The person trade-off technique is a way of estimating the social value of different health care interventions. Basically it consists in asking people how many outcomes of one kind they consider equivalent in social value to X outcomes of another kind. The paper outlines a number of the author's previous studies using the technique. The studies suggest that while the technique is theoretically appealing for resource allocation purposes, it is in practice quite demanding. It needs to be applied in fairly large groups of subjects to keep random measurement error at an acceptable level. Possible framing effects include the effects of argument presentation and the choice of start points in numerical exercises. To control for these effects, it seems important to take subjects through a multistep procedure, in which they are induced to carefully consider the various arguments that might be relevant in each exercise and to reconsider initial responses in the light of their implications. The investigator must also think through which decision context he/she wishes to study and make his/her choice of context very clear when reporting the results.