OBJECTIVES: In economic evaluation it is assumed that utility is a linear function of longevity. The aim of this study was first to test the assumption of linearity, and second to test whether there exists a minimum threshold below which life extensions provide no marginal utility. METHODS: A representative sample of the population (n=2,400) was asked to imagine that they had a limited remaining lifetime (1 or 10 years) and were offered a treatment that would extend life by 1, 2, 4, 8, 16, 32 weeks (1 year perspective) or 2, 3, 8, 12, 32 or 52 weeks (10 year perspective). In each perspective, the price of treatment was constant per week of life extension. Additionally, respondents were asked about their maximum WTP for the life extension. RESULTS: The proportions that accepted the treatment offer increased monotonously with increasing extensions (1 year: 44%, 51%, 52%, 57%, 57%, 59%; 10 year: 53%, 57%, 60%, 62%, 63%, 75%) indicating a convex utility function (increasing marginal utility). The proportions with zero WTP for the treatment declined from 34% for one week life extension to 13% for 50 weeks. These results were confirmed in various econometric models. CONCLUSIONS: While most health care programmes offer life extensions of less than one year, considerable proportions of the population may be unwilling to pay anything for such gains. For life extensions up to one year, the utility function does not seem to be linear, but rather implies increasing marginal utility of longevity. The results of the study challenge conventional methods for economic evaluation.