Violations of procedural invariance, epitomised by ‘classic’ preference reversals, have been observed for more than 40 years, and yet the study of this phenomenon in health remains nascent. This is an oversight because such violations pose a challenge to health economics, where choice and valuation methodologies often are used interchangeably. This article reports two experiments that aim to test for preference reversals over outcomes defined by health status, using both ‘open’ and ‘assisted’ valuation procedures. Although systematic preference reversals in the direction generally reported in the literature were not observed, the rates of non-systematic reversal were substantial, measuring 35–40%. By analysing the respondents’ explanations for their answers, it is clear that many of them used heuristics to answer the questions, and they may have good reasons to do so, which undermines the notion that preferences are always fixed and stable. These results, and others like it, pose a challenge to those who unquestioningly assume procedural invariance and maintain that population preferences should be used to inform health policy. Copyright © 2012 John Wiley & Sons, Ltd.