 points higher than the rescaled, directly observed VAS scores. Some differences in the rank order of health states were seen in the two sets of values. However, a generally linear pattern was seen between the computed and observed VAS values with a Pearson correlation coefficient as 0.969. OLS regression confirmed this relationship between estimated and observed VAS values with the estimated value = VAS value*1.079 + 5.527. CONCLUSION: Convergence between the two sets of values confirms the interval scale properties of the VAS value set, confirming the necessary cardinal measurement property to be legitimately used in economic evaluation