OBJECTIVES: There is some debate about whether utilities can exceed 1.0.  This assumption has important ramifications regarding which statistical methods are appropriate.  It is convention that utilities are anchored at 0.0 (death) and 1.0 (full health).  The descriptive health state corresponding to ‘full health’ differs across instruments.  The objective of this research is to examine the extent to which the value for ‘full health’ differs between the SF-6D and the EQ-5D and to explore the existence of a ceiling effect for utilities. METHODS: The 2000-2003 Medical Expenditure Panel Survey was used to estimate predicted SF-6D and EQ-5D (UK) scores using Tobit, CLAD and OLS regressions.  EQ-5D (and SF-6D) scores were regressed on SF-12 scores and sociodemographic characteristics.  The regression equations were then used to predict EQ-5D and SF-6D scores.  The two series of predicted scores were compared to examine ceiling and floor effects. RESULTS: 47% of individuals had a score of 1.0 on the EQ-5D and only 7% on the SF-6D.  The SF-6D exhibited significant floor effects with the lowest value at 0.344 compared to -0.594 for the EQ-5D.  Based on the Tobit predicted scores, a value of 1.0 on the SF-6D corresponded to a value of 1.2 on the EQ-5D.  This result suggests that a large portion of the 46% of individuals with a score of 1.0 on the EQ-5D would actually have utility scores greater than 1.0 on a scale without such a pronounced ceiling effect.  Likewise a 0.34 on the SF-6D corresponded to 0.03 on the EQ-5D, suggesting a floor effect for the SF-6D. CONCLUSIONS: Statistical estimation should incorporate censored regression methods to address ceiling and floor effects evident both theoretically and empirically.  Also, future descriptive systems and tariff valuation processes should incorporate values that exceed ‘full health’ (1.0) as is currently done for 'death' (0.0).