In a pilot study we investigate whether the inferences we draw about people's preferences towards health care treatments are altered if we vary the procedure that is used to elicit these preferences. In a conventional time trade-off (TTO) question, respondents express their preferences towards treatment by comparing a period of ill-health with a shorter period in a higher quality of life. In our less conventional TTO question, we vary the procedure by asking respondents their preferences towards treatment by comparing a period of ill-health with a longer period in a lower quality of life. The quantitative data are equivocal about whether preferences for treatment differ between the conventional and unconventional questions. The qualitative data support the notion of contrasting issues in the questions that involve prolonging time in a lower quality of life and appear to account for a failure to find quantitative differences in all of the questions. Copyright © 2002 John Wiley & Sons, Ltd.