Discrete choice experiments have the advantage that they can study preferences in health care where revealed preference data is not readily available. However, as a substitute for actual observed market led data, the experimental set-up for hypothetical situations must mimic the circumstances under which actual choices are made. One situation that a consumer/patient might face is an opt-out option. They might not choose to accept any of the positive actions available and as such will be a non-demander of the health care on offer. This paper explores issues raised in the modelling of such data within an experiment looking at women's preferences for cervical screening services. Copyright © 2003 John Wiley & Sons, Ltd.