OBJECTIVES: Differences in the evaluation of health states depending on the evaluation method are widely acknowledged. However, the impact of reference points on evaluation results have not been investigated so far. METHODS: In the framework of Prospect Theory, risk attitude with respect to quality of life can be taken as an indicator for reference points. Given a certain health state (here: Tinnitus), participants were asked whether they would undergo a treatment that could either improve or worsen their health condition, both with an equal chance. Five possible answers (in no case, unlikely, maybe, likely, in any case) allow five reference levels to be distinguished. Health evaluations of 210 Tinnitus patients and 210 unaffected persons were elicited. Time Tradeoff (TTO) and Standard Gamble (SG) evaluations were compared. Differences were adjusted with a one-parametric (a) power function. RESULTS: In accordance to the literature, the values assigned to the health state Tinnitus were higher with Standard Gamble than with Time Tradeoff (Patients: SG = 0.879, TTO = 0.827; Non-patients: SG = 0.807, TTO = 0.780). However, the parameter values a varied substantially depending on the reference level. Those Tinnitus patients who answered 'in no case' to a possible treatment exhibited on average an adjustment parameter of 0.398 (Non-patients: a = 0.626). However, participants who answered 'in any case' assigned with the SG method even lower values to Tinnitus than with the TTO method leading to an a of 1.102 (Non-patients: a = 1.254). CONCLUSION: Any adjustment procedure depends on underlying reference points. This result deserves attention in the framework of "Who's preferences count?" since patients and the general public evaluate health states with respect to different reference levels.