The paper re-examines the issue of the appropriate unit for measuring output in cost utility analysis and the technique that will measure it. There are two main themes. The first is that utility, as it is often conceived and quantified, is not an appropriate basis for measurement. Consequently, a question arises concerning the selection of an appropriate unit of measurement. The second theme is that there is a need to establish criteria for the evaluation of measurement units. Four criteria are proposed which follow from commonly accepted social objectives and from the requirements of a measurement unit. It is concluded that, as judged by these criteria, the measurement units produced by the time trade-off and person trade-off (equivalence) techniques are more satisfactory than the units produced by the rating scale, magnitude estimation or the standard gamble.