Many economists are becoming supportive of 'soft' paternalistic interventions that help people to avoid common decision errors without curtailing individual autonomy. To identify when such interventions could be beneficial, and to assess their success, requires a welfare criterion. However, traditional preference or choice-based criteria cannot serve this function because they assume that whatever people choose makes them better off. An alternative criterion that bases welfare on happiness rather than choice avoids this problem but has several of its own drawbacks. Most notably, people often adapt to serious chronic health conditions, and exhibit high levels of happiness, even though both those with and those without the condition agree that it is much preferable to be healthy. After reviewing different lines of research that shed light on the pros and cons of these alternative welfare criteria, we argue that no simple criterion based on either concept can surmount these problems. Instead, evaluations of welfare will inevitably have to be informed by a combination of both approaches, patched together in a fashion that depends on the specific context. © 2008.