Despite becoming increasingly common in evaluations of health care, different methods of quantitatively measuring health status appear to produce different valuations for identical descriptions of health. This paper reports on a study that elicited health state valuations from the general public using three different methods: the visual analogue scale (VAS), the standard gamble (SG) and the time trade-off (TTO). Two variants of the SG and TTO were tested: Props (using specially designed boards and cards); and No Props (using a self-completion booklet). This paper focuses on empirical relationships between health state valuations from the VAS and the (four) other methods. The relationships were estimated using Tobit regression of individual-level data. In contrast to a priori expectations, the mapping functions estimated suggest that differences are more pronounced across variant than across method. Furthermore, relationships with VAS scores are found to depend on the severity of the state: TTO Props valuations are higher than VAS responses for mild states and lower for more severe states; SG Props valuations are broadly similar to VAS scores over a wide range; and No Props responses are consistently higher than VAS valuations, particularly for more severe states. Explanations are proposed for these findings.