Human biologists have shown increasing interest in the use of cortisol as an objective marker of stress in recent years. This paper reviews both practical and theoretical problems raised by this approach. The methodology of cortisol assessment using urine, saliva, and plasma is considered. There are many practical problems involved in obtaining good measures of cortisol, although the availability of salivary assays offers a promising way forward. Theoretical issues are of greater concern. Results of laboratory studies led to the hypothesis that cortisol is elevated when an individual is distressed. However, relatively little work has been done on cortisol variation in “real life,” and the results of such studies do not, in general, conform to this hypothesis. It is concluded that the determinants of cortisol variation are not well understood. For example, cortisol level may be as much influenced by positive emotions as by negative emotions usually identified with stress. Further documentation of cortisol variation in everyday life is needed, and at present it is premature to use cortisol level as a marker of stress. © 1995 Wiley-Liss, Inc.